Evaluating Cognitive Load in Clinical Workflows Highlights Leverage Points for Guideline-Concordant Statin Initiation

Evaluating Cognitive Load in Clinical Workflows Highlights Leverage Points for Guideline-Concordant Statin Initiation | 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 Cognitive Load in Clinical Workflows Highlights Leverage Points for Guideline-Concordant Statin Initiation Ratnalekha V. N. Viswanadham, Yuhan Cui, Priyanka Solanki, Nicole Redfern, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7632374/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 02 Apr, 2026 Read the published version in BMC Medical Informatics and Decision Making → Version 1 posted 11 You are reading this latest preprint version Abstract Introduction Linking EHR use to care quality offers insights for interventions to improve guideline adherence and close care gaps. We examine how EHR metadata can measure cognitive load in primary care providers during statin prescribing and identify points of cognitive load in the EHR workflow. Methods EHR primary care encounter data from a large academic health system in 2024 were retrospectively extracted. We identified adult patients who met the criteria for statin initiation and calculated their ASCVD risk scores. Cognitive load metrics were derived from EHR metadata. Logistic regressions evaluate associations between cognitive load and statin initiation, adjusting for patient covariates and provider fixed effects. Gradient-boosted forests and SHAP values identified key EHR events and cognitive load associated with the initiation of statin therapy. Results Longer encounter duration increased the likelihood of statin initiation, whereas more time spent per EHR event decreased it. Non-linear effects were observed for loop count and distinct event count: the probability of initiation decreased with increasing loop counts up to 93.9 loops, then increased beyond this threshold. For distinct events, the initiation probability increased up to approximately 18 events and declined at higher counts. In a gradient-boosted decision tree model, average time per event was the strongest predictor (72.2% relative contribution). Additional positive predictors included the time spent reviewing lab results and on suggested medication order sets. Modifying the order list and looping back to it were negatively associated with statin initiation. Discussion EHR metadata can associate cognitive load with appropriate clinical behavior, finding nonlinear relations between cognitive load and statin initiation rates. This work highlights the need to optimize EHR systems to reduce cognitive burden and support clinical decision-making. Connecting cognitive load to prescribing behavior gives insight into how workflow adjustments and enhanced decision support can improve adherence to guidelines and patient care. statin clinical decision support EHR process mining cognitive load audit log data machine learning Figures Figure 1 Figure 2 1 Introduction 1.0 Research Goals and Hypotheses The objectives of this study were to assess how electronic health record (EHR) metadata can be used to measure cognitive load within primary care workflows related to initiating guideline-concordant statin therapy. We pursued two aims. First, we evaluated whether cognitive load metrics were associated with the likelihood of appropriate statin initiation. Second, we explored whether specific EHR workflow events could be linked to cognitive load patterns that predict statin initiation. We hypothesized that providers adhering to statin guidelines would exhibit higher cognitive effort and distinct EHR interaction patterns compared to those who did not. We expected this increased cognitive load to manifest as: (a) repeated actions to gather necessary information, (b) longer time spent per EHR activity, (c) more overall actions within the EHR, and (d) extended encounter duration. The second aim was exploratory, focusing on developing methods to connect EHR event–level cognitive load metrics with statin initiation behavior. 1.1 Literature contribution (Table 1 ) Table 1 Literature Contribution Table Problem or Issue How can electronic health record (EHR) metadata identify ways to close clinical care gaps? What are EHR workflow cognitive load differences between providers who give appropriate care and providers who do not? What is Already Known Statins reduce cardiovascular risk but are under prescribed despite guidelines. Clinical decision support systems aim to improve prescribing but have low adoption. Cognitive load in providers can lead to errors and decision delays. EHR metadata captures clinician behavior and cognitive load, but few studies link it to specific clinical actions. What this Paper Adds A use case to associate EHR metadata metrics with care quality in healthcare. Who would benefit from the new knowledge in this paper Healthcare systems aiming to leverage EHR metadata to improve healthcare delivery. Health informaticians interested in leveraging data-driven approaches to improve clinical workflows. Behavioral scientists interested in understanding process workflows and sources of cognitive load in workflows. Policymakers improving quality of care through targeted interventions in EHR design and clinical guidelines. 2 Related Work 2.1 Clinical Context: Statin Prescribing for ASCVD Risk Prevention Cardiovascular disease remains the leading cause of mortality globally, and failure to appropriately prescribe statins represents a missed opportunity for primary prevention at a population level.[ 1 ] Statins reduce the risk of major adverse cardiovascular events and mortality.[ 2 ] However, primary care providers fail to prescribe statin therapy for about half of patients meeting guideline criteria for initiation.[ 3 – 8 ] Among patients prescribed a statin, about two-thirds receive a lower-than-optimal dose.[ 5 ] To ensure patients are appropriately prescribed statins, providers must recognize that a patient meets one of four indications, often requiring calculating a patient’s risk of atherosclerotic cardiovascular disease (ASCVD) and prescribing a guideline-approved statin at the correct dose. Individualized risk assessment requires collecting elements from the patient’s medical history, vital signs, and laboratory results, and then using a risk calculator. The EHR presents opportunities to develop clinical decision support systems (CDSSs) that support the recognition, assessment, and management of ASCVD risk, like alerts and pre-populated medication orders. However, generally low provider adoption has limited the clinical impact of current CDSSs designed to improve guideline-concordant statin prescribing.[ 9 – 13 ] Additionally, evolving guidelines and varying thresholds for treatment based on age and comorbidities further complicate the prescribing process, especially without automated support. 2.2 EHR Metadata to Evaluate EHR Utilization Cognitive load can influence the quality of healthcare decision-making by increasing the likelihood of delayed action, reliance on cognitive shortcuts, and misalignment with evidence-based care.[ 14 , 15 ] In the EHR environment, poorly integrated workflows and excessive demands on attention have been associated with clinician burnout and suboptimal patient outcomes.[ 16 ] Quantifying cognitive load in routine practice is therefore critical for understanding its role in decision-making and identifying opportunities for improvement.[ 17 ] Traditional approaches to studying clinician-EHR interactions, such as time and motion studies, direct observation, surveys, and interviews, are resource-intensive and prone to biases like the Hawthorne Effect, where awareness of being observed alters behavior.[ 18 ] These methods may also fail to capture subtle but important real-time decision points. EHR metadata is an automated tracking feature that logs behavior within the system (such as adding, deleting, or querying information) along with contextual information (date/time, user, and patient). EHR certification mandates this tracking for security and privacy purposes. As of 2017, 96 percent of hospitals and 80 percent of private practices that used EHRs were compliant with these criteria.[ 19 ] Specifically, EHR audit log metadata, collected for HIPAA compliance to track what providers view and click in the EHR, can serve as a non-invasive method to observe how providers and patients interact within the EHR without imposing additional burdens on clinicians.[ 20 ] For instance, audit log data can be used to measure how many times a provider clicks between the lab results and medication order screen during a visit, offering a proxy for task-switching or uncertainty. Using EHR metadata provides an analytical approach that not only captures natural interaction patterns but also enables scalable and objective measurement of cognitive load and its link to prescribing behavior, laying a foundation for future interventions that can be assessed through real-world practice patterns. EHR audit log metadata offers expanding opportunities for health services research. [ 21 ] The clinical informatics literature has identified three metadata metrics that directly capture EHR user behavior related to measuring cognitive load: counts of actions captured by metadata, activity durations, and activity sequences.[ 22 ] However, extracting and interpreting these metrics to meaningfully reflect cognitive load is technically challenging. Raw EHR audit log metadata are high-volume, inconsistently structured across systems, and not collected with research in mind, making it challenging to align events with specific clinical decision-making moments. While the literature is limited in linking these metrics to specific healthcare research questions and in measuring the effectiveness of healthcare delivery, [ 23 ] process mining can bridge EHR metadata to desired behaviors and reveal system changes that may drive behavior change. By analyzing how these metrics relate to guideline adherence, health systems can identify workflow inefficiencies or decision points where additional support may be warranted. This study aims to fill a critical gap by evaluating whether cognitive load, as measured through EHR metadata, differs between providers who prescribe statins to eligible patients and those who do not—an insight that could inform workflow-based interventions to close persistent care gaps in ASCVD prevention. Additionally, we aim to determine which EHR activities (e.g., medication orders, patient communications, and the use of pre-existing tools) are associated with a higher cognitive load in encounters that result in guideline-concordant statin prescribing. 3 Methods We conducted a retrospective, cross-sectional, observational study using EHR audit log metadata and clinical encounter data to evaluate the association between cognitive load during primary care visits and initiation of statin therapy. This study was designed and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.[ 24 ] This study was approved by the NYULH Institutional Review Board (IRB #i23-01349), which granted a waiver of authorization. 3.1 Population and Number of Observations Data were retrospectively pulled from the EHR of a large academic health system that uses Epic Systems between January 1, 2024, and December 31, 2024, for primary care encounters of patients who qualified for statin initiation during their clinic visit. 3.2 Data Inclusion and Exclusion To determine statin eligibility of adult patients between 20 and 75 years of age, we extracted laboratory data, including their most recent lipid panels and hemoglobin A1c, age, self-identified race and ethnicity, sex, smoking history, latest collected systolic blood pressure, medication history, and any previous history of ASCVD events to calculate each patient's cardiovascular risk based on the 2013 American College of Cardiology/American Heart Association (ACC/AHA) ASCVD 10-year risk score [ 25 ] We identified each eligible patient's most recent primary care visit in 2024 and retrieved encounter-level metadata, including provider identifiers. We included adult patients with at least one documented primary care encounter in 2024 who met clinical guidelines for statin therapy. [ 26 ] Patients were excluded if they lacked sufficient data to calculate a valid risk score, had no primary care encounters during the study period, indicated an allergy or contraindication to a statin, or were pregnant in 2024. Patients who qualified for statin therapy were identified based on the guidelines in Table 2 , which are based on the ACC/AHA Task Force on Clinical Practice’s guidelines on the management of blood cholesterol. [ 27 ] Patients were then classified based on their statin prescribing status at their latest visit: initiated, defined as those with a statin medication order documented during the encounter, and not prescribed, defined as those without a corresponding statin order despite meeting guideline-based indications. Statins initiated in 2024 can be found in Table A.1. Table 2 Inclusion and Exclusion Criteria for Statin Therapy Eligibility Inclusion Criteria Exclusion Criteria Patients with any of the following guidelines: • Age 20 to 75, with LDL ≥ 190 mg/dL and diagnosis of Type 2 Diabetes Mellitus 1 • Age 40 to 75, with Type 2 Diabetes Mellitus 1 and a history of myocardial infarction (MI) 2 or stroke 2 • Age 40 to 75, with an estimated 10-year ASCVD risk ≥ 7.5% • Age 20 to 75, with a history of myocardial infarction (MI) 2 or stroke 2 Patients with any of the following conditions: • Pregnant • Have a known allergy to statins or a contraindication to statin use • Have any of the following diagnoses: o Hepatitis 2 o Cirrhosis 2 o End-Stage Renal Disease (ESRD) 2 o Chronic Kidney Disease (CKD) 2 1 Type 2 Diabetes Mellitus identified via ICD-10 codes (Table A.2.) or HbA1c ≥ 6.5% in their latest blood draw. 2 Diagnoses identified using ICD-10 codes listed in Table A.2. 3.3 Data and Data Structure EHR audit log metadata and relevant patient clinical information were extracted from the healthcare system’s data lake, which was built on a Hadoop distributed file system, and the relevant data were queried using MySQL databases. For patients initiated on statins in 2024, we extracted metadata from the encounter linked to the initiation. For those not initiated, we used metadata from their most recent annual wellness visit, which occurred in 2024. We defined an encounter as a unique interaction between a statin-prescribing provider (physician, physician assistant [PA], nurse practitioner [NP]) and a patient. An encounter may consist of multiple, non-contiguous time segments during which the provider engages with the patient’s data or performs related actions. For example, a provider might review the patient’s chart regarding an encounter from 1:00 PM to 1:03 PM, conduct a face-to-face visit from 1:40 PM to 2:10 PM, and later complete documentation after work hours (between 7:00 PM and 7:00 AM) or even the next day. These time segments are combined into a single encounter to reflect the total effort involved with the encounter. From the encounter's metadata, we derived four cognitive load metrics based on the user interactions (an “event”) logged by the EHR. Events in the EHRs include instances when a provider accesses patient data, their actions related to the data, and other workflow activities associated with an ID tied to the patient encounter. The sequence of timestamped events enables the creation of metrics that represent the effort exerted by the provider when interacting with the EHR, which we refer to as cognitive load metrics. Each of these metrics was chosen because a stakeholder can introduce changes in the EHR based on a desire to alter these metrics, which would, in turn, affect cognitive load and lead to meaningful changes in clinical behavior. These metrics are elaborated on in Table 3 . Table 3 Cognitive Load Metric Definitions Cognitive Load Metric Definition Example of Cognitive Load Metric Associations between Metric and Opportunity for Behavior Change Number of Loops (N) The number of times an event is revisited during an encounter. User moves from Event A → Event B → Event C → Event A → Event B → Event D (2 loops total – one loop back to Event A and one loop back to Event B) More loops could increase cognitive effort due to redundancy or confusion. More loops suggest redesigning workflows to remove unnecessary repetition. Average Time Spent on an EHR Event (seconds) The average time spent on an EHR event. 8.7 seconds per event (e.g., Event A, Event B, Event C) More time spent per event indicates complexity and more fixating, suggesting opportunities for simplification or support. Number of Distinct EHR Events The number of unique events in the EHR that a provider visits during their encounter. 15 unique events (Event A, Event B, Event C, …, Event O) More distinct events increase task complexity and intrinsic load. More distinct events suggest chunking related tasks or adding grouping mechanisms. Duration of an Encounter (minutes) The total active time an agent spends engaging with data or actions related to a specific customer interaction, summed across all non-contiguous time segments associated with the same encounter ID. 17.9 minutes from first event (A) to last event (C) related to an encounter Longer encounter duration indicates longer time engaging with the EHR for reviewing patient data, which could introduce fatigue or signal focused patient engagement. Shorter encounter duration may be related to goal-oriented behaviors, resulting in decreased cognitive load. 3.4 Analytical Methods All analyses were completed in RStudio Version 2024.12.0 + 467. 3.4.1 Descriptive Statistics We collected means, standard deviations, medians, and ranges of cognitive load metrics, patient characteristics during their encounter with the provider, and provider sex. We used the interquartile range (IQR) method to identify potential outliers in measures of EHR interaction complexity, including the number of loops, number of distinct events, total duration, and average time per event. Following the Tukey IQR rule, [ 28 ] values falling more than 1.5 times the IQR below the first quartile or above the third quartile were classified as outliers. A record was flagged as an outlier if any of the cognitive load metrics exceeded the Tukey thresholds. 3.4.2 Associations Between Statin Prescribing and Cognitive Load Metrics The primary outcome of interest is the likelihood of a cognitive load metric influencing a provider’s decision to initiate a statin for a patient. To do so, we ran a logistic regression analysis, with the dependent variable being whether a provider had initiated a statin for a patient who was eligible for statin therapy during an encounter in 2024. The independent variables of interest were each of the four cognitive load metrics: number of loops, average time spent on an EHR event, number of distinct EHR events, and duration of an encounter (Table 3 ). The controlled patient-level covariates included the patient’s sex, self-identified race and ethnicity, insurance status (aggregated as commercial or non-commercial), Elixhauser Comorbidity Index [ 29 ] during their visit, the number of active diagnoses during their encounter, and age. We included provider fixed effects to account for unobserved, time-invariant characteristics of each provider that may influence initiation decisions (e.g., years of active practice and number of other patients with similar conditions). In this healthcare system, physicians can see patients independently; however, NPs and PAs, despite having prescribing authority, must have a physician involved in the encounter. To adjust for multiple providers with prescribing abilities seeing a patient in the same encounter, we included clustered standard errors by the encounter ID. This combined approach helps isolate the effect of interest while producing conservative and robust standard errors. However, cognitive load may have a nonlinear relationship with the probability of initiating a statin [ 30 – 32 ]. In some cases, clinical decision-making improves as cognitive challenge or task engagement increases, up to a saturation point, after which it declines, suggesting an optimal “middle ground” of effort. In other cases, low-effort task sequences may reflect automation of desired behaviors, whereas high-effort sequences may signal increased deliberation. In contrast, certain intermediate levels of effort could lead to inertia, resulting in no action being taken. Therefore, we introduced quadratic terms of each cognitive load metric to capture potential threshold effects flexibly. If the quadratic term did not statistically contribute to statin initiation, we tested whether excluding it would significantly impact the model fit by comparing the adjusted R 2 . If excluding the nonlinear term did not significantly impact the model fit, the quadratic term was removed to reduce the chance of overfitting the model. We calculated the average marginal effect of the metric on the likelihood of statin initiation for the cognitive load metrics with nonlinear terms to understand how much the likelihood of statin initiation changes for a particular amount of cognitive load. This approach enabled us to quantify both the direction and magnitude of the relationship across different levels of cognitive load, identifying points at which additional effort began to hinder rather than support statin initiation. 