Effect of Passive Choice and Active Choice Interventions in the Electronic Health Record to Cardiologists on Statin Prescribing: A Cluster Randomized Clinical Trial

Key Points

Question  Do passive choice or active choice decision support interventions in the electronic health record directed to cardiologists improve guideline-directed statin prescribing?

Findings  In this cluster randomized clinical trial of 82 cardiologists from 16 practices including 11 693 patients, neither intervention resulted in a significant change in statin prescribing for the overall sample. Among the subset of patients with an atherosclerotic cardiovascular disease clinical diagnosis, the active choice intervention led to a significant increase in statin prescribing at the optimal dose, but the passive choice intervention did not.

Meaning  Nudges to cardiologists through the electronic health record may be more effective for increasing statin prescribing if they clearly describe their indication, are designed to require action, and are targeted to high-risk patients with existing atherosclerotic cardiovascular disease; these approaches may need to be combined with other interventions to increase their effect.

Importance  Statin therapy is underused for many patients who could benefit.

Objective  To evaluate the effect of passive choice and active choice interventions in the electronic health record (EHR) to promote guideline-directed statin therapy.

Design, Setting, and Participants  Three-arm randomized clinical trial with a 6-month preintervention period and 6-month intervention. Randomization conducted at the cardiologist level at 16 cardiology practices in Pennsylvania and New Jersey. The study included 82 cardiologists and 11 693 patients. Data were analyzed between May 8, 2019, and January 9, 2020.

Interventions  In passive choice, cardiologists had to manually access an alert embedded in the EHR to select options to initiate or increase statin therapy. In active choice, an interruptive EHR alert prompted the cardiologist to accept or decline guideline-directed statin therapy. Cardiologists in the control group were informed of the trial but received no other interventions.

Main Outcomes and Measures  Primary outcome was statin therapy at optimal dose based on clinical guidelines. Secondary outcome was statin therapy at any dose.

Results  The sample comprised 11 693 patients with a mean (SD) age of 63.8 (9.1) years; 58% were male (n = 6749 of 11 693), 66% were White (n = 7683 of 11 693), and 24% were Black (n = 2824 of 11 693). The mean (SD) 10-year atherosclerotic cardiovascular disease (ASCVD) risk score was 15.4 (10.0); 68% had an ASVCD clinical diagnosis. Baseline statin prescribing rates at the optimal dose were 40.3% in the control arm, 39.1% in the passive choice arm, and 41.2% in the active choice arm. In adjusted analyses, the change in statin prescribing rates at optimal dose over time was not significantly different from control for passive choice (adjusted difference in percentage points, 0.2; 95% CI, −2.9 to 2.8; P = .86) or active choice (adjusted difference in percentage points, 2.4; 95% CI, −0.6 to 5.0; P = .08). In adjusted analyses of the subset of patients with clinical ASCVD, the active choice intervention resulted in a significant increase in statin prescribing at optimal dose relative to control (adjusted difference in percentage points, 3.8; 95% CI, 1.0-6.4; P = .008). No other subset analyses were significant. There were no significant changes in statin prescribing at any dose for either intervention.

Conclusions and Relevance  The passive choice and active choice interventions did not change statin prescribing. In the subgroup of patients with clinical ASCVD, the active choice intervention led to a small increase in statin prescribing at the optimal dose, which could inform the design or targeting of future interventions.

Trial Registration  ClinicalTrials.gov Identifier: NCT03271931

Atherosclerotic cardiovascular disease (ASCVD) remains the leading cause of adult morbidity and mortality in the United States and globally.¹ Statins have been demonstrated to reduce the risk of major adverse cardiovascular events.^2,3 However, about half of patients meeting guideline criteria for a statin have not been prescribed one.^4-10 Among patients prescribed a statin, about two-thirds are receiving a lower than optimal dose.⁶

Appropriate statin therapy relies on the clinician to recognize that a patient could benefit from it, prescribe the medication, and then use guideline-recommended dosage levels. Patients must understand, accept, and adhere to statin medications. In a multicentered registry across the United States, nearly 60% of patients who were eligible for a statin but not taking one reported that it was never offered to them, while only 10% stated they declined it.⁵ Intervening to influence clinician prescribing of statins for patients at elevated risk for ASCVD represents a significant opportunity to reduce cardiovascular risk.

