A pragmatic randomized controlled trial of artificial intelligence (AI)-based predictive analytics monitoring for early detection of clinical deterioration

A pragmatic randomized controlled trial of artificial intelligence (AI)-based predictive analytics monitoring for early detection of clinical deterioration | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (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];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search A pragmatic randomized controlled trial of artificial intelligence (AI)-based predictive analytics monitoring for early detection of clinical deterioration View ORCID Profile Jessica Keim-Malpass , Sarah J Ratcliffe , Matthew T Clark , Katy N Krahn , Oliver J Monfredi , Susan Hamil , Gholamreza Yousevfand , Marieke K Jones , Ashley Nelson , Liza P Moorman , View ORCID Profile J. Randall Moorman , Jamieson M Bourque doi: https://doi.org/10.1101/2025.01.20.25320838 Jessica Keim-Malpass 1 Center for Advanced Medical Analytics, School of Medicine, University of Virginia , Charlottesville, VA USA 2 Department of Pediatrics, Hematology-Oncology Division, School of Medicine, University of Virginia , Charlottesville, VA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jessica Keim-Malpass For correspondence: jlk2t{at}virginia.edu Sarah J Ratcliffe 3 Department of Public Health Sciences, Biostatistics Division, University of Virginia , Charlottesville, VA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Matthew T Clark 4 Nihon Kohden Digital Health Solutions , Irvine, CA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Katy N Krahn 1 Center for Advanced Medical Analytics, School of Medicine, University of Virginia , Charlottesville, VA USA 5 Department of Internal Medicine, Cardiovascular Division, School of Medicine, University of Virginia , Charlottesville, VA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Oliver J Monfredi 6 Department of Internal Medicine, Cardiovascular Division, School of Medicine, Johns Hopkins University , Baltimore, MD, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Susan Hamil 1 Center for Advanced Medical Analytics, School of Medicine, University of Virginia , Charlottesville, VA USA 5 Department of Internal Medicine, Cardiovascular Division, School of Medicine, University of Virginia , Charlottesville, VA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Gholamreza Yousevfand 1 Center for Advanced Medical Analytics, School of Medicine, University of Virginia , Charlottesville, VA USA 5 Department of Internal Medicine, Cardiovascular Division, School of Medicine, University of Virginia , Charlottesville, VA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Marieke K Jones 3 Department of Public Health Sciences, Biostatistics Division, University of Virginia , Charlottesville, VA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Ashley Nelson 5 Department of Internal Medicine, Cardiovascular Division, School of Medicine, University of Virginia , Charlottesville, VA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Liza P Moorman 4 Nihon Kohden Digital Health Solutions , Irvine, CA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site J. Randall Moorman 1 Center for Advanced Medical Analytics, School of Medicine, University of Virginia , Charlottesville, VA USA 5 Department of Internal Medicine, Cardiovascular Division, School of Medicine, University of Virginia , Charlottesville, VA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for J. Randall Moorman Jamieson M Bourque 1 Center for Advanced Medical Analytics, School of Medicine, University of Virginia , Charlottesville, VA USA 5 Department of Internal Medicine, Cardiovascular Division, School of Medicine, University of Virginia , Charlottesville, VA USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Abstract Full Text Info/History Metrics Data/Code Preview PDF ABSTRACT Background This pragmatic randomized controlled trial aimed to assess the effect of a passive display of artificial intelligence (AI)-based predictive analytics on hours free of clinical deterioration events among medical and surgical patients in an acute care cardiology medical-surgical ward. Methods 10,422 inpatient visits were randomly assigned by cluster to the intervention group of a display of risk trajectories or to a control group of usual medical care. The trial was undertaken on an 85-bed inpatient cardiology and cardiac surgery ward of an academic hospital with a substantial implementation and education plan. This was a passive display with no specific response mandated. The primary analysis compared events of clinical deterioration (death, emergent ICU transfer, emergent endotracheal intubation, cardiac arrest, or emergent surgery) and compared mortality 21 days after admission. Results Patients with a large spike in risk score had, on average, twice the length of hospital stay (6.8 compared to 3.4 days). There was no change in the primary outcome between groups. Among those who had a clinical event, there were more event-free hours in the intervention/display-on group compared to the standard-of-care/display-off, but this did not approach statistical significance. 