Evaluating the performance of an artificial intelligence-based electronic reader for malaria rapid diagnostic tests across four sub-Saharan African countries

Evaluating the performance of an artificial intelligence-based electronic reader for malaria rapid diagnostic tests across four sub-Saharan African countries | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Evaluating the performance of an artificial intelligence-based electronic reader for malaria rapid diagnostic tests across four sub-Saharan African countries Kim A. Lindblade, Corine Ngufor, William Yavo, Sunday Atobatele, and 12 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6645811/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 30 Sep, 2025 Read the published version in Malaria Journal → Version 1 posted 12 You are reading this latest preprint version Abstract Background The introduction of malaria rapid diagnostic tests (RDTs) has expanded parasitologic confirmation of malaria at all levels of health systems in sub-Saharan Africa (SSA), improving case management and surveillance. However, concerns persist about healthcare worker adherence to results and the accuracy of results recorded in health facility registers. Electronic RDT readers have been proposed to improve the consistency of diagnosis and reporting, though their performance relative to expert human interpretation varies. We assessed the performance of the HealthPulse (Audere, Seattle, WA USA) smartphone application, an artificial intelligence (AI)-based RDT reader, across four countries in SSA. Methods In 2023, the Malaria Rapid Diagnostic Test Capture and Reporting Assessment (MaCRA) was implemented in health facilities in Benin, Côte d’Ivoire, Nigeria, and Uganda. Study staff collected images of malaria RDTs using the HealthPulse app after healthcare workers performed and interpreted the tests. A trained panel of external reviewers interpreted the RDT images, serving as the reference standard. RDTs were classified as positive, negative, invalid or indeterminate. We evaluated classification accuracy using recall, precision, and F1 scores (harmonic mean of recall and precision), and applied logistic regression to assess factors influencing AI performance across countries, RDT products, presence of faint lines and anomalies. Results Out of 110,843 RDT images collected, 110,231 (99.4%) were included in the analysis. The AI algorithm demonstrated high overall accuracy (96.8%) and a F1 score of 96.6% compared to panel interpretations. Recall and precision were >96% for positive and negative outcomes but much lower for invalid (recall: 84.5%; precision: 42.9%) and indeterminate classifications (recall: 0.7%; precision: 2.3%). AI performance varied by country, RDT product, and presence of faint lines. When test lines were faint, the OR of both positive recall (adjusted OR 0.01; 95% CI 0.00, 0.01) and negative recall (adjusted OR 0.20; 95% CI 0.11, 0.35) by the AI algorithm were reduced. Conclusions The HealthPulse AI algorithm demonstrated high agreement with a trained panel in interpreting malaria RDT images across diverse settings. However, reduced performance for invalid and indeterminate results and varying performance by country and RDT product highlights the need for further refinement. The HealthPulse app shows potential as a supportive tool in research and training. Malaria rapid diagnostic test artificial intelligence digital health diagnostic accuracy electronic RDT reader BACKGROUND The development of visually interpreted, clinical laboratory tests using immunoassay-based detection methods, known as rapid diagnostic tests (RDTs), has expanded access to diagnostic confirmation for many infectious diseases. These tests are now used not only in hospitals and primary care facilities but also at the community and household levels. Malaria RDTs were first developed in the early 1990s. However, widespread adoption followed only after the World Health Organization (WHO) and other agencies established standards for product and lot testing [ 1 ]. In 2010, with increasing availability of WHO-prequalified RDTs, WHO recommended that all suspected malaria cases be confirmed using a quality-assured parasitologic test, either microscopy or an RDT [ 2 ]. Since then, the use of malaria RDTs has increased from 20 million tests in 2010 to more than 328 million in 2023. In sub-Saharan Africa, RDTs now confirm nearly four times as many malaria infections as microscopy [ 3 ]. Malaria RDTs are easy to use and interpret, helping clinicians and community health workers diagnose and manage febrile patients more accurately. Their widespread use has improved malaria case management by reducing or eliminating presumptive treatment. RDTs have also enhanced the quality of malaria surveillance data by reducing misclassification of cases. The availability of malaria RDTs at the community level through community health workers has expanded access to parasitological confirmation in remote areas. Despite RDTs advantages, clinicians do not always adhere to RDT results when making treatment decisions [ 4 ]. While diagnostic tools are meant to guide care decisions, they do not override clinical judgment, and several factors may lead clinicians to diverge from RDT findings. For example, clinicians may doubt negative RDT results due to concerns about low-density infections falling below the test’s limit of detection [ 5 ]. Additionally, a common malaria antigen used in RDTs, the histidine-rich protein 2 (HRP2), can persist in the blood stream for days or weeks after infection has cleared, leading to false positives [ 6 ]. In other cases, parasites lacking the genes encoding HRP2 may go undetected [ 7 ]. Beyond biological considerations, clinicians may also face external pressures, including patient or caregiver demands for antimalarial treatment despite negative test results [ 8 ]. Although clinicians may have legitimate clinical grounds to disregard an RDT result when making their treatment decisions, overtreatment of patients with negative results wastes critical medicines and may delay appropriate care by preventing further diagnostic testing. Malaria-affected countries actively monitor overuse of antimalarial medicines through different approaches, including tracking the ratio of antimalarial treatments to confirmed cases over time and assessing adherence to negative RDT results in outreach supervision checklists [ 9 – 11 ] and end-user verification surveys [ 12 ]. As a result of the emphasis on reducing unnecessary use of antimalarials, some healthcare workers may intentionally misrecord negative RDT results as positive in health facility registers to justify antimalarial prescription. In a previous publication, we found that between 5.0% and 7.1% of RDT results across four sub-Saharan African countries were originally negative results but were misrecorded as positive in health facility registers [ 13 ]. Although some of these discrepancies could be due to unintentional errors, the fact that originally positive results were misrecorded as negative at a much lower rate (0.7–3.5%) suggests that at least some misrecording may be deliberate. In response to concerns over the accuracy of recorded RDT results and to facilitate rapid reporting, electronic readers of RDTs were developed to promote more consistent test interpretation and accurate and timely reporting [ 14 ]. RDT readers may be either a stand-alone, dedicated device or a digital application that uses the camera in a phone or tablet to interpret an RDT image. WHO produced a target product profile (TPP) for electronic RDT readers that includes operational and performance characteristics, specifying a minimum of ≥ 95% (optimal ≥ 98%) agreement between the electronic reader and expert, in-person visual interpretation of the test by a panel of skilled operators who directly view the RDT [ 15 ]. The WHO TPP also notes that an electronic reader is unlikely to improve on expert visual interpretation in terms of key metrics of diagnostic performance (sensitivity, specificity, limit of detection) and that the performance of electronic readers should be assessed specifically with faint lines (low positives), invalid, and indeterminate results. We evaluated the performance characteristics of the HealthPulse digital application (HealthPulse, Audere, Seattle WA USA) in a multi-country study by comparing results generated by its artificial intelligence (AI) algorithms to those interpreted by a panel of trained RDT readers who reviewed images of the RDTs. METHODS Malaria Rapid Diagnostic Test Capture and Reporting Assessment (MaCRA) We implemented the Malaria Rapid Diagnostic Test Capture and Reporting Assessment (MaCRA) in public health facilities in Benin, Côte d’Ivoire, Nigeria and Uganda in 2023. The methods have been described elsewhere in detail [ 13 ]. Briefly, study staff used the HealthPulse application on a smartphone to take images of malaria RDTs as soon as possible after the RDT was performed and interpreted by a healthcare worker (HCW). A trained external panelist reviewed each RDT image and recorded their interpretation. In the background of the HealthPulse application, an AI algorithm interpreted the result from the RDT image. Development of the HealthPulse app The HealthPulse digital application takes an image of an RDT cassette and uses an AI algorithm to interpret the result based on identification of the presence or absence of the control and one or more test lines. The app has an image quality assurance (IQA) component that leverages computer vision and machine learning processes to assess image quality. The app immediately flags images that do not meet quality standards (such as those with blur or skew) and prompts the user to retake the photo. However, if the photo is not retaken, the original image is used for interpretation. The HealthPulse app works on Android phones and can function either as a non-medical (the AI interpretation of the result is not shared with the user) or medical (the AI interpretation is shared with the user) electronic RDT reader. For the MaCRA study, the app was used as a non-medical RDT reader and the AI outcome was not shared with the study staff or HCWs during the course of the study nor was it used to inform clinical management of patients. The RDT products purchased by the ministries of health for public facilities in Benin, Côte d’Ivoire, Nigeria and Uganda were identified prior to the start of the MaCRA study and used to develop the AI algorithm (Supplementary Table 1). Sets of RDT images that were generated through laboratory production, synthetic imagery and collection of RDT cassettes in health facilities in South Africa were used to train the AI algorithm to interpret results in diverse conditions such as varying lighting, focal lengths and RDT misadministration. The HealthPulse AI algorithm was further trained on RDT images collected during the first two weeks of the study in each country; images and data from this period were not included in the analytical dataset. Collection of RDT images through the HealthPulse app Trained study staff collected RDT images over the course of implementing the MaCRA study, which is described more fully in a previous publication [ 13 ]. Briefly, RDTs performed in 16 health facilities over a four to six month period in each country were photographed by study staff using the digital HealthPulse application as soon as possible after a HCW administered the test and interpreted the result. The images, along with the RDT result recorded in the health facility register, basic data on the patient and a unique identifier for the HCW who recorded the RDT result, were automatically uploaded to a cloud server whenever internet connection was available. Study staff used Android-based smartphones that met minimum requirements prespecified by Audere for the HealthPulse app. All had rear cameras of at least 8 megapixels (MP); however, Côte d’Ivoire used a smartphone with a 50 MP camera and Uganda used one with a 64 MP camera. Interpretation of rapid diagnostic tests by a trained panel Metadata were removed from RDT images captured using the HealthPulse app before review by a trained panel. Panelists, trained on both high- and poor-quality images and tested for accuracy prior to employment, reviewed images independently and were physically separated to minimize bias. For each image, they recorded the presence or absence of the control and test lines, identified the RDT product, and flagged quality issues such as blur, glare, darkness, skew, multiple RDTs per image, and excess blood in the result window. Faint test lines suggestive of low-positive results were also flagged; however, panelists did not indicate which specific test line was faint. Each image interpreted by the panel was submitted to a quality control specialist who reviewed the image and passed or failed either the interpretation or the flagged conditions. Images with rejected interpretations or flagged conditions were sent back to the pool and reviewed by a new panelist. In a second quality control step, 30% of all images were randomly selected for review; rejected images re-entered the pool and were re-reviewed as with the first quality control step. At least the first 2200 images in each country were reviewed by three panelists to determine the interrater reliability. As measurement of interrater agreement using Fleiss’ kappa was between 0.998–1.0 for each country, indicating almost perfect agreement among the three panelists, we moved to a single reviewer for the remaining images to save costs. Data management and statistical analysis Data were cleaned to remove records with RDT products that had not been used to develop the AI model. For analyses stratified by RDT product, RDT products with fewer than 100 records were dropped. We evaluated the performance of the HealthPulse AI algorithm using the panel RDT result as the reference standard. Both the panel and the AI algorithm classified results into one of four categories: positive, negative, invalid or indeterminate. RDT products with two test lines were classified as positive if either line was present, and negative when both test lines were absent. Tests were considered invalid when the control line was not visible. Indeterminate results referred to cases where the test could not be interpreted due to an obstruction such as blood or another object obscuring the control or test lines, or when multiple RDTs appeared in the image. We categorized results using a multiclass confusion matrix with four outcomes [ 16 ]. The columns represent the true classifications according to the panel and the rows represent the AI classifications (Table 1 ). Each cell represents a true or false outcome with respect to the panel results. Table 1 Example of a multiclass confusion matrix comparing the panel results to the rapid diagnostic test outcomes predicted by the artificial intelligence algorithm AI classification Panel results Positive Negative Invalid Indeterminate Positive True positive (TP) False positive (FP) FP FP Negative False negative (FN) True negative (TN) FN FN Invalid False invalid (FIv) FIv True invalid (TIv) FIv Indeterminate False indeterminate (FId) FId FId True indeterminate (TId) AI: artificial intelligence; FId : False indeterminate ; FIv : False invalid ; FN : false negative FP : false positive ; TId: True indeterminate; TIv: True invalid; TN : true negative ; TP : true positive. We calculated several diagnostic accuracy metrics specific to measuring the performance of classification models in machine learning [ 17 , 18 ]. Recall is analogous to sensitivity and specificity for positive and negative outcomes, respectively, and measures the proportion of true outcomes that were correctly predicted by the HealthPulse AI algorithm (Table 2 ). Similarly, precision is similar to positive and negative predictive values (PPV and NPV) and reflects the likelihood that an outcome predicted by the AI algorithm is actually correct. The F1 score integrates precision and recall into a single metric to summarize model performance and will be high only if both the recall and precision are high, making it a suitable metric for evaluating models that need to balance between minimizing false positives and false negatives. The overall F1 score was calculated as the weighted average of the F1 scores for each outcome. Table 2 Formulas to calculate performance metrics for the artificial intelligence classification Metric Description Formula Accuracy The percentage of tests correctly predicted out of all the tests \(\:\frac{TP+TN+TIv+TId}{TP+TN+TIv+TId+FP+FN+FIv+FId}\:x\:100\) Recall The percentage of tests with a specific outcome that were correctly predicted by the AI algorithm (analogous to sensitivity and specificity for positive and negative outcomes, respectively) Positive tests : \(\:\frac{TP}{TP+FN+FIv+FId}\) x 100 Negative tests : \(\:\frac{TN}{TN+FP+FIv+FId}\) x 100 Invalid tests : \(\:\frac{TIv}{TIv+FP+FN+FId}\) x 100 Indeterminate tests : \(\:\frac{TId}{TId+FP+FN+FIv}\) x 100 Precision The percentage of tests predicted by the AI algorithm for a specific outcome that were classified with that outcome by the panel (analogous to positive predictive value and negative predictive value for positive and negative outcomes, respectively) Positive tests : \(\:\frac{TP}{TP+FP}\) x 100 Negative tests : \(\:\frac{TN}{TN+FN}\) x 100 Invalid tests : \(\:\frac{TIv}{TIv+FIv}\) x 100 Indeterminate tests : \(\:\frac{TId}{TId+FId}\) x 100 F1 score A weighted harmonic mean of recall and precision for each test outcome \(\:\frac{2\:\times\:Precision\:\times\:Recall}{Precision+Recall}\) x 100 AI: artificial intelligence; FId : false indeterminate ; FIv : false invalid ; FN : false negative FP : false positive ; TId: true indeterminate; Tiv: true invalid; TN : true negative ; TP : true positive. We used logistic regression to compare the odds of the HealthPulse AI algorithm correctly recalling positive cases (i.e. sensitivity) and negative cases (i.e. specificity) across countries, RDT products, presence of faint lines, blood in the window and photographic anomalies. Separate logistic regression models were constructed for data subsets containing only true positives or only true negatives using the glm function in R (R Foundation for Statistical Computing, Vienna, Austria) to measure association between various factors and accurate recall. We calculated unadjusted odds ratios with 95% confidence intervals (CIs), as well as adjusted odds ratios from multivariable models. RESULTS We collected 110,843 RDT images between June and December 2023. Of these, 612 images (0.6%) were discarded because the RDT product was not among those the AI algorithm had been trained to interpret. Of the 110,231 images remaining in the analytical data set, Benin contributed 36,411 (33.0%), Côte d’Ivoire 11,597 (10.5%), Nigeria 22,285 (20.2%) and Uganda 39,938 (36.2%) (Table 3 ). Table 3 Number and proportion of RDTs observed in the study by product type and other characteristics, 2023 Characteristic Overall n = 110,231 N (%) Country Faint line Benin n = 36,411 N (%) Côte d’Ivoire n = 11,597 N (%) Nigeria n = 22,285 N (%) Uganda n = 39,938 N (%) Yes n = 9714 N (%) No n = 100,517 N (%) RDT products AdvDx Malaria Pf 25,339 (23.0) 7602 (65.6) 17,734 (79.6) 3 (0.0) 2153 (22.2) 23,186 (23.1) Bioline Malaria Pf 56,888 (51.6) 36,194 (99.4) 3 (0.0) 189 (0.8) 20,502 (51.3) 3814 (39.3) 53,074 (93.3) Bioline Malaria Pf (HRP2/pLDH) * 204 (0.2) 204 (0.5) 61 (29.9) 143 (70.1) Bioline Malaria Pf Pan* 4641 (4.2) 4641 (11.6) 1003 (21.6) 3638 (78.4) CareStart Malaria Pf 31 (0.0) 31 (0.1) 14 (45.2) 17 (54.8) First Response Malaria Pf 10,697 (9.7) 3 (0.0) 3846 (33.2) 4331 (19.4) 2517 (6.3) 890 (8.3) 9807 (91.7) First Response Malaria Pf Ag (pLDH/HRP2) * 3504 (3.2) 3504 (8.6) 874 (24.9) 2630 (75.1) ParaHIT Malaria Pf 1142 (1.0) 214 (0.6) 145 (1.3) 783 (2.0) 124 (10.9) 1018 (89.1) Standard Q Malaria Pf 7785 (7.1) 1 (0.0) 7784 (19.5) 781 (10.0) 7004 (90.0) Faint line Yes 9714 (8.8) 1608 (4.4) 845 (7.3) 1938 (8.7) 5323 (13.3) No 100,517 (91.2) 34,803 (95.6) 10,752 (92.7) 20,347 (91.3) 34,615 (86.7) Blood present Yes 15,371 (13.9) 4463 (12.3) 1599 (13.8) 2451 (11.0) 6858 (17.2) No 94,860 (86.1) 31,948 (87.7) 9998 (86.2) 19,834 (89) 33,080 (82.8) Photo anomaly Yes 3002 (2.7) 655 (1.8) 411 (3.5) 765 (3.4) 1171 (2.9) No 107,229 (97.3) 35,756 (98.2) 11,186 (96.5) 21,520 (96.6) 38,767 (97.1) Image flagged for retake Yes 109,999 (99.8) 36,396 (100) 11,584 (99.9) 22,230 (99.8) 39,789 (99.6) No 232 (0.2) 15 (0.0) 13 (0.1) 55 (0.2) 149 (0.4) HRP2: histidine-rich protein 2; Pf: Plasmodium falciparum ; pLDH: Plasmodium lactate dehydrogenase; RDT: rapid diagnostic test *RDT products that include a pLDH line in addition to the P. falciparum -specific HRP2 line. There were nine RDT products identified during the course of the study (Table 3 ). All RDT products observed in the study were prequalified by WHO, although the CareStart Malaria Pf (Access Bio Inc, NJ USA) product has been delisted (Supplementary Table 1) [ 19 ]. The Bioline Malaria Pf (Abbott, IL USA) test was the most common RDT product in Benin and Uganda, while Côte d’Ivoire and Nigeria used predominantly AdvDx Malaria Pf (Advy Chemical, Mumbai, India) and First Response Malaria Pf (Premier Medical Corporation Ltd, Gujarat, India). Six RDT products have a single test line that detects Plasmodium falciparum -specific histidine-rich protein 2 (HRP2): AdvDx Malaria Pf, Bioline Malaria Pf, CareStart Malaria Pf, First Response Malaria Pf, ParaHIT (Arkray Healthcare Prvt Ltd, Mumbai, India) and Standard Q Malaria Pf (SD Biosensor, Gyeonggi-do, Republic of Korea). The remaining three RDTs include a second test line to detect Plasmodium lactate dehydrogenase (pLDH), an enzyme found in all Plasmodium spp that can also be used to differentiate species (Bioline Malaria Pf [HRP2/pLDH], Bioline Malaria Pf Pan and First Response Malaria Pf Ag [pLDH/HRP2]). The pLDH line in the Bioline Malaria Pf (HRP2/pLDH) RDT detects only P. falciparum while the other two tests use a pan- Plasmodium pLDH antigen that detects non-falciparum species as well. The proportion of RDTs with faint lines varied by country and ranged from 4.4% in Benin to 13.3% in Uganda. Faint lines were more common among the RDT products with a second pLDH test line (21.6–29.9%) compared to those a single HRP2 test line (6.7–10.9%), except for CareStart Malaria Pf where 45.2% (14/31) of the RDTs included in the study had faint lines. The presence of blood in the RDT well ranged from 11.0% in Nigeria to 17.2% in Uganda. Photo anomalies were more rare, affecting between 1.8% (Benin) and 3.5% (Côte d’Ivoire) of RDTs. Images were flagged for retake in 0.4% of RDTs in Uganda but less than 0.1% in Benin. Performance characteristics of the HealthPulse artificial intelligence algorithm The panel classified 56,434 (51.2%) RDT outcomes as positive, 52,963 (48.0%) as negative, 271 (0.2%) as invalid and 563 (0.5%) as indeterminate (Table 4 ). The AI algorithm predicted 56,404 (51.2%) outcomes as positive, 53,117 (48.2%) as negative, 534 (0.5%) as invalid and 176 (0.2) as indeterminate. Table 4 Confusion matrix with the RDT outcomes classified by the panel and the HealthPulse RDT reader artificial intelligence algorithm (N = 110,231) AI classification Panel results Total N (%) Positive N (%) Negative N (%) Invalid N (%) Indeterminate N (%) Positive 54,933 (97.3) 1336 (2.5) 15 (5.4) 120 (21.3) 56,404 (51.2) Negative 1300 (2.3) 51,446 (97.1) 21 (7.6) 350 (62.1) 53,117 (48.2) Invalid 126 (0.2) 90 (0.2) 229 (82.4) 89 (15.8) 534 (0.5) Indeterminate 75 (0.1) 91 (0.2) 6 (2.2) 4 (0.7) 176 (0.2) Total 56,434 (51.2) 52,963 (48.0) 271 (0.2) 563 (0.5) 110,231 AI: artificial intelligence. Overall, we found a high level of accuracy (96.8%) in the AI classification of RDT outcomes and the overall F1 score (96.6%) was similarly high (Table 5 ). The recall statistics for positive (97.3%) and negative (97.1%) outcomes indicated that a high proportion of the true positive and negative outcomes were identified by the AI algorithm. However, recall was lower for invalid (84.5%) and indeterminate outcomes (0.7%). Precision measures were high for both positive (97.4%) and negative (96.9%) outcomes but low for invalid (42.9%) and indeterminate (2.3%) outcomes, suggesting that outcomes classified as invalid or indeterminate by the AI algorithm were very likely to be incorrect. Table 5. Performance characteristics of the HealthPulse RDT reader artificial intelligence algorithm, 2023 Characteristic Accuracy % Overall F1 % Positive Negative Invalid Indeterminate Recall % Precision % Recall % Precision % Recall % Precision % Recall % Precision % Overall 96.8 96.6 97.3 97.4 97.1 96.9 84.5 42.9 0.7 2.3 Country Benin 97.1 96.6 98.4 96.7 97.1 97.5 85.7 20.3 0.3 25.0 Côte d’Ivoire 98.0 97.9 98.8 98.5 97.8 97.9 82.4 57.1 0.0 0.0 Nigeria 96.4 96.2 96.4 96.6 97.4 96.8 81.5 51.4 2.4 15.8 Uganda 96.4 96.5 96.4 98.0 96.8 96.0 87.0 40.7 0.0 0.0 RDT product Advdx Malaria Pf 97.0 96.7 97.5 96.9 97.4 97.3 84.2 32.0 0.8 7.7 Bioline Malaria Pf 97.0 96.7 97.5 97.4 97.6 96.7 86.4 26.4 0.3 25.0 Bioline Malaria Pf (HRP2/pLDH)* 99.0 99.0 99.3 99.3 98.4 98.4 ~ ~ ~ ~ Bioline Malaria Pf Pan* 92.3 94.1 95.8 95.3 88.7 98.6 90.0 11.1 0.0 0.0 First Response Malaria Pf 97.1 96.9 97.6 98.7 98.0 96.4 78.7 60.6 3.8 15.4 First Response Malaria Pf Ag (pLDH/HRP2)* 95.1 93.8 97.7 95.8 92.6 96.9 91.4 54.1 0.0 ~ ParaHIT Malaria Pf 95.6 95.3 96.9 95.0 95.6 96.2 ~ ~ 0.0 ~ Standard Q Malaria Pf 97.6 97.6 96.6 98.8 98.7 96.4 ~ 0.0 0.0 ~ Faint line Yes 85.6 90.8 86.0 99.6 68.6 6.0 71.4 18.0 0.0 0.0 No 97.9 97.7 99.7 97.0 97.2 99.2 86.0 49.4 0.8 2.8 Blood present Yes 91.7 89.7 95.8 92.5 93.2 92.0 84.1 51.1 0.4 12.5 No 97.6 97.7 97.6 98.2 97.7 97.6 87.5 19.9 5.7 1.2 Photo anomaly Yes 95.9 95.0 97.0 97.4 97.3 95.6 88.9 38.1 2.0 16.7 No 96.8 96.6 97.4 97.4 97.1 96.9 84.2 43.3 0.6 1.8 Image flagged for retake Yes 97.0 96.7 97.5 97.4 97.3 96.9 86.6 43.2 0.0 ~ No 20.3 30.6 22.2 95.7 17.7 80.0 22.2 25.0 36.4 2.3 HRP2: histidine-rich protein 2; Pf: Plasmodium falciparum ; pLDH: Plasmodium lactate dehydrogenase; RDT: rapid diagnostic test *RDT products that include a pLDH line in addition to the P. falciparum -specific HRP2 line. ~Could not be calculated. The AI scored above 96% in accuracy in all countries but only scored ≥ 98% in Côte d’Ivoire (Table 5 ). The F1 score was highest in Côte d’Ivoire (97.9%) and lowest in Nigeria (96.2%) but differences were minor. The performance of the HealthPulse AI algorithm was high across all RDT products with F1 scores ranging from 93.8–99.0%. The RDT product for which the AI performed the best overall was the Bioline Malaria Pf (HRP2/pLDH) test, which has two test lines. However, only 204 instances of this RDT product were included in the dataset. The AI algorithm had the worst performances with Bioline Malaria Pf Pan (F1 score: 94.1%) and First Response Malaria Pf Ag (pLDH/HRP2) (F1 score: 93.8%), the other two RDT products with two test lines. Restricting the analysis to data from Uganda, where multiple RDT products were used, did not alter the F1 scores, suggesting minimal bias from country-specific RDT usage patterns (data not shown). AI performance was lower when faint lines were present on the RDT. The F1 score was 97.7% in the absence of faint lines but dropped to 90.8% when faint lines were present (Table 5 ). The presence of blood in the RDT cassette similarly reduced performance, with the F1 score declining from 97.7–89.7%. Photographic anomalies affected accuracy, although to a lesser extent, with the F1 score decreasing from 96.6–95.0%. The AI performance was importantly reduced for images that were flagged for retake, with the F1 score declining from 96.7–30.6%. Odds of correct recall of positive and negative results The unadjusted odds of the AI correctly recalling a positive result differed significantly by country, RDT product, and the presence of faint lines. Compared with Benin, the odds of positive recall were higher in Côte d’Ivoire (OR 1.77, 95% CI 1.33, 2.33) but lower in Nigeria (OR 0.49, 95% CI 0.41, 0.57) and Uganda (OR 0.52, 95% CI 0.45, 0.59) (Table 6 ). Using AdvDx Malaria Pf as the reference RDT, the odds of correct positive recall were higher for Bioline Malaria Pf Pan (OR 3.21, 95% CI 1.99, 5.19), First Response Malaria Pf Ag (pLDH/HRP2) (OR 3.04, 95% CI 1.86, 4.96), and First Response Malaria Pf (OR 1.40, 95% CI 1.11, 1.75), and lower for Standard Q Malaria Pf (OR 0.72, 95% CI 0.58, 0.89). No statistically significant differences were observed for Bioline Malaria Pf, Bioline Malaria Pf (HRP2/pLDH), or ParaHIT Malaria Pf compared to the reference RDT product. The odds of correct positive recall were significantly reduced in the presence of a faint line (OR 0.01, 95% CI 0.00, 0.01) or blood in the result window (OR 0.65, 95% CI 0.57, 0.75), but were not significantly affected by the presence of photographic anomalies (OR 1.04, 95% CI 0.74, 1.47) or images flagged for retake (OR 1.93, 95% CI 0.26, 14.31). Table 6 Unadjusted and adjusted odds ratios for the HealthPulse RDT reader artificial intelligence algorithm to correctly recall positive results by characteristics of the RDTs (n = 56,419) Characteristic N Unadjusted Adjusted* Odds ratio (95% CI) Odds ratio (95% CI) Country Benin 18,004 Ref. Ref. Côte d’Ivoire 6873 1.77 (1.34, 2.33) 0.99 (0.61, 1.62) Nigeria 9423 0.49 (0.41, 0.57) 0.39 (0.25, 0.61) Uganda 22119 0.51 (0.45, 0.59) 0.80 (0.68, 0.95) RDT product AdvDx Malaria Pf 11,649 Ref. Ref. Bioline Malaria Pf 29053 0.95 (0.82, 1.09) 0.35 (0.23, 0.55) Bioline Malaria Pf (HRP2/pLDH) ** 143 3.45 (0.48, 24.76) 4.92 (0.65, 37.24) Bioline Malaria Pf Pan** 2481 3.21 (1.99, 5.18) 4.54 (2.40, 8.59) First Response Malaria Pf 6314 1.40 (1.12, 1.75) 0.82 (0.62, 1.08) First Response Malaria Pf Ag (pLDH/HRP2) ** 2173 3.03 (1.86, 4.96) 4.59 (2.40, 8.78) ParaHIT Malaria Pf 524 0.77 (0.46, 1.29) 0.62 (0.31, 1.21) Standard Q Malaria Pf 4082 0.72 (0.58, 0.89) 0.44 (0.28, 0.71) Faint line Yes 9526 0.01 (0.00, 0.01) 0.01 (0.00, 0.01) No 46,893 Ref. Ref. Blood present Yes 7432 0.65 (0.57, 0.75) 0.64 (0.54, 0.75) No 48,987 Ref. Ref. Photo anomaly Yes 1544 1.04 (0.74, 1.47) No 54,875 Ref. Image flagged for retake Yes 56,320 1.93 (0.26, 14.31) No 99 Ref. CI: confidence interval; HRP2: histidine-rich protein 2; Pf: Plasmodium falciparum ; pLDH: Plasmodium lactate dehydrogenase; Ref.: reference; RDT: rapid diagnostic test *The model included all variables except photo anomaly and whether it was flagged for retake. **RDT products that include a pLDH line in addition to the P. falciparum -specific HRP2 line. In the adjusted model, which included RDT product, faint lines, and presence of blood, the odds of correct positive recall remained significantly lower in Nigeria (OR 0.39, 95% CI 0.25, 0.61) and Uganda (OR 0.80, 95% CI 0.68, 0.95) compared with Benin (Table 6 ). No significant difference was observed between Benin and Côte d’Ivoire (OR 0.99, 95% CI 0.61, 1.62). In the adjusted model, the AI was more likely to correctly recall a positive test with Bioline Malaria Pf Pan (OR 4.54, 95% CI 2.40, 8.59) and First Response Malaria Pf Ag (pLDH/HRP2) (OR 4.59, 95% CI 2.40, 8.78) compared with AdvDx Malaria Pf, and less likely with Bioline Malaria Pf (OR 0.35, 95% CI 0.23, 0.55) and Standard Q Malaria Pf (OR 0.44, 95% CI 0.28, 0.71). The ORs for the presence of a faint line or blood were not affected by controlling for country and RDT product. The unadjusted odds of correctly recalling negative test results were higher in Côte d’Ivoire (OR 1.35, 95% CI 1.09, 1.68), Nigeria (OR 1.20, 95% CI 1.04, 1.39), and Uganda (OR 1.23, 95% CI 1.08, 1.41) compared with Benin (Table 7 ). Correct recall of negative tests was more likely for First Response Malaria Pf (OR 1.61, 95% CI 1.24, 2.10) and Standard Q Malaria Pf (OR 2.04, 95% CI 1.50, 2.77) and less likely for Bioline Malaria Pf Pan (OR 0.45, 95% CI 0.36, 0.56), First Response Malaria Pf Ag (pLDH/HRP2) (OR 0.37, 95% CI 0.29, 0.48), and ParaHIT Malaria Pf (OR 0.57, 95% CI 0.38, 0.85) relative to AdvDx Malaria Pf. As with positive results, the odds of correctly recalling negative tests were significantly reduced when a faint line was present (OR 0.12, 95% CI 0.07, 0.20) or when blood was present (OR 0.30, 95% CI 0.27, 0.34). Photo anomalies did not significantly affect performance (OR 1.18, 95% CI 0.82, 1.70), and the odds ratio for images flagged for retake could not be calculated. Table 7 Unadjusted and adjusted odds ratios for the HealthPulse RDT reader artificial intelligence algorithm to correctly recall negative results by characteristics of the RDTs (n = 52,947) Characteristic N Unadjusted Adjusted* Odds ratio (95% CI) Odds ratio (95% CI) Country Benin 18,102 Ref. Ref. Côte d’Ivoire 4652 1.35 (1.09, 1.68) 2.25 (1.28, 3.95) Nigeria 12,612 1.20 (1.04, 1.39) 2.09 (1.19, 3.65) Uganda 17,581 1.23 (1.08, 1.41) 1.93 (1.61, 2.32) RDT product AdvDx Malaria Pf 13,549 Ref. Ref. Bioline Malaria Pf 27,502 1.07 (0.94, 1.22) 1.99 (1.14, 3.47) Bioline Malaria Pf (HRP2/pLDH) ** 61 1.57 (0.22, 11.39) 1.54 (0.20, 11.98) Bioline Malaria Pf Pan** 2135 0.45 (0.36, 0.56) 0.47 (0.26, 0.83) First Response Malaria Pf 4203 1.61 (1.24, 2.10) 1.81 (1.35, 2.41) First Response Malaria Pf Ag (pLDH/HRP2) ** 1185 0.37 (0.29, 0.48) 0.65 (0.36, 1.18) ParaHIT Malaria Pf 611 0.57 (0.38, 0.85) 0.81 (0.43, 1.53) Standard Q Malaria Pf 3701 2.04 (1.50, 2.77) 2.50 (1.35, 4.61) Faint line Yes 114 0.12 (0.07, 0.20) 0.20 (0.11, 0.35) No 52,833 Ref. Ref. Blood present Yes 7156 0.30 (0.27, 0.34) 0.31 (0.27, 0.34) No 45,791 Ref. Ref. Photo anomaly Yes 1390 1.18 (0.82, 1.70) No 51,557 Ref. Image flagged for retake‡ Yes 52,834 Not estimated No 113 Ref. CI: confidence interval; HRP2: histidine-rich protein 2; Pf: Plasmodium falciparum ; pLDH: Plasmodium lactate dehydrogenase; Ref.: reference; RDT: rapid diagnostic test *The model included all variables except photo anomaly and whether it was flagged for retake. **RDT products that include a pLDH line in addition to the P. falciparum -specific HRP2 line. ‡In the subset of true negative RDTs, the AI algorithm did not classify any tests that failed the photo filters as positive. Consequently, the odds ratio could not be estimated. In the adjusted model, there were no meaningful changes in the ORs for correct recall of negative results by country or the presence of faint lines. However, negative recall was statistically more likely for Bioline Malaria Pf (OR 1.99, 95% CI 1.14, 3.47), while First Response Malaria Pf Ag (pLDH/HRP2) was no longer significantly associated with poor negative recall (OR 0.65, 95% CI 0.36, 1.18). DISCUSSION We found that the HealthPulse application, an electronic malaria RDT reader powered by an AI algorithm, demonstrated a high degree of accuracy (> 96%) when compared to a trained panel reviewing images of the RDTs. The AI-predicted RDT results showed high positive and negative recall (i.e. sensitivity and specificity, respectively) and high positive and negative precision (i.e. PPV and NPV, respectively) relative to the panel. However, the AI’s recall and precision were notably lower for results the panel identified as invalid or indeterminate. Since invalid and indeterminate results were rare, the AI’s F1 scores remained high, ranging from 96.2–97.9% across Benin, Côte d’Ivoire, Nigeria and Uganda. In multivariable models, positive recall by the AI algorithm was significantly lower for images collected in Nigeria and Uganda compared to Benin and Côte d’Ivoire. Although lighting conditions within health facilities may have contributed to differences in AI performance, they are unlikely to fully account for the observed country-level variations. Differences in smartphone camera performance may have played a role: while all devices had at least 8-megapixel resolution, variability in features such as low-light performance of cameras could have influenced the algorithm’s ability to detect test and control lines. As each country used a different smartphone model, we were unable to separate the effects of country from those of the specific device. The HealthPulse AI algorithm’s recall also varied across RDT products. Consistently, the AI algorithm performed best with the Bioline Malaria Pf (HRP2/pLDH) product and worst with the Bioline Malaria Pf Pan test. We could not find a consistent feature to explain this variability. The HealthPulse AI algorithm was trained on these RDT products at a specific test positivity rate (~ 25%) to balance positive and negative recall and achieve an F1 score of at least 95%. Variation between the test positivity rates in the training data and those in this study (~ 50%), and variability by RDT product due to use patterns by country, may have affected the algorithm’s performance. Nonetheless, the F1 score exceeded 96% for six out of the nine RDT products included in this study. The presence of faint lines was associated with a significant reduction in both positive and negative recall by the AI algorithm. Faint lines were more frequently observed on RDT products that included a pLDH line. Although we could not determine whether the faint line corresponded specifically to the pLDH target, pLDH is typically present at lower concentrations in the blood than HRP2, resulting in weaker or fainter test lines on RDTs [ 20 , 21 ]. After adjusting for the presence of faint lines, the OR for RDT products containing a pLDH line increased, improving the AI algorithm’s positive recall in comparison to AdvDx as a reference, while the OR for RDT products not containing a pLDH line decreased. With respect to negative recall, there was no clear pattern of changes after adjustment. Because RDTs with pLDH lines also include an HRP2 test line, it is likely that the AI’s improved positive recall after adjustment for faint lines was influenced by the availability of additional visual evidence of a positive result. While multiple RDT readers have been assessed in prior studies, cross-study comparisons of performance are difficult due to variations in use cases and evaluation methodologies. One laboratory-based study compared five RDT readers using samples with known parasite concentrations, including negative samples [ 20 ]. They found that the human eye was better at detecting positive lines at low parasite densities (i.e. faint lines) compared to the RDT readers, though most RDT readers performed well with parasite densities typically associated with clinical illness (i.e. ≥ 100 parasites per µl). Other studies comparing the human eye to RDT readers have reported high levels of agreement. In Tanzania, agreement between an RDT reader and visual inspection was 98.4% (95% CI 97.6, 99.0) among febrile patients using the Bioline Malaria Pf/Pan RDT [ 22 ]. Using the same RDT product among Ugandan children selected from the community, agreement with visual inspection was 98.9% (95% CI 93.2, 99.8) [ 23 ]. Originally developed to improve the consistency and accuracy of RDT