Health systems are increasingly implementing changes to the design of electronic health records (EHRs) to influence medical decision-making.^11,12 One example is the use of a Best Practice Advisory (BPA) to remind clinicians to address gaps in patient care. These advisories can be designed to be interruptive, thereby forcing the clinician to respond before continuing (active choice).^13-18 This leverages insights from behavioral economics in that the intervention may cause clinicians to more heavily weight the long-term effect of their present actions and prompt them to avoid future regret by making decisions that adhere to guideline recommendations.^11,19-21 While this results in an immediate decision, it can also lead to alert fatigue when clinicians receive too many alerts.^22-24 These decision support interventions can also be passively embedded into the EHR in a specific location without a forced choice (passive choice). While this approach leads to less alert fatigue, it also risks being ignored because manual effort from the clinician is needed to access and address it. To our knowledge, these 2 approaches have not been well tested head-to-head to evaluate their effect on clinician behavior.

The objective of this cluster randomized clinical trial was to evaluate the effect of passive choice and active choice interventions in the EHR relative to usual care to promote guideline-directed statin prescribing by cardiologists. We hypothesized that active choice would significantly increase statin prescribing but that passive choice would not. This pragmatic trial was conducted in a large, academic health system and included 16 cardiology practices in Pennsylvania and New Jersey.

This was a 3-arm cluster randomized clinical trial conducted among cardiologists from 16 practices at the University of Pennsylvania Health System (Penn Medicine). The trial included a 6-month preintervention period (March 24, 2018, to September 23, 2018) and a 6-month intervention period (September 24, 2018, to March 23, 2019). Analyses were conducted between May 8, 2019, and January 9, 2020. The trial compared guideline-directed statin prescribing among cardiologists randomized to usual care and 2 interventions in the electronic health record: either a passive choice using a noninterruptive alert or an active choice using an interrupted alert that required action before the clinician could proceed further in using the EHR. The trial protocol was approved by the University of Pennsylvania institutional review board and can be found in Supplement 1. Informed consent by both physicians and patients was waived because this was a pragmatic evaluation of a health system initiative that posed minimal risk to participants. Neither physicians nor patients were compensated for their participation. This trial followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline (Supplement 1).

Eligible cardiologists at Penn Medicine included those who were attending physicians and provided care for patients with general cardiovascular disease. Cardiologists were excluded if they had fewer than 50 patient encounters within the previous year.

Eligible patients had a visit with a Penn Medicine cardiologist during the study period and were candidates for statin prescribing by either the 2013 American College of Cardiology/American Heart Association (AHA/ACC) Guideline on the Treatment of Blood Cholesterol to Reduce Atherosclerotic Cardiovascular Risk in Adults or the 2015 National Lipid Association (NLA) Recommendations for Patient-Centered Management of Dyslipidemia.^25,26 These guidelines included patients who were (1) age 21 to 75 years with any form of ASCVD and with most recent low-density lipoprotein cholesterol (LDL-C) level greater than 70 mg/dL (to convert to millimoles per liter, multiply by 0.0259); (2) age 21 to 75 years and most recent LDL-C level at least 190 mg/dL; (3) age 40 to 75 years with a history of any type of diabetes and most recent LDL-C level ranging from 70 to 189 mg/dL; and (4) age 40 to 75 years with a pooled cohort risk equation 10-year ASCVD risk score of at least 7.5%. Patients were excluded if they had a documented allergy to statin or statin intolerance/adverse reaction, severe kidney insufficiency defined as glomerular filtration rate less than 30 mL/min or receiving dialysis, history of rhabdomyolysis from any etiology, history of hepatitis, were currently pregnant, or were already receiving a proprotein convertase subtilisin/kexin type 9 inhibitor for lipid therapy.