11% of patients were transferred into or out of display beds, a censoring event removing them from the analysis, thereby undermining aspects of the randomized nature of the study. Conclusion Predictive analytics monitoring incorporating continuous cardiorespiratory monitoring and displays of risk trajectories coupled with an education plan did not improve patient outcomes or reduce deaths. While necessary to conduct the study, the pragmatic design allowed for significant movement towards intervention/displayed beds for sicker patients. Design considerations in the future must focus on understanding clinicians’ interpretation, care processes, and communication practices. Trial registration NCT04359641 Introduction Unanticipated clinical deterioration of hospital patients remains a pernicious problem that contributes to in-hospital morbidity and mortality. 1 – 4 , 5 Often following the clinical deterioration event, clinicians who review these cases can recognize trends in vital signs, laboratory tests, and other electronic health record elements that, in retrospect, might have raised clinical suspicion of illness. 6 For example, the coincidence of rising heart rate and falling blood pressure should suggest incipient shock, but often the changes in physiological parameters are subtle and insidious. Automated detection of such signatures of illness might inform clinicians of patients who could benefit from earlier assessment and proactive clinical action. 1 , 7 Since its inception more than 20 years ago in the neonatal intensive care unit (ICU), 8 , 9 predictive analytics monitoring has been increasingly recognized for its potential to improve patient outcomes by early detection of deterioration. Artificial intelligence (AI) and machine learning methods support synthesizing unprocessed cardiorespiratory monitoring data, laboratory results and clinician-entered vital signs into computational models that can visually represent the estimation of the risk of clinical deterioration. 10 Since AI’s new and sometimes opaque methodologies lie at its core, predictive analytics monitoring has been held to high standards for transparency. For example, there are numerous guidelines and opinion pieces to inform the developer and the user. 11 – 13 , 14 Incontrovertible proof of the benefit of predictive analytics monitoring in the form of large randomized clinical trials has been lacking. While a meta-analysis of artificial intelligence-based sepsis alerts showed an average mortality reduction of 30%, it included only six randomized trials with 2938 patients. 15 To this list, we can add the original trial in this field, that of heart rate characteristics monitoring for neonatal sepsis in 3003 premature infants, which showed a 22% relative mortality reduction when risk estimates were displayed to clinicians. 16 Here, we extend the work in neonates to evaluate the impact of predictive analytics monitoring on events of clinical deterioration in a heterogeneous group of adults in acute care cardiac medical-surgical wards. We test the use of an AI-based risk analytic, Continuous Monitoring of Event Trajectories (CoMET, Nihon Kohden Digital Health Solutions, Irvine, CA), that visually displays risk estimates of multiple events of clinical deterioration every 15 minutes. We hypothesized that the passive display of a visual risk analytic would facilitate early identification of patients at risk for impending adverse outcomes and increase the hours free from clinical deterioration. Methods Study design We conducted a cluster-randomized controlled trial (RCT) among patients admitted to 85-bed acute care cardiology medical-surgical wards. Eleven clusters were defined by room number, and patients were randomized to receive either the intervention (passive display of CoMET) or usual care (standard of care). Randomization began on January 4, 2021, and ended on October 4, 2022. Due to the pragmatic nature and replication of real-world patient contexts, all patients admitted to these floors during the study period were enrolled in the study. Clusters were re-randomized every 2 months following a Latin square design within each stratum to ensure a balanced treatment assignment. Full pre-registered study protocol details and additional randomization details can be found in 17 , and the clinical trial was registered ( NCT04359641 ). The University of Virginia (UVA) IRB approved the study with a waiver of informed consent (IRB#22196). A Data Safety and Monitoring Board reviewed intercurrent results. The results are reported in accordance with the CONSORT-AI guidelines. 