3.4.3 Exploratory Analyses – Sources of Cognitive Load within the EHR We aimed to link cognitive load measures within the EHR to statin initiation to identify opportunities for integrating clinical decision support. Using machine learning, we identified which EHR events were most strongly associated with statin initiation and the types of cognitive load associated with those events. We structured the EHR metadata so that each variable represented a cognitive load metric for a specific EHR event. This means, for example, we measured the average time a provider spent at Event A, how many times a provider looped back to Event B, and whether the provider visited Event C within an encounter. Because not every provider interacts with every EHR event, many of these variables contain zero values, reflecting events that were skipped or that did not have the cognitive load metric. To explore how these detailed interaction patterns relate to statin initiation, we used a gradient-boosted forest model that took these variables as inputs to predict whether a statin was initiated during the encounter. This machine learning method is well-suited for handling large, complex datasets with many variables, especially when relationships may be nonlinear (as per the earlier investigation) or involve interactions between events. It allowed us to identify which specific EHR events’ cognitive load metrics were most strongly associated with the decision to prescribe a statin. We tuned the forest to a learning rate ( \(\:\eta\:\) ) of 0.1 and a train-validation-test split of 70%-15%-15%. The procedure was repeated 200 times to generate a distribution of performance metrics, from which 95% confidence intervals were derived, offering a reliable assessment of the model's performance and variability. To interpret the model outputs, we applied Shapley Additive explanations (SHAP) values, which provide a theoretically grounded approach to attributing feature contributions to individual predictions. SHAP values offer both local interpretability, revealing how specific features influence each decision, and global interpretability, summarizing the overall importance of features across the dataset. [ 33 ] Mean SHAP values for each cognitive load-metric covariate and global mean SHAP values were computed to quantify the importance of each feature within the model in terms of direction (negatively or positively contributing to the prediction) and magnitude (how much the metric at the event contributes to overall prediction). Isolating the most significant contributors to the desired clinical behavior (i.e., statin initiation) with quantified contributions enables the future justification of introducing clinical decision support in the EHR. 4 Results 4.1 Descriptive Results We analyzed 20,376 patient encounters across the academic healthcare system. The mean age of patients eligible for statin initiation was 58.8 years (SD = 10.8), and about half were male. The cohort was predominantly non-Hispanic/Latinx White, and approximately 80% had commercial insurance. Each patient had an average of 6 active diagnoses during their primary care provider encounter. Most encounters were handled by physicians (82.41%), with fewer by nurse practitioners (NP, 11.66%) and physician assistants (PAs, 5.93%). Encounters with NPs had the highest statin initiation rate (50.72% of NP encounters resulted in a statin initiation) compared to 46.86% for physicians and 46.6% for PAs. Encounters with statin initiation showed slightly lower cognitive load metrics on average than encounters with no statin. These differences were small but statistically significant: encounters with initiation had fewer loops, fewer distinct events, shorter duration, and a lower average time per event. Overall, 17.6% of records had at least one extreme value in the cognitive load metrics (flagged as outliers per the Tukey IQR rule). A chi-squared test showed no significant difference in statin initiation rates between encounters with vs. without such outlier metrics (p = 0.20) – 45.7% vs 44.5% initiation, respectively. Risk analysis revealed that outlier encounters were 3% less likely to involve statin initiation (RR = 0.97; OR = 0.95, 95% CI 0.89–1.03), with no significant difference in the likelihood of statin prescribing between the two groups. In other words, extreme cognitive load metrics do not make an encounter more or less likely to result in a prescription. Therefore, we retained all observations in subsequent analyses. All descriptive analyses are presented in Table 4 . Table 4 Descriptive statistics of patients, providers, and cognitive load metrics No Initiation Initiation P-value (N = 11105 Patients) (N = 9271 Patients) Patient Characteristics Age (years) Mean (SD) 58.9 (10.8) 58.7 (10.8) 0.211 Median [Min, Max] 61.0 [20.0, 75.0] 60.0 [20.0, 75.0] Sex Male 5542 (49.9%) 4788 (51.6%) 0.0141 Female 5560 (50.1%) 4483 (48.4%) Unknown 3 (0.0%) 0 (0%) Race Ethnicity Non-Hispanic/Latinx White 6612 (59.5%) 4366 (47.1%) < 0.001 Non-Hispanic/Latinx Black/African American 1434 (12.9%) 1098 (11.8%) Hispanic/Latinx 1091 (9.8%) 1791 (19.3%) Asian 844 (7.6%) 723 (7.8%) Other 1124 (10.1%) 1293 (13.9%) Financial Class Commercial 8907 (80.2%) 7263 (78.3%) 0.00111 Non-commercial 2198 (19.8%) 2008 (21.7%) Number of Active Diagnosis Mean (SD) 6.40 (3.07) 6.09 (3.51) < 0.001 Median [Min, Max] 6.00 [1.00, 24.0] 5.00 [1.00, 32.0] Elixhauser Comorbidity Index Mean (SD) 1.68 (5.42) 1.42 (5.37) < 0.001 Median [Min, Max] 0 [-14.0, 42.0] 0 [-18.0, 58.0] Number of Providers Involved in Encounter One 8133 (73.2%) 6866 (74.1%) Two 2637 (23.7%) 2154 (23.2%) Three 335 (3.0%) 251 (2.7%) Provider Type Physician Nurse Practitioner Physician Assistant Number of Encounters 33281 (82.41%) 4709 (11.66%) 2393 (5.93%) Initiation Rate Mean (SD) 46.86 (0.77) 50.72 (1.47) 46.60 (2.00) Median [Min, Max] 46.15 [0, 100] 50.00 [0, 100] 41.05 [0, 100] No Initiation Initiation Cognitive Load Metrics (N = 11105 Encounters) (N = 9271 Encounters) P-value Number of Loops Mean (SD) 75.0 (34.2) 71.1 (39.4) < 0.001 Median [Min, Max] 70.0 [1.00, 311] 63.0 [0, 465] Number of Distinct Events Mean (SD) 34.0 (6.01) 33.4 (7.00) < 0.001 Median [Min, Max] 34.0 [8.00, 69.0] 33.0 [9.00, 80.0] Encounter Duration (Minutes) Mean (SD) 38.8 (20.0) 37.8 (22.1) < 0.001 Median [Min, Max] 35.0 [1.17, 183] 33.0 [0.333, 246] Average Time per EHR Event (Seconds) Mean (SD) 44.5 (61.3) 40.9 (55.9) < 0.001 Median [Min, Max] 22.7 [1.84, 970] 22.0 [1.11, 791] 4.2 Regression Analyses between Cognitive Load Metrics and Statin Initiation We examined whether the cognitive load metrics were associated with the likelihood of statin initiation in a multivariate logistic regression model, adjusting for patient characteristics and provider effects. Initially, we tested a model with only linear terms for each cognitive load metric; however, two of the cognitive load metrics—encounter duration (in minutes) and the average time spent per EHR event (in seconds)—did not show a significant linear relationship with statin initiation. We then examined non-linear relationships by adding quadratic terms. Model comparison revealed that quadratic terms for the number of loops and the number of distinct events significantly improved the fit (Wald test, p < 0.001). Thus, our final model included non-linear (quadratic) terms for loops and distinct events, as well as linear terms for duration and time per event. Comparisons of the linear and nonlinear models are presented in Table A.3. 4.2.1 Linear Effect of Cognitive Metrics on Statin Initiation Regression results are presented in Table 5 . In the final regression model, the average time spent per EHR event and total encounter duration showed significant linear relationships with the initiation of statins. The average time spent per event showed a significant negative association with the likelihood of statin initiation, where spending more time per event was associated with a lower likelihood of initiation (β = -5.017 × 10 − 4 , p < 0.001). On the other hand, the total duration of the encounter showed a significant positive association with the odds of initiating a statin, where longer encounter duration was associated with increased odds (β = 8.223 × 10 − 3 , p < 0.001). These findings may reflect provider time management strategies, where extended engagement across multiple sections of the EHR is more conducive to initiating guideline-recommended treatment. 4.2.2 Nonlinear Effects of Cognitive Metrics on Statin Initiation Figure 1 and Fig. 2 illustrate the two cognitive load metrics which exhibited significant nonlinear associations with statin initiation. For the number of loops, coefficients indicated a U-shaped relationship (linear β = −0.0124, p < 0.001; quadratic β = 0.597×10⁻⁵, p < 0.001). Encounters with either very low or very high loop counts were more likely to result in initiation than those with intermediate counts (Fig. 1 , dashed green line). The inflection point occurred at 93.9 loops (Fig. 1 , vertical dashed red line); below this threshold, each additional loop reduced the probability of initiation by 0.21%, whereas above it, each additional loop increased the probability by 0.00046%. The peak effect was observed at 237 loops, with a predicted initiation probability of 0.684. The number of distinct events showed an inverse U-shaped relationship (linear model: β = 0.0124, p < 0.001; quadratic model: β = −0.597×10 − 5 , p < 0.001), with the highest probability of initiation observed when encounters have moderate event counts. The inflection point occurred at approximately 18 events (Fig. 2 , vertical dashed red line); below this threshold, each additional event increased the probability of initiation by 1.35%, whereas above it, the effect was negligible, and the probability gradually declined. This suggests an optimal range of distinct EHR actions that may facilitate decision-making regarding guideline-concordant statin initiation, with diminishing returns once complexity exceeds this threshold. Patient race and sex were also significant predictors of statin initiation, as shown in Table 5 . Compared to non-Hispanic/Latinx White patients, patients of racial and ethnic minorities were more likely to be initiated on a statin, as well as older patients. Female patients, as well as those with a greater number of active diagnoses or higher comorbidity burden, were less likely to receive initiation. Table 5 Coefficients of the nonlinear model associating cognitive load with statin initiation. + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 Reference Level Coefficient Std. Err Statistic Number of Loops -- −1.239e − 02*** (2.123e − 03) -5.84 Number of Loops² 6.597e − 05*** (1.021e − 05) 6.46 Number of Distinct Events -- 6.412e − 02*** (8.381e − 03) 7.65 Distinct Events² −1.787e − 03*** (1.762e − 04) -10.1 Total Duration (min) -- 8.223e − 03*** (1.436e − 03) 5.73 Avg Time Per Event (sec) -- −5.017e − 04*** (1.314e − 04) -3.82 Race: Black/African American Non-Hispanic/Latinx White 1.967e − 01** (6.567e − 02) 3 Race: Hispanic/Latino -- 5.906e − 01*** (6.508e − 02) 9.08 Race: Asian -- 4.425e − 01*** (8.083e − 02) 5.47 Race: Other 6.839e − 01*** (6.363e − 02) 10.7 Patient Age 1.023e − 02*** (2.101e − 03) 4.87 Patient Sex: Female Male −8.218e − 02* (4.103e − 02) -2 Patient Sex: Unknown −2.016e + 01*** (5.316e − 02) -379 Insurance: Non-commercial Commercial 5.738e − 02 (5.466e − 02) 1.05 Number of Active Diagnoses −4.879e − 02*** (7.546e − 03) -6.47 Elixhauser Comorbidity Index −9.656e − 03** (3.742e − 03) -2.58 N Observations 36131 R 2 0.2 R 2 Adj. 0.128 RMSE 0.43 4.3 Exploratory Analyses for EHR Events Associated with Cognitive Load Metrics We analyzed the EHR metadata with a machine learning model to pinpoint which specific EHR activities and associated cognitive load metrics were most strongly linked to statin initiation. We trained a gradient-boosted decision tree model. We computed SHAP values to quantify the contribution of each EHR interaction feature to the model’s predictions. The model achieved an area under the receiver operating characteristic curve (AUROC) of 0.91, indicating that using cognitive load metrics has strong performance in predicting statin initiation. Table 6 presents the 10 EHR event-specific features that contributed most to the model’s prediction, ordered by their global mean SHAP values and importance. A full list of contributions by cognitive load-EHR events can be found in Table A.4. The average time per event (in seconds) had the highest relative contribution to predicting statin initiation, accounting for 72.2% of the model’s relative contribution to its predictions. There was a positive average SHAP value contribution, indicating a positive association between average time spent per event and statin initiation The number of loops accounted for 15.1% of the relative contribution but had a small negative average SHAP value contribution, indicating a negative association between loops and statin initiation. The presence of specific EHR events contributed 12.6% to the model’s predictions and had a small positive average contribution, indicating a positive association between the presence of an event and statin initiation. The average time spent on viewing patient communications had a positive mean SHAP value of 0.0874, indicating that spending more time on a patient’s results was associated with a higher probability of statin initiation. This feature accounted for 23.1% of the model’s relative contribution to its predictions, making it the most influential predictor. The average time spent on suggested medication order sets had a positive mean SHAP value of 0.0129, indicating that spending more time on suggested medication order sets was associated with a higher probability of statin initiation. This feature accounted for 12.7% of the model’s relative contribution to its predictions, making it the second most influential predictor. The presence of the order set for testing contributed positively to predicting statin initiation, accounting for a 6.5% increase in the model’s predictive accuracy. Conversely, the average time spent on changing the list of orders showed a negative mean SHAP value of -0.0115, implying that modifying the order list was associated with a decreased likelihood of initiation, contributing approximately 7.74% to the model’s relative influence. Looping back to the order list contributed 3.74% to the model’s relative influence. It had a negative average SHAP value, indicating that the number of loops to change an order list was negatively associated with statin initiation. These results suggest that interactions with the patient's lab results and spending time on medication order sets positively influence prescribing behavior, while spending time on order modifications and looping back to change the orders may signal complexity or uncertainty, which reduces the likelihood of initiation. Table 6 Top 10 cognitive load features associated with statin initiation, ranked by mean absolute SHAP value. The relative contribution of these features is 70.06% of the model’s predictions. Rank EHR Event Cognitive Load Metric Mean SHAP Mean Absolute SHAP 95% CI Lower 95% CI Upper Contribution Direction Relative Contribution 1 Patient communications viewed Average Time Per Event (sec) 0.0874 0.7205 0.1014 0.0734 Increasing 0.2313 2 Order set suggested Average Time Per Event (sec) 0.0129 0.3967 0.0200 0.0059 Increasing 0.1274 3 Order modified Average Time Per Event (sec) -0.0115 0.2410 -0.0064 -0.0166 Decreasing 0.0774 4 Order set suggested Event Shown (1/0) 0.0115 0.2030 0.0151 0.0078 Increasing 0.0652 5 Patient’s lab results reviewed Average Time Per Event (sec) 0.0293 0.1321 0.0322 0.0264 Increasing 0.0424 6 Order list changed Number of Loops -0.0090 0.1164 -0.0068 -0.0112 Decreasing 0.0374 7 Order set activity selected Average Time Per Event (sec) 0.0043 0.1091 0.0066 0.0019 Increasing 0.0350 8 Alert displayed Average Time Per Event (sec) -0.0051 0.1004 -0.0034 -0.0069 Decreasing 0.0322 9 Problem List modified Number of Loops -0.0021 0.0826 -0.0002 -0.0041 Decreasing 0.0265 10 Chart Review Encounter report viewed Average Time Per Event (sec) -0.0024 0.0807 -0.0010 -0.0038 Decreasing 0.0259 5 Discussion 5.1 Summary We identified statistically significant relationships between cognitive load metrics and the likelihood of initiating statin therapy during primary care encounters. First, we identified that cognitive load does not have a linear relationship to desired clinical behavior. Some types of cognitive load (in this paper’s case, engagement via repetitive behaviors) suggest that a “middle ground” of effort by providers increases the likelihood of a clinical behavior, that some effort, but not too much or too little effort, was associated with more appropriate prescribing. We also found that under-engagement (involving too few interactions) and over-engagement (characterized by excessive interactions) can positively impact clinical decision outcomes in statin initiation, possibly suggesting that either more active engagement or short, goal-directed engagement with patients leads to more appropriate prescribing behavior. These results highlight the dynamic nature of cognitive load in clinical environments. By applying machine learning techniques to EHR metadata, we identified key behaviors that current CDSS can leverage to support providers in delivering appropriate clinical care, as well as behaviors where existing EHR features may inadvertently hinder such decisions. It has been well documented that EHR nudges can result in alert fatigue; however, our work suggests that aptly designed CDS can optimize intended effect while reducing cognitive load. Specifically, CDS can create changes to interfaces when providers interact with a patient’s lab results, order medications, and work with alerts in the EHR that facilitate target actions. These findings underscore the potential of CDSS to enhance clinical decision-making and improve patient outcomes. [ 34 ] 5.2 Implications This study provides actionable insights into how cognitive load in EHR workflows affects clinical decision-making behavior and presents a framework for how to improve clinical decision-making. By measuring cognitive load indicators within the EHR and identifying specific interaction patterns associated with prescribing decisions, we highlight key moments in the workflow that can be targeted with more effective CDS tools. These findings lay the groundwork for designing interventions that reduce unnecessary complexity and better support clinicians in following guideline-recommended therapies. Our methodological approach provides both immediate and long-term benefits for health systems. In the short term, these insights can help health systems identify workflow gaps that hinder the provision of effective decision support. Over the long term, this approach has the potential to transform how health systems monitor and improve provider behavior, enabling data-driven quality initiatives. These methods can be further refined to distinguish between helpful and unhelpful cognitive load for providers, ultimately leading to better patient care through intentional, thoughtful digital actions. For patients, improved prescribing practices can have a direct and meaningful impact on outcomes, particularly by increasing the use of appropriate statins to lower primary ASCVD risk. The developed methodology is scalable and adaptable to various clinical settings and EHR systems. Our data processing and workflow characterization techniques can support a wide range of efforts to enhance clinical decision support and promote evidence-based care at scale. Firstly, our approach to extracting audit log metadata and clinical behavior data, and linking these to clinical data, can be easily adapted for other behavior-related questions, such as specialty referrals, shared decision-making, and screening choices. Secondly, the identification of behaviors within the EHR and the creation of cognitive load metrics are independent of any specific EHR, as they rely on process mining and data science techniques. The analytical methods used to examine the relationship between cognitive load and clinical behavior, along with machine learning approaches that connect cognitive load to EHR events, offer a systematic way to evaluate CDSS designed to improve clinical practices and identify behaviors that need adjustment at specific interfaces. 