Similar to prior work,^4,14 Clarity, an Epic reporting database, was used to obtain data on patients, cardiologists, and clinic visits. Data on patients included demographic characteristics, insurance, comorbidities, and baseline measures such as ASCVD 10-year pooled cohort equation risk score, clinical ASCVD diagnoses, laboratory results, and body mass index. Data on cardiologists included sex, practice location, statin prescriptions, and patient encounters. Patient’s household income level was obtained using US Census data on median household income based on zip code. Health insurance claim data and pharmacy medication fill data were not available.

Cardiologists at Penn Medicine (and their respective patients) were electronically randomized to a study arm using block sizes of 3, stratifying on quartile of baseline statin prescribing rate at optimal dose and number of patients eligible for a statin. All investigators, statisticians, and data analysts were blinded to arm assignments until the trial and all analyses were completed.

Prior to initiation of the trial, cardiologists in all arms were sent an email describing the goals of the study. Cardiologists in the usual care arm received no other interventions. Cardiologists randomized to 1 of 2 intervention arms were prompted during patient visits through the EHR that the patient was not taking appropriate statin therapy with options to prescribe statins that met guideline recommendations. After the trial completed, cardiologists in the 2 intervention arms were asked to complete an online survey describing their experiences with the interventions.

Cardiologists in the active choice arm were forced to accept or dismiss options for statin guideline-concordant statin therapy. This was implemented through an interruptive BPA that alerted the cardiologist while they were entering orders in the EHR during a patient encounter (eFigures 1 and 2 in Supplement 2). The BPA was triggered based on the patient’s problem list, medical history elements, and encounter diagnoses entered before and during the visit, laboratory values, and ASCVD risk score in the EHR at the time of order entry. The BPA begun with the following text: “This patient is not achieving ACC/AHA or National Lipid Association guidelines for cholesterol management. A [moderate or high] intensity statin is recommended for this patient.” The appropriate dose of atorvastatin was preselected. Additional options included rosuvastatin and the following laboratory tests: creatinine kinase (ordered at time of visit), comprehensive metabolic panel (ordered at time of visit), thyroid-stimulating hormone (ordered at time of visit), and lipid screen (ordered 3 months from the time of the visit). If the cardiologist decided to dismiss the BPA, they had the option to acknowledge one of the following reasons: patient refused, not clinically indicated, or alternative treatment. If the cardiologist accepted the BPA, he or she needed to electronically sign the statin prescription order.

Cardiologists in the passive choice arm had the same BPA appear embedded within the context of their EHR’s visit navigator while documenting within a patient’s encounter (eFigures 3 and 4 in Supplement 2). It was noninterruptive, and cardiologists had to manually access the visit navigator section to see the BPA. Therefore, cardiologists were not forced to view or act on it.

The primary outcome measure was the change in percentage of patients prescribed statin therapy at a dose that meets evidence-based guidelines (optimal dose). The secondary outcome measure was the change in percentage of eligible patients prescribed any dose of a statin (any dose). Patients were considered at goal if they met either the 2013 AHA/ACC or 2015 NLA guidelines.

A priori power calculations used data from Penn Medicine health system’s EHR clinical data warehouse (Clarity) to estimate baseline statin prescribing rates. Assuming that 40% of patients were receiving statin therapy at the optimal dose, we estimated that a sample size of 78 cardiologists (26 per arm) would provide at least 80% power to detect a 7–percentage point difference between each intervention arm and the control arm, using a conservative Bonferroni adjustment of the type I error rate, with a 2-sided α of .025 as our threshold for statistical significance.

After randomization but before the intervention began, 4 cardiologists stopped practicing at the study site (1 in control, 1 in the passive choice arm, and 2 in the active choice arm). All other randomly assigned cardiologists and their patients were included in the modified intention-to-treat analysis. We used the patient as the unit of analysis and evaluated the first visit with a cardiologist in each of the 2 study periods.