12 Patient care units and populations The Medical-Cardiology unit predominantly comprised acute cardiology patients admitted with decompensated congestive heart failure, acute coronary syndrome, arrhythmias such as atrial fibrillation, and other cardiac diagnoses. The Medical-Surgical unit was historically the same as Medical-Cardiology. Immediately preceding the study, though, 11 of the 20 beds were repurposed for patients with solid organ transplants under the care of a dedicated transplant team. They differed from the patients and the practices of the cardiology and cardiac surgery beds. The baseline distributions of vital signs and laboratory test results, for example, differed, with a bias toward findings of chronic kidney and liver failure in the transplant patients. Due to the pragmatic nature of this study design in real-world contexts, they were left within the study population. The Cardio-Thoracic Surgery unit predominantly comprised patients awaiting discharge. Risk displays We used the CoMET system to show two risk trajectories, calculated from previously published logistic regression machine learning analyses relating continuous cardiorespiratory monitoring parameters, nurse-charted vital signs, and laboratory test results to outcome events on these wards from 2013 to 2015. 18 The unit of measurement of the risk estimate is the fold-increase in the probability of the event in the next 24 hours compared to the average probability. Thus, a patient with a score of 1 has the average risk, a patient with a score of 2 has twice the average risk, and so on. The plot shows a comet-like icon for each patient with the bed number in the head; the tail is 3 hours long. Additional details on the model inputs and model performance, including excellent calibration, are in the following publication. 19 Intervention implementation details We followed principles derived from the prior implementation of predictive analytics monitoring in ICUs and non-ICUs at the UVA Hospital and a regional hospital. 20 Each of the three units identified an interprofessional team of clinical champions to serve as super users, oversee implementation, and ensure staff education. Our team of educators (PIs, CoMET educators, clinical research coordinators (CRCs), and other study investigators) provided sessions with the clinical champions and other clinician groups in each unit before randomization and during the first quarter. We encouraged the integration of the CoMET scores into routine nurse vital sign assessments, rounding discussions, hand-off reports, and escalation protocols. Our recommended protocol for reporting and responding to changes in the CoMET scores included nurse notification of physicians or advanced practice providers of an increase in score by 2 units over a three-hour period and physician evaluation of the patient as appropriate. A CRC attended morning rounds on each unit on a rotating basis to show the display and answer questions about interpreting score trajectories. The CoMET scores flowed through to the electronic medical record every hour and were recorded within the vital sign flowsheet. 17 Study outcomes The primary outcome was the number of hours free of clinical deterioration (adverse clinical events including: an emergent ICU transfer, emergent intubation, cardiac arrest, emergent surgery, and death) within 21 days of admission. A maximum score was 21 event-free days (504 hours), and patients who were successfully discharged from the hospital prior to the 21 days without an event were awarded all 504 hours. Patients who died during admission were counted as having 0 event-free days/0 hours. Patients with non-emergent ICU transfers, non-emergent/planned surgeries, or changes in bed assignments that caused a change in arm assignment were censored. Dedicated clinician CRCs adjudicated each case of clinical deterioration/study endpoint event. For any events that were ambiguous, the PIs of the study (JKM and JMB) blindly adjudicated outcomes through a weekly meeting. Statistical analysis We used a generalized estimating equation with a zero-inflated negative binomial to analyze event-free hours and a Cox proportional hazards model to analyze time-until-first-critical events. The model handled the correlation of multiple hospital admissions from the same patient, within-unit clusters within the study design, and differences in the amount of time at risk for censored