5.3 Limitations Several limitations should be considered when interpreting our findings. First, the cognitive load and workflow measures used in this study were intentionally simple. While more sophisticated process mining techniques, such as clustering workflow sequences or calculating transition probabilities between EHR events, could yield deeper insights, we prioritized interpretable, actionable metrics aligned with behavior change theory. These simpler measures are more readily translatable into tools within the EHR to support clinical decision-making. Moreover, the application of EHR metadata for healthcare quality improvement remains an emerging area within clinical informatics. There is a pressing need for standardization of EHR metadata metrics by organizations such as the Office of the National Coordinator for Health Information Technology (ONC), which could also enable validation of more complex measures through direct observation or qualitative research. Second, our findings are based on data from a single EHR platform within a single academic health center’s EHR system. While the specific events captured may vary across institutions, our computational approach was designed to be generalizable by focusing on abstracted, system-agnostic event types. Thus, while implementation may require adaptation to local data structures, the methodology should be transferable to other EHR platforms. Third, the cognitive load metrics used in this study do not distinguish between intrinsic, extrinsic, and germane cognitive load. As a result, we cannot determine whether the measured cognitive load metrics reflect the inherent complexity of the clinical decision (intrinsic load), inefficiencies or distractions introduced by the EHR interface or workflow (extrinsic load), or the cognitive effort invested in understanding and integrating information for decision-making (germane load). Without this differentiation, it is difficult to target design interventions precisely, like reducing extrinsic load without inadvertently removing beneficial germane load. Future work incorporating methods such as task analysis, think-aloud protocols, or real-time workload assessment could help disentangle these components and guide more tailored improvements to CDSS and EHR. Fourth, our current methods do not link each EHR action to a specific medication, lab, or diagnosis. This allowed us to compare all encounters with and without statin initiation in a general way. In future work, we could focus on encounters that involve the same medications and labs and see how cognitive effort on each specific action—like reviewing cholesterol lab results or adjusting a statin order—affects whether a statin is prescribed. Fifth, this study focused on one clinical use case: initiating maintenance medications (statins) in primary care. Although our population identification and data linkage methods were tailored to this context, the overall approach can be applied to other clinical domains. Importantly, the metadata metrics were derived independently of the clinical content, which supports broader applicability. With further development, EHR vendors and health systems could aggregate and standardize such metrics to inform workflow improvement and decision support across various clinical scenarios. For example, the scope of the study was intentionally limited to initiation and did not include decisions regarding the titration of statins. Further research can elaborate on differentiating the cognitive load of workflows for initiation versus titration, given the distinct behavioral mechanisms and decision-making processes associated with each. This study also limited its analysis to the behavior of providers, despite evidence of patient hesitancy to start statin therapy.[ 35 ] The methods of this study are equally applicable to the EHR audit log metadata of patient-facing interfaces and patient portals, to enhance the digital patient experience. 6 Conclusion Our findings suggest that the likelihood of initiating statins is not simply determined by the amount of effort in the EHR but is influenced by a nonlinear relationship between cognitive load and decision-making. Both too little and too much engagement with the EHR can decrease the chances of initiating appropriate therapy, while an optimal level of engagement seems to promote guideline-concordant prescribing. Similarly, extreme engagement behaviors in the EHR can support clinical actions, whereas a moderate approach may lead to low engagement. By identifying specific EHR interaction patterns and workloads that either facilitate or hinder statin initiation, our results highlight practical opportunities to redesign clinical decision support and workflows to better balance cognitive demands. Efficient, targeted EHR use—rather than maximal or minimal effort—may be the most effective way to ensure timely and appropriate treatment decisions. Importantly, our study addresses and overcomes significant barriers and challenges associated with using EHR metadata. We demonstrate that these complex metrics can be extracted, processed, and mapped to detailed clinical workflows, enabling robust analysis of how cognitive load impacts clinical decision-making. This advancement provides a framework for health systems to better understand and optimize EHR interactions, ultimately supporting improved guideline adherence and patient outcomes. Abbreviations Term / Abbreviation Definition Atherosclerotic cardiovascular disease (ASCVD) Condition caused by plaque buildup (atherosclerosis) in the arteries ASCVD 10-year risk score Estimate of a person’s chance of having a heart attack or stroke within the next 10 years Clinical Decision Support Systems (CDSS) Tools or systems within the EHR that provide clinicians with information, alerts, or recommendations to improve patient care decisions. Cognitive Load The mental effort required to process information and complete tasks, influenced by the complexity of the information, the environment, and the interface design. Electronic Health Record (EHR) A digital system for storing and managing patient medical records and clinical workflows. EHR Metadata Data automatically generated and stored by the EHR that describes user interactions, such as keystrokes, clicks, screen time, and navigation patterns. Gradient-boosted forests Machine learning method that combines decision trees built in sequence in which new trees correct errors of previous trees to improve accuracy Hadoop distributed file system System for storing large data sets across multiple machines, allowing processing in parallel Hawthorne Effect When people change their behavior because they are aware they are being observed AUC-ROC (Area Under the Curve – Receiver Operating Characteristic) A measure of a binary classifier’s performance; higher values indicate better discrimination between positive and negative cases. Random Forest A machine learning method that uses multiple decision trees to make predictions or classifications; in this study, used to predict statin initiation from cognitive load metrics. Shapley additive explanation (SHAP) values Method to explain machine learning model’s predictions by quantifying each feature’s contribution to a specific prediction Statin A class of cholesterol-lowering medications used to reduce cardiovascular risk, often prescribed for patients with elevated ASCVD risk, by blocking the production of low-density lipids (LDL) in the liver, helping prevent heart attacks and strokes. STROBE The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for reporting observational studies Tukey IQR rule A statistical method for identifying outliers by defining data points lying more than 1.5 times the interquartile range above the third quartile or below the first quartile. Declarations 7.2 Ethics Approval and Consent to Participate This study was reviewed and approved by the NYU Langone Health Office of Science and Research Institutional Review Board (IRB #i23-01349). The NYU Langone IRB granted a waiver of informed consent. Confidential patient data were accessed and analyzed within the secure hospital network in compliance with institutional policies. The study was conducted in accordance with the principles of the Declaration of Helsinki. 7.3 Consent for Publication Not applicable. 7.4 Availability of Data and Materials Due to the use of patient data in this study, deidentified versions of the data can be made available upon reasonable request. Dr. Safiya Richardson can be contacted at [email protected] for data requests. 7.5 Competing Interests The authors declare that they have no competing interests. 7.6 Funding This work was supported by the National Institutes of Health’s Heart, Lung, and Blood Institute (1R01HL171292-01). 7.7 Author Contributions (CRediT Taxonomy) Contributor Role Role Definition Author Conceptualization Ideas; formulation or evolution of overarching research goals and aims. RVNV, SIR Data Curation Management activities to annotate (produce metadata), scrub data and maintain research data (including software code, where it is necessary for interpreting the data itself) for initial use and later reuse. RVNV, YC Formal Analysis Application of statistical, mathematical, computational, or other formal techniques to analyze or synthesize study data. RVNV, YC Funding Acquisition Acquisition of the financial support for the project leading to this publication. RVNV, DMM, SIR Investigation Conducting a research and investigation process, specifically performing the experiments, or data/evidence collection. RVNV, YC, PS Methodology Development or design of methodology; creation of models RVNV, YC, PS, SIR Project Administration Management and coordination responsibility for the research activity planning and execution. RVNV, NR, SIR Resources Provision of study materials, reagents, materials, patients, laboratory samples, animals, instrumentation, computing resources, or other analysis tools. DMM Software Programming, software development; designing computer programs; implementation of the computer code and supporting algorithms; testing of existing code components. RVNV, YC Supervision Oversight and leadership responsibility for the research activity planning and execution, including mentorship external to the core team. RVNV, NR, SIR, DMM Validation Verification, whether as a part of the activity or separate, of the overall replication/reproducibility of results/experiments and other research outputs. N/A Visualization Preparation, creation and/or presentation of the published work, specifically visualization/data presentation. RVNV, YC Writing – Original Draft Preparation Creation and/or presentation of the published work, specifically writing the initial draft (including substantive translation). RVNV, YC, PS, NR, AM, DLL, ARS, SIR, DMM Writing – Review & Editing Preparation, creation and/or presentation of the published work by those from the original research group, specifically critical review, commentary or revision – including pre- or post-publication stages. RVNV, YC, PS, NR, AM, DLL, ARS, SIR, DMM 7.8 Acknowledgements Not applicable. References Sparrow RT, Khan AM, Ferreira-Legere LE, Ko DT, Jackevicius CA, Goodman SG, et al. Effectiveness of Interventions Aimed at Increasing Statin-Prescribing Rates in Primary Cardiovascular Disease Prevention: A Systematic Review of Randomized Clinical Trials. JAMA Cardiol. 2019;4:1160-9. Baigent C, Blackwell L, Emberson J, Holland LE, Reith C, Bhala N, et al. Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials. Lancet. 2010;376:1670-81. Kannan S, Asch DA, Kurtzman GW, Honeywell Jr. S, Day SC, Patel MS. Patient and Physician Predictors of Hyperlipidemia Screening and Statin Prescription. Am J Manag care. 2018;24:241-8. Bradley CK, Wang TY, Li S, Robinson JG, Roger VL, Goldberg AC, et al. Patient-Reported Reasons for Declining or Discontinuing Statin Therapy: Insights From the PALM Registry. J Am Heart Assoc. 2019;8:e011765. Salami JA, Warraich H, Valero-Elizondo J, Spatz ES, Desai NR, Rana JS, et al. National Trends in Statin Use and Expenditures in the US Adult Population From 2002 to 2013: Insights From the Medical Expenditure Panel Survey. JAMA cardiology. 2017;2:56-65. Austin PC, Mamdani MM, Juurlink DN, Alter DA, Tu JV. Missed opportunities in the secondary prevention of myocardial infarction: an assessment of the effects of statin underprescribing on mortality. Am Heart J. 2006;151:969-75. Pauff BR, Jiroutek MR, Holland MA, Sutton BS. Statin Prescribing Patterns: An Analysis of Data From Patients With Diabetes in the National Hospital Ambulatory Medical Care Survey Outpatient Department and National Ambulatory Medical Care Survey Databases, 2005-2010. Clin Ther. 2015;37:1329-39. Berthold HK, Gouni-Berthold I, Bohm M, Krone W, Bestehorn KP. Patterns and predictors of statin prescription in patients with type 2 diabetes. Cardiovasc Diabetol. 2009;8:25. Gold R, Larson AE, Sperl-Hillen JM, Boston D, Sheppler CR, Heintzman J, et al. Effect of Clinical Decision Support at Community Health Centers on the Risk of Cardiovascular Disease: A Cluster Randomized Clinical Trial. JAMA Netw Open. 2022;5:e2146519. Alameddine R, Seifeddine S, Ishak H, Antoun J. Improving statin prescription through the involvement of nurses in the provision of ASCVD score: a quality improvement initiative in primary care. Postgrad Med. 2020;132:479-84. Peiris D, Usherwood T, Panaretto K, Harris M, Hunt J, Redfern J, et al. Effect of a computer-guided, quality improvement program for cardiovascular disease risk management in primary health care: the treatment of cardiovascular risk using electronic decision support cluster-randomized trial. Circ Cardiovasc Qual Outcomes. 2015;8:87-95. Derose SF, Dudl JR, Benson VM, Contreras R, Nakahiro RK, Ziel FH. Point-of-Service reminders for prescribing cardiovascular medications. Am J Manag Care. 2005;11:298-304. Jacobson TA, Gutkin SW, Harper CR. Effects of a global risk educational tool on primary coronary prevention: the Atherosclerosis Assessment Via Total Risk (AVIATOR) study. Current Medical Research and Opinion. 2006;22:1065-73. Willcox M. Shortcuts versus Analysis –Ignoring is Decisional Bliss. The Business of Choice: How Human Instinct Influences Everyone’s Decisions. 2020. Singh M. Heuristics in the delivery room. Science. 2021;374:324-9. Adler-Milstein J, Zhao W, Willard-Grace R, Knox M, Grumbach K. Electronic health records and burnout: Time spent on the electronic health record after hours and message volume associated with exhaustion but not with cynicism among primary care clinicians. J Am Med Inform Assoc. 2020;27:531-8. Asgari E, Kaur J, Nuredini G, Balloch J, Taylor AM, Sebire N, et al. Impact of Electronic Health Record Use on Cognitive Load and Burnout Among Clinicians: Narrative Review. JMIR Med Inform. 2024;12:e55499. Sedgwick P, Greenwood N. Understanding the Hawthorne effect. BMJ. 2015;351:h4672. Cohen G, Brown L, Fitzgerald M, Somplasky A. To Measure The Burden Of EHR Use, Audit Logs Offer Promise—But Not Without Further Collaboration. 2020. Adler-Milstein J, Adelman JS, Tai-Seale M, Patel VL, Dymek C. EHR audit logs: A new goldmine for health services research? J Biomed Inform. 2020;101:103343. Wu DT, Smart N, Ciemins EL, Lanham HJ, Lindberg C, Zheng K. Using EHR audit trail logs to analyze clinical workflow: a case study from community-based ambulatory clinics. AMIA Annual Symposium Proceedings2018. p. 1820. Rule A, Chiang MF, Hribar MR. Using electronic health record audit logs to study clinical activity: a systematic review of aims, measures, and methods. J Am Med Inform Assoc. 2020;27:480-90. De Roock E, Martin N. Process mining in healthcare - An updated perspective on the state of the art. J Biomed Inform. 2022;127:103995. Von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. The lancet. 2007;370:1453-7. Goff DC, Jr., Lloyd-Jones DM, Bennett G, Coady S, D'Agostino RB, Gibbons R, et al. 2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. Circulation. 2014;129:S49-73. Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, et al. 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA Guideline on the Management of Blood Cholesterol: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Circulation. 2019;139:e1082-e143. Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, et al. 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA guideline on the management of blood cholesterol: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Journal of the American College of Cardiology. 2019;73:e285-e350. Tukey JW. Exploratory data analysis: Springer; 1977. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity Measures for Use with Administrative Data. Medical Care. 1998;36:8-27. Sweller J. Measuring cognitive load. Perspect Med Educ. 2018;7:1-2. Wason PC, Evans JSBT. Dual processes in reasoning? Cognition. 1974;3:141-54. Yerkes RM, Dodson JD. The relation of strength of stimulus to rapidity of habit‐formation. Journal of Comparative Neurology and Psychology. 1908;18:459-82. Merrick L, Taly A. The explanation game: Explaining machine learning models using shapley values. Machine Learning and Knowledge Extraction: 4th IFIP TC 5, TC 12, WG 84, WG 89, WG 129 International Cross-Domain Conference, CD-MAKE 2020, Dublin, Ireland, August 25–28, 2020, Proceedings 4: Springer; 2020. p. 17-38. Kawamoto K, Houlihan CA, Balas EA, Lobach DF. Improving clinical practice using clinical decision support systems: a systematic review of trials to identify features critical to success. BMJ. 2005;330:765. Xie M, Martin SS, Turchin A. Reasons for non-acceptance of statin therapy by patients at high cardiovascular risk. Sci Rep. 2025;15:17014. Additional Declarations No competing interests reported. Supplementary Files AuditLogManuscriptAppendixTables.xlsx Cite Share Download PDF Status: Published Journal Publication published 02 Apr, 2026 Read the published version in BMC Medical Informatics and Decision Making → Version 1 posted Editorial decision: Revision requested 12 Nov, 2025 Reviews received at journal 11 Nov, 2025 Reviews received at journal 03 Nov, 2025 Reviewers agreed at journal 27 Oct, 2025 Reviewers agreed at journal 26 Oct, 2025 Reviewers agreed at journal 20 Oct, 2025 Reviewers invited by journal 13 Oct, 2025 Editor assigned by journal 13 Oct, 2025 Editor invited by journal 12 Oct, 2025 Submission checks completed at journal 08 Oct, 2025 First submitted to journal 08 Oct, 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-7632374","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":534145714,"identity":"8a30274f-ae37-46ab-8f13-1d39e3a0d32f","order_by":0,"name":"Ratnalekha V. N. Viswanadham","email":"","orcid":"","institution":"NYU Grossman School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Ratnalekha","middleName":"V. N.","lastName":"Viswanadham","suffix":""},{"id":534145715,"identity":"9780e2a9-d641-4708-9e65-97ea62806db6","order_by":1,"name":"Yuhan Cui","email":"","orcid":"","institution":"NYU Grossman School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yuhan","middleName":"","lastName":"Cui","suffix":""},{"id":534145717,"identity":"624c5c39-43ae-4b69-9ac6-ce31b52ae5af","order_by":2,"name":"Priyanka Solanki","email":"","orcid":"","institution":"NYU Langone Health","correspondingAuthor":false,"prefix":"","firstName":"Priyanka","middleName":"","lastName":"Solanki","suffix":""},{"id":534145718,"identity":"7f4fe976-3388-417a-9b24-f4a519968392","order_by":3,"name":"Nicole Redfern","email":"","orcid":"","institution":"NYU Grossman School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Nicole","middleName":"","lastName":"Redfern","suffix":""},{"id":534145720,"identity":"794db842-d241-4cba-95a4-60b1dc103535","order_by":4,"name":"Amelia Shunk","email":"","orcid":"","institution":"Tulane University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Amelia","middleName":"","lastName":"Shunk","suffix":""},{"id":534145722,"identity":"0eec6db1-a105-4f9c-a299-ffc7fe62266b","order_by":5,"name":"Angela Mastrianni","email":"","orcid":"","institution":"NYU Grossman School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Angela","middleName":"","lastName":"Mastrianni","suffix":""},{"id":534145724,"identity":"6928b248-5b9e-407e-81d0-47528568ee4a","order_by":6,"name":"Defne L. Levine","email":"","orcid":"","institution":"NYU Grossman School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Defne","middleName":"L.","lastName":"Levine","suffix":""},{"id":534145725,"identity":"c73a4697-3648-4fde-b319-dd19b70c71bd","order_by":7,"name":"Devin M. Mann","email":"","orcid":"","institution":"NYU Grossman School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Devin","middleName":"M.","lastName":"Mann","suffix":""},{"id":534145728,"identity":"4d781978-86a7-4a76-add7-f6c1aa0678c5","order_by":8,"name":"Safiya I. Richardson","email":"data:image/png;base64,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","orcid":"","institution":"NYU Grossman School of Medicine","correspondingAuthor":true,"prefix":"","firstName":"Safiya","middleName":"I.","lastName":"Richardson","suffix":""}],"badges":[],"createdAt":"2025-09-16 15:53:41","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7632374/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7632374/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12911-026-03392-6","type":"published","date":"2026-04-02T15:58:58+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":94481464,"identity":"ccc817eb-4865-4f62-9081-61e06910666b","added_by":"auto","created_at":"2025-10-27 16:13:26","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":465023,"visible":true,"origin":"","legend":"","description":"","filename":"AuditLogManuscriptBMCMIDM100825.docx","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/3c4158b81280cc43f8225168.docx"},{"id":94489935,"identity":"1d2a842c-4b47-48ff-a0bd-fb279bbc500d","added_by":"auto","created_at":"2025-10-27 17:06:34","extension":"png","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":151546,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.nloops.