interpretation, digital RDT readers gained momentum during the COVID-19 pandemic alongside the expansion of in-home testing. Over the past 15 years, malaria case management and surveillance have improved substantially due to the widespread use of RDTs [ 3 ]. However, persistent challenges, including nonadherence to negative RDT results and misrecording of RDT outcomes in facility registers, underscore the continued need for data quality assessments and sustainable strategies to enhance the accuracy of RDT reporting [ 13 ]. Although RDT readers have been proposed as a potential solution, limited incentives exist for healthcare workers in public health settings to adopt these tools unless they reduce clinical workload. In the private sector, RDT readers have been explored as a means to improve data quality and to provide verification of test results to support reimbursement and subsidy processes. However, the performance standards required for such applications remain to be clearly defined [ 25 , 26 ]. RDT readers may also serve as valuable training tools or job aids, but improvements in their ability to detect invalid and indeterminate results are needed. Research settings that require consistent and objective interpretation of RDT results represent another potential use case for high-performing RDT readers, such as HealthPulse. This study has several limitations. First, the comparator used to evaluate the HealthPulse AI algorithm was a trained human panel interpreting images of RDTs, whereas the WHO TPP requires expert interpretation of the physical RDT cassette immediately after the incubation period as the reference standard if the reader is to be used as a medical device. While this reference standard is likely to be more accurate, it is less feasible for large-scale studies. Importantly, because we did not apply the WHO-recommended comparator, we do not make conclusions regarding the suitability of the HealthPulse app as a clinical or medical RDT reader. Despite robust quality control processes to ensure the accuracy of the human panel’s interpretations, some misclassification may have occurred. Any such errors would likely reduce associations between predictors and AI performance outcomes. In conclusion, we found that the HealthPulse RDT reader AI algorithm accurately interpreted malaria RDT results when compared to a trained panel reviewing images of RDTs. The HealthPulse AI algorithm exhibited a high F1 score, balancing recall and precision. However, several factors may have influenced the algorithm’s performance, including the country setting, the type of smartphone used, the RDT product and presence of faint lines, warranting further investigation. While the HealthPulse AI shows promise for use in research or training contexts, improvements are needed, particularly in the detection of invalid and indeterminate results. Abbreviations AI Artificial intelligence CI Confidence interval HCW Healthcare worker HRP2 Histidine-rich protein 2 IQA Image quality assurance N Number OR Odds ratio Pf Plasmodium falciparum pLDH Parasite lactate dehydrogenase PMI President’s Malaria Initiative RDT Rapid diagnostic test WHO World Health Organization Declarations ACKNOWLEDGEMENTS We are grateful to the Audere team (Shawna Cooper, Sasha Frade and Sam Smedinghoff) for providing technical support related to the HealthPulse application. Audere had no involvement in the analysis or interpretation of the data or the conclusions presented in this study. We acknowledge the contributions of Audere’s AI data creation and labeling teams at the Centre for HIV-AIDS Prevention Studies (South Africa) and Indivillage (India). We are grateful to the HCWs who gave generously of their time to participate in this evaluation. Many staff members of the organizations implementing the study in country made important contributions: Manfred Accrombessi, Corneille Hueha (CREC, Benin); M. Anatole Mian, Orphée Kangah-Kouakou, Valérie Bedia-Tanoh (INSP, Côte d’Ivoire); Hilary Okagbue, Evelyn Orya, Shiva Gab-Deedam, Olufisayo Bademosi (Sydani Group, Nigeria); Anne Katahoire, Jane Frances Namuganga, Jenipher Musoke (CHDC, Uganda). We are grateful to members of the national malaria programs in each country for their support: Cyriaque Affoukou and Julien Aissan (Benin); Jacques Agnon (Côte d’Ivoire); Onyebuchi Okoro, Chukwu Okoronkwo and Nnenna Ogbulafor (Nigeria); and Bosco Agaba, Catherine Maiteki Sebuguzi and Gerald Rukundo (Uganda). We received excellent research support from Saadjo Sow (PMI Insights). Megan Littrell, Taj Munson and Kim Vu provided overall direction and administrative support to this and the other PMI Insights projects. Aysu Uygur (The Gates Foundation) is greatly appreciated for her contributions during the design of the evaluation. Thank you to the PMI staff in country: Raoul Oloukoi, Virgile Gnanguenon (Benin); Pascal Zinzindohoue, Patricia Yepassis-Zembrou, Blaise Kouadio, Melaine Tape, Yao Stephane, Christie Billingsley (Côte d’Ivoire); Jules Mihigo, Cassandra Elagbaje, Veronica Momoh, Valerie Bampoe (Nigeria); Edgar Agba, Grace Appiah, Patrick Condo (Uganda). FUNDING This evaluation was co-funded by PMI Insights and the Bill & Melinda Gates Foundation (INV-043942). PMI Insights was the global operational research and program evaluation project of the U.S. President’s Malaria Initiative (PMI). Funding for this evaluation is made possible by the generous support of the American people through the United States Agency for International Development (USAID) through cooperative agreement No. 7200AA20CA00031. The contents are the responsibility of the authors and do not necessarily reflect the views of USAID or the United States Government. AUTHOR INFORMATION Authors and Affiliations PMI Insights Project/PATH, Geneva, Switzerland Kim A. Lindblade Centre de Recherche Entomologique de Cotonou, Cotonou, Benin Idelphonse Ahogni Corine Ngufor Institut National de Santé Publique, Abidjan, Côte d’Ivoire Abibatou Konate-Toure William Yavo Sydani Group, Abuja, Nigeria Ese Akpiroroh Sunday Atobatele Child Health and Development Centre, Makerere University, Kampala, Uganda Arthur Mpimbaza Nelson Ssewante U.S. President’s Malaria Initiative, USAID, Washington, DC USA Kevin Griffith Michael Humes Programme National de Lutte contre le Paludisme, Cotonou, Benin Augustin Kpemasse Programme National de Lutte contre le Paludisme, Abidjan, Côte d’Ivoire Antoine Tanoh National Malaria Elimination Programme, Abuja, Nigeria Godwin Ntadom National Malaria Control Division, Ministry of Health, Kampala, Uganda Jimmy Opigo Contributions KL, MH and KG conceived and designed the evaluation. AK, AKT, AM, AMT, AK-T, CN, EA, GN, IA, JO, NS, SA, and WY oversaw data collection activities. KL and SZ drafted the manuscript. KL analyzed the data. All authors critically reviewed the manuscript. All authors read and approved the final manuscript. Corresponding author Correspondence to: Kim A. Lindblade, Rue du Grand-Pré, 1202 Geneva, Switzerland ; [email protected] . Ethical approval and consent to participate Ethical approval was obtained from: the Comité National d’Ethique pour la Recherche en Santé (Benin); the Comité National d’Ethique des Sciences de la Vie et de la Santé (Côte d’Ivoire); the National Health Research Ethics Committee (Nigeria); the Oyo State Health Research Ethics Committee (Nigeria); the Sokoto State Health Research Ethics Committee (Nigeria); the Uganda National Council on Science and Technology (Uganda); the Vector Control Division Research and Ethics Committee (Uganda); and the WGC IRB in the USA. All participants provided written, informed consent to participate. Consent for publication Not applicable. Availability of data and materials The datasets used and/or analyzed during the current evaluation can be provided by the corresponding author on reasonable request. Competing interests The authors declare that they have no competing interests. References Cunningham J, Jones S, Gatton ML, Barnwell JW, Cheng Q, Chiodini PL, et al. A review of the WHO malaria rapid diagnostic test product testing programme (2008–2018): performance, procurement and policy. Malar J. 2019;18:387. WHO. Guidelines for the treatment of malaria. Second edition. [Internet]. Geneva, Switzerland: World Health Organization; 2010. Available from: https://www.paho.org/sites/default/files/TreatmentGuidelines-2nd-ed-2010-eng.pdf WHO. World Malaria Report 2024. Geneva, Switzerland: World Health Organization; 2024. Kabaghe AN, Visser BJ, Spijker R, Phiri KS, Grobusch MP, van Vugt M. Health workers’ compliance to rapid diagnostic tests (RDTs) to guide malaria treatment: a systematic review and meta-analysis. Malar J. 2016;15:163. Wu L, van den Hoogen LL, Slater H, Walker PGT, Ghani AC, Drakeley CJ, et al. Comparison of diagnostics for the detection of asymptomatic Plasmodium falciparum infections to inform control and elimination strategies. Nature. 2015;528:S86–93. Ntuku H, Whittemore B, Dausab L, Jang IK, Golden A, Sheahan W, et al. Post-treatment duration of positivity for standard and ultra-sensitive Plasmodium falciparum antigen-based rapid diagnostic tests, a cohort study from a low-endemic setting in Namibia. EBioMedicine. 2025;111:105489. Berhane A, Anderson K, Mihreteab S, Gresty K, Rogier E, Mohamed S, et al. Major Threat to Malaria Control Programs by Plasmodium falciparum Lacking Histidine-Rich Protein 2, Eritrea. Emerg Infect Dis. 2018;24:462–70. Altaras R, Nuwa A, Agaba B, Streat E, Tibenderana JK, Strachan CE. Why do health workers give anti-malarials to patients with negative rapid test results? A qualitative study at rural health facilities in western Uganda. Malar J. 2016;15:23. Altaras R, Worges M, La Torre S, Audu BM, Mwangi G, Zeh-Meka A, et al. Outreach Training and Supportive Supervision for Quality Malaria Service Delivery: A Qualitative Evaluation in 11 Sub-Saharan African Countries. Am J Trop Med Hyg. 2024;110:20–34. Agbemafle EE, Addo-Lartey A, Odikro MA, Frimpong JA, Kubio C, Ameme DK, et al. Adherence to the test, treat and track strategy for malaria control among prescribers, Mfantseman Municipality, Central Region, Ghana. PLOS ONE. 2023;18:e0279712. Koko D, Arouna D, Bernard Y-M, Ba T, Mostel J, Abdou Y, et al. How Outreach Training and Supportive Supervision (OTSS) Affect Health Facility Readiness and Health-Care Worker Competency to Prevent and Treat Malaria in Niger: A Secondary Analysis of OTSS Data. Am J Trop Med Hyg. 2024;110:50–5. USAID Global Health Supply Chain Program. EUV Survey Question Guide [Internet]. Washington, DC: USAID Global Health Supply Chain Program; 2020. Available from: https://www.ghsupplychain.org/euv-survey-question-guide Lindblade, KA, Mpimbaza, A, Ngufor, C, Yavo, W, Atobatele, S, Akpiroroh, E, et al. Assessing the accuracy of the recording and reporting of malaria rapid diagnostic test results in four African countries: Methods and key results. under review. White, W., Korir, G. Landscape of RDT-reading apps [Internet]. Geneva, Switzerland: FIND; 2023 Jan. Available from: https://www.finddx.org/wp-content/uploads/2023/02/20230203_rep_market_rdt_reading_app_FV_EN.pdf WHO. Target product profile for readers of rapid diagnostic tests [Internet]. Geneva, Switzerland: World Health Organization; 2023. Available from: https://iris.who.int/bitstream/handle/10665/365980/9789240067172 Confusion Matrix - an overview | ScienceDirect Topics [Internet]. [cited 2024 Nov 29]. Available from: https://www.sciencedirect.com/topics/computer-science/confusion-matrix Hicks SA, Strümke I, Thambawita V, Hammou M, Riegler MA, Halvorsen P, et al. On evaluation metrics for medical applications of artificial intelligence. Sci Rep. 2022;12:5979. Saito T, Rehmsmeier M. The Precision-Recall Plot Is More Informative than the ROC Plot When Evaluating Binary Classifiers on Imbalanced Datasets. PLOS ONE. 2015;10:e0118432. Prequalified In Vitro Diagnostics | WHO - Prequalification of Medical Products (IVDs, Medicines, Vaccines and Immunization Devices, Vector Control) [Internet]. [cited 2024 Oct 24]. Available from: https://extranet.who.int/prequal/vitro-diagnostics/prequalified-vitro-diagnostics Visser T, Ramachandra S, Pothin E, Jacobs J, Cunningham J, Menach AL, et al. A comparative evaluation of mobile medical APPS (MMAS) for reading and interpreting malaria rapid diagnostic tests. Malar J. 2021;20:39. Martiáñez-Vendrell X, Jiménez A, Vásquez A, Campillo A, Incardona S, González R, et al. Quantification of malaria antigens PfHRP2 and pLDH by quantitative suspension array technology in whole blood, dried blood spot and plasma. Malar J. 2020;19:12. Shekalaghe S, Cancino M, Mavere C, Juma O, Mohammed A, Abdulla S, et al. Clinical performance of an automated reader in interpreting malaria rapid diagnostic tests in Tanzania. Malar J. 2013;12:141. Oyet C, Roh ME, Kiwanuka GN, Orikiriza P, Wade M, Parikh S, et al. Evaluation of the Deki Reader TM , an automated RDT reader and data management device, in a household survey setting in low malaria endemic southwestern Uganda. Malar J. 2017;16:449. Kalinga AK, Mwanziva C, Chiduo S, Mswanya C, Ishengoma DI, Francis F, et al. Comparison of visual and automated Deki Reader interpretation of malaria rapid diagnostic tests in rural Tanzanian military health facilities. Malar J. 2018;17:214. CHAI. Case Study: Scoping digital solutions for improving quality of care in the informal private sector [Internet]. Clint. Health Access Initiat. 2022 [cited 2025 Apr 9]. Available from: https://www.clintonhealthaccess.org/case-study/case-study-scoping-digital-solutions-for-improving-quality-of-care-in-the-informal-private-sector/ van Duijn SMC, Siteyi AK, Smith S, Milimo E, Stijvers L, Oguttu M, et al. Connected diagnostics to improve accurate diagnosis, treatment, and conditional payment of malaria services in Kenya. BMC Med Inform Decis Mak. 2021;21:233. WHO. WHO prequalification of in vitro diagnostics programme: AdvDx Malaria Pf Rapid Malaria Ag Detection Test [Internet]. Geneva, Switzerland: WHO; 2019. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0345-101-00_AdvDxMalariaDetectionTest_v2.pdf WHO. WHO prequalification of diagnostics programme: Bioline Malaria Ag P.f [Internet]. Geneva, Switzerland; 2021. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0031-012-01_BiolineMalaria_Ag_P-f-v7.pdf WHO. WHO prequalification of in vitro diagnostics: First Response Malaria Antigen P. falciparum (HRP2) Card Test [Internet]. Geneva, Switzerland: WHO; 2018. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0283-010-00_FirstResponseMalaria_v1.pdf WHO. WHO prequalification of in vitro diagnostics programme: ParaHIT f ver. 1.0 rapid test for P. falciparum malaria device [Internet]. Geneva, Switzerland: WHO; 2016. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0062-023-00_ParaHIT-f_RapidTest_v3_0.pdf WHO. WHO prequalification of in vitro diagnostics: STANDARD Q Malaria P.f Ag Test [Internet]. Geneva, Switzerland: WHO; 2021. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0346-117-00_STANDARD-Q-MalariaP-fAg_Test_PQ_v2.0.pdf WHO. WHO prequalificaiton of in vitro diagnostics programme: Bioline Malaria Ag P.f (HRP2/pLDH) [Internet]. Geneva, Switzerland: World Health Organization; 2025. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/bioline-malaria-ag-p.f-hrp2-pldh-pqdx-0209-012-00-public-report-v-7.0.pdf WHO. WHO prequalification of in vitro diagnostics programme: Bioline Malaria Ag P.f/Pan [Internet]. Geneva, Switzerland: WHO; 2022. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0030-012-01_BiolineMalariaAgPfPan_v11.0.pdf WHO. WHO prequalification of in vitro diagnostics: First Response Malaria Ag.pLDH HRP2 Combo Card Test [Internet]. Geneva, Switzerland: World Health Organization; 2025. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/First%20Response%20Malaria%20Ag.pLDH%20HRP2%20Combo%20Card%20Test_PQDx%200285-010-00_v2.pdf Additional Declarations No competing interests reported. Supplementary Files SupplementaryTable1.docx Cite Share Download PDF Status: Published Journal Publication published 30 Sep, 2025 Read the published version in Malaria Journal → Version 1 posted Editorial decision: Revision requested 10 Jun, 2025 Reviews received at journal 10 Jun, 2025 Reviews received at journal 06 Jun, 2025 Reviewers agreed at journal 30 May, 2025 Reviews received at journal 26 May, 2025 Reviewers agreed at journal 22 May, 2025 Reviewers agreed at journal 22 May, 2025 Reviewers agreed at journal 21 May, 2025 Reviewers invited by journal 21 May, 2025 Editor assigned by journal 13 May, 2025 Submission checks completed at journal 13 May, 2025 First submitted to journal 12 May, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6645811","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":469210380,"identity":"397ccc86-a41f-4cba-95cf-062cea9f6dc6","order_by":0,"name":"Kim A. Lindblade","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9UlEQVRIiWNgGAWjYBACCQbmhgMMbAdkGBgSGB8ABXj4CGthBGvhAWphNgBpYSNGCwNUC5sESISgFsn2g42HC8ru8PC3Jx+r/JpjJ8PGwHzs4xc8WqR5EhsOzzj3jEfizLO027LbkoEOY0ueLYNHixwDUAtv22Eehhs5ZrcltzEDtfAYM0vg08L/EKJF/kb+t2LJbfWEtUhLQG0xuJHDxvhx22GwFsYP+Lw/A2gLD9AvhmeeGUszbjvOw8bMlsyMRweDxPnkw595yu7IyR1Pfvjx57Zqe3725sOMP/DpQQbMPGASxiAGwA0n3pZRMApGwSgYCQAAFmxLkxny9VAAAAAASUVORK5CYII=","orcid":"","institution":"PATH","correspondingAuthor":true,"prefix":"","firstName":"Kim","middleName":"A.","lastName":"Lindblade","suffix":""},{"id":469210381,"identity":"ad165e40-e653-4dd0-b0a2-0885d86ad0e8","order_by":1,"name":"Corine Ngufor","email":"","orcid":"","institution":"Centre de Recherche Entomologique de Cotonou","correspondingAuthor":false,"prefix":"","firstName":"Corine","middleName":"","lastName":"Ngufor","suffix":""},{"id":469210382,"identity":"74126e4f-b40d-43bb-8f6f-361f4201adb1","order_by":2,"name":"William Yavo","email":"","orcid":"","institution":"Institut National de Santé Publique","correspondingAuthor":false,"prefix":"","firstName":"William","middleName":"","lastName":"Yavo","suffix":""},{"id":469210383,"identity":"85ffdb30-7c49-489f-b301-0a192fdbae8f","order_by":3,"name":"Sunday Atobatele","email":"","orcid":"","institution":"Sydani Group","correspondingAuthor":false,"prefix":"","firstName":"Sunday","middleName":"","lastName":"Atobatele","suffix":""},{"id":469210384,"identity":"581a9a62-3a0e-4d0b-a6e9-3243129cc1b5","order_by":4,"name":"Arthur Mpimbaza","email":"","orcid":"","institution":"Makerere University","correspondingAuthor":false,"prefix":"","firstName":"Arthur","middleName":"","lastName":"Mpimbaza","suffix":""},{"id":469210385,"identity":"7db5a38d-5530-41d8-92a3-c1d6e8373f2e","order_by":5,"name":"Nelson Ssewante","email":"","orcid":"","institution":"Makerere University","correspondingAuthor":false,"prefix":"","firstName":"Nelson","middleName":"","lastName":"Ssewante","suffix":""},{"id":469210386,"identity":"772a7838-09bc-4a12-9f70-e909cbbe177b","order_by":6,"name":"Ese Akpiroroh","email":"","orcid":"","institution":"Sydani Group","correspondingAuthor":false,"prefix":"","firstName":"Ese","middleName":"","lastName":"Akpiroroh","suffix":""},{"id":469210387,"identity":"6aab1866-3179-45af-982f-73342fd49479","order_by":7,"name":"Abibatou Konaté-Toure","email":"","orcid":"","institution":"Sydani Group","correspondingAuthor":false,"prefix":"","firstName":"Abibatou","middleName":"","lastName":"Konaté-Toure","suffix":""},{"id":469210388,"identity":"8a167961-4990-4cc8-b2fa-5c13338e04d2","order_by":8,"name":"Idelphonse Ahogni","email":"","orcid":"","institution":"Centre de Recherche Entomologique de