Similar to prior work,^14,27 we used the SAS procedure PROC GENMOD (SAS Institute Inc) to fit models for the 2 outcome measures (statin prescribed at optimal dose or any dose) based on generalized estimating equations with a logit link and an independence correlation structure using cardiologist as the clustering unit.²⁸ The model used patient-level observations and included baseline statin prescribing rate at the cardiologist level, month fixed effects, treatment arm, period (preintervention or intervention), and an interaction term for treatment arm and period. Models were fit to compare the active choice arm with the control arm and to compare the passive choice arm with the control arm. To obtain the adjusted difference and 95% confidence intervals in the percentage of patients prescribed a statin between arms, we used the bootstrap method, resampling patients 500 times.^29,30 Resampling of patients was conducted by the cardiologist variable to maintain clustering at the cardiologist level. Subgroup analyses were conducted to evaluate the effect of the interventions on patients with different statin indications. The same models were fit for patients with a clinical ASCVD condition, diabetes, LDL-C level of at least 190 mg/dL, and ASCVD 10-year risk score of at least 7.5%. Two-sided hypothesis tests used a significance level of P = .03. All analyses were conducted using SAS, version 9.4 (SAS Institute Inc).

The sample comprised 82 cardiologists from 16 practices (Figure). Cardiologists were 76.8% male (n = 63 of 82); 56.1% were from general cardiology (n = 46 of 82), 22.0% were heart failure specialists (n = 18 of 82), and 22.0% were interventional cardiologists (n = 18 of 82). Baseline statin prescribing rates were similar among the 3 arms (Table 1). During the intervention, there were 11 693 patients with a mean (SD) age of 63.8 (9.1) years, 57.7% were male (n = 6749 of 11 693), 66% were White (n = 7683 of 11 693), and 24% were Black (n = 2824 of 11 693). The mean (SD) 10-year ASCVD risk score was 15.5 (8.8) for patients without an ASCVD clinical diagnosis; 68.2% had an ASDCVD clinical diagnosis (n = 7980 of 11 693), 27.1 had diabetes (n = 3172 of 11 693), and 6.1% had an LDL-C level of at least 190 mg/dL (n = 712 of 11 693) (Table 2). Patient characteristics during the preintervention period are available in the eTable in Supplement 2.

Among all patients during the intervention, the percentage of patients prescribed a statin at an optimal dose was 42.6% in the control arm, 40.6% in passive choice, and 44.5% in active choice. In adjusted analyses of statin prescribing at optimal dose, the change in statin prescribing rates over time was not significantly different from control for the passive choice arm (adjusted difference in percentage points, 0.2; 95% CI, −2.9 to 2.8; P = .86) or the active choice arm (adjusted difference in percentage points, 2.4; 95% CI, −0.6 to 5.0; P = .08). In adjusted subset analyses among patients with an ASCVD clinical diagnosis, the active choice arm achieved a significant increase in statin prescribing at optimal dose relative to the control arm (adjusted difference in percentage points, 3.8; 95% CI, 1.0-6.4; P = .008). No other subset analyses were significant (Table 3).

Among all patients during the intervention, the percentage of patients prescribed a statin at any dose was 71.6% in the control arm (n = 2335 of 3263), 70.9% in the passive choice arm (n = 3086 of 4383), and 73.1% in the active choice arm (n = 2979 of 4077). In adjusted analyses of statin prescribing at any dose, the change in statin prescribing over time was not significantly different relative to control for the passive choice arm (adjusted difference in percentage points, −0.5; 95% CI, −2.6 to 1.6; P = .60) or the active choice arm (adjusted difference in percentage points, −0.05; 95% CI, −2.6 to 2.2; P = .96). There were no significant changes in any subset analyses (Table 4).

The postintervention survey was completed by 42.3% of cardiologists (n = 11 of 26) who received the passive choice intervention and 69.2% of cardiologists (n = 18 of 26) who completed the active choice intervention (overall response, 55.8% [n = 29 of 52]). Only 27.3% of cardiologists (n = 3 of 11) in the passive choice arm responded that they remembered seeing the alert in the EHR, compared with 100% of cardiologists (n = 18 of 18) who responded from the active choice arm. Among respondents in the passive choice arm, 18.2% (n = 2 of 11) stated the intervention helped them prescribe a statin to patients not already receiving one, and 0 of 8 respondents responded that it helped to increase the dose of patients at suboptimal dosage. Among respondents in the active choice arm,23.5% (n = 4 of 17) stated the intervention helped them prescribe a statin to patients not already receiving one, and 11.8% (n = 2 of 17) responded that it helped to increase the dose of patients at suboptimal dosage.