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/9e8d967e54ee463a9108859c.png"},{"id":94481461,"identity":"34678a40-6a7c-46da-87bc-f37bc8fe884c","added_by":"auto","created_at":"2025-10-27 16:13:25","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":151879,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.distinctevents.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/7c83004ef5d483020f444118.png"},{"id":94481707,"identity":"02a6da7f-662e-4fb5-bfd5-a5f040ef80bf","added_by":"auto","created_at":"2025-10-27 16:14:33","extension":"json","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14448,"visible":true,"origin":"","legend":"","description":"","filename":"62671030ebd14fae8d1082e8e8b4fc48.json","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/5288a38d9d7d0fd21d9805f6.json"},{"id":94481631,"identity":"156a3935-404e-43a4-9983-056d97ab4bf1","added_by":"auto","created_at":"2025-10-27 16:13:59","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40968,"visible":true,"origin":"","legend":"","description":"","filename":"AuditLogManuscriptAppendixTables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/fd9bf6adab5d619199ee81f4.xlsx"},{"id":94481459,"identity":"799475ba-edb5-49f3-848f-937012885c29","added_by":"auto","created_at":"2025-10-27 16:13:25","extension":"xml","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":159992,"visible":true,"origin":"","legend":"","description":"","filename":"62671030ebd14fae8d1082e8e8b4fc481enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/ff017ee8757963808b7f4c1b.xml"},{"id":94481700,"identity":"a6c19c88-1675-4435-b14c-2f1ca57af724","added_by":"auto","created_at":"2025-10-27 16:14:31","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":151546,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.nloops.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/3d4dda80212739998571ed53.png"},{"id":94481462,"identity":"3a03c289-732c-42d9-b564-ab760600c04b","added_by":"auto","created_at":"2025-10-27 16:13:25","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":151879,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.distinctevents.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/82916d02fc598c3bbb83e4d0.png"},{"id":94481479,"identity":"7ed85da3-fb27-4818-91fe-8022496db2b7","added_by":"auto","created_at":"2025-10-27 16:13:30","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":151546,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.nloops.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/ee221fa3ebc8ad7e75c0ce07.png"},{"id":94481319,"identity":"23c68685-92f1-4f66-8d31-0b5f9061dc81","added_by":"auto","created_at":"2025-10-27 16:13:07","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":151879,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.distinctevents.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/1e121c0000880d8c0f1b9d47.png"},{"id":94481446,"identity":"a84f88d1-be97-4465-b3ba-1e854df84b24","added_by":"auto","created_at":"2025-10-27 16:13:23","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53488,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure1.nloops.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/3edcc607c41f4fc9c2147413.png"},{"id":94481616,"identity":"18971468-d105-4941-8906-ae3275131624","added_by":"auto","created_at":"2025-10-27 16:13:55","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":45557,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure2.distinctevents.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/af34d0a52db5ec829e26a941.png"},{"id":94481705,"identity":"fe2e6ad2-d56e-487d-acde-6d0d1b4458b2","added_by":"auto","created_at":"2025-10-27 16:14:33","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53488,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure1.nloops.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/c4df98a41ff157373930394c.png"},{"id":94481113,"identity":"5fdc7ff9-5cdc-48aa-89d0-ed5d91fee6b1","added_by":"auto","created_at":"2025-10-27 16:12:41","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":45557,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure2.distinctevents.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/5597f47d0043698dc6143df1.png"},{"id":94481111,"identity":"8d972a0a-eb8f-4849-bcc9-1be1d86f780e","added_by":"auto","created_at":"2025-10-27 16:12:41","extension":"xml","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":152076,"visible":true,"origin":"","legend":"","description":"","filename":"62671030ebd14fae8d1082e8e8b4fc481structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/4b33b3163d8759f289f1578d.xml"},{"id":94481219,"identity":"ae8905ac-bea2-4ede-9a50-e06013f906a0","added_by":"auto","created_at":"2025-10-27 16:12:53","extension":"html","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":168568,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/fe8772049b699d8fcfcde784.html"},{"id":94481150,"identity":"3182034a-6a2e-4dc6-a25f-65fa45134576","added_by":"auto","created_at":"2025-10-27 16:12:48","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":106278,"visible":true,"origin":"","legend":"\u003cp\u003ePlotted predicted probabilities and marginal effects of initiating a statin relative to the number of EHR event loops\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/1e261e6e0c2ed523de2b343d.png"},{"id":94481455,"identity":"966af5f2-84de-41c6-ac95-0ce858e4224f","added_by":"auto","created_at":"2025-10-27 16:13:24","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":105782,"visible":true,"origin":"","legend":"\u003cp\u003ePlotted predicted probabilities and marginal effects of initiating a statin relative to the number of distinct EHR events\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/3df921af48442a1e630b1a23.png"},{"id":106343400,"identity":"c08d9c5a-9ea8-4a5c-ab33-7596bedc3cc7","added_by":"auto","created_at":"2026-04-07 16:05:18","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1914996,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/5b50d94a-7e16-4bcf-b753-96191eac09c8.pdf"},{"id":94481124,"identity":"19534183-ae55-4bda-86b5-d1bfcb1aee93","added_by":"auto","created_at":"2025-10-27 16:12:43","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":40968,"visible":true,"origin":"","legend":"","description":"","filename":"AuditLogManuscriptAppendixTables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7632374/v1/ebc3cc1a8e59358e6a6fe6a1.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Evaluating Cognitive Load in Clinical Workflows Highlights Leverage Points for Guideline-Concordant Statin Initiation","fulltext":[{"header":"1 Introduction","content":"\u003cdiv id=\"Sec2\" class=\"Section2\"\u003e\u003ch2\u003e1.0 Research Goals and Hypotheses\u003c/h2\u003e\u003cp\u003e The objectives of this study were to assess how electronic health record (EHR) metadata can be used to measure cognitive load within primary care workflows related to initiating guideline-concordant statin therapy. We pursued two aims. First, we evaluated whether cognitive load metrics were associated with the likelihood of appropriate statin initiation. Second, we explored whether specific EHR workflow events could be linked to cognitive load patterns that predict statin initiation. We hypothesized that providers adhering to statin guidelines would exhibit higher cognitive effort and distinct EHR interaction patterns compared to those who did not. We expected this increased cognitive load to manifest as: (a) repeated actions to gather necessary information, (b) longer time spent per EHR activity, (c) more overall actions within the EHR, and (d) extended encounter duration. The second aim was exploratory, focusing on developing methods to connect EHR event\u0026ndash;level cognitive load metrics with statin initiation behavior.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e1.1 Literature contribution (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/h2\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\u003eLiterature Contribution Table\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\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\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eProblem or Issue\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eHow can electronic health record (EHR) metadata identify ways to close clinical care gaps? What are EHR workflow cognitive load differences between providers who give appropriate care and providers who do not?\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWhat is Already Known\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eStatins reduce cardiovascular risk but are under prescribed despite guidelines.\u003c/p\u003e\u003cp\u003eClinical decision support systems aim to improve prescribing but have low adoption.\u003c/p\u003e\u003cp\u003eCognitive load in providers can lead to errors and decision delays.\u003c/p\u003e\u003cp\u003eEHR metadata captures clinician behavior and cognitive load, but few studies link it to specific clinical actions.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWhat this Paper Adds\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eA use case to associate EHR metadata metrics with care quality in healthcare.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWho would benefit from the new knowledge in this paper\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eHealthcare systems aiming to leverage EHR metadata to improve healthcare delivery.\u003c/p\u003e\u003cp\u003eHealth informaticians interested in leveraging data-driven approaches to improve clinical workflows.\u003c/p\u003e\u003cp\u003eBehavioral scientists interested in understanding process workflows and sources of cognitive load in workflows.\u003c/p\u003e\u003cp\u003ePolicymakers improving quality of care through targeted interventions in EHR design and clinical guidelines.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"2 Related Work","content":"\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Clinical Context: Statin Prescribing for ASCVD Risk Prevention\u003c/h2\u003e\u003cp\u003eCardiovascular disease remains the leading cause of mortality globally, and failure to appropriately prescribe statins represents a missed opportunity for primary prevention at a population level.[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e] Statins reduce the risk of major adverse cardiovascular events and mortality.[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e] However, primary care providers fail to prescribe statin therapy for about half of patients meeting guideline criteria for initiation.[\u003cspan additionalcitationids=\"CR4 CR5 CR6 CR7\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] Among patients prescribed a statin, about two-thirds receive a lower-than-optimal dose.[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e] To ensure patients are appropriately prescribed statins, providers must recognize that a patient meets one of four indications, often requiring calculating a patient\u0026rsquo;s risk of atherosclerotic cardiovascular disease (ASCVD) and prescribing a guideline-approved statin at the correct dose. Individualized risk assessment requires collecting elements from the patient\u0026rsquo;s medical history, vital signs, and laboratory results, and then using a risk calculator. The EHR presents opportunities to develop clinical decision support systems (CDSSs) that support the recognition, assessment, and management of ASCVD risk, like alerts and pre-populated medication orders. However, generally low provider adoption has limited the clinical impact of current CDSSs designed to improve guideline-concordant statin prescribing.[\u003cspan additionalcitationids=\"CR10 CR11 CR12\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e] Additionally, evolving guidelines and varying thresholds for treatment based on age and comorbidities further complicate the prescribing process, especially without automated support.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.2 EHR Metadata to Evaluate EHR Utilization\u003c/h2\u003e\u003cp\u003eCognitive load can influence the quality of healthcare decision-making by increasing the likelihood of delayed action, reliance on cognitive shortcuts, and misalignment with evidence-based care.[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] In the EHR environment, poorly integrated workflows and excessive demands on attention have been associated with clinician burnout and suboptimal patient outcomes.[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] Quantifying cognitive load in routine practice is therefore critical for understanding its role in decision-making and identifying opportunities for improvement.[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] Traditional approaches to studying clinician-EHR interactions, such as time and motion studies, direct observation, surveys, and interviews, are resource-intensive and prone to biases like the Hawthorne Effect, where awareness of being observed alters behavior.[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] These methods may also fail to capture subtle but important real-time decision points.\u003c/p\u003e\u003cp\u003eEHR metadata is an automated tracking feature that logs behavior within the system (such as adding, deleting, or querying information) along with contextual information (date/time, user, and patient). EHR certification mandates this tracking for security and privacy purposes. As of 2017, 96 percent of hospitals and 80 percent of private practices that used EHRs were compliant with these criteria.[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] Specifically, EHR audit log metadata, collected for HIPAA compliance to track what providers view and click in the EHR, can serve as a non-invasive method to observe how providers and patients interact within the EHR without imposing additional burdens on clinicians.[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] For instance, audit log data can be used to measure how many times a provider clicks between the lab results and medication order screen during a visit, offering a proxy for task-switching or uncertainty. Using EHR metadata provides an analytical approach that not only captures natural interaction patterns but also enables scalable and objective measurement of cognitive load and its link to prescribing behavior, laying a foundation for future interventions that can be assessed through real-world practice patterns.\u003c/p\u003e\u003cp\u003eEHR audit log metadata offers expanding opportunities for health services research. [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] The clinical informatics literature has identified three metadata metrics that directly capture EHR user behavior related to measuring cognitive load: counts of actions captured by metadata, activity durations, and activity sequences.[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e] However, extracting and interpreting these metrics to meaningfully reflect cognitive load is technically challenging. Raw EHR audit log metadata are high-volume, inconsistently structured across systems, and not collected with research in mind, making it challenging to align events with specific clinical decision-making moments. While the literature is limited in linking these metrics to specific healthcare research questions and in measuring the effectiveness of healthcare delivery, [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e] process mining can bridge EHR metadata to desired behaviors and reveal system changes that may drive behavior change. By analyzing how these metrics relate to guideline adherence, health systems can identify workflow inefficiencies or decision points where additional support may be warranted.\u003c/p\u003e\u003cp\u003eThis study aims to fill a critical gap by evaluating whether cognitive load, as measured through EHR metadata, differs between providers who prescribe statins to eligible patients and those who do not\u0026mdash;an insight that could inform workflow-based interventions to close persistent care gaps in ASCVD prevention. Additionally, we aim to determine which EHR activities (e.g., medication orders, patient communications, and the use of pre-existing tools) are associated with a higher cognitive load in encounters that result in guideline-concordant statin prescribing.\u003c/p\u003e\u003c/div\u003e"},{"header":"3 Methods","content":"\u003cp\u003eWe conducted a retrospective, cross-sectional, observational study using EHR audit log metadata and clinical encounter data to evaluate the association between cognitive load during primary care visits and initiation of statin therapy. This study was designed and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e] This study was approved by the NYULH Institutional Review Board (IRB #i23-01349), which granted a waiver of authorization.\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Population and Number of Observations\u003c/h2\u003e\u003cp\u003eData were retrospectively pulled from the EHR of a large academic health system that uses Epic Systems between January 1, 2024, and December 31, 2024, for primary care encounters of patients who qualified for statin initiation during their clinic visit.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Data Inclusion and Exclusion\u003c/h2\u003e\u003cp\u003eTo determine statin eligibility of adult patients between 20 and 75 years of age, we extracted laboratory data, including their most recent lipid panels and hemoglobin A1c, age, self-identified race and ethnicity, sex, smoking history, latest collected systolic blood pressure, medication history, and any previous history of ASCVD events to calculate each patient's cardiovascular risk based on the 2013 American College of Cardiology/American Heart Association (ACC/AHA) ASCVD 10-year risk score [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e] We identified each eligible patient's most recent primary care visit in 2024 and retrieved encounter-level metadata, including provider identifiers.