Cotonou","correspondingAuthor":false,"prefix":"","firstName":"Idelphonse","middleName":"","lastName":"Ahogni","suffix":""},{"id":469210389,"identity":"e7ea483e-ec70-4242-a4a9-d77318f6c22c","order_by":9,"name":"Augustin Kpemasse","email":"","orcid":"","institution":"Programme National de Lutte contre le Paludisme","correspondingAuthor":false,"prefix":"","firstName":"Augustin","middleName":"","lastName":"Kpemasse","suffix":""},{"id":469210391,"identity":"c6ccf497-7cc3-439e-a684-8ada4ab4a221","order_by":10,"name":"Antoine Mea Tanoh","email":"","orcid":"","institution":"Programme National de Lutte contre le Paludisme","correspondingAuthor":false,"prefix":"","firstName":"Antoine","middleName":"Mea","lastName":"Tanoh","suffix":""},{"id":469210394,"identity":"91420f51-8389-401e-bab6-4d4b5e7437fa","order_by":11,"name":"Godwin Ntadom","email":"","orcid":"","institution":"National Malaria Elimination Programme","correspondingAuthor":false,"prefix":"","firstName":"Godwin","middleName":"","lastName":"Ntadom","suffix":""},{"id":469210396,"identity":"02c81d73-ce9b-4ca7-b746-b44119def42a","order_by":12,"name":"Jimmy Opigo","email":"","orcid":"","institution":"National Malaria Control Division","correspondingAuthor":false,"prefix":"","firstName":"Jimmy","middleName":"","lastName":"Opigo","suffix":""},{"id":469210398,"identity":"ed06aa71-0698-47c3-935f-571cd8839e02","order_by":13,"name":"Stephanie Zobrist","email":"","orcid":"","institution":"PATH","correspondingAuthor":false,"prefix":"","firstName":"Stephanie","middleName":"","lastName":"Zobrist","suffix":""},{"id":469210399,"identity":"a2393a0f-ab01-466f-9bea-2c600a25fe58","order_by":14,"name":"Kevin Griffith","email":"","orcid":"","institution":"U.S. President’s Malaria Initiative","correspondingAuthor":false,"prefix":"","firstName":"Kevin","middleName":"","lastName":"Griffith","suffix":""},{"id":469210400,"identity":"81de05bb-cad0-4816-937b-f09b454b5bfc","order_by":15,"name":"Michael Humes","email":"","orcid":"","institution":"U.S. President’s Malaria Initiative","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"","lastName":"Humes","suffix":""}],"badges":[],"createdAt":"2025-05-12 10:53:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6645811/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6645811/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12936-025-05522-3","type":"published","date":"2025-09-30T15:56:59+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":92883992,"identity":"2b57562a-2ec2-438b-8a02-ad0bd513675f","added_by":"auto","created_at":"2025-10-06 16:11:50","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1568264,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6645811/v1/8bca6841-236a-4f56-9325-39100cddd91e.pdf"},{"id":87880575,"identity":"ef042475-6284-4a8f-b9d2-4e830222782a","added_by":"auto","created_at":"2025-07-30 04:00:37","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":111868,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable1.docx","url":"https://assets-eu.researchsquare.com/files/rs-6645811/v1/a2a1f51445c0d62567a5d1df.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Evaluating the performance of an artificial intelligence-based electronic reader for malaria rapid diagnostic tests across four sub-Saharan African countries","fulltext":[{"header":"BACKGROUND","content":"\u003cp\u003eThe development of visually interpreted, clinical laboratory tests using immunoassay-based detection methods, known as rapid diagnostic tests (RDTs), has expanded access to diagnostic confirmation for many infectious diseases. These tests are now used not only in hospitals and primary care facilities but also at the community and household levels. Malaria RDTs were first developed in the early 1990s. However, widespread adoption followed only after the World Health Organization (WHO) and other agencies established standards for product and lot testing [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. In 2010, with increasing availability of WHO-prequalified RDTs, WHO recommended that all suspected malaria cases be confirmed using a quality-assured parasitologic test, either microscopy or an RDT [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Since then, the use of malaria RDTs has increased from 20\u0026nbsp;million tests in 2010 to more than 328\u0026nbsp;million in 2023. In sub-Saharan Africa, RDTs now confirm nearly four times as many malaria infections as microscopy [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eMalaria RDTs are easy to use and interpret, helping clinicians and community health workers diagnose and manage febrile patients more accurately. Their widespread use has improved malaria case management by reducing or eliminating presumptive treatment. RDTs have also enhanced the quality of malaria surveillance data by reducing misclassification of cases. The availability of malaria RDTs at the community level through community health workers has expanded access to parasitological confirmation in remote areas.\u003c/p\u003e \u003cp\u003eDespite RDTs advantages, clinicians do not always adhere to RDT results when making treatment decisions [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. While diagnostic tools are meant to guide care decisions, they do not override clinical judgment, and several factors may lead clinicians to diverge from RDT findings. For example, clinicians may doubt negative RDT results due to concerns about low-density infections falling below the test\u0026rsquo;s limit of detection [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Additionally, a common malaria antigen used in RDTs, the histidine-rich protein 2 (HRP2), can persist in the blood stream for days or weeks after infection has cleared, leading to false positives [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. In other cases, parasites lacking the genes encoding HRP2 may go undetected [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Beyond biological considerations, clinicians may also face external pressures, including patient or caregiver demands for antimalarial treatment despite negative test results [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAlthough clinicians may have legitimate clinical grounds to disregard an RDT result when making their treatment decisions, overtreatment of patients with negative results wastes critical medicines and may delay appropriate care by preventing further diagnostic testing. Malaria-affected countries actively monitor overuse of antimalarial medicines through different approaches, including tracking the ratio of antimalarial treatments to confirmed cases over time and assessing adherence to negative RDT results in outreach supervision checklists [\u003cspan additionalcitationids=\"CR10\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] and end-user verification surveys [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. As a result of the emphasis on reducing unnecessary use of antimalarials, some healthcare workers may intentionally misrecord negative RDT results as positive in health facility registers to justify antimalarial prescription. In a previous publication, we found that between 5.0% and 7.1% of RDT results across four sub-Saharan African countries were originally negative results but were misrecorded as positive in health facility registers [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Although some of these discrepancies could be due to unintentional errors, the fact that originally positive results were misrecorded as negative at a much lower rate (0.7\u0026ndash;3.5%) suggests that at least some misrecording may be deliberate.\u003c/p\u003e \u003cp\u003eIn response to concerns over the accuracy of recorded RDT results and to facilitate rapid reporting, electronic readers of RDTs were developed to promote more consistent test interpretation and accurate and timely reporting [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. RDT readers may be either a stand-alone, dedicated device or a digital application that uses the camera in a phone or tablet to interpret an RDT image. WHO produced a target product profile (TPP) for electronic RDT readers that includes operational and performance characteristics, specifying a minimum of \u0026ge;\u0026thinsp;95% (optimal\u0026thinsp;\u0026ge;\u0026thinsp;98%) agreement between the electronic reader and expert, in-person visual interpretation of the test by a panel of skilled operators who directly view the RDT [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. The WHO TPP also notes that an electronic reader is unlikely to improve on expert visual interpretation in terms of key metrics of diagnostic performance (sensitivity, specificity, limit of detection) and that the performance of electronic readers should be assessed specifically with faint lines (low positives), invalid, and indeterminate results.\u003c/p\u003e \u003cp\u003eWe evaluated the performance characteristics of the HealthPulse digital application (HealthPulse, Audere, Seattle WA USA) in a multi-country study by comparing results generated by its artificial intelligence (AI) algorithms to those interpreted by a panel of trained RDT readers who reviewed images of the RDTs.\u003c/p\u003e"},{"header":"METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eMalaria Rapid Diagnostic Test Capture and Reporting Assessment (MaCRA)\u003c/h2\u003e \u003cp\u003eWe implemented the Malaria Rapid Diagnostic Test Capture and Reporting Assessment (MaCRA) in public health facilities in Benin, C\u0026ocirc;te d\u0026rsquo;Ivoire, Nigeria and Uganda in 2023. The methods have been described elsewhere in detail [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Briefly, study staff used the HealthPulse application on a smartphone to take images of malaria RDTs as soon as possible after the RDT was performed and interpreted by a healthcare worker (HCW). A trained external panelist reviewed each RDT image and recorded their interpretation. In the background of the HealthPulse application, an AI algorithm interpreted the result from the RDT image.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eDevelopment of the HealthPulse app\u003c/h3\u003e\n\u003cp\u003eThe HealthPulse digital application takes an image of an RDT cassette and uses an AI algorithm to interpret the result based on identification of the presence or absence of the control and one or more test lines. The app has an image quality assurance (IQA) component that leverages computer vision and machine learning processes to assess image quality. The app immediately flags images that do not meet quality standards (such as those with blur or skew) and prompts the user to retake the photo. However, if the photo is not retaken, the original image is used for interpretation. The HealthPulse app works on Android phones and can function either as a non-medical (the AI interpretation of the result is not shared with the user) or medical (the AI interpretation is shared with the user) electronic RDT reader. For the MaCRA study, the app was used as a non-medical RDT reader and the AI outcome was not shared with the study staff or HCWs during the course of the study nor was it used to inform clinical management of patients.\u003c/p\u003e \u003cp\u003eThe RDT products purchased by the ministries of health for public facilities in Benin, C\u0026ocirc;te d\u0026rsquo;Ivoire, Nigeria and Uganda were identified prior to the start of the MaCRA study and used to develop the AI algorithm (Supplementary Table\u0026nbsp;1). Sets of RDT images that were generated through laboratory production, synthetic imagery and collection of RDT cassettes in health facilities in South Africa were used to train the AI algorithm to interpret results in diverse conditions such as varying lighting, focal lengths and RDT misadministration. The HealthPulse AI algorithm was further trained on RDT images collected during the first two weeks of the study in each country; images and data from this period were not included in the analytical dataset.\u003c/p\u003e\n\u003ch3\u003eCollection of RDT images through the HealthPulse app\u003c/h3\u003e\n\u003cp\u003eTrained study staff collected RDT images over the course of implementing the MaCRA study, which is described more fully in a previous publication [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Briefly, RDTs performed in 16 health facilities over a four to six month period in each country were photographed by study staff using the digital HealthPulse application as soon as possible after a HCW administered the test and interpreted the result. The images, along with the RDT result recorded in the health facility register, basic data on the patient and a unique identifier for the HCW who recorded the RDT result, were automatically uploaded to a cloud server whenever internet connection was available.\u003c/p\u003e \u003cp\u003eStudy staff used Android-based smartphones that met minimum requirements prespecified by Audere for the HealthPulse app. All had rear cameras of at least 8 megapixels (MP); however, C\u0026ocirc;te d\u0026rsquo;Ivoire used a smartphone with a 50 MP camera and Uganda used one with a 64 MP camera.\u003c/p\u003e\n\u003ch3\u003eInterpretation of rapid diagnostic tests by a trained panel\u003c/h3\u003e\n\u003cp\u003eMetadata were removed from RDT images captured using the HealthPulse app before review by a trained panel. Panelists, trained on both high- and poor-quality images and tested for accuracy prior to employment, reviewed images independently and were physically separated to minimize bias. For each image, they recorded the presence or absence of the control and test lines, identified the RDT product, and flagged quality issues such as blur, glare, darkness, skew, multiple RDTs per image, and excess blood in the result window. Faint test lines suggestive of low-positive results were also flagged; however, panelists did not indicate which specific test line was faint.\u003c/p\u003e \u003cp\u003eEach image interpreted by the panel was submitted to a quality control specialist who reviewed the image and passed or failed either the interpretation or the flagged conditions. Images with rejected interpretations or flagged conditions were sent back to the pool and reviewed by a new panelist. In a second quality control step, 30% of all images were randomly selected for review; rejected images re-entered the pool and were re-reviewed as with the first quality control step.\u003c/p\u003e \u003cp\u003eAt least the first 2200 images in each country were reviewed by three panelists to determine the interrater reliability. As measurement of interrater agreement using Fleiss\u0026rsquo; kappa was between 0.998\u0026ndash;1.0 for each country, indicating almost perfect agreement among the three panelists, we moved to a single reviewer for the remaining images to save costs.\u003c/p\u003e\n\u003ch3\u003eData management and statistical analysis\u003c/h3\u003e\n\u003cp\u003eData were cleaned to remove records with RDT products that had not been used to develop the AI model. For analyses stratified by RDT product, RDT products with fewer than 100 records were dropped.\u003c/p\u003e \u003cp\u003eWe evaluated the performance of the HealthPulse AI algorithm using the panel RDT result as the reference standard. Both the panel and the AI algorithm classified results into one of four categories: positive, negative, invalid or indeterminate. RDT products with two test lines were classified as positive if either line was present, and negative when both test lines were absent. Tests were considered invalid when the control line was not visible. Indeterminate results referred to cases where the test could not be interpreted due to an obstruction such as blood or another object obscuring the control or test lines, or when multiple RDTs appeared in the image.\u003c/p\u003e \u003cp\u003eWe categorized results using a multiclass confusion matrix with four outcomes [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. The columns represent the true classifications according to the panel and the rows represent the AI classifications (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Each cell represents a true or false outcome with respect to the panel results.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eExample of a multiclass confusion matrix comparing the panel results to the rapid diagnostic test outcomes predicted by the artificial intelligence algorithm\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAI classification\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003ePanel results\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePositive\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNegative\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInvalid\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eIndeterminate\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePositive\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTrue positive (TP)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFalse positive (FP)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFP\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eNegative\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFalse negative (FN)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTrue negative (TN)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFN\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eInvalid\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFalse invalid (FIv)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFIv\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTrue invalid (TIv)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFIv\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIndeterminate\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFalse indeterminate (FId)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFId\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFId\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTrue indeterminate (TId)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003eAI: artificial intelligence; FId : False indeterminate ; FIv : False invalid ; FN : false negative FP : false positive ; TId: True indeterminate; TIv: True invalid; TN : true negative ; TP : true positive.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eWe calculated several diagnostic accuracy metrics specific to measuring the performance of classification models in machine learning [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Recall is analogous to sensitivity and specificity for positive and negative outcomes, respectively, and measures the proportion of true outcomes that were correctly predicted by the HealthPulse AI algorithm (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Similarly, precision is similar to positive and negative predictive values (PPV and NPV) and reflects the likelihood that an outcome predicted by the AI algorithm is actually correct. The F1 score integrates precision and recall into a single metric to summarize model performance and will be high only if both the recall and precision are high, making it a suitable metric for evaluating models that need to balance between minimizing false positives and false negatives. The overall F1 score was calculated as the weighted average of the F1 scores for each outcome.