Qualitative feedback from open-ended responses among cardiologists in the active choice arm indicated 2 main opportunities to improve the intervention. First, several cardiologists stated it would have been helpful to know the indications that triggered the alert. One cardiologist stated, “My perception is that I am already prescribing statins appropriately for my patients. When I see this alert, the first question that enters my mind is how this message was generated, ie, what were the determinants that went into this recommendation? It would be more helpful if the reasons for the recommendation were more transparent–hard to just ‘buy into’ something that contradicts my preceding judgment.” Another cardiologist said, “Often unable to figure out why it fired.” Second, some cardiologists were confused when the alert was triggered for patients receiving a suboptimal dose because they thought it was indicating the patient was not taking a statin at all, rather than that the dose could be increased. For example, 1 cardiologist said, “the alert nearly always showed up on patient who were already on a statin.” Another cardiologist said, “Already on a statin.”

In this pragmatic, randomized clinical trial, we found that passive choice and active choice interventions embedded within the EHR did not change overall statin prescribing among cardiologists at 16 practice sites. In subgroup analysis of patients with an ASCVD clinical diagnosis, there was a small but significant increase (3.8 percentage points) in prescribing the optimal statin dosage. This indicates that there may have been a benefit for this group of patients with clinical ASCVD but warrants evaluation in a future study.

Our findings reveal several important insights for health systems considering these types of interventions. First, the passive choice intervention had no effect on statin prescribing for either outcome measure or any of the subgroup analyses. This alert was embedded within the EHR and required the cardiologist to manually access it in a section not always within routine workflow. While this design prevents fatigue from pop-up alerts that require clinician action, it also appears to have resulted in clinicians not always noticing the alert. Among cardiologists in the passive choice arm who completed the post-trial survey, 8 of 11 (72.7%) stated that they did not remember seeing the alert during the 6-month study period. Given their reduced exposure, passive alerts may be more appropriate for lower-priority actions. Future interventions that use passive alerts might consider using peer comparison feedback on performance or another type of approach to increase the salience of the intervention and thereby motivate clinicians to access these alerts as part of their usual workflow.^14,31,32

Second, the active choice intervention had a small effect only for patients with an ASCVD clinical diagnosis who were already taking statins but at suboptimal doses. While active choice has been used previously in primary care settings,^13-16,18 there is less study of its use among cardiologists. A previous randomized clinical trial¹⁷ tested the use of a similar EHR intervention to improve referral for implantable cardiac defibrillator and found a similar magnitude effect with referral rates at 35.9% for patients who visited physicians that received the alert and 30.3% for patients who visited physicians who were in the control group. Qualitative feedback revealed 2 opportunities to improve the design of these alerts that could lead to better adoption in future interventions. First, clinicians wanted to know why the alerts were being triggered. Given evolving practice guidelines and that there are several indications for statin therapy, greater transparency on the patient’s indications for statin therapy could have improved clinicians’ trust in the intervention. Second, despite an initial email communication that described the goals of the trial and design of the intervention, many clinicians reported that they were unsure why alerts were being shown for patients already taking a statin. Future interventions could increase general communications before and after the intervention is implemented, as well as within the alert, so that cardiologists are aware that some alerts are focused on increasing the dose of a statin that the patient is already taking.

Third, statin therapy requires shared decision-making between the clinician and the patient. In a randomized clinical trial testing the use of financial incentives to improve LDL levels, interventions using incentives to both clinicians and patients were more effective than interventions offering incentives to either clinicians or patients alone.³³ Future studies could consider ways to test nudges to both clinicians and patients, as well as using interventions to improve the use of shared decision-making tools.

Fourth, the design of these types of approaches involves several components, each of which are likely critical to intervention success. It may have been that the implementation, software, user interface, or user experience were not sufficient to change prescribing behavior. Future studies should use other approaches and compare them with the types of interventions evaluated in this trial.