\u003c/p\u003e\u003cp\u003e We included adult patients with at least one documented primary care encounter in 2024 who met clinical guidelines for statin therapy. [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e] Patients were excluded if they lacked sufficient data to calculate a valid risk score, had no primary care encounters during the study period, indicated an allergy or contraindication to a statin, or were pregnant in 2024. Patients who qualified for statin therapy were identified based on the guidelines in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, which are based on the ACC/AHA Task Force on Clinical Practice\u0026rsquo;s guidelines on the management of blood cholesterol. [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e] Patients were then classified based on their statin prescribing status at their latest visit: initiated, defined as those with a statin medication order documented during the encounter, and not prescribed, defined as those without a corresponding statin order despite meeting guideline-based indications. Statins initiated in 2024 can be found in Table A.1.\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\u003eInclusion and Exclusion Criteria for Statin Therapy Eligibility\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\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\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInclusion Criteria\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eExclusion Criteria\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePatients with any of the following guidelines:\u003c/p\u003e\u003cp\u003e\u0026bull; Age 20 to 75, with LDL\u0026thinsp;\u0026ge;\u0026thinsp;190 mg/dL and diagnosis of Type 2 Diabetes Mellitus\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e\u003cp\u003e\u0026bull; Age 40 to 75, with Type 2 Diabetes Mellitus\u003csup\u003e1\u003c/sup\u003e and a history of myocardial infarction (MI)\u003csup\u003e2\u003c/sup\u003e or stroke\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\u003cp\u003e\u0026bull; Age 40 to 75, with an estimated 10-year ASCVD risk\u0026thinsp;\u0026ge;\u0026thinsp;7.5%\u003c/p\u003e\u003cp\u003e\u0026bull; Age 20 to 75, with a history of myocardial infarction (MI)\u003csup\u003e2\u003c/sup\u003e or stroke\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePatients with any of the following conditions:\u003c/p\u003e\u003cp\u003e\u0026bull; Pregnant\u003c/p\u003e\u003cp\u003e\u0026bull; Have a known allergy to statins or a contraindication to statin use\u003c/p\u003e\u003cp\u003e\u0026bull; Have any of the following diagnoses:\u003c/p\u003e\u003cp\u003eo Hepatitis\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eo Cirrhosis\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eo End-Stage Renal Disease (ESRD)\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eo Chronic Kidney Disease (CKD)\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003e\u003csup\u003e1\u003c/sup\u003eType 2 Diabetes Mellitus identified via ICD-10 codes (Table A.2.) or HbA1c\u0026thinsp;\u0026ge;\u0026thinsp;6.5% in their latest blood draw.\u003c/p\u003e\u003cp\u003e\u003csup\u003e2\u003c/sup\u003e Diagnoses identified using ICD-10 codes listed in Table A.2.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Data and Data Structure\u003c/h2\u003e\u003cp\u003eEHR audit log metadata and relevant patient clinical information were extracted from the healthcare system\u0026rsquo;s data lake, which was built on a Hadoop distributed file system, and the relevant data were queried using MySQL databases. For patients initiated on statins in 2024, we extracted metadata from the encounter linked to the initiation. For those not initiated, we used metadata from their most recent annual wellness visit, which occurred in 2024. We defined an encounter as a unique interaction between a statin-prescribing provider (physician, physician assistant [PA], nurse practitioner [NP]) and a patient. An encounter may consist of multiple, non-contiguous time segments during which the provider engages with the patient\u0026rsquo;s data or performs related actions. For example, a provider might review the patient\u0026rsquo;s chart regarding an encounter from 1:00 PM to 1:03 PM, conduct a face-to-face visit from 1:40 PM to 2:10 PM, and later complete documentation after work hours (between 7:00 PM and 7:00 AM) or even the next day. These time segments are combined into a single encounter to reflect the total effort involved with the encounter. From the encounter's metadata, we derived four cognitive load metrics based on the user interactions (an \u0026ldquo;event\u0026rdquo;) logged by the EHR. Events in the EHRs include instances when a provider accesses patient data, their actions related to the data, and other workflow activities associated with an ID tied to the patient encounter. The sequence of timestamped events enables the creation of metrics that represent the effort exerted by the provider when interacting with the EHR, which we refer to as cognitive load metrics. Each of these metrics was chosen because a stakeholder can introduce changes in the EHR based on a desire to alter these metrics, which would, in turn, affect cognitive load and lead to meaningful changes in clinical behavior. These metrics are elaborated on in 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\u003eCognitive Load Metric Definitions\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\"\u003e\u003cp\u003eCognitive Load Metric\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDefinition\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eExample of Cognitive Load Metric\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAssociations between Metric and Opportunity for Behavior Change\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNumber of Loops (N)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eThe number of times an event is revisited during an encounter.\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eUser moves from Event A \u0026rarr; Event B \u0026rarr; Event C \u0026rarr; Event A \u0026rarr; Event B \u0026rarr; Event D (2 loops total \u0026ndash; one loop back to Event A and one loop back to Event B)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMore loops could increase cognitive effort due to redundancy or confusion.\u003c/p\u003e\u003cp\u003eMore loops suggest redesigning workflows to remove unnecessary repetition.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAverage Time Spent on an EHR Event (seconds)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eThe average time spent on an EHR event.\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8.7 seconds per event (e.g., Event A, Event B, Event C)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMore time spent per event indicates complexity and more fixating, suggesting opportunities for simplification or support.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNumber of Distinct EHR Events\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eThe number of unique events in the EHR that a provider visits during their encounter.\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e15 unique events (Event A, Event B, Event C, \u0026hellip;, Event O)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMore distinct events increase task complexity and intrinsic load. More distinct events suggest chunking related tasks or adding grouping mechanisms.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDuration of an Encounter (minutes)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eThe total active time an agent spends engaging with data or actions related to a specific customer interaction, summed across all non-contiguous time segments associated with the same encounter ID.\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e17.9 minutes from first event (A) to last event (C) related to an encounter\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eLonger encounter duration indicates longer time engaging with the EHR for reviewing patient data, which could introduce fatigue or signal focused patient engagement. Shorter encounter duration may be related to goal-oriented behaviors, resulting in decreased cognitive load.\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Analytical Methods\u003c/h2\u003e\u003cp\u003eAll analyses were completed in RStudio Version 2024.12.0\u0026thinsp;+\u0026thinsp;467.\u003c/p\u003e\u003cdiv id=\"Sec12\" class=\"Section3\"\u003e\u003ch2\u003e3.4.1 Descriptive Statistics\u003c/h2\u003e\u003cp\u003eWe collected means, standard deviations, medians, and ranges of cognitive load metrics, patient characteristics during their encounter with the provider, and provider sex. We used the interquartile range (IQR) method to identify potential outliers in measures of EHR interaction complexity, including the number of loops, number of distinct events, total duration, and average time per event. Following the Tukey IQR rule, [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e] values falling more than 1.5 times the IQR below the first quartile or above the third quartile were classified as outliers. A record was flagged as an outlier if any of the cognitive load metrics exceeded the Tukey thresholds.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section3\"\u003e\u003ch2\u003e3.4.2 Associations Between Statin Prescribing and Cognitive Load Metrics\u003c/h2\u003e\u003cp\u003eThe primary outcome of interest is the likelihood of a cognitive load metric influencing a provider\u0026rsquo;s decision to initiate a statin for a patient. To do so, we ran a logistic regression analysis, with the dependent variable being whether a provider had initiated a statin for a patient who was eligible for statin therapy during an encounter in 2024. The independent variables of interest were each of the four cognitive load metrics: number of loops, average time spent on an EHR event, number of distinct EHR events, and duration of an encounter (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). The controlled patient-level covariates included the patient\u0026rsquo;s sex, self-identified race and ethnicity, insurance status (aggregated as commercial or non-commercial), Elixhauser Comorbidity Index [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e] during their visit, the number of active diagnoses during their encounter, and age. We included provider fixed effects to account for unobserved, time-invariant characteristics of each provider that may influence initiation decisions (e.g., years of active practice and number of other patients with similar conditions). In this healthcare system, physicians can see patients independently; however, NPs and PAs, despite having prescribing authority, must have a physician involved in the encounter. To adjust for multiple providers with prescribing abilities seeing a patient in the same encounter, we included clustered standard errors by the encounter ID. This combined approach helps isolate the effect of interest while producing conservative and robust standard errors.\u003c/p\u003e\u003cp\u003eHowever, cognitive load may have a nonlinear relationship with the probability of initiating a statin [\u003cspan additionalcitationids=\"CR31\" citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. In some cases, clinical decision-making improves as cognitive challenge or task engagement increases, up to a saturation point, after which it declines, suggesting an optimal \u0026ldquo;middle ground\u0026rdquo; of effort. In other cases, low-effort task sequences may reflect automation of desired behaviors, whereas high-effort sequences may signal increased deliberation. In contrast, certain intermediate levels of effort could lead to inertia, resulting in no action being taken. Therefore, we introduced quadratic terms of each cognitive load metric to capture potential threshold effects flexibly. If the quadratic term did not statistically contribute to statin initiation, we tested whether excluding it would significantly impact the model fit by comparing the adjusted R\u003csup\u003e2\u003c/sup\u003e. If excluding the nonlinear term did not significantly impact the model fit, the quadratic term was removed to reduce the chance of overfitting the model. We calculated the average marginal effect of the metric on the likelihood of statin initiation for the cognitive load metrics with nonlinear terms to understand how much the likelihood of statin initiation changes for a particular amount of cognitive load. This approach enabled us to quantify both the direction and magnitude of the relationship across different levels of cognitive load, identifying points at which additional effort began to hinder rather than support statin initiation.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section3\"\u003e\u003ch2\u003e3.4.3 Exploratory Analyses \u0026ndash; Sources of Cognitive Load within the EHR\u003c/h2\u003e\u003cp\u003eWe aimed to link cognitive load measures within the EHR to statin initiation to identify opportunities for integrating clinical decision support. Using machine learning, we identified which EHR events were most strongly associated with statin initiation and the types of cognitive load associated with those events. We structured the EHR metadata so that each variable represented a cognitive load metric for a specific EHR event. This means, for example, we measured the average time a provider spent at Event A, how many times a provider looped back to Event B, and whether the provider visited Event C within an encounter. Because not every provider interacts with every EHR event, many of these variables contain zero values, reflecting events that were skipped or that did not have the cognitive load metric. To explore how these detailed interaction patterns relate to statin initiation, we used a gradient-boosted forest model that took these variables as inputs to predict whether a statin was initiated during the encounter. This machine learning method is well-suited for handling large, complex datasets with many variables, especially when relationships may be nonlinear (as per the earlier investigation) or involve interactions between events. It allowed us to identify which specific EHR events\u0026rsquo; cognitive load metrics were most strongly associated with the decision to prescribe a statin. We tuned the forest to a learning rate (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\eta\\:\\)\u003c/span\u003e\u003c/span\u003e ) of 0.1 and a train-validation-test split of 70%-15%-15%. The procedure was repeated 200 times to generate a distribution of performance metrics, from which 95% confidence intervals were derived, offering a reliable assessment of the model's performance and variability.\u003c/p\u003e\u003cp\u003eTo interpret the model outputs, we applied Shapley Additive explanations (SHAP) values, which provide a theoretically grounded approach to attributing feature contributions to individual predictions. SHAP values offer both local interpretability, revealing how specific features influence each decision, and global interpretability, summarizing the overall importance of features across the dataset. [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e] Mean SHAP values for each cognitive load-metric covariate and global mean SHAP values were computed to quantify the importance of each feature within the model in terms of direction (negatively or positively contributing to the prediction) and magnitude (how much the metric at the event contributes to overall prediction). Isolating the most significant contributors to the desired clinical behavior (i.e., statin initiation) with quantified contributions enables the future justification of introducing clinical decision support in the EHR.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"4 Results","content":"\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e4.1 Descriptive Results\u003c/h2\u003e\u003cp\u003eWe analyzed 20,376 patient encounters across the academic healthcare system. The mean age of patients eligible for statin initiation was 58.8 years (SD\u0026thinsp;=\u0026thinsp;10.8), and about half were male. The cohort was predominantly non-Hispanic/Latinx White, and approximately 80% had commercial insurance. Each patient had an average of 6 active diagnoses during their primary care provider encounter. Most encounters were handled by physicians (82.41%), with fewer by nurse practitioners (NP, 11.66%) and physician assistants (PAs, 5.93%). Encounters with NPs had the highest statin initiation rate (50.72% of NP encounters resulted in a statin initiation) compared to 46.86% for physicians and 46.6% for PAs. Encounters with statin initiation showed slightly lower cognitive load metrics on average than encounters with no statin. These differences were small but statistically significant: encounters with initiation had fewer loops, fewer distinct events, shorter duration, and a lower average time per event.\u003c/p\u003e\u003cp\u003eOverall, 17.6% of records had at least one extreme value in the cognitive load metrics (flagged as outliers per the Tukey IQR rule). A chi-squared test showed no significant difference in statin initiation rates between encounters with vs. without such outlier metrics (p\u0026thinsp;=\u0026thinsp;0.20) \u0026ndash; 45.7% vs 44.5% initiation, respectively. Risk analysis revealed that outlier encounters were 3% less likely to involve statin initiation (RR\u0026thinsp;=\u0026thinsp;0.97; OR\u0026thinsp;=\u0026thinsp;0.95, 95% CI 0.89\u0026ndash;1.03), with no significant difference in the likelihood of statin prescribing between the two groups. In other words, extreme cognitive load metrics do not make an encounter more or less likely to result in a prescription. Therefore, we retained all observations in subsequent analyses. All descriptive analyses are presented in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e.\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\u003eDescriptive statistics of patients, providers, and cognitive load metrics\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\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo Initiation\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eInitiation\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eP-value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;11105 Patients)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;9271 Patients)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePatient Characteristics\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAge (years)\u003c/b\u003e\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\u003eMean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e58.9 (10.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e58.7 (10.8)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.211\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedian [Min, Max]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e61.0 [20.0, 75.0]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e60.0 [20.0, 75.0]\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\u003e\u003cb\u003eSex\u003c/b\u003e\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\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5542 (49.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4788 (51.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.0141\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFemale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5560 (50.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4483 (48.4%)\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\u003eUnknown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3 (0.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0 (0%)\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\u003e\u003cb\u003eRace Ethnicity\u003c/b\u003e\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\u003eNon-Hispanic/Latinx White\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6612 (59.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4366 (47.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNon-Hispanic/Latinx Black/African American\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1434 (12.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1098 (11.8%)\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\u003eHispanic/Latinx\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1091 (9.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1791 (19.3%)\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\u003eAsian\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e844 (7.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e723 (7.8%)\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\u003eOther\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1124 (10.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1293 (13.9%)\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\u003e\u003cb\u003eFinancial Class\u003c/b\u003e\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\u003eCommercial\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8907 (80.