patients. An intent-to-treat analytic strategy was utilized where all randomized patients were included for the primary endpoint comparison between the intervention and usual care. As a post-hoc analysis, we also analyzed the primary outcome in an ‘at-risk’ cohort, or those patients who experienced a rise in 2 on either axis of the CoMET display during their hospitalization. Results Over the 22 months of randomization, there were 10,422 patient visits within our study cohort ( Figure 1 ). They were majority male, White, non-Hispanic, and on average 65 years of age with relatively low hospital length of stays ( Table 1 ). Model performance is reported separately 19 , but overall the predictive models performed well and outperformed a common early warning score, qSOFA. View this table: View inline View popup Download powerpoint Table 1: Admission Characteristics of the Cohort and At-Risk Subgroup by Arm Download figure Open in new tab Figure 1. CONSORT randomization information The vast majority of participants were discharged successfully and had the maximum amount of event-free hours ( Table 2 ). There was no statistically significant difference in the primary outcome by study arm in both the full cohort and those at risk for clinical deterioration. Among those who had a clinical event, there were more event-free hours in the intervention/display-on group compared to the standard-of-care/display-off, but this did not approach statistical significance. Censoring events and components meeting the endpoint of clinical deterioration are found in Table 2 . The estimated marginal means from the negative binomial generalized estimating equation are found in Figure 2 . View this table: View inline View popup Table 2: Study Outcomes for the Cohort and At Risk Subjects Download figure Open in new tab Figure 2: Estimated marginal mean (and 95% CI) event-free hours by arm and population Bed movement was common within this RCT (n=382 went from usual care to CoMET display, and n=419 went from CoMET display to usual care). In a post-hoc analysis, we found higher relative risk of clinical deterioration among those transferred from a non-display to a display bed. This finding indicates that clinicians made decisions that undermined the cluster RCT design by preferentially placing sicker patients into the CoMET display beds, presumably to obtain additional information about their underlying physiological instability. ( Figure 3 ). Download figure Open in new tab Figure 3: Time zero is transfer from one arm to another. The line is the median and the shaded region is the IQR. Blue are those patients transferred from a display-off to display-on bed and their overall relative risk of clinical deterioration is higher, representing a sicker population. Discussion We conducted a large randomized controlled trial to test the impact of predictive analytics monitoring on cardiology and cardiac surgery patients in in-patient acute care wards. Our major findings are that a display of the trajectories of the risk of imminent deterioration without alerts was not associated with an increase in hours free of clinical deterioration. We found that there was an increase in relocating sicker patients from usual care to CoMET display-on, thereby distorting the RCT design. We note this study was undertaken during the early stages of the COVID-19 pandemic in our area, which disrupted staffing and resource utilization 10 and was associated with a lower event rate than our pre-COVID baseline. We justify using continuous cardiorespiratory monitoring in the predictive models because of the causality leakage of clinician-ordered tests. For example, the presence of a serum lactate measurement in the chart bespeaks clinical suspicion of sepsis. It excels as a predictor, but not because it detects changes before clinicians. Likewise, the use of pressors, fluid administration, and blood cultures as predictors rather than endpoints allows causality leakage, artificially inflating model performance. Kamran and coworkers recently showed that the Epic Sepsis Model had no predictive capability at all when assessed prior to any clinician suspicion or action, that is, before fluids, antibiotics, and lactate tests. 21 We justify the randomized clinical trial approach to the study of AI-based risk analytics. We find irreducible uncertainty in other trial designs and results, even our own demonstration of a 50% reduction in septic shock in a surgical ICU after the introduction of CoMET trained for urgent unplanned intubation for acute respiratory failure and hemorrhage requiring large transfusion. 