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eFormulas to calculate performance metrics for the artificial intelligence classification\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMetric\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDescription\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFormula\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe percentage of tests correctly predicted out of all the tests\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TP+TN+TIv+TId}{TP+TN+TIv+TId+FP+FN+FIv+FId}\\:x\\:100\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe percentage of tests with a specific outcome that were correctly predicted by the AI algorithm (analogous to sensitivity and specificity for positive and negative outcomes, respectively)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePositive tests\u0026nbsp;: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TP}{TP+FN+FIv+FId}\\)\u003c/span\u003e\u003c/span\u003e x 100\u003c/p\u003e \u003cp\u003eNegative tests\u0026nbsp;: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TN}{TN+FP+FIv+FId}\\)\u003c/span\u003e\u003c/span\u003e x 100\u003c/p\u003e \u003cp\u003eInvalid tests\u0026nbsp;: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TIv}{TIv+FP+FN+FId}\\)\u003c/span\u003e\u003c/span\u003e x 100\u003c/p\u003e \u003cp\u003eIndeterminate tests\u0026nbsp;: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TId}{TId+FP+FN+FIv}\\)\u003c/span\u003e\u003c/span\u003e x 100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe percentage of tests predicted by the AI algorithm for a specific outcome that were classified with that outcome by the panel (analogous to positive predictive value and negative predictive value for positive and negative outcomes, respectively)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePositive tests\u0026nbsp;: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TP}{TP+FP}\\)\u003c/span\u003e\u003c/span\u003e x 100\u003c/p\u003e \u003cp\u003eNegative tests\u0026nbsp;: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TN}{TN+FN}\\)\u003c/span\u003e\u003c/span\u003e x 100\u003c/p\u003e \u003cp\u003eInvalid tests\u0026nbsp;: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TIv}{TIv+FIv}\\)\u003c/span\u003e\u003c/span\u003e x 100\u003c/p\u003e \u003cp\u003eIndeterminate tests\u0026nbsp;: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{TId}{TId+FId}\\)\u003c/span\u003e\u003c/span\u003e x 100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eF1 score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eA weighted harmonic mean of recall and precision for each test outcome\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{2\\:\\times\\:Precision\\:\\times\\:Recall}{Precision+Recall}\\)\u003c/span\u003e\u003c/span\u003e x 100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003eAI: artificial intelligence; FId : false indeterminate ; FIv : false invalid ; FN : false negative FP : false positive ; TId: true indeterminate; Tiv: true invalid; TN : true negative ; TP : true positive.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eWe used logistic regression to compare the odds of the HealthPulse AI algorithm correctly recalling positive cases (i.e. sensitivity) and negative cases (i.e. specificity) across countries, RDT products, presence of faint lines, blood in the window and photographic anomalies. Separate logistic regression models were constructed for data subsets containing only true positives or only true negatives using the \u003cem\u003eglm\u003c/em\u003e function in R (R Foundation for Statistical Computing, Vienna, Austria) to measure association between various factors and accurate recall. We calculated unadjusted odds ratios with 95% confidence intervals (CIs), as well as adjusted odds ratios from multivariable models.\u003c/p\u003e"},{"header":"RESULTS","content":"\u003cp\u003eWe collected 110,843 RDT images between June and December 2023. Of these, 612 images (0.6%) were discarded because the RDT product was not among those the AI algorithm had been trained to interpret. Of the 110,231 images remaining in the analytical data set, Benin contributed 36,411 (33.0%), C\u0026ocirc;te d\u0026rsquo;Ivoire 11,597 (10.5%), Nigeria 22,285 (20.2%) and Uganda 39,938 (36.2%) (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eNumber and proportion of RDTs observed in the study by product type and other characteristics, 2023\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eOverall \u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;110,231\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c6\" namest=\"c3\"\u003e \u003cp\u003eCountry\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c8\" namest=\"c7\"\u003e \u003cp\u003eFaint line\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eBenin\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;36,411\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eC\u0026ocirc;te d\u0026rsquo;Ivoire\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;11,597\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNigeria\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;22,285\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eUganda\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;39,938\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;9714\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;100,517\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRDT products\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAdvDx Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e25,339 (23.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7602 (65.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e17,734 (79.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e2153 (22.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e23,186 (23.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBioline Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e56,888 (51.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e36,194 (99.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e189 (0.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e20,502 (51.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e3814 (39.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e53,074 (93.3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBioline Malaria Pf (HRP2/pLDH)\u0026nbsp;*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e204 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e204 (0.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e61 (29.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e143 (70.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBioline Malaria Pf Pan*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4641 (4.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4641 (11.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e1003 (21.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3638 (78.4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCareStart Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e31 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e31 (0.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e14 (45.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e17 (54.8)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFirst Response Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e10,697 (9.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3846 (33.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4331 (19.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2517 (6.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e890 (8.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e9807 (91.7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFirst Response Malaria Pf Ag (pLDH/HRP2)\u0026nbsp;*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3504 (3.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3504 (8.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e874 (24.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e2630 (75.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParaHIT Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1142 (1.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e214 (0.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e145 (1.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e783 (2.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e124 (10.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e1018 (89.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStandard Q Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7785 (7.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e7784 (19.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e781 (10.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e7004 (90.0)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFaint line\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e9714 (8.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1608 (4.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e845 (7.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1938 (8.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e5323 (13.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e100,517 (91.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e34,803 (95.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e10,752 (92.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e20,347 (91.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e34,615 (86.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBlood present\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e15,371 (13.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4463 (12.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1599 (13.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2451 (11.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e6858 (17.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e94,860 (86.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e31,948 (87.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9998 (86.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e19,834 (89)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e33,080 (82.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhoto anomaly\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3002 (2.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e655 (1.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e411 (3.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e765 (3.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1171 (2.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e107,229 (97.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e35,756 (98.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e11,186 (96.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e21,520 (96.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e38,767 (97.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eImage flagged for retake\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e109,999 (99.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e36,396 (100)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e11,584 (99.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e22,230 (99.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e39,789 (99.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e232 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15 (0.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e13 (0.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e55 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e149 (0.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003eHRP2: histidine-rich protein 2; Pf: \u003cem\u003ePlasmodium falciparum\u003c/em\u003e; pLDH: \u003cem\u003ePlasmodium\u003c/em\u003e lactate dehydrogenase; RDT: rapid diagnostic test\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003e*RDT products that include a pLDH line in addition to the \u003cem\u003eP. falciparum\u003c/em\u003e-specific HRP2 line.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThere were nine RDT products identified during the course of the study (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). All RDT products observed in the study were prequalified by WHO, although the CareStart Malaria Pf (Access Bio Inc, NJ USA) product has been delisted (Supplementary Table\u0026nbsp;1) [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. The Bioline Malaria Pf (Abbott, IL USA) test was the most common RDT product in Benin and Uganda, while C\u0026ocirc;te d\u0026rsquo;Ivoire and Nigeria used predominantly AdvDx Malaria Pf (Advy Chemical, Mumbai, India) and First Response Malaria Pf (Premier Medical Corporation Ltd, Gujarat, India).\u003c/p\u003e \u003cp\u003eSix RDT products have a single test line that detects \u003cem\u003ePlasmodium falciparum\u003c/em\u003e-specific histidine-rich protein 2 (HRP2): AdvDx Malaria Pf, Bioline Malaria Pf, CareStart Malaria Pf, First Response Malaria Pf, ParaHIT (Arkray Healthcare Prvt Ltd, Mumbai, India) and Standard Q Malaria Pf (SD Biosensor, Gyeonggi-do, Republic of Korea). The remaining three RDTs include a second test line to detect \u003cem\u003ePlasmodium\u003c/em\u003e lactate dehydrogenase (pLDH), an enzyme found in all \u003cem\u003ePlasmodium\u003c/em\u003e spp that can also be used to differentiate species (Bioline Malaria Pf [HRP2/pLDH], Bioline Malaria Pf Pan and First Response Malaria Pf Ag [pLDH/HRP2]). The pLDH line in the Bioline Malaria Pf (HRP2/pLDH) RDT detects only \u003cem\u003eP. falciparum\u003c/em\u003e while the other two tests use a pan-\u003cem\u003ePlasmodium\u003c/em\u003e pLDH antigen that detects non-falciparum species as well.\u003c/p\u003e \u003cp\u003eThe proportion of RDTs with faint lines varied by country and ranged from 4.4% in Benin to 13.3% in Uganda. Faint lines were more common among the RDT products with a second pLDH test line (21.6\u0026ndash;29.9%) compared to those a single HRP2 test line (6.7\u0026ndash;10.9%), except for CareStart Malaria Pf where 45.2% (14/31) of the RDTs included in the study had faint lines. The presence of blood in the RDT well ranged from 11.0% in Nigeria to 17.2% in Uganda. Photo anomalies were more rare, affecting between 1.8% (Benin) and 3.5% (C\u0026ocirc;te d\u0026rsquo;Ivoire) of RDTs. Images were flagged for retake in 0.4% of RDTs in Uganda but less than 0.1% in Benin.\u003c/p\u003e\n\u003ch3\u003ePerformance characteristics of the HealthPulse artificial intelligence algorithm\u003c/h3\u003e\n\u003cp\u003eThe panel classified 56,434 (51.2%) RDT outcomes as positive, 52,963 (48.0%) as negative, 271 (0.2%) as invalid and 563 (0.5%) as indeterminate (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). The AI algorithm predicted 56,404 (51.2%) outcomes as positive, 53,117 (48.2%) as negative, 534 (0.5%) as invalid and 176 (0.2) as indeterminate.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eConfusion matrix with the RDT outcomes classified by the panel and the HealthPulse RDT reader artificial intelligence algorithm (N\u0026thinsp;=\u0026thinsp;110,231)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAI classification\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003ePanel results\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePositive\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNegative\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInvalid\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eIndeterminate\u003c/p\u003e \u003cp\u003eN (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePositive\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e54,933 (97.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1336 (2.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e15 (5.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e120 (21.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e56,404 (51.2)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1300 (2.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e51,446 (97.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e21 (7.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e350 (62.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e53,117 (48.2)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInvalid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e126 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e90 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e229 (82.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e89 (15.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e534 (0.5)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIndeterminate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e75 (0.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e91 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6 (2.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e4 (0.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e176 (0.2)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTotal\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e56,434 (51.2)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e52,963 (48.0)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e271 (0.2)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e563 (0.5)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e110,231\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"6\"\u003eAI: artificial intelligence.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eOverall, we found a high level of accuracy (96.8%) in the AI classification of RDT outcomes and the overall F1 score (96.6%) was similarly high (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). The recall statistics for positive (97.3%) and negative (97.1%) outcomes indicated that a high proportion of the true positive and negative outcomes were identified by the AI algorithm. However, recall was lower for invalid (84.5%) and indeterminate outcomes (0.7%). Precision measures were high for both positive (97.4%) and negative (96.9%) outcomes but low for invalid (42.9%) and indeterminate (2.3%) outcomes, suggesting that outcomes classified as invalid or indeterminate by the AI algorithm were very likely to be incorrect.\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;5. Performance characteristics of the HealthPulse RDT reader artificial intelligence algorithm, 2023\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" valign=\"bottom\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCharacteristic\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAccuracy\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOverall F1\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 19.788%;\" colspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003ePositive\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 13.9089%;\"\u003e\u003cp\u003e\u003cstrong\u003eNegative\u003c/strong\u003e\u003cp\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 13.9089%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInvalid\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 13.9089%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eIndeterminate\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRecall\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePrecision\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRecall\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePrecision\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRecall\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePrecision\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRecall\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePrecision\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e%\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eOverall\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e97.