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e7263 (78.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.00111\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNon-commercial\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2198 (19.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2008 (21.7%)\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\u003e\u003cb\u003eNumber of Active Diagnosis\u003c/b\u003e\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\u003eMean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.40 (3.07)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6.09 (3.51)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedian [Min, Max]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.00 [1.00, 24.0]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5.00 [1.00, 32.0]\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\u003e\u003cb\u003eElixhauser Comorbidity Index\u003c/b\u003e\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\u003eMean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1.68 (5.42)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.42 (5.37)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedian [Min, Max]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0 [-14.0, 42.0]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0 [-18.0, 58.0]\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\u003e\u003cb\u003eNumber of Providers Involved in Encounter\u003c/b\u003e\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\u003eOne\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8133 (73.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6866 (74.1%)\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\u003eTwo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2637 (23.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2154 (23.2%)\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\u003eThree\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e335 (3.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e251 (2.7%)\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\u003e\u003cb\u003eProvider Type\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003ePhysician\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003eNurse Practitioner\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003ePhysician Assistant\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNumber of Encounters\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e33281 (82.41%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4709 (11.66%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e2393 (5.93%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eInitiation Rate\u003c/b\u003e\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\u003eMean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e46.86 (0.77)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e50.72 (1.47)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e46.60 (2.00)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedian [Min, Max]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e46.15 [0, 100]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e50.00 [0, 100]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e41.05 [0, 100]\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003eNo Initiation\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003eInitiation\u003c/b\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\u003e\u003cb\u003eCognitive Load Metrics\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e(N\u0026thinsp;=\u0026thinsp;11105 Encounters)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e(N\u0026thinsp;=\u0026thinsp;9271 Encounters)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003eP-value\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eNumber of Loops\u003c/b\u003e\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\u003eMean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e75.0 (34.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e71.1 (39.4)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedian [Min, Max]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e70.0 [1.00, 311]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e63.0 [0, 465]\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\u003e\u003cb\u003eNumber of Distinct Events\u003c/b\u003e\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\u003eMean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e34.0 (6.01)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33.4 (7.00)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedian [Min, Max]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e34.0 [8.00, 69.0]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33.0 [9.00, 80.0]\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\u003e\u003cb\u003eEncounter Duration (Minutes)\u003c/b\u003e\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\u003eMean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e38.8 (20.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e37.8 (22.1)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedian [Min, Max]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e35.0 [1.17, 183]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33.0 [0.333, 246]\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\u003e\u003cb\u003eAverage Time per EHR Event (Seconds)\u003c/b\u003e\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\u003eMean (SD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e44.5 (61.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e40.9 (55.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedian [Min, Max]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e22.7 [1.84, 970]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e22.0 [1.11, 791]\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\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e4.2 Regression Analyses between Cognitive Load Metrics and Statin Initiation\u003c/h2\u003e\u003cp\u003eWe examined whether the cognitive load metrics were associated with the likelihood of statin initiation in a multivariate logistic regression model, adjusting for patient characteristics and provider effects. Initially, we tested a model with only linear terms for each cognitive load metric; however, two of the cognitive load metrics\u0026mdash;encounter duration (in minutes) and the average time spent per EHR event (in seconds)\u0026mdash;did not show a significant linear relationship with statin initiation. We then examined non-linear relationships by adding quadratic terms. Model comparison revealed that quadratic terms for the number of loops and the number of distinct events significantly improved the fit (Wald test, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Thus, our final model included non-linear (quadratic) terms for loops and distinct events, as well as linear terms for duration and time per event. Comparisons of the linear and nonlinear models are presented in Table A.3.\u003c/p\u003e\u003cdiv id=\"Sec18\" class=\"Section3\"\u003e\u003ch2\u003e4.2.1 Linear Effect of Cognitive Metrics on Statin Initiation\u003c/h2\u003e\u003cp\u003eRegression results are presented in Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. In the final regression model, the average time spent per EHR event and total encounter duration showed significant linear relationships with the initiation of statins. The average time spent per event showed a significant negative association with the likelihood of statin initiation, where spending more time per event was associated with a lower likelihood of initiation (β = -5.017 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). On the other hand, the total duration of the encounter showed a significant positive association with the odds of initiating a statin, where longer encounter duration was associated with increased odds (β\u0026thinsp;=\u0026thinsp;8.223 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). These findings may reflect provider time management strategies, where extended engagement across multiple sections of the EHR is more conducive to initiating guideline-recommended treatment.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section3\"\u003e\u003ch2\u003e4.2.2 Nonlinear Effects of Cognitive Metrics on Statin Initiation\u003c/h2\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrate the two cognitive load metrics which exhibited significant nonlinear associations with statin initiation. For the number of loops, coefficients indicated a U-shaped relationship (linear β = \u0026minus;0.0124, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; quadratic β\u0026thinsp;=\u0026thinsp;0.597\u0026times;10⁻⁵, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Encounters with either very low or very high loop counts were more likely to result in initiation than those with intermediate counts (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, dashed green line). The inflection point occurred at 93.9 loops (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, vertical dashed red line); below this threshold, each additional loop reduced the probability of initiation by 0.21%, whereas above it, each additional loop increased the probability by 0.00046%. The peak effect was observed at 237 loops, with a predicted initiation probability of 0.684.\u003c/p\u003e\u003cp\u003eThe number of distinct events showed an inverse U-shaped relationship (linear model: β\u0026thinsp;=\u0026thinsp;0.0124, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; quadratic model: β = \u0026minus;0.597\u0026times;10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), with the highest probability of initiation observed when encounters have moderate event counts. The inflection point occurred at approximately 18 events (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, vertical dashed red line); below this threshold, each additional event increased the probability of initiation by 1.35%, whereas above it, the effect was negligible, and the probability gradually declined. This suggests an optimal range of distinct EHR actions that may facilitate decision-making regarding guideline-concordant statin initiation, with diminishing returns once complexity exceeds this threshold.\u003c/p\u003e\u003cp\u003ePatient race and sex were also significant predictors of statin initiation, as shown in Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. Compared to non-Hispanic/Latinx White patients, patients of racial and ethnic minorities were more likely to be initiated on a statin, as well as older patients. Female patients, as well as those with a greater number of active diagnoses or higher comorbidity burden, were less likely to receive initiation.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eCoefficients of the nonlinear model associating cognitive load with statin initiation. + p\u0026thinsp;\u0026lt;\u0026thinsp;0.1, * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01, *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001\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\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eReference Level\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCoefficient\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eStd. Err\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eStatistic\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNumber of Loops\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e--\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026minus;1.239e\u0026thinsp;\u0026minus;\u0026thinsp;02***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(2.123e\u0026thinsp;\u0026minus;\u0026thinsp;03)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-5.84\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNumber of Loops\u0026sup2;\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6.597e\u0026thinsp;\u0026minus;\u0026thinsp;05***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(1.021e\u0026thinsp;\u0026minus;\u0026thinsp;05)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e6.46\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNumber of Distinct Events\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e--\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6.412e\u0026thinsp;\u0026minus;\u0026thinsp;02***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(8.381e\u0026thinsp;\u0026minus;\u0026thinsp;03)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e7.65\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDistinct Events\u0026sup2;\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026minus;1.787e\u0026thinsp;\u0026minus;\u0026thinsp;03***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(1.762e\u0026thinsp;\u0026minus;\u0026thinsp;04)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-10.1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTotal Duration (min)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e--\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e8.223e\u0026thinsp;\u0026minus;\u0026thinsp;03***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(1.436e\u0026thinsp;\u0026minus;\u0026thinsp;03)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e5.73\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAvg Time Per Event (sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e--\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026minus;5.017e\u0026thinsp;\u0026minus;\u0026thinsp;04***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(1.314e\u0026thinsp;\u0026minus;\u0026thinsp;04)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-3.82\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRace: Black/African American\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNon-Hispanic/Latinx White\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.967e\u0026thinsp;\u0026minus;\u0026thinsp;01**\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(6.567e\u0026thinsp;\u0026minus;\u0026thinsp;02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRace: Hispanic/Latino\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e--\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5.906e\u0026thinsp;\u0026minus;\u0026thinsp;01***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(6.508e\u0026thinsp;\u0026minus;\u0026thinsp;02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e9.08\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRace: Asian\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e--\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.425e\u0026thinsp;\u0026minus;\u0026thinsp;01***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(8.083e\u0026thinsp;\u0026minus;\u0026thinsp;02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e5.47\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRace: Other\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6.839e\u0026thinsp;\u0026minus;\u0026thinsp;01***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(6.363e\u0026thinsp;\u0026minus;\u0026thinsp;02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e10.7\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePatient Age\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1.023e\u0026thinsp;\u0026minus;\u0026thinsp;02***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(2.101e\u0026thinsp;\u0026minus;\u0026thinsp;03)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e4.87\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePatient Sex: Female\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMale\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026minus;8.218e\u0026thinsp;\u0026minus;\u0026thinsp;02*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(4.103e\u0026thinsp;\u0026minus;\u0026thinsp;02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-2\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePatient Sex: Unknown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026minus;2.016e\u0026thinsp;+\u0026thinsp;01***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(5.316e\u0026thinsp;\u0026minus;\u0026thinsp;02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-379\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInsurance: Non-commercial\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCommercial\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e5.738e\u0026thinsp;\u0026minus;\u0026thinsp;02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(5.466e\u0026thinsp;\u0026minus;\u0026thinsp;02)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1.05\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNumber of Active Diagnoses\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026minus;4.879e\u0026thinsp;\u0026minus;\u0026thinsp;02***\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(7.546e\u0026thinsp;\u0026minus;\u0026thinsp;03)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-6.47\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eElixhauser Comorbidity Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u0026minus;9.656e\u0026thinsp;\u0026minus;\u0026thinsp;03**\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e(3.742e\u0026thinsp;\u0026minus;\u0026thinsp;03)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-2.58\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eN Observations\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e36131\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e Adj.\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.128\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRMSE\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003e4.3 Exploratory Analyses for EHR Events Associated with Cognitive Load Metrics\u003c/h2\u003e\u003cp\u003eWe analyzed the EHR metadata with a machine learning model to pinpoint which specific EHR activities and associated cognitive load metrics were most strongly linked to statin initiation. We trained a gradient-boosted decision tree model. We computed SHAP values to quantify the contribution of each EHR interaction feature to the model\u0026rsquo;s predictions. The model achieved an area under the receiver operating characteristic curve (AUROC) of 0.91, indicating that using cognitive load metrics has strong performance in predicting statin initiation. Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e presents the 10 EHR event-specific features that contributed most to the model\u0026rsquo;s prediction, ordered by their global mean SHAP values and importance. A full list of contributions by cognitive load-EHR events can be found in Table A.4.\u003c/p\u003e\u003cp\u003eThe average time per event (in seconds) had the highest relative contribution to predicting statin initiation, accounting for 72.2% of the model\u0026rsquo;s relative contribution to its predictions. There was a positive average SHAP value contribution, indicating a positive association between average time spent per event and statin initiation The number of loops accounted for 15.1% of the relative contribution but had a small negative average SHAP value contribution, indicating a negative association between loops and statin initiation. The presence of specific EHR events contributed 12.6% to the model\u0026rsquo;s predictions and had a small positive average contribution, indicating a positive association between the presence of an event and statin initiation.\u003c/p\u003e\u003cp\u003eThe average time spent on viewing patient communications had a positive mean SHAP value of 0.0874, indicating that spending more time on a patient\u0026rsquo;s results was associated with a higher probability of statin initiation. This feature accounted for 23.1% of the model\u0026rsquo;s relative contribution to its predictions, making it the most influential predictor. The average time spent on suggested medication order sets had a positive mean SHAP value of 0.0129, indicating that spending more time on suggested medication order sets was associated with a higher probability of statin initiation. This feature accounted for 12.7% of the model\u0026rsquo;s relative contribution to its predictions, making it the second most influential predictor. The presence of the order set for testing contributed positively to predicting statin initiation, accounting for a 6.5% increase in the model\u0026rsquo;s predictive accuracy. Conversely, the average time spent on changing the list of orders showed a negative mean SHAP value of -0.0115, implying that modifying the order list was associated with a decreased likelihood of initiation, contributing approximately 7.74% to the model\u0026rsquo;s relative influence. Looping back to the order list contributed 3.74% to the model\u0026rsquo;s relative influence. It had a negative average SHAP value, indicating that the number of loops to change an order list was negatively associated with statin initiation.