22 While a control medical ICU without CoMET monitoring had no consistent benefit over the same time period, the absence of randomization has allowed skepticism of the before-and-after study results. We especially endorse the simple and pragmatic nature of the current trial -we enrolled everyone to mimic complex real-world acute care environments and used important endpoints to clinicians and patients. Other studies have used non-randomized designs. For example, Winslow, Edelson, Churpek, and coworkers found reduced mortality throughout the hospital after initiating their early warning system. 23 The magnitude was substantial but was the same even for patients with low risks who were never alerted. Adams, Henry, Saria, and coworkers used a machine learning-based sepsis early warning system that included mechanical ventilation, lactate, and vasopressors as sepsis predictors and placed passive alerts in the EHR, but only after verifiable symptoms were present. They found reduced mortality in the 1.2% of patients who met retrospective sepsis criteria and whose alerts were acknowledged within three hours but did not present outcomes of the other patients. 24 , 25 Groups led by Shimabukuro 26 and Watkinson 27 performed small randomized trials with divergent results. Escobar, Liu and coworkers used a non-randomized stepped wedge approach to show a convincing mortality reduction with an EHR-based early warning score in 19 hospitals. 28 They employed an informed intermediary to receive alerts and to review the clinical record before alerting bedside clinicians. Cluster randomization was chosen because clinicians needed to easily recognize who was in the intervention group. We did not want to add additional burden to the complexity of the acute care environment and did not want to falsely reassure a clinician who did not see a rising score to assume the patient was in the intervention group. Although we have previously tested the positive predictive value of creating a threshold-based alert from CoMET 1 , we did not want to add to clinician burden or create additional markers of alarm fatigue and decided to offer suggestions for response without an active alert strategy. We approached education and implementation of novel analytics in real-world environments within the context of the COVID-19 pandemic and wanted to minimize any contribution to clinician burnout. 29 Implementation considerations focused on promoting trust, transparency, integration with workflow, prolonged stakeholder engagement, tailored education embedded within hospital norms, supporting actionability through integration within clinical care, and sustainability through ongoing supportive interactions. 29 AI clinical trials call for an understanding of new and separate design elements not previously contemplated. There are several elements of this type of intervention such asthe use of passive displays that warrant further study and have RCT design implications when studying AI-based predictive analytics: the proactive nature of the clinical action response, the rarity of the event of interest, and the complexity of implementation within evolving acute care environments. While the pragmatic design allowed for implementation in real-world environments, this also meant there was heterogeneity in patient populations, unit work environments, and clinician responses. The real-world context of clinical trials such as this implies that the implementation of AI-based risk analytics must also be adaptive to ensure long-term integration and practice flexibility across different work environments. Understanding the cognitive interpretation of scores in real-time, the response through diverse care processes, and communication patterns involving AI scores preceding clinical events is an area in particular need of study. In hindsight, hours free of events of clinical deterioration may not have been the most sensitive primary outcome. Outcomes specific to the at-risk population and subsequent intensive care unit trajectories may have been more appropriate, given the overall rarity of our event. Data Availability All data produced in the present study are available upon reasonable request to the authors Footnotes Funding: This study was funded by the Frederick Thomas Advanced Medical Analytics Fund, University of Virginia and AHRQ R01HS028803 (Keim-Malpass/Bourque MPI). The investigators received in-kind support from Nihon Kohden Digital Health Solutions for use and support of the CoMET system. The funders had no direct role/oversight of this published work. COI: MC is an employee of Nihon Kohden Digital Health Solutions. LPM and JRM are consultants for Nihon Kohden Digital Health Solutions. JKM owns equity shares of ArteraAI and JRM owns equity shares of Medical Predictive Science Corporation, whose products are not discussed in this unrelated work. The other authors do not