Our study has limitations. First, this trial was conducted within a single health system, which limits generalizability. However, cardiologists were from 16 practices across 2 states. Second, we evaluated data on statin prescriptions but did not have data on prescription fill rates or adherence. Third, interventions and analysis relied on EHR data and did not account for patient information outside of our health system. Fourth, AHA/ACC released new statin prescribing guidelines in March 2019. However, these were released in the final week of our 6-month study and were unlikely to effect the overall results. Fifth, randomization was conducted at the level of the cardiologist, not the practice, and therefore, there could have been some contamination of intervention among cardiologists who were in the same practice as other cardiologists who were in different arms. Sixth, our study was 6 months in duration, which, to our knowledge, is the longest trial of its kind among cardiologists but limits our ability to make conclusions about longer-term outcomes.

In a cluster randomized clinical trial of 2 decision support interventions within an EHR, the passive choice and active choice interventions did not change statin prescribing among cardiologists. In the subgroup of patients with clinical ASCVD, the active choice intervention led to a small but significant increase in statin prescribing at the optimal dose. Further study is needed to evaluate the active choice intervention among patients with clinical ASCVD, and future interventions could focus on ways to improve the design of active choice and combine it with other approaches to further improve statin prescribing.

Corresponding Author: Mitesh S. Patel, MD, MBA, Penn Medicine, 3400 Civic Center Blvd, 14-176 S Pavilion, Philadelphia, PA 19104 (mpatel@pennmedicine.upenn.edu).

Accepted for Publication: June 10, 2020.

Published Online: October 7, 2020. doi:10.1001/jamacardio.2020.4730

Author Contributions: Drs Adusumalli and Patel had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Adusumalli, Westover, Jacoby, VanZandbergen, Chen, Cavella, Pepe, Volpp, Asch, Patel.

Acquisition, analysis, or interpretation of data: Adusumalli, Westover, Small, Chen, Pepe, Rareshide, Snider, Patel.

Drafting of the manuscript: Adusumalli, Patel.

Critical revision of the manuscript for important intellectual content: Adusumalli, Westover, Jacoby, Small, VanZandbergen, Chen, Cavella, Pepe, Rareshide, Snider, Volpp, Asch.

Statistical analysis: Small, Cavella, Rareshide.

Obtained funding: Patel.

Administrative, technical, or material support: Adusumalli, Westover, Jacoby, VanZandbergen, Chen, Cavella, Pepe, Volpp, Patel.

Supervision: Adusumalli, Jacoby, Volpp, Patel.

Conflict of Interest Disclosures: Dr Adusumalli reported being a member of the Epic Cardiology Specialty Steering Board. Dr Volpp reported personal fees and other from VAL Health; grants from Humana, Hawaii Medical Services Association, Vitality/Discovery, CVS Caremark, WW, and Oscar; and personal fees from Center for Corporate Innovation, Lehigh Valley Medical Center, Vizient, Greater Philadelphia Business Coalition on Health, American Gastroenterological Association Tech Conference, Bridge to Population Health meeting, and Irish Medtech Summit outside the submitted work. Dr Asch reported personal fees and other support from VAL Health; personal fees from GSK, Meeting Designs, Capital Consulting, Healthcare Financial Management Association, and National Alliance of Health Care Purchaser Coalitions; and personal fees and nonfinancial support from Cosmetic Boot Camp, Alliance for Continuing Education in the Health Professions, Deloitte, and NACCME outside the submitted work. Dr Patel reported personal fees and other from Catalyst Health, HealthMine Services, and Holistic Industries and other support from Life.io. No other disclosures were reported.

Funding/Support: This study was supported by the University of Pennsylvania Health System through the Penn Medicine Nudge Unit.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 3.

Additional Contributions: We thank Harvey Waxman, MD, Andrew Litwack, MD, Robert Norris, MD, Janet Zhang, MD, Susan Weissman, MBA, Karen Logan, BS, Lynn Cardona, MSN, RN, Natasha Rush, MHA, and Tom Falkowski, BS, from the Penn Cardiology Access and Operations group at the University of Pennsylvania for their contributions to the design and implementation of the interventions. No compensation was received for these contributions.