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e97.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e84.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e42.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e2.3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eCountry\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eBenin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e98.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e96.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e85.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e20.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e25.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eC\u0026ocirc;te d\u0026rsquo;Ivoire\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e98.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e98.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e98.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e82.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e57.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eNigeria\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e96.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e96.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e81.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e51.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e2.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e15.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eUganda\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e96.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e98.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e96.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e87.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e40.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eRDT product\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eAdvdx Malaria Pf\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e97.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e96.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e84.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e32.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e7.7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eBioline Malaria Pf\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e97.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e97.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e86.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e26.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e25.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eBioline Malaria Pf (HRP2/pLDH)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e99.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e99.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e99.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e99.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e98.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e98.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eBioline Malaria Pf Pan*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e92.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e94.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e95.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e95.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e88.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e98.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e90.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e11.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eFirst Response Malaria Pf\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e97.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e98.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e98.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e78.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e60.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e3.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e15.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eFirst Response Malaria Pf Ag (pLDH/HRP2)* \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e95.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e93.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e97.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e95.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e92.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e91.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e54.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eParaHIT Malaria Pf\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e95.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e95.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e96.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e95.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e95.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eStandard Q Malaria Pf\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e96.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e98.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e98.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eFaint line\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e85.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e90.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e86.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e99.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e68.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e6.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e71.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e18.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e99.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e97.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e99.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e86.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e49.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e2.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eBlood present\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e91.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e89.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e95.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e92.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e93.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e92.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e84.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e51.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e12.5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e97.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e98.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e87.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e19.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e5.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e1.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003ePhoto anomaly\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e95.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e95.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e97.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e97.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e95.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e88.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e38.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e2.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e16.7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e97.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e97.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e84.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e43.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e1.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eImage flagged for retake\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e97.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e97.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e97.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e97.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e96.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e86.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e43.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e~\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 19.4023%;\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e20.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e30.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.896%;\"\u003e\n \u003cp\u003e22.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 12.7401%;\"\u003e\n \u003cp\u003e95.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.4804%;\"\u003e\n \u003cp\u003e17.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e80.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 6.7791%;\"\u003e\n \u003cp\u003e22.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e25.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 5.961%;\"\u003e\n \u003cp\u003e36.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 7.9479%;\"\u003e\n \u003cp\u003e2.3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 19.4023%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width: 7.9479%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width: 7.9479%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width: 6.896%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width: 5.6103%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width: 7.4804%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width: 7.9479%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width: 6.7791%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width: 7.9479%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width: 5.961%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd style=\"width: 7.9479%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eHRP2: histidine-rich protein 2; Pf: \u003cem\u003ePlasmodium falciparum\u003c/em\u003e; pLDH: \u003cem\u003ePlasmodium\u0026nbsp;\u003c/em\u003elactate dehydrogenase; RDT: rapid diagnostic test\u003c/p\u003e\n\u003cp\u003e*RDT products that include a pLDH line in addition to the \u003cem\u003eP. falciparum\u003c/em\u003e-specific HRP2 line.\u003c/p\u003e\n\u003cp\u003e~Could not be calculated.\u003c/p\u003e\u003cp\u003eThe AI scored above 96% in accuracy in all countries but only scored\u0026thinsp;\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e\u0026ge;\u003c/span\u003e\u0026thinsp;98% in C\u0026ocirc;te d\u0026rsquo;Ivoire (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). The F1 score was highest in C\u0026ocirc;te d\u0026rsquo;Ivoire (97.9%) and lowest in Nigeria (96.2%) but differences were minor. The performance of the HealthPulse AI algorithm was high across all RDT products with F1 scores ranging from 93.8\u0026ndash;99.0%. The RDT product for which the AI performed the best overall was the Bioline Malaria Pf (HRP2/pLDH) test, which has two test lines. However, only 204 instances of this RDT product were included in the dataset. The AI algorithm had the worst performances with Bioline Malaria Pf Pan (F1 score: 94.1%) and First Response Malaria Pf Ag (pLDH/HRP2) (F1 score: 93.8%), the other two RDT products with two test lines. Restricting the analysis to data from Uganda, where multiple RDT products were used, did not alter the F1 scores, suggesting minimal bias from country-specific RDT usage patterns (data not shown).\u003c/p\u003e \u003cp\u003eAI performance was lower when faint lines were present on the RDT. The F1 score was 97.7% in the absence of faint lines but dropped to 90.8% when faint lines were present (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). The presence of blood in the RDT cassette similarly reduced performance, with the F1 score declining from 97.7\u0026ndash;89.7%. Photographic anomalies affected accuracy, although to a lesser extent, with the F1 score decreasing from 96.6\u0026ndash;95.0%. The AI performance was importantly reduced for images that were flagged for retake, with the F1 score declining from 96.7\u0026ndash;30.6%.\u003c/p\u003e\n\u003ch3\u003eOdds of correct recall of positive and negative results\u003c/h3\u003e\n\u003cp\u003eThe unadjusted odds of the AI correctly recalling a positive result differed significantly by country, RDT product, and the presence of faint lines. Compared with Benin, the odds of positive recall were higher in C\u0026ocirc;te d\u0026rsquo;Ivoire (OR 1.77, 95% CI 1.33, 2.33) but lower in Nigeria (OR 0.49, 95% CI 0.41, 0.57) and Uganda (OR 0.52, 95% CI 0.45, 0.59) (Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). Using AdvDx Malaria Pf as the reference RDT, the odds of correct positive recall were higher for Bioline Malaria Pf Pan (OR 3.21, 95% CI 1.99, 5.19), First Response Malaria Pf Ag (pLDH/HRP2) (OR 3.04, 95% CI 1.86, 4.96), and First Response Malaria Pf (OR 1.40, 95% CI 1.11, 1.75), and lower for Standard Q Malaria Pf (OR 0.72, 95% CI 0.58, 0.89). No statistically significant differences were observed for Bioline Malaria Pf, Bioline Malaria Pf (HRP2/pLDH), or ParaHIT Malaria Pf compared to the reference RDT product. The odds of correct positive recall were significantly reduced in the presence of a faint line (OR 0.01, 95% CI 0.00, 0.01) or blood in the result window (OR 0.65, 95% CI 0.57, 0.75), but were not significantly affected by the presence of photographic anomalies (OR 1.04, 95% CI 0.74, 1.47) or images flagged for retake (OR 1.93, 95% CI 0.26, 14.31).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eUnadjusted and adjusted odds ratios for the HealthPulse RDT reader artificial intelligence algorithm to correctly recall positive results by characteristics of the RDTs (n\u0026thinsp;=\u0026thinsp;56,419)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eUnadjusted\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAdjusted*\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOdds ratio (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eOdds ratio (95% CI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCountry\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBenin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18,004\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eC\u0026ocirc;te d\u0026rsquo;Ivoire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6873\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.77 (1.34, 2.33)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.99 (0.61, 1.62)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNigeria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9423\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.49 (0.41, 0.57)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.39 (0.25, 0.61)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUganda\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e22119\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.51 (0.45, 0.59)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.80 (0.68, 0.95)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRDT product\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAdvDx Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11,649\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBioline Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e29053\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.95 (0.82, 1.09)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.35 (0.23, 0.55)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBioline Malaria Pf (HRP2/pLDH)\u0026nbsp;**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e143\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.45 (0.48, 24.76)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.92 (0.65, 37.24)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBioline Malaria Pf Pan**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2481\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.21 (1.99, 5.18)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.54 (2.40, 8.59)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFirst Response Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6314\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.40 (1.12, 1.75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.82 (0.62, 1.08)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFirst Response Malaria Pf Ag (pLDH/HRP2) \u0026nbsp;**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2173\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.03 (1.86, 4.96)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.59 (2.40, 8.78)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParaHIT Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e524\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.77 (0.46, 1.29)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.62 (0.31, 1.21)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStandard Q Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4082\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.72 (0.58, 0.89)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.44 (0.28, 0.71)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFaint line\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9526\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.01 (0.00, 0.01)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.01 (0.00, 0.01)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e46,893\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBlood present\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7432\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.65 (0.57, 0.75)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.64 (0.54, 0.75)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e48,987\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhoto anomaly\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1544\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.04 (0.74, 1.47)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e54,875\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eImage flagged for retake\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e56,320\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.93 (0.26, 14.31)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003eCI: confidence interval; HRP2: histidine-rich protein 2; Pf: \u003cem\u003ePlasmodium falciparum\u003c/em\u003e; pLDH: \u003cem\u003ePlasmodium\u003c/em\u003e lactate dehydrogenase; Ref.: reference; RDT: rapid diagnostic test\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e*The model included all variables except photo anomaly and whether it was flagged for retake.\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e**RDT products that include a pLDH line in addition to the \u003cem\u003eP. falciparum\u003c/em\u003e-specific HRP2 line.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eIn the adjusted model, which included RDT product, faint lines, and presence of blood, the odds of correct positive recall remained significantly lower in Nigeria (OR 0.39, 95% CI 0.25, 0.61) and Uganda (OR 0.80, 95% CI 0.68, 0.95) compared with Benin (Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). No significant difference was observed between Benin and C\u0026ocirc;te d\u0026rsquo;Ivoire (OR 0.99, 95% CI 0.61, 1.62). In the adjusted model, the AI was more likely to correctly recall a positive test with Bioline Malaria Pf Pan (OR 4.54, 95% CI 2.40, 8.59) and First Response Malaria Pf Ag (pLDH/HRP2) (OR 4.59, 95% CI 2.40, 8.78) compared with AdvDx Malaria Pf, and less likely with Bioline Malaria Pf (OR 0.35, 95% CI 0.23, 0.55) and Standard Q Malaria Pf (OR 0.44, 95% CI 0.28, 0.71). The ORs for the presence of a faint line or blood were not affected by controlling for country and RDT product.