express conflicts of interest. Bibliography 1. ↵ Keim-Malpass J , Clark MT , Lake DE , Moorman JR . Towards development of alert thresholds for clinical deterioration using continuous predictive analytics monitoring . J Clin Monit Comput 2020 ; 34 ( 4 ): 797 – 804 . OpenUrl PubMed 2. Smith MEB , Chiovaro JC , O’Neil M , et al. Early warning system scores for clinical deterioration in hospitalized patients: a systematic review . Ann Am Thorac Soc 2014 ; 11 ( 9 ): 1454 – 65 . OpenUrl CrossRef PubMed 3. Escobar GJ , LaGuardia JC , Turk BJ , Ragins A , Kipnis P , Draper D. Early detection of impending physiologic deterioration among patients who are not in intensive care: development of predictive models using data from an automated electronic medical record . J Hosp Med 2012 ; 7 ( 5 ): 388 – 95 . OpenUrl CrossRef PubMed 4. ↵ Escobar GJ , Greene JD , Gardner MN , Marelich GP , Quick B , Kipnis P. Intra-hospital transfers to a higher level of care: contribution to total hospital and intensive care unit (ICU) mortality and length of stay (LOS) . J Hosp Med 2011 ; 6 ( 2 ): 74 – 80 . OpenUrl CrossRef PubMed 5. ↵ Curtis K , Sivabalan P , Bedford DS , et al. Treatments costs associated with inpatient clinical deterioration . Resuscitation 2021 ; 166 : 49 – 54 . OpenUrl PubMed 6. ↵ Blackwell JN , Keim-Malpass J , Clark MT , et al. Early Detection of In-Patient Deterioration: One Prediction Model Does Not Fit All . Crit Care Explor 2020 ; 2 ( 5 ): e0116 . OpenUrl 7. ↵ Moss TJ , Lake DE , Calland JF , et al. Signatures of subacute potentially catastrophic illness in the ICU: model development and validation . Crit Care Med 2016 ; 44 ( 9 ): 1639 – 48 . OpenUrl CrossRef PubMed 8. ↵ Randall Moorman J. The principles of whole-hospital predictive analytics monitoring for clinical medicine originated in the neonatal ICU . npj Digital Med 2022 ; 5 ( 1 ): 41 . OpenUrl 9. ↵ Fairchild KD , Schelonka RL , Kaufman DA , et al. Septicemia mortality reduction in neonates in a heart rate characteristics monitoring trial . Pediatr Res 2013 ; 74 ( 5 ): 570 – 5 . OpenUrl CrossRef PubMed Web of Science 10. ↵ Keim-Malpass J , Moorman LP . Nursing and precision predictive analytics monitoring in the acute and intensive care setting: An emerging role for responding to COVID-19 and beyond . International Journal of Nursing Studies Advances 2021 ; 3 : 100019 . OpenUrl PubMed 11. ↵ Collins GS , Dhiman P , Andaur Navarro CL , et al. Protocol for development of a reporting guideline (TRIPOD-AI) and risk of bias tool (PROBAST-AI) for diagnostic and prognostic prediction model studies based on artificial intelligence . BMJ Open 2021 ; 11 ( 7 ): e048008 . OpenUrl Abstract / FREE Full Text 12. ↵ Liu X , Cruz Rivera S , Moher D , Calvert MJ , Denniston AK , SPIRIT-AI and CONSORT-AI Working Group. Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension . Nat Med 2020 ; 26 ( 9 ): 1364 – 74 . OpenUrl CrossRef PubMed 13. ↵ Rivera SC , Liu X , Chan A-W , Denniston AK , Calvert MJ , SPIRIT-AI and CONSORT-AI Working Group. Guidelines for clinical trial protocols for interventions involving artificial intelligence: the SPIRIT-AI Extension . BMJ 2020 ; 370 : m3210 . OpenUrl Abstract / FREE Full Text 14. ↵ Vasey B , Nagendran M , Campbell B , et al. Reporting guideline for the early-stage clinical evaluation of decision support systems driven by artificial intelligence: DECIDE-AI . Nat Med 2022 ; 28 ( 5 ): 924 – 33 . OpenUrl CrossRef PubMed 15. ↵ Zhang Z , Chen L , Xu P , et al. Effectiveness of automated alerting system compared to usual care for the management of sepsis . npj Digital Med 2022 ; 5 ( 1 ): 101 . OpenUrl 16. ↵ Moorman JR , Carlo WA , Kattwinkel J , et al. Mortality reduction by heart rate characteristic monitoring in very low birth weight neonates: a randomized trial . J Pediatr 2011 ; 159 ( 6 ): 900 - 6.e1 . OpenUrl CrossRef PubMed Web of Science 17. ↵ Keim-Malpass J , Ratcliffe SJ , Moorman LP , et al. Predictive Monitoring-Impact in Acute Care Cardiology Trial (PM-IMPACCT): Protocol for a Randomized Controlled Trial . JMIR Res Protoc 2021 ; 10 ( 7 ): e29631 . OpenUrl 18. ↵ Moss TJ , Clark MT , Calland JF , et al. Cardiorespiratory dynamics measured from continuous ECG monitoring improves detection of deterioration in acute care patients: A retrospective cohort study . PLoS ONE 2017 ; 12 ( 8 ): e0181448 . OpenUrl CrossRef PubMed 19. ↵ Keim-Malpass J , Moorman LP , Moorman JR , et al. Prospective validation of clinical deterioration predictive models prior to intensive care unit transfer among patients admitted to acute care cardiology wards . Physiol Meas 2024 ; 45 ( 6 ). 