\u003c/p\u003e\u003cp\u003eThese results suggest that interactions with the patient's lab results and spending time on medication order sets positively influence prescribing behavior, while spending time on order modifications and looping back to change the orders may signal complexity or uncertainty, which reduces the likelihood of initiation.\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\u003eTop 10 cognitive load features associated with statin initiation, ranked by mean absolute SHAP value. The relative contribution of these features is 70.06% of the model\u0026rsquo;s predictions.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"9\"\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\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRank\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEHR Event\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCognitive Load Metric\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMean SHAP\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eMean Absolute SHAP\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003e95% CI Lower\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003e95% CI Upper\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eContribution Direction\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u003cp\u003eRelative Contribution\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\u003e1\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePatient communications viewed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAverage Time Per Event (sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.0874\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.7205\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.1014\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.0734\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eIncreasing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.2313\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e2\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOrder set\u0026nbsp;suggested\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAverage Time Per Event (sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.0129\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.3967\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.0200\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.0059\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eIncreasing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.1274\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e3\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOrder modified\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAverage Time Per Event (sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.0115\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.2410\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e-0.0064\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e-0.0166\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eDecreasing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.0774\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e4\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOrder set suggested\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eEvent Shown (1/0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.0115\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.2030\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.0151\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.0078\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eIncreasing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.0652\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e5\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePatient\u0026rsquo;s lab results\u0026nbsp;reviewed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAverage Time Per Event (sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.0293\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.1321\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.0322\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.0264\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eIncreasing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.0424\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e6\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOrder\u0026nbsp;list\u0026nbsp;changed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNumber of Loops\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.0090\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.1164\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e-0.0068\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e-0.0112\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eDecreasing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.0374\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e7\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOrder set\u0026nbsp;activity\u0026nbsp;selected\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAverage Time Per Event (sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.0043\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.1091\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.0066\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e0.0019\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eIncreasing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.0350\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e8\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAlert\u0026nbsp;displayed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAverage Time Per Event (sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.0051\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.1004\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e-0.0034\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e-0.0069\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eDecreasing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.0322\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e9\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eProblem\u0026nbsp;List\u0026nbsp;modified\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNumber of Loops\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.0021\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.0826\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e-0.0002\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e-0.0041\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eDecreasing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.0265\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003e10\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eChart\u0026nbsp;Review\u0026nbsp;Encounter\u0026nbsp;report\u0026nbsp;viewed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAverage Time Per Event (sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e-0.0024\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0.0807\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e-0.0010\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e-0.0038\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eDecreasing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e0.0259\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"5 Discussion","content":"\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\u003ch2\u003e5.1 Summary\u003c/h2\u003e\u003cp\u003eWe identified statistically significant relationships between cognitive load metrics and the likelihood of initiating statin therapy during primary care encounters. First, we identified that cognitive load does not have a linear relationship to desired clinical behavior. Some types of cognitive load (in this paper\u0026rsquo;s case, engagement via repetitive behaviors) suggest that a \u0026ldquo;middle ground\u0026rdquo; of effort by providers increases the likelihood of a clinical behavior, that some effort, but not too much or too little effort, was associated with more appropriate prescribing. We also found that under-engagement (involving too few interactions) and over-engagement (characterized by excessive interactions) can positively impact clinical decision outcomes in statin initiation, possibly suggesting that either more active engagement or short, goal-directed engagement with patients leads to more appropriate prescribing behavior. These results highlight the dynamic nature of cognitive load in clinical environments. By applying machine learning techniques to EHR metadata, we identified key behaviors that current CDSS can leverage to support providers in delivering appropriate clinical care, as well as behaviors where existing EHR features may inadvertently hinder such decisions. It has been well documented that EHR nudges can result in alert fatigue; however, our work suggests that aptly designed CDS can optimize intended effect while reducing cognitive load. Specifically, CDS can create changes to interfaces when providers interact with a patient\u0026rsquo;s lab results, order medications, and work with alerts in the EHR that facilitate target actions. These findings underscore the potential of CDSS to enhance clinical decision-making and improve patient outcomes. [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec23\" class=\"Section2\"\u003e\u003ch2\u003e5.2 Implications\u003c/h2\u003e\u003cp\u003eThis study provides actionable insights into how cognitive load in EHR workflows affects clinical decision-making behavior and presents a framework for how to improve clinical decision-making. By measuring cognitive load indicators within the EHR and identifying specific interaction patterns associated with prescribing decisions, we highlight key moments in the workflow that can be targeted with more effective CDS tools. These findings lay the groundwork for designing interventions that reduce unnecessary complexity and better support clinicians in following guideline-recommended therapies.\u003c/p\u003e\u003cp\u003eOur methodological approach provides both immediate and long-term benefits for health systems. In the short term, these insights can help health systems identify workflow gaps that hinder the provision of effective decision support. Over the long term, this approach has the potential to transform how health systems monitor and improve provider behavior, enabling data-driven quality initiatives. These methods can be further refined to distinguish between helpful and unhelpful cognitive load for providers, ultimately leading to better patient care through intentional, thoughtful digital actions. For patients, improved prescribing practices can have a direct and meaningful impact on outcomes, particularly by increasing the use of appropriate statins to lower primary ASCVD risk.\u003c/p\u003e\u003cp\u003eThe developed methodology is scalable and adaptable to various clinical settings and EHR systems. Our data processing and workflow characterization techniques can support a wide range of efforts to enhance clinical decision support and promote evidence-based care at scale. Firstly, our approach to extracting audit log metadata and clinical behavior data, and linking these to clinical data, can be easily adapted for other behavior-related questions, such as specialty referrals, shared decision-making, and screening choices. Secondly, the identification of behaviors within the EHR and the creation of cognitive load metrics are independent of any specific EHR, as they rely on process mining and data science techniques. The analytical methods used to examine the relationship between cognitive load and clinical behavior, along with machine learning approaches that connect cognitive load to EHR events, offer a systematic way to evaluate CDSS designed to improve clinical practices and identify behaviors that need adjustment at specific interfaces.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec24\" class=\"Section2\"\u003e\u003ch2\u003e5.3 Limitations\u003c/h2\u003e\u003cp\u003eSeveral limitations should be considered when interpreting our findings. First, the cognitive load and workflow measures used in this study were intentionally simple. While more sophisticated process mining techniques, such as clustering workflow sequences or calculating transition probabilities between EHR events, could yield deeper insights, we prioritized interpretable, actionable metrics aligned with behavior change theory. These simpler measures are more readily translatable into tools within the EHR to support clinical decision-making. Moreover, the application of EHR metadata for healthcare quality improvement remains an emerging area within clinical informatics. There is a pressing need for standardization of EHR metadata metrics by organizations such as the Office of the National Coordinator for Health Information Technology (ONC), which could also enable validation of more complex measures through direct observation or qualitative research.\u003c/p\u003e\u003cp\u003eSecond, our findings are based on data from a single EHR platform within a single academic health center\u0026rsquo;s EHR system. While the specific events captured may vary across institutions, our computational approach was designed to be generalizable by focusing on abstracted, system-agnostic event types. Thus, while implementation may require adaptation to local data structures, the methodology should be transferable to other EHR platforms.\u003c/p\u003e\u003cp\u003eThird, the cognitive load metrics used in this study do not distinguish between intrinsic, extrinsic, and germane cognitive load. As a result, we cannot determine whether the measured cognitive load metrics reflect the inherent complexity of the clinical decision (intrinsic load), inefficiencies or distractions introduced by the EHR interface or workflow (extrinsic load), or the cognitive effort invested in understanding and integrating information for decision-making (germane load). Without this differentiation, it is difficult to target design interventions precisely, like reducing extrinsic load without inadvertently removing beneficial germane load. Future work incorporating methods such as task analysis, think-aloud protocols, or real-time workload assessment could help disentangle these components and guide more tailored improvements to CDSS and EHR.\u003c/p\u003e\u003cp\u003eFourth, our current methods do not link each EHR action to a specific medication, lab, or diagnosis. This allowed us to compare all encounters with and without statin initiation in a general way. In future work, we could focus on encounters that involve the same medications and labs and see how cognitive effort on each specific action\u0026mdash;like reviewing cholesterol lab results or adjusting a statin order\u0026mdash;affects whether a statin is prescribed.\u003c/p\u003e\u003cp\u003eFifth, this study focused on one clinical use case: initiating maintenance medications (statins) in primary care. Although our population identification and data linkage methods were tailored to this context, the overall approach can be applied to other clinical domains. Importantly, the metadata metrics were derived independently of the clinical content, which supports broader applicability. With further development, EHR vendors and health systems could aggregate and standardize such metrics to inform workflow improvement and decision support across various clinical scenarios. For example, the scope of the study was intentionally limited to initiation and did not include decisions regarding the titration of statins. Further research can elaborate on differentiating the cognitive load of workflows for initiation versus titration, given the distinct behavioral mechanisms and decision-making processes associated with each. This study also limited its analysis to the behavior of providers, despite evidence of patient hesitancy to start statin therapy.[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e] The methods of this study are equally applicable to the EHR audit log metadata of patient-facing interfaces and patient portals, to enhance the digital patient experience.\u003c/p\u003e\u003c/div\u003e"},{"header":"6 Conclusion","content":"\u003cp\u003eOur findings suggest that the likelihood of initiating statins is not simply determined by the amount of effort in the EHR but is influenced by a nonlinear relationship between cognitive load and decision-making. Both too little and too much engagement with the EHR can decrease the chances of initiating appropriate therapy, while an optimal level of engagement seems to promote guideline-concordant prescribing. Similarly, extreme engagement behaviors in the EHR can support clinical actions, whereas a moderate approach may lead to low engagement. By identifying specific EHR interaction patterns and workloads that either facilitate or hinder statin initiation, our results highlight practical opportunities to redesign clinical decision support and workflows to better balance cognitive demands. Efficient, targeted EHR use\u0026mdash;rather than maximal or minimal effort\u0026mdash;may be the most effective way to ensure timely and appropriate treatment decisions.\u003c/p\u003e\u003cp\u003eImportantly, our study addresses and overcomes significant barriers and challenges associated with using EHR metadata. We demonstrate that these complex metrics can be extracted, processed, and mapped to detailed clinical workflows, enabling robust analysis of how cognitive load impacts clinical decision-making. This advancement provides a framework for health systems to better understand and optimize EHR interactions, ultimately supporting improved guideline adherence and patient outcomes.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"672\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTerm / Abbreviation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 390px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDefinition\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eAtherosclerotic cardiovascular disease (ASCVD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 390px;\"\u003e\n \u003cp\u003eCondition caused by plaque buildup (atherosclerosis) in the arteries\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eASCVD 10-year risk score\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 390px;\"\u003e\n \u003cp\u003eEstimate of a person\u0026rsquo;s chance of having a heart attack or stroke within the next 10 years\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eClinical Decision Support Systems (CDSS)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 390px;\"\u003e\n \u003cp\u003eTools or systems within the EHR that provide clinicians with information, alerts, or recommendations to improve patient care decisions.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eCognitive Load\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 390px;\"\u003e\n \u003cp\u003eThe mental effort required to process information and complete tasks, influenced by the complexity of the information, the environment, and the interface design.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eElectronic Health Record (EHR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 390px;\"\u003e\n \u003cp\u003eA digital system for storing and managing patient medical records and clinical workflows.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eEHR Metadata\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 390px;\"\u003e\n \u003cp\u003eData automatically generated and stored by the EHR that describes user interactions, such as keystrokes, clicks, screen time, and navigation patterns.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eGradient-boosted forests\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 390px;\"\u003e\n \u003cp\u003eMachine learning method that combines decision trees built in sequence in which new trees correct errors of previous trees to improve accuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eHadoop distributed file system\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 390px;\"\u003e\n \u003cp\u003eSystem for storing large data sets across multiple machines, allowing processing in parallel\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eHawthorne Effect\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 390px;\"\u003e\n \u003cp\u003eWhen people change their behavior because they are aware they are being observed\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eAUC-ROC (Area Under the Curve \u0026ndash; Receiver Operating Characteristic)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 390px;\"\u003e\n \u003cp\u003eA measure of a binary classifier\u0026rsquo;s performance; higher values indicate better discrimination between positive and negative cases.