Benjamin D Horne1, Joseph B Muhlestein2, Donald L Lappé2, Heidi T May3, Viet T Le4, Tami L Bair3, Daniel Babcock3, Daniel Bride3, Kirk U Knowlton5, Jeffrey L Anderson2. 1. Intermountain Medical Center Heart Institute, Salt Lake City, Utah, USA; Division of Cardiovascular Medicine, Department of Medicine, Stanford University, Stanford, California, USA. Electronic address: benjamin.horne@imail.org. 2. Intermountain Medical Center Heart Institute, Salt Lake City, Utah, USA; Cardiology Division, Department of Internal Medicine, University of Utah, Salt Lake City, Utah, USA. 3. Intermountain Medical Center Heart Institute, Salt Lake City, Utah, USA. 4. Intermountain Medical Center Heart Institute, Salt Lake City, Utah, USA; Rocky Mountain University of Health Professions, Provo, Utah, USA. 5. Intermountain Medical Center Heart Institute, Salt Lake City, Utah, USA; Division of Cardiovascular Medicine, Department of Medicine, University of California San Diego, La Jolla, California, USA.
Abstract
BACKGROUND: Medication adherence is generally low and challenging to address because patient actions control healthcare delivery outside of medical environments. Behavioral nudging changes clinician behavior, but nudging patient decision-making requires further testing. This trial evaluated whether behavioral nudges can increase statin adherence, measured as the proportion of days covered (PDC). METHODS: In a 12-month parallel-group, unblinded, randomized controlled trial, adult patients in Intermountain Healthcare cardiology clinics were enrolled. Inclusion required an indication for statins and membership in SelectHealth insurance. Subjects were randomized 1:1 to control or nudges. Nudge content, timing, frequency, and delivery route were personalized by CareCentra using machine learning of subject motivations and abilities from psychographic assessment, demographics, social determinants, and the Intermountain Mortality Risk Score. PDC calculation used SelectHealth claims data. RESULTS: Among 182 subjects, age averaged 63.2±8.5 years, 25.8% were female, baseline LDL-C was 82.5±32.7 mg/dL, and 93.4% had coronary disease. Characteristics were balanced between nudge (n = 89) and control arms (n = 93). The statin PDC was greater at 12 months in the nudge group (PDC: 0.742±0.318) compared to controls (PDC: 0.639±0.358, P = 0.042). Adherent subjects (PDC ≥80%) were more concentrated in the nudge group (66.3% vs controls: 50.5%, P = 0.036) while a composite of death, myocardial infarction, stroke, and revascularization was non-significant (nudges: 6.7% vs control: 10.8%, P = 0.44). CONCLUSIONS: Persuasive behavioral nudges driven by artificial intelligence resulted in a clinically important increase in statin adherence in general cardiology patients. This precision patient decision support utilized computerized nudge design and delivery with minimal on-going human input.
BACKGROUND: Medication adherence is generally low and challenging to address because patient actions control healthcare delivery outside of medical environments. Behavioral nudging changes clinician behavior, but nudging patient decision-making requires further testing. This trial evaluated whether behavioral nudges can increase statin adherence, measured as the proportion of days covered (PDC). METHODS: In a 12-month parallel-group, unblinded, randomized controlled trial, adult patients in Intermountain Healthcare cardiology clinics were enrolled. Inclusion required an indication for statins and membership in SelectHealth insurance. Subjects were randomized 1:1 to control or nudges. Nudge content, timing, frequency, and delivery route were personalized by CareCentra using machine learning of subject motivations and abilities from psychographic assessment, demographics, social determinants, and the Intermountain Mortality Risk Score. PDC calculation used SelectHealth claims data. RESULTS: Among 182 subjects, age averaged 63.2±8.5 years, 25.8% were female, baseline LDL-C was 82.5±32.7 mg/dL, and 93.4% had coronary disease. Characteristics were balanced between nudge (n = 89) and control arms (n = 93). The statin PDC was greater at 12 months in the nudge group (PDC: 0.742±0.318) compared to controls (PDC: 0.639±0.358, P = 0.042). Adherent subjects (PDC ≥80%) were more concentrated in the nudge group (66.3% vs controls: 50.5%, P = 0.036) while a composite of death, myocardial infarction, stroke, and revascularization was non-significant (nudges: 6.7% vs control: 10.8%, P = 0.44). CONCLUSIONS: Persuasive behavioral nudges driven by artificial intelligence resulted in a clinically important increase in statin adherence in general cardiology patients. This precision patient decision support utilized computerized nudge design and delivery with minimal on-going human input.