\u003c/p\u003e \u003cp\u003eThe unadjusted odds of correctly recalling negative test results were higher in C\u0026ocirc;te d\u0026rsquo;Ivoire (OR 1.35, 95% CI 1.09, 1.68), Nigeria (OR 1.20, 95% CI 1.04, 1.39), and Uganda (OR 1.23, 95% CI 1.08, 1.41) compared with Benin (Table\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). Correct recall of negative tests was more likely for First Response Malaria Pf (OR 1.61, 95% CI 1.24, 2.10) and Standard Q Malaria Pf (OR 2.04, 95% CI 1.50, 2.77) and less likely for Bioline Malaria Pf Pan (OR 0.45, 95% CI 0.36, 0.56), First Response Malaria Pf Ag (pLDH/HRP2) (OR 0.37, 95% CI 0.29, 0.48), and ParaHIT Malaria Pf (OR 0.57, 95% CI 0.38, 0.85) relative to AdvDx Malaria Pf. As with positive results, the odds of correctly recalling negative tests were significantly reduced when a faint line was present (OR 0.12, 95% CI 0.07, 0.20) or when blood was present (OR 0.30, 95% CI 0.27, 0.34). Photo anomalies did not significantly affect performance (OR 1.18, 95% CI 0.82, 1.70), and the odds ratio for images flagged for retake could not be calculated.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab7\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eUnadjusted and adjusted odds ratios for the HealthPulse RDT reader artificial intelligence algorithm to correctly recall negative results by characteristics of the RDTs (n\u0026thinsp;=\u0026thinsp;52,947)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eUnadjusted\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAdjusted*\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOdds ratio (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eOdds ratio (95% CI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCountry\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBenin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18,102\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eC\u0026ocirc;te d\u0026rsquo;Ivoire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4652\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.35 (1.09, 1.68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.25 (1.28, 3.95)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNigeria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12,612\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.20 (1.04, 1.39)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.09 (1.19, 3.65)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUganda\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e17,581\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.23 (1.08, 1.41)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.93 (1.61, 2.32)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRDT product\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAdvDx Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13,549\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBioline Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e27,502\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.07 (0.94, 1.22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.99 (1.14, 3.47)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBioline Malaria Pf (HRP2/pLDH)\u0026nbsp;**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.57 (0.22, 11.39)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.54 (0.20, 11.98)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBioline Malaria Pf Pan**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2135\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.45 (0.36, 0.56)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.47 (0.26, 0.83)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFirst Response Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4203\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.61 (1.24, 2.10)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.81 (1.35, 2.41)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFirst Response Malaria Pf Ag (pLDH/HRP2) **\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1185\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.37 (0.29, 0.48)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.65 (0.36, 1.18)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParaHIT Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e611\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.57 (0.38, 0.85)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.81 (0.43, 1.53)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStandard Q Malaria Pf\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3701\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.04 (1.50, 2.77)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.50 (1.35, 4.61)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFaint line\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e114\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.12 (0.07, 0.20)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.20 (0.11, 0.35)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e52,833\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBlood present\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7156\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.30 (0.27, 0.34)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.31 (0.27, 0.34)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e45,791\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhoto anomaly\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1390\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.18 (0.82, 1.70)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e51,557\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eImage flagged for retake\u0026Dagger;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e52,834\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eNot estimated\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e113\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRef.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003eCI: confidence interval; HRP2: histidine-rich protein 2; Pf: \u003cem\u003ePlasmodium falciparum\u003c/em\u003e; pLDH: \u003cem\u003ePlasmodium\u003c/em\u003e lactate dehydrogenase; Ref.: reference; RDT: rapid diagnostic test\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e*The model included all variables except photo anomaly and whether it was flagged for retake.\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e**RDT products that include a pLDH line in addition to the \u003cem\u003eP. falciparum\u003c/em\u003e-specific HRP2 line.\u003c/td\u003e\u003c/tr\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e\u0026Dagger;In the subset of true negative RDTs, the AI algorithm did not classify any tests that failed the photo filters as positive. Consequently, the odds ratio could not be estimated.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eIn the adjusted model, there were no meaningful changes in the ORs for correct recall of negative results by country or the presence of faint lines. However, negative recall was statistically more likely for Bioline Malaria Pf (OR 1.99, 95% CI 1.14, 3.47), while First Response Malaria Pf Ag (pLDH/HRP2) was no longer significantly associated with poor negative recall (OR 0.65, 95% CI 0.36, 1.18).\u003c/p\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eWe found that the HealthPulse application, an electronic malaria RDT reader powered by an AI algorithm, demonstrated a high degree of accuracy (\u0026gt;\u0026thinsp;96%) when compared to a trained panel reviewing images of the RDTs. The AI-predicted RDT results showed high positive and negative recall (i.e. sensitivity and specificity, respectively) and high positive and negative precision (i.e. PPV and NPV, respectively) relative to the panel. However, the AI\u0026rsquo;s recall and precision were notably lower for results the panel identified as invalid or indeterminate. Since invalid and indeterminate results were rare, the AI\u0026rsquo;s F1 scores remained high, ranging from 96.2\u0026ndash;97.9% across Benin, C\u0026ocirc;te d\u0026rsquo;Ivoire, Nigeria and Uganda.\u003c/p\u003e \u003cp\u003eIn multivariable models, positive recall by the AI algorithm was significantly lower for images collected in Nigeria and Uganda compared to Benin and C\u0026ocirc;te d\u0026rsquo;Ivoire. Although lighting conditions within health facilities may have contributed to differences in AI performance, they are unlikely to fully account for the observed country-level variations. Differences in smartphone camera performance may have played a role: while all devices had at least 8-megapixel resolution, variability in features such as low-light performance of cameras could have influenced the algorithm\u0026rsquo;s ability to detect test and control lines. As each country used a different smartphone model, we were unable to separate the effects of country from those of the specific device.\u003c/p\u003e \u003cp\u003eThe HealthPulse AI algorithm\u0026rsquo;s recall also varied across RDT products. Consistently, the AI algorithm performed best with the Bioline Malaria Pf (HRP2/pLDH) product and worst with the Bioline Malaria Pf Pan test. We could not find a consistent feature to explain this variability. The HealthPulse AI algorithm was trained on these RDT products at a specific test positivity rate (~\u0026thinsp;25%) to balance positive and negative recall and achieve an F1 score of at least 95%. Variation between the test positivity rates in the training data and those in this study (~\u0026thinsp;50%), and variability by RDT product due to use patterns by country, may have affected the algorithm\u0026rsquo;s performance. Nonetheless, the F1 score exceeded 96% for six out of the nine RDT products included in this study.\u003c/p\u003e \u003cp\u003eThe presence of faint lines was associated with a significant reduction in both positive and negative recall by the AI algorithm. Faint lines were more frequently observed on RDT products that included a pLDH line. Although we could not determine whether the faint line corresponded specifically to the pLDH target, pLDH is typically present at lower concentrations in the blood than HRP2, resulting in weaker or fainter test lines on RDTs [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. After adjusting for the presence of faint lines, the OR for RDT products containing a pLDH line increased, improving the AI algorithm\u0026rsquo;s positive recall in comparison to AdvDx as a reference, while the OR for RDT products not containing a pLDH line decreased. With respect to negative recall, there was no clear pattern of changes after adjustment. Because RDTs with pLDH lines also include an HRP2 test line, it is likely that the AI\u0026rsquo;s improved positive recall after adjustment for faint lines was influenced by the availability of additional visual evidence of a positive result.\u003c/p\u003e \u003cp\u003eWhile multiple RDT readers have been assessed in prior studies, cross-study comparisons of performance are difficult due to variations in use cases and evaluation methodologies. One laboratory-based study compared five RDT readers using samples with known parasite concentrations, including negative samples [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. They found that the human eye was better at detecting positive lines at low parasite densities (i.e. faint lines) compared to the RDT readers, though most RDT readers performed well with parasite densities typically associated with clinical illness (i.e. \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e\u0026ge;\u003c/span\u003e100 parasites per \u0026micro;l). Other studies comparing the human eye to RDT readers have reported high levels of agreement. In Tanzania, agreement between an RDT reader and visual inspection was 98.4% (95% CI 97.6, 99.0) among febrile patients using the Bioline Malaria Pf/Pan RDT [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Using the same RDT product among Ugandan children selected from the community, agreement with visual inspection was 98.9% (95% CI 93.2, 99.8) [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eOriginally developed to improve the consistency and accuracy of RDT interpretation, digital RDT readers gained momentum during the COVID-19 pandemic alongside the expansion of in-home testing. Over the past 15 years, malaria case management and surveillance have improved substantially due to the widespread use of RDTs [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. However, persistent challenges, including nonadherence to negative RDT results and misrecording of RDT outcomes in facility registers, underscore the continued need for data quality assessments and sustainable strategies to enhance the accuracy of RDT reporting [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Although RDT readers have been proposed as a potential solution, limited incentives exist for healthcare workers in public health settings to adopt these tools unless they reduce clinical workload. In the private sector, RDT readers have been explored as a means to improve data quality and to provide verification of test results to support reimbursement and subsidy processes. However, the performance standards required for such applications remain to be clearly defined [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. RDT readers may also serve as valuable training tools or job aids, but improvements in their ability to detect invalid and indeterminate results are needed. Research settings that require consistent and objective interpretation of RDT results represent another potential use case for high-performing RDT readers, such as HealthPulse.\u003c/p\u003e \u003cp\u003eThis study has several limitations. First, the comparator used to evaluate the HealthPulse AI algorithm was a trained human panel interpreting images of RDTs, whereas the WHO TPP requires expert interpretation of the physical RDT cassette immediately after the incubation period as the reference standard if the reader is to be used as a medical device. While this reference standard is likely to be more accurate, it is less feasible for large-scale studies. Importantly, because we did not apply the WHO-recommended comparator, we do not make conclusions regarding the suitability of the HealthPulse app as a clinical or medical RDT reader. Despite robust quality control processes to ensure the accuracy of the human panel\u0026rsquo;s interpretations, some misclassification may have occurred. Any such errors would likely reduce associations between predictors and AI performance outcomes.\u003c/p\u003e \u003cp\u003eIn conclusion, we found that the HealthPulse RDT reader AI algorithm accurately interpreted malaria RDT results when compared to a trained panel reviewing images of RDTs. The HealthPulse AI algorithm exhibited a high F1 score, balancing recall and precision. However, several factors may have influenced the algorithm\u0026rsquo;s performance, including the country setting, the type of smartphone used, the RDT product and presence of faint lines, warranting further investigation. While the HealthPulse AI shows promise for use in research or training contexts, improvements are needed, particularly in the detection of invalid and indeterminate results.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"636\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003eAI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003eArtificial intelligence\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003eCI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003eConfidence interval\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003eHCW\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003eHealthcare worker\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003eHRP2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003eHistidine-rich protein 2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003eIQA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003eImage quality assurance\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003eN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003eNumber\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003eOR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003eOdds ratio\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003ePf\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003e\u003cem\u003ePlasmodium falciparum\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003epLDH\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003eParasite lactate dehydrogenase\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003ePMI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003ePresident\u0026rsquo;s Malaria Initiative\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003eRDT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003eRapid diagnostic test\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 78px;\"\u003e\n \u003cp\u003eWHO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 558px;\"\u003e\n \u003cp\u003eWorld Health Organization\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eACKNOWLEDGEMENTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe are grateful to the Audere team (Shawna Cooper, Sasha Frade and Sam Smedinghoff) for providing technical support related to the HealthPulse application. Audere had no involvement in the analysis or interpretation of the data or the conclusions presented in this study. We acknowledge the contributions of Audere\u0026rsquo;s AI data creation and labeling teams at the Centre for HIV-AIDS Prevention Studies (South Africa) and Indivillage (India). We are grateful to the HCWs who gave generously of their time to participate in this evaluation. Many staff members of the organizations implementing the study in country made important contributions: Manfred Accrombessi, Corneille Hueha (CREC, Benin); M. Anatole Mian, Orph\u0026eacute;e Kangah-Kouakou, Val\u0026eacute;rie Bedia-Tanoh (INSP, C\u0026ocirc;te d\u0026rsquo;Ivoire); Hilary Okagbue, Evelyn Orya, Shiva Gab-Deedam, Olufisayo Bademosi (Sydani Group, Nigeria); Anne Katahoire, Jane Frances Namuganga, Jenipher Musoke (CHDC, Uganda). We are grateful to members of the national malaria programs in each country for their support: Cyriaque Affoukou and Julien Aissan (Benin); Jacques Agnon (C\u0026ocirc;te d\u0026rsquo;Ivoire); Onyebuchi Okoro, Chukwu Okoronkwo and Nnenna Ogbulafor (Nigeria); and Bosco Agaba, Catherine Maiteki Sebuguzi and Gerald Rukundo (Uganda). We received excellent research support from Saadjo Sow (PMI Insights). Megan Littrell, Taj Munson and Kim Vu provided overall direction and administrative support to this and the other PMI Insights projects. Aysu Uygur (The Gates Foundation) is greatly appreciated for her contributions during the design of the evaluation. Thank you to the PMI staff in country: Raoul Oloukoi, Virgile Gnanguenon (Benin); Pascal Zinzindohoue, Patricia Yepassis-Zembrou, Blaise Kouadio, Melaine Tape, Yao Stephane, Christie Billingsley (C\u0026ocirc;te d\u0026rsquo;Ivoire); Jules Mihigo, Cassandra Elagbaje, Veronica Momoh, Valerie Bampoe (Nigeria); Edgar Agba, Grace Appiah, Patrick Condo (Uganda).