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eRandom Forest\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 390px;\"\u003e\n \u003cp\u003eA machine learning method that uses multiple decision trees to make predictions or classifications; in this study, used to predict statin initiation from cognitive load metrics.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eShapley additive explanation (SHAP) values\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 390px;\"\u003e\n \u003cp\u003eMethod to explain machine learning model\u0026rsquo;s predictions by quantifying each feature\u0026rsquo;s contribution to a specific prediction\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eStatin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 390px;\"\u003e\n \u003cp\u003eA class of cholesterol-lowering medications used to reduce cardiovascular risk, often prescribed for patients with elevated ASCVD risk, by blocking the production of low-density lipids (LDL) in the liver, helping prevent heart attacks and strokes.\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eSTROBE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 390px;\"\u003e\n \u003cp\u003eThe Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for reporting observational studies\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 282px;\"\u003e\n \u003cp\u003eTukey IQR rule\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 390px;\"\u003e\n \u003cp\u003eA statistical method for identifying outliers by defining data points lying more than 1.5 times the interquartile range above the third quartile or below the first quartile.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"Declarations","content":"\u003ch2\u003e7.2 Ethics Approval and Consent to Participate\u003c/h2\u003e\n\u003cp\u003eThis study was reviewed and approved by the NYU Langone Health Office of Science and Research Institutional Review Board (IRB #i23-01349). The NYU Langone\u0026nbsp;IRB granted a waiver of informed consent. Confidential patient data were accessed and analyzed within the secure hospital network in compliance with institutional policies. The study was conducted in accordance with the principles of the Declaration of Helsinki.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e7.3 Consent for Publication\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003ch2\u003e7.4 Availability of Data and Materials\u003c/h2\u003e\n\u003cp\u003eDue to the use of patient data in this study, deidentified versions of the data can be made available upon reasonable request. Dr. Safiya Richardson can be contacted at [email protected] for data requests.\u003c/p\u003e\n\u003ch2\u003e7.5 Competing Interests\u003c/h2\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003ch2\u003e7.6 Funding\u003c/h2\u003e\n\u003cp\u003eThis work was supported by the National Institutes of Health\u0026rsquo;s Heart, Lung, and Blood Institute (1R01HL171292-01).\u003c/p\u003e\n\u003ch2\u003e7.7 Author Contributions (CRediT Taxonomy)\u003c/h2\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eContributor Role\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRole Definition\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAuthor\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eConceptualization\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eIdeas; formulation or evolution of overarching research goals and aims.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, SIR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eData Curation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eManagement activities to annotate (produce metadata), scrub data and maintain research data (including software code, where it is necessary for interpreting the data itself) for initial use and later reuse.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, YC\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eFormal Analysis\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eApplication of statistical, mathematical, computational, or other formal techniques to analyze or synthesize study data.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, YC\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eFunding Acquisition\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAcquisition of the financial support for the project leading to this publication.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, DMM, SIR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eInvestigation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eConducting a research and investigation process, specifically performing the experiments, or data/evidence collection.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, YC, PS\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMethodology\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eDevelopment or design of methodology; creation of models\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, YC, PS, SIR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eProject Administration\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eManagement and coordination responsibility for the research activity planning and execution.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, NR, SIR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eResources\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eProvision of study materials, reagents, materials, patients, laboratory samples, animals, instrumentation, computing resources, or other analysis tools.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eDMM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eSoftware\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eProgramming, software development; designing computer programs; implementation of the computer code and supporting algorithms; testing of existing code components.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, YC\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eSupervision\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eOversight and leadership responsibility for the research activity planning and execution, including mentorship external to the core team.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, NR, SIR, DMM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eValidation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eVerification, whether as a part of the activity or separate, of the overall replication/reproducibility of results/experiments and other research outputs.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eN/A\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eVisualization\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003ePreparation, creation and/or presentation of the published work, specifically visualization/data presentation.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, YC\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eWriting \u0026ndash; Original Draft Preparation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eCreation and/or presentation of the published work, specifically writing the initial draft (including substantive translation).\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, YC, PS, NR, AM, DLL, ARS, SIR, DMM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eWriting \u0026ndash; Review \u0026amp; Editing\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003ePreparation, creation and/or presentation of the published work by those from the original research group, specifically critical review, commentary or revision \u0026ndash; including pre- or post-publication stages.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eRVNV, YC, PS, NR, AM, DLL, ARS, SIR, DMM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch2\u003e7.8 Acknowledgements\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSparrow RT, Khan AM, Ferreira-Legere LE, Ko DT, Jackevicius CA, Goodman SG, et al. Effectiveness of Interventions Aimed at Increasing Statin-Prescribing Rates in Primary Cardiovascular Disease Prevention: A Systematic Review of Randomized Clinical Trials. JAMA Cardiol. 2019;4:1160-9.\u003c/li\u003e\n\u003cli\u003eBaigent C, Blackwell L, Emberson J, Holland LE, Reith C, Bhala N, et al. Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials. Lancet. 2010;376:1670-81.\u003c/li\u003e\n\u003cli\u003eKannan S, Asch DA, Kurtzman GW, Honeywell Jr. S, Day SC, Patel MS. Patient and Physician Predictors of Hyperlipidemia Screening and Statin Prescription. Am J Manag care. 2018;24:241-8.\u003c/li\u003e\n\u003cli\u003eBradley CK, Wang TY, Li S, Robinson JG, Roger VL, Goldberg AC, et al. Patient-Reported Reasons for Declining or Discontinuing Statin Therapy: Insights From the PALM Registry. J Am Heart Assoc. 2019;8:e011765.\u003c/li\u003e\n\u003cli\u003eSalami JA, Warraich H, Valero-Elizondo J, Spatz ES, Desai NR, Rana JS, et al. National Trends in Statin Use and Expenditures in the US Adult Population From 2002 to 2013: Insights From the Medical Expenditure Panel Survey. JAMA cardiology. 2017;2:56-65.\u003c/li\u003e\n\u003cli\u003eAustin PC, Mamdani MM, Juurlink DN, Alter DA, Tu JV. Missed opportunities in the secondary prevention of myocardial infarction: an assessment of the effects of statin underprescribing on mortality. Am Heart J. 2006;151:969-75.\u003c/li\u003e\n\u003cli\u003ePauff BR, Jiroutek MR, Holland MA, Sutton BS. Statin Prescribing Patterns: An Analysis of Data From Patients With Diabetes in the National Hospital Ambulatory Medical Care Survey Outpatient Department and National Ambulatory Medical Care Survey Databases, 2005-2010. Clin Ther. 2015;37:1329-39.\u003c/li\u003e\n\u003cli\u003eBerthold HK, Gouni-Berthold I, Bohm M, Krone W, Bestehorn KP. Patterns and predictors of statin prescription in patients with type 2 diabetes. Cardiovasc Diabetol. 2009;8:25.\u003c/li\u003e\n\u003cli\u003eGold R, Larson AE, Sperl-Hillen JM, Boston D, Sheppler CR, Heintzman J, et al. Effect of Clinical Decision Support at Community Health Centers on the Risk of Cardiovascular Disease: A Cluster Randomized Clinical Trial. JAMA Netw Open. 2022;5:e2146519.\u003c/li\u003e\n\u003cli\u003eAlameddine R, Seifeddine S, Ishak H, Antoun J. Improving statin prescription through the involvement of nurses in the provision of ASCVD score: a quality improvement initiative in primary care. Postgrad Med. 2020;132:479-84.\u003c/li\u003e\n\u003cli\u003ePeiris D, Usherwood T, Panaretto K, Harris M, Hunt J, Redfern J, et al. Effect of a computer-guided, quality improvement program for cardiovascular disease risk management in primary health care: the treatment of cardiovascular risk using electronic decision support cluster-randomized trial. Circ Cardiovasc Qual Outcomes. 2015;8:87-95.\u003c/li\u003e\n\u003cli\u003eDerose SF, Dudl JR, Benson VM, Contreras R, Nakahiro RK, Ziel FH. Point-of-Service reminders for prescribing cardiovascular medications. Am J Manag Care. 2005;11:298-304.\u003c/li\u003e\n\u003cli\u003eJacobson TA, Gutkin SW, Harper CR. Effects of a global risk educational tool on primary coronary prevention: the Atherosclerosis Assessment Via Total Risk (AVIATOR) study. Current Medical Research and Opinion. 2006;22:1065-73.\u003c/li\u003e\n\u003cli\u003eWillcox M. Shortcuts versus Analysis \u0026ndash;Ignoring is Decisional Bliss. The Business of Choice: How Human Instinct Influences Everyone\u0026rsquo;s Decisions. 2020.\u003c/li\u003e\n\u003cli\u003eSingh M. Heuristics in the delivery room. Science. 2021;374:324-9.\u003c/li\u003e\n\u003cli\u003eAdler-Milstein J, Zhao W, Willard-Grace R, Knox M, Grumbach K. Electronic health records and burnout: Time spent on the electronic health record after hours and message volume associated with exhaustion but not with cynicism among primary care clinicians. J Am Med Inform Assoc. 2020;27:531-8.\u003c/li\u003e\n\u003cli\u003eAsgari E, Kaur J, Nuredini G, Balloch J, Taylor AM, Sebire N, et al. Impact of Electronic Health Record Use on Cognitive Load and Burnout Among Clinicians: Narrative Review. JMIR Med Inform. 2024;12:e55499.\u003c/li\u003e\n\u003cli\u003eSedgwick P, Greenwood N. Understanding the Hawthorne effect. BMJ. 2015;351:h4672.\u003c/li\u003e\n\u003cli\u003eCohen G, Brown L, Fitzgerald M, Somplasky A. To Measure The Burden Of EHR Use, Audit Logs Offer Promise\u0026mdash;But Not Without Further Collaboration. 2020.\u003c/li\u003e\n\u003cli\u003eAdler-Milstein J, Adelman JS, Tai-Seale M, Patel VL, Dymek C. EHR audit logs: A new goldmine for health services research? J Biomed Inform. 2020;101:103343.\u003c/li\u003e\n\u003cli\u003eWu DT, Smart N, Ciemins EL, Lanham HJ, Lindberg C, Zheng K. Using EHR audit trail logs to analyze clinical workflow: a case study from community-based ambulatory clinics. AMIA Annual Symposium Proceedings2018. p. 1820.\u003c/li\u003e\n\u003cli\u003eRule A, Chiang MF, Hribar MR. Using electronic health record audit logs to study clinical activity: a systematic review of aims, measures, and methods. J Am Med Inform Assoc. 2020;27:480-90.\u003c/li\u003e\n\u003cli\u003eDe Roock E, Martin N. Process mining in healthcare - An updated perspective on the state of the art. J Biomed Inform. 2022;127:103995.\u003c/li\u003e\n\u003cli\u003eVon Elm E, Altman DG, Egger M, Pocock SJ, G\u0026oslash;tzsche PC, Vandenbroucke JP. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. The lancet. 2007;370:1453-7.\u003c/li\u003e\n\u003cli\u003eGoff DC, Jr., Lloyd-Jones DM, Bennett G, Coady S, D\u0026apos;Agostino RB, Gibbons R, et al. 2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. Circulation. 2014;129:S49-73.\u003c/li\u003e\n\u003cli\u003eGrundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, et al. 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA Guideline on the Management of Blood Cholesterol: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Circulation. 2019;139:e1082-e143.\u003c/li\u003e\n\u003cli\u003eGrundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, et al. 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA guideline on the management of blood cholesterol: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Journal of the American College of Cardiology. 2019;73:e285-e350.\u003c/li\u003e\n\u003cli\u003eTukey JW. Exploratory data analysis: Springer; 1977.\u003c/li\u003e\n\u003cli\u003eElixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity Measures for Use with Administrative Data. Medical Care. 1998;36:8-27.\u003c/li\u003e\n\u003cli\u003eSweller J. Measuring cognitive load. Perspect Med Educ. 2018;7:1-2.\u003c/li\u003e\n\u003cli\u003eWason PC, Evans JSBT. Dual processes in reasoning? Cognition. 1974;3:141-54.\u003c/li\u003e\n\u003cli\u003eYerkes RM, Dodson JD. The relation of strength of stimulus to rapidity of habit‐formation. Journal of Comparative Neurology and Psychology. 1908;18:459-82.\u003c/li\u003e\n\u003cli\u003eMerrick L, Taly A. The explanation game: Explaining machine learning models using shapley values. Machine Learning and Knowledge Extraction: 4th IFIP TC 5, TC 12, WG 84, WG 89, WG 129 International Cross-Domain Conference, CD-MAKE 2020, Dublin, Ireland, August 25\u0026ndash;28, 2020, Proceedings 4: Springer; 2020. p. 17-38.\u003c/li\u003e\n\u003cli\u003eKawamoto K, Houlihan CA, Balas EA, Lobach DF. Improving clinical practice using clinical decision support systems: a systematic review of trials to identify features critical to success. BMJ. 2005;330:765.\u003c/li\u003e\n\u003cli\u003eXie M, Martin SS, Turchin A. Reasons for non-acceptance of statin therapy by patients at high cardiovascular risk. Sci Rep. 2025;15:17014.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"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":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"statin, clinical decision support, EHR, process mining, cognitive load, audit log data, machine learning","lastPublishedDoi":"10.21203/rs.3.rs-7632374/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7632374/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cb\u003eIntroduction\u003c/b\u003e\u003c/p\u003e\u003cp\u003e Linking EHR use to care quality offers insights for interventions to improve guideline adherence and close care gaps. We examine how EHR metadata can measure cognitive load in primary care providers during statin prescribing and identify points of cognitive load in the EHR workflow.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMethods\u003c/b\u003e\u003c/p\u003e\u003cp\u003eEHR primary care encounter data from a large academic health system in 2024 were retrospectively extracted. We identified adult patients who met the criteria for statin initiation and calculated their ASCVD risk scores. Cognitive load metrics were derived from EHR metadata. Logistic regressions evaluate associations between cognitive load and statin initiation, adjusting for patient covariates and provider fixed effects. Gradient-boosted forests and SHAP values identified key EHR events and cognitive load associated with the initiation of statin therapy.\u003c/p\u003e\u003cp\u003e\u003cb\u003eResults\u003c/b\u003e\u003c/p\u003e\u003cp\u003eLonger encounter duration increased the likelihood of statin initiation, whereas more time spent per EHR event decreased it. Non-linear effects were observed for loop count and distinct event count: the probability of initiation decreased with increasing loop counts up to 93.9 loops, then increased beyond this threshold. For distinct events, the initiation probability increased up to approximately 18 events and declined at higher counts. In a gradient-boosted decision tree model, average time per event was the strongest predictor (72.2% relative contribution). Additional positive predictors included the time spent reviewing lab results and on suggested medication order sets. Modifying the order list and looping back to it were negatively associated with statin initiation.\u003c/p\u003e\u003cp\u003e\u003cb\u003eDiscussion\u003c/b\u003e\u003c/p\u003e\u003cp\u003eEHR metadata can associate cognitive load with appropriate clinical behavior, finding nonlinear relations between cognitive load and statin initiation rates. This work highlights the need to optimize EHR systems to reduce cognitive burden and support clinical decision-making. Connecting cognitive load to prescribing behavior gives insight into how workflow adjustments and enhanced decision support can improve adherence to guidelines and patient care.\u003c/p\u003e","manuscriptTitle":"Evaluating Cognitive Load in Clinical Workflows Highlights Leverage Points for Guideline-Concordant Statin Initiation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-27 15:25:20","doi":"10.21203/rs.3.rs-7632374/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-11-12T09:38:46+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-11T08:56:02+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-03T20:55:27+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"178372978545307321308613556406325986056","date":"2025-10-27T15:34:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"97101769014130669725616915823925379168","date":"2025-10-26T19:50:30+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"307652742064187786946124268872244047226","date":"2025-10-20T17:39:45+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-13T19:52:24+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-13T19:36:42+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-10-12T15:46:09+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-10-08T21:01:50+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Informatics and Decision Making","date":"2025-10-08T20:58:57+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"66a86e7c-5fec-4ef7-801c-7b8464209df8","owner":[],"postedDate":"October 27th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-04-07T16:02:27+00:00","versionOfRecord":{"articleIdentity":"rs-7632374","link":"https://doi.org/10.1186/s12911-026-03392-6","journal":{"identity":"bmc-medical-informatics-and-decision-making","isVorOnly":false,"title":"BMC Medical Informatics and Decision Making"},"publishedOn":"2026-04-02 15:58:58","publishedOnDateReadable":"April 2nd, 2026"},"versionCreatedAt":"2025-10-27 15:25:20","video":"","vorDoi":"10.1186/s12911-026-03392-6","vorDoiUrl":"https://doi.org/10.1186/s12911-026-03392-6","workflowStages":[]},"version":"v1","identity":"rs-7632374","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7632374","identity":"rs-7632374","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}