20. ↵ Prudente Moorman L. Principles for Real-World Implementation of Bedside Predictive Analytics Monitoring . Applied Clinical Informatics [Internet] 2021 [cited 2021 Sep 20];Available from : DOI 10.1055/s-0041-1735183 OpenUrl CrossRef 21. ↵ Unable to find information for 16016370 . 22. ↵ Ruminski CM , Clark MT , Lake DE , et al. Impact of predictive analytics based on continuous cardiorespiratory monitoring in a surgical and trauma intensive care unit . J Clin Monit Comput 2019 ; 33 ( 4 ): 703 – 11 . OpenUrl CrossRef PubMed 23. ↵ Winslow CJ , Edelson DP , Churpek MM , et al. The impact of a machine learning early warning score on hospital mortality: A multicenter clinical intervention trial . Crit Care Med 2022 ; 50 ( 9 ): 1339 – 47 . OpenUrl PubMed 24. ↵ Adams R , Henry KE , Sridharan A , et al. Prospective, multi-site study of patient outcomes after implementation of the TREWS machine learning-based early warning system for sepsis . Nat Med 2022 ; 28 ( 7 ): 1455 – 60 . OpenUrl CrossRef PubMed 25. ↵ Henry KE , Kornfield R , Sridharan A , et al. Human-machine teaming is key to AI adoption: clinicians’ experiences with a deployed machine learning system . npj Digital Med 2022 ; 5 ( 1 ): 97 . OpenUrl 26. ↵ Shimabukuro DW , Barton CW , Feldman MD , Mataraso SJ , Das R. Effect of a machine learning-based severe sepsis prediction algorithm on patient survival and hospital length of stay: a randomised clinical trial . BMJ Open Respir Res 2017 ; 4 ( 1 ): e000234 . OpenUrl Abstract / FREE Full Text 27. ↵ Watkinson PJ , Barber VS , Price JD , Hann A , Tarassenko L , Young JD . A randomised controlled trial of the effect of continuous electronic physiological monitoring on the adverse event rate in high risk medical and surgical patients . Anaesthesia 2006 ; 61 ( 11 ): 1031 – 9 . OpenUrl CrossRef PubMed Web of Science 28. ↵ Escobar GJ , Liu VX , Schuler A , Lawson B , Greene JD , Kipnis P. Automated Identification of Adults at Risk for In-Hospital Clinical Deterioration . N Engl J Med 2020 ; 383 ( 20 ): 1951 – 60 . OpenUrl CrossRef PubMed 29. ↵ Moorman LP . Principles for real-world implementation of bedside predictive analytics monitoring . Appl Clin Inform 2021 ; 12 ( 4 ): 888 – 96 . OpenUrl PubMed View the discussion thread. Back to top Previous Next Posted January 22, 2025. Download PDF Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following A pragmatic randomized controlled trial of artificial intelligence (AI)-based predictive analytics monitoring for early detection of clinical deterioration Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share A pragmatic randomized controlled trial of artificial intelligence (AI)-based predictive analytics monitoring for early detection of clinical deterioration Jessica Keim-Malpass , Sarah J Ratcliffe , Matthew T Clark , Katy N Krahn , Oliver J Monfredi , Susan Hamil , Gholamreza Yousevfand , Marieke K Jones , Ashley Nelson , Liza P Moorman , J. Randall Moorman , Jamieson M Bourque medRxiv 2025.01.20.25320838; doi: https://doi.org/10.1101/2025.01.20.25320838 Share This Article: Copy Citation Tools A pragmatic randomized controlled trial of artificial intelligence (AI)-based predictive analytics monitoring for early detection of clinical deterioration Jessica Keim-Malpass , Sarah J Ratcliffe , Matthew T Clark , Katy N Krahn , Oliver J Monfredi , Susan Hamil , Gholamreza Yousevfand , Marieke K Jones , Ashley Nelson , Liza P Moorman , J. Randall Moorman , Jamieson M Bourque medRxiv 2025.01.20.25320838; doi: https://doi.org/10.1101/2025.01.20.25320838 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Health Informatics Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (299) Cardiovascular Medicine (4423) Dentistry and Oral Medicine (443) Dermatology (382) Emergency Medicine (607) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1507) Epidemiology (15219) Forensic Medicine (30) Gastroenterology (1123) Genetic and Genomic Medicine (6587) Geriatric Medicine (667) Health Economics (997) Health Informatics (4524) Health Policy (1368) Health Systems and Quality Improvement (1612) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15910) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (145) Nephrology (667) Neurology (6588) Nursing (345) Nutrition (998) Obstetrics and Gynecology (1143) Occupational and Environmental Health (956) Oncology (3331) Ophthalmology (970) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (663) Pediatrics (1690) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5438) Public and Global Health (9218) Radiology and Imaging (2195) Rehabilitation Medicine and Physical Therapy (1369) Respiratory Medicine (1195) Rheumatology (593) Sexual and Reproductive Health (709) Sports Medicine (529) Surgery (709) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ff8439e1cf4300f',t:'MTc3OTQxNTYxMw=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();