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFUNDING\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis evaluation was co-funded by PMI Insights and the Bill \u0026amp; Melinda Gates Foundation (INV-043942). PMI Insights was the global operational research and program evaluation project of the U.S. President\u0026rsquo;s Malaria Initiative (PMI). Funding for this evaluation is made possible by the generous support of the American people through the United States Agency for International Development (USAID) through cooperative agreement No. 7200AA20CA00031. The contents are the responsibility of the authors and do not necessarily reflect the views of USAID or the United States Government.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAUTHOR INFORMATION\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors and Affiliations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ePMI Insights Project/PATH, Geneva, Switzerland\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eKim A. Lindblade\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCentre de Recherche Entomologique de Cotonou, Cotonou, Benin\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eIdelphonse Ahogni\u003c/p\u003e\n\u003cp\u003eCorine Ngufor\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eInstitut National de Sant\u0026eacute; Publique, Abidjan, C\u0026ocirc;te d\u0026rsquo;Ivoire\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAbibatou Konate-Toure\u003c/p\u003e\n\u003cp\u003eWilliam Yavo\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSydani Group, Abuja, Nigeria\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eEse Akpiroroh\u003c/p\u003e\n\u003cp\u003eSunday Atobatele\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eChild Health and Development Centre, Makerere University, Kampala, Uganda\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eArthur Mpimbaza\u003c/p\u003e\n\u003cp\u003eNelson Ssewante\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eU.S. President\u0026rsquo;s Malaria Initiative, USAID, Washington, DC USA\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eKevin Griffith\u003c/p\u003e\n\u003cp\u003eMichael Humes\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eProgramme National de Lutte contre le Paludisme, Cotonou, Benin\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAugustin Kpemasse\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eProgramme National de Lutte contre le Paludisme, Abidjan, C\u0026ocirc;te d\u0026rsquo;Ivoire\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAntoine Tanoh\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNational Malaria Elimination Programme, Abuja, Nigeria\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eGodwin Ntadom\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNational Malaria Control Division, Ministry of Health, Kampala, Uganda\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eJimmy Opigo\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eContributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eKL, MH\u0026nbsp;\u003c/em\u003eand\u003cem\u003e\u0026nbsp;KG conceived and designed the evaluation. AK, AKT, AM, AMT, AK-T, CN, EA, GN, IA, JO, NS, SA, and WY oversaw data collection activities. KL and SZ drafted the manuscript. KL analyzed the data. All authors critically reviewed the manuscript. All authors read and approved the final manuscript.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCorresponding author\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCorrespondence to: Kim A. Lindblade, Rue du Grand-Pr\u0026eacute;, 1202 Geneva, Switzerland ; [email protected].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEthical approval was obtained from: the Comit\u0026eacute; National d\u0026rsquo;Ethique pour la Recherche en Sant\u0026eacute; (Benin); the Comit\u0026eacute; National d\u0026rsquo;Ethique des Sciences de la Vie et de la Sant\u0026eacute; (C\u0026ocirc;te d\u0026rsquo;Ivoire); the National Health Research Ethics Committee (Nigeria); the Oyo State Health Research Ethics Committee (Nigeria); the Sokoto State Health Research Ethics Committee (Nigeria); the Uganda National Council on Science and Technology (Uganda); the Vector Control Division Research and Ethics Committee (Uganda); and the WGC IRB in the USA. All participants provided written, informed consent to participate.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analyzed during the current evaluation can be provided by the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eCunningham J, Jones S, Gatton ML, Barnwell JW, Cheng Q, Chiodini PL, et al. A review of the WHO malaria rapid diagnostic test product testing programme (2008\u0026ndash;2018): performance, procurement and policy. Malar J. 2019;18:387. \u003c/li\u003e\n\u003cli\u003eWHO. Guidelines for the treatment of malaria. Second edition. [Internet]. Geneva, Switzerland: World Health Organization; 2010. Available from: https://www.paho.org/sites/default/files/TreatmentGuidelines-2nd-ed-2010-eng.pdf\u003c/li\u003e\n\u003cli\u003eWHO. World Malaria Report 2024. Geneva, Switzerland: World Health Organization; 2024. \u003c/li\u003e\n\u003cli\u003eKabaghe AN, Visser BJ, Spijker R, Phiri KS, Grobusch MP, van Vugt M. Health workers\u0026rsquo; compliance to rapid diagnostic tests (RDTs) to guide malaria treatment: a systematic review and meta-analysis. Malar J. 2016;15:163. \u003c/li\u003e\n\u003cli\u003eWu L, van den Hoogen LL, Slater H, Walker PGT, Ghani AC, Drakeley CJ, et al. Comparison of diagnostics for the detection of asymptomatic Plasmodium falciparum infections to inform control and elimination strategies. Nature. 2015;528:S86\u0026ndash;93. \u003c/li\u003e\n\u003cli\u003eNtuku H, Whittemore B, Dausab L, Jang IK, Golden A, Sheahan W, et al. Post-treatment duration of positivity for standard and ultra-sensitive Plasmodium falciparum antigen-based rapid diagnostic tests, a cohort study from a low-endemic setting in Namibia. EBioMedicine. 2025;111:105489. \u003c/li\u003e\n\u003cli\u003eBerhane A, Anderson K, Mihreteab S, Gresty K, Rogier E, Mohamed S, et al. Major Threat to Malaria Control Programs by Plasmodium falciparum Lacking Histidine-Rich Protein 2, Eritrea. Emerg Infect Dis. 2018;24:462\u0026ndash;70. \u003c/li\u003e\n\u003cli\u003eAltaras R, Nuwa A, Agaba B, Streat E, Tibenderana JK, Strachan CE. Why do health workers give anti-malarials to patients with negative rapid test results? A qualitative study at rural health facilities in western Uganda. Malar J. 2016;15:23. \u003c/li\u003e\n\u003cli\u003eAltaras R, Worges M, La Torre S, Audu BM, Mwangi G, Zeh-Meka A, et al. Outreach Training and Supportive Supervision for Quality Malaria Service Delivery: A Qualitative Evaluation in 11 Sub-Saharan African Countries. Am J Trop Med Hyg. 2024;110:20\u0026ndash;34. \u003c/li\u003e\n\u003cli\u003eAgbemafle EE, Addo-Lartey A, Odikro MA, Frimpong JA, Kubio C, Ameme DK, et al. Adherence to the test, treat and track strategy for malaria control among prescribers, Mfantseman Municipality, Central Region, Ghana. PLOS ONE. 2023;18:e0279712. \u003c/li\u003e\n\u003cli\u003eKoko D, Arouna D, Bernard Y-M, Ba T, Mostel J, Abdou Y, et al. How Outreach Training and Supportive Supervision (OTSS) Affect Health Facility Readiness and Health-Care Worker Competency to Prevent and Treat Malaria in Niger: A Secondary Analysis of OTSS Data. Am J Trop Med Hyg. 2024;110:50\u0026ndash;5. \u003c/li\u003e\n\u003cli\u003eUSAID Global Health Supply Chain Program. EUV Survey Question Guide [Internet]. Washington, DC: USAID Global Health Supply Chain Program; 2020. Available from: https://www.ghsupplychain.org/euv-survey-question-guide\u003c/li\u003e\n\u003cli\u003eLindblade, KA, Mpimbaza, A, Ngufor, C, Yavo, W, Atobatele, S, Akpiroroh, E, et al. Assessing the accuracy of the recording and reporting of malaria rapid diagnostic test results in four African countries: Methods and key results. under review. \u003c/li\u003e\n\u003cli\u003eWhite, W., Korir, G. Landscape of RDT-reading apps [Internet]. Geneva, Switzerland: FIND; 2023 Jan. Available from: https://www.finddx.org/wp-content/uploads/2023/02/20230203_rep_market_rdt_reading_app_FV_EN.pdf\u003c/li\u003e\n\u003cli\u003eWHO. Target product profile for readers of rapid diagnostic tests [Internet]. Geneva, Switzerland: World Health Organization; 2023. Available from: https://iris.who.int/bitstream/handle/10665/365980/9789240067172\u003c/li\u003e\n\u003cli\u003eConfusion Matrix - an overview | ScienceDirect Topics [Internet]. [cited 2024 Nov 29]. Available from: https://www.sciencedirect.com/topics/computer-science/confusion-matrix\u003c/li\u003e\n\u003cli\u003eHicks SA, Str\u0026uuml;mke I, Thambawita V, Hammou M, Riegler MA, Halvorsen P, et al. On evaluation metrics for medical applications of artificial intelligence. Sci Rep. 2022;12:5979. \u003c/li\u003e\n\u003cli\u003eSaito T, Rehmsmeier M. The Precision-Recall Plot Is More Informative than the ROC Plot When Evaluating Binary Classifiers on Imbalanced Datasets. PLOS ONE. 2015;10:e0118432. \u003c/li\u003e\n\u003cli\u003ePrequalified In Vitro Diagnostics | WHO - Prequalification of Medical Products (IVDs, Medicines, Vaccines and Immunization Devices, Vector Control) [Internet]. [cited 2024 Oct 24]. Available from: https://extranet.who.int/prequal/vitro-diagnostics/prequalified-vitro-diagnostics\u003c/li\u003e\n\u003cli\u003eVisser T, Ramachandra S, Pothin E, Jacobs J, Cunningham J, Menach AL, et al. A comparative evaluation of mobile medical APPS (MMAS) for reading and interpreting malaria rapid diagnostic tests. Malar J. 2021;20:39. \u003c/li\u003e\n\u003cli\u003eMarti\u0026aacute;\u0026ntilde;ez-Vendrell X, Jim\u0026eacute;nez A, V\u0026aacute;squez A, Campillo A, Incardona S, Gonz\u0026aacute;lez R, et al. Quantification of malaria antigens PfHRP2 and pLDH by quantitative suspension array technology in whole blood, dried blood spot and plasma. Malar J. 2020;19:12. \u003c/li\u003e\n\u003cli\u003eShekalaghe S, Cancino M, Mavere C, Juma O, Mohammed A, Abdulla S, et al. Clinical performance of an automated reader in interpreting malaria rapid diagnostic tests in Tanzania. Malar J. 2013;12:141. \u003c/li\u003e\n\u003cli\u003eOyet C, Roh ME, Kiwanuka GN, Orikiriza P, Wade M, Parikh S, et al. Evaluation of the Deki Reader\u003csup\u003eTM\u003c/sup\u003e, an automated RDT reader and data management device, in a household survey setting in low malaria endemic southwestern Uganda. Malar J. 2017;16:449. \u003c/li\u003e\n\u003cli\u003eKalinga AK, Mwanziva C, Chiduo S, Mswanya C, Ishengoma DI, Francis F, et al. Comparison of visual and automated Deki Reader interpretation of malaria rapid diagnostic tests in rural Tanzanian military health facilities. Malar J. 2018;17:214. \u003c/li\u003e\n\u003cli\u003eCHAI. Case Study: Scoping digital solutions for improving quality of care in the informal private sector [Internet]. Clint. Health Access Initiat. 2022 [cited 2025 Apr 9]. Available from: https://www.clintonhealthaccess.org/case-study/case-study-scoping-digital-solutions-for-improving-quality-of-care-in-the-informal-private-sector/\u003c/li\u003e\n\u003cli\u003evan Duijn SMC, Siteyi AK, Smith S, Milimo E, Stijvers L, Oguttu M, et al. Connected diagnostics to improve accurate diagnosis, treatment, and conditional payment of malaria services in Kenya. BMC Med Inform Decis Mak. 2021;21:233. \u003c/li\u003e\n\u003cli\u003eWHO. WHO prequalification of in vitro diagnostics programme: AdvDx Malaria Pf Rapid Malaria Ag Detection Test [Internet]. Geneva, Switzerland: WHO; 2019. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0345-101-00_AdvDxMalariaDetectionTest_v2.pdf\u003c/li\u003e\n\u003cli\u003eWHO. WHO prequalification of diagnostics programme: Bioline Malaria Ag P.f [Internet]. Geneva, Switzerland; 2021. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0031-012-01_BiolineMalaria_Ag_P-f-v7.pdf\u003c/li\u003e\n\u003cli\u003eWHO. WHO prequalification of in vitro diagnostics: First Response Malaria Antigen P. falciparum (HRP2) Card Test [Internet]. Geneva, Switzerland: WHO; 2018. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0283-010-00_FirstResponseMalaria_v1.pdf\u003c/li\u003e\n\u003cli\u003eWHO. WHO prequalification of in vitro diagnostics programme: ParaHIT f ver. 1.0 rapid test for P. falciparum malaria device [Internet]. Geneva, Switzerland: WHO; 2016. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0062-023-00_ParaHIT-f_RapidTest_v3_0.pdf\u003c/li\u003e\n\u003cli\u003eWHO. WHO prequalification of in vitro diagnostics: STANDARD Q Malaria P.f Ag Test [Internet]. Geneva, Switzerland: WHO; 2021. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0346-117-00_STANDARD-Q-MalariaP-fAg_Test_PQ_v2.0.pdf\u003c/li\u003e\n\u003cli\u003eWHO. WHO prequalificaiton of in vitro diagnostics programme: Bioline Malaria Ag P.f (HRP2/pLDH) [Internet]. Geneva, Switzerland: World Health Organization; 2025. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/bioline-malaria-ag-p.f-hrp2-pldh-pqdx-0209-012-00-public-report-v-7.0.pdf\u003c/li\u003e\n\u003cli\u003eWHO. WHO prequalification of in vitro diagnostics programme: Bioline Malaria Ag P.f/Pan [Internet]. Geneva, Switzerland: WHO; 2022. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/PQDx_0030-012-01_BiolineMalariaAgPfPan_v11.0.pdf\u003c/li\u003e\n\u003cli\u003eWHO. WHO prequalification of in vitro diagnostics: First Response Malaria Ag.pLDH HRP2 Combo Card Test [Internet]. Geneva, Switzerland: World Health Organization; 2025. Available from: https://extranet.who.int/prequal/sites/default/files/whopr_files/First%20Response%20Malaria%20Ag.pLDH%20HRP2%20Combo%20Card%20Test_PQDx%200285-010-00_v2.pdf\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":true,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"malaria-journal","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"malj","sideBox":"Learn more about [Malaria Journal](http://malariajournal.biomedcentral.com/)","snPcode":"12936","submissionUrl":"https://submission.nature.com/new-submission/12936/3","title":"Malaria Journal","twitterHandle":"@malariajournal","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Malaria, rapid diagnostic test, artificial intelligence, digital health, diagnostic accuracy, electronic RDT reader","lastPublishedDoi":"10.21203/rs.3.rs-6645811/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6645811/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe introduction of malaria rapid diagnostic tests (RDTs) has expanded parasitologic confirmation of malaria at all levels of health systems in sub-Saharan Africa (SSA), improving case management and surveillance. However, concerns persist about healthcare worker adherence to results and the accuracy of results recorded in health facility registers. Electronic RDT readers have been proposed to improve the consistency of diagnosis and reporting, though their performance relative to expert human interpretation varies. We assessed the performance of the HealthPulse (Audere, Seattle, WA USA) smartphone application, an artificial intelligence (AI)-based RDT reader, across four countries in SSA.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn 2023, the Malaria Rapid Diagnostic Test Capture and Reporting Assessment (MaCRA) was implemented in health facilities in Benin, Côte d’Ivoire, Nigeria, and Uganda. Study staff collected images of malaria RDTs using the HealthPulse app after healthcare workers performed and interpreted the tests. A trained panel of external reviewers interpreted the RDT images, serving as the reference standard. RDTs were classified as positive, negative, invalid or indeterminate. We evaluated classification accuracy using recall, precision, and F1 scores (harmonic mean of recall and precision), and applied logistic regression to assess factors influencing AI performance across countries, RDT products, presence of faint lines and anomalies.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOut of 110,843 RDT images collected, 110,231 (99.4%) were included in the analysis. The AI algorithm demonstrated high overall accuracy (96.8%) and a F1 score of 96.6% compared to panel interpretations. Recall and precision were \u0026gt;96% for positive and negative outcomes but much lower for invalid (recall: 84.5%; precision: 42.9%) and indeterminate classifications (recall: 0.7%; precision: 2.3%). AI performance varied by country, RDT product, and presence of faint lines. When test lines were faint, the OR of both positive recall (adjusted OR 0.01; 95% CI 0.00, 0.01) and negative recall (adjusted OR 0.20; 95% CI 0.11, 0.35) by the AI algorithm were reduced.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe HealthPulse AI algorithm demonstrated high agreement with a trained panel in interpreting malaria RDT images across diverse settings. However, reduced performance for invalid and indeterminate results and varying performance by country and RDT product highlights the need for further refinement. The HealthPulse app shows potential as a supportive tool in research and training.\u003c/p\u003e","manuscriptTitle":"Evaluating the performance of an artificial intelligence-based electronic reader for malaria rapid diagnostic tests across four sub-Saharan African countries","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-30 03:52:33","doi":"10.21203/rs.3.rs-6645811/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-06-10T11:58:51+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-10T09:06:01+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-06T12:15:13+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"194816377008852291898956198012675129415","date":"2025-05-30T15:03:10+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-05-26T12:07:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"200774686644382207071681488976017362089","date":"2025-05-22T22:06:37+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"28307365685175452559597945452000427600","date":"2025-05-22T06:11:58+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"55932173306428118317757807329594397135","date":"2025-05-22T00:52:16+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-05-21T15:54:33+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-05-13T15:57:19+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-05-13T15:52:54+00:00","index":"","fulltext":""},{"type":"submitted","content":"Malaria Journal","date":"2025-05-12T10:37:51+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"malaria-journal","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"malj","sideBox":"Learn more about [Malaria Journal](http://malariajournal.biomedcentral.com/)","snPcode":"12936","submissionUrl":"https://submission.nature.com/new-submission/12936/3","title":"Malaria Journal","twitterHandle":"@malariajournal","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c87f6b26-f321-42a4-af54-7800a1e9dab9","owner":[],"postedDate":"July 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-10-06T16:06:12+00:00","versionOfRecord":{"articleIdentity":"rs-6645811","link":"https://doi.org/10.1186/s12936-025-05522-3","journal":{"identity":"malaria-journal","isVorOnly":false,"title":"Malaria Journal"},"publishedOn":"2025-09-30 15:56:59","publishedOnDateReadable":"September 30th, 2025"},"versionCreatedAt":"2025-07-30 03:52:33","video":"","vorDoi":"10.1186/s12936-025-05522-3","vorDoiUrl":"https://doi.org/10.1186/s12936-025-05522-3","workflowStages":[]},"version":"v1","identity":"rs-6645811","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6645811","identity":"rs-6645811","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}