- Despite guideline recommendations, both statin use overall and high intensity statin (HIS) use specifically remains low in patients with established atherosclerotic cardiovascular disease (ASCVD). Both therapeutic inertia (TI) and concerns for statin associated side effects (SASE) have been identified as contributors to under-prescribing.
- In this cluster randomized controlled trial performed in the Department of Veterans Affairs (VA), clinicians at the intervention sites received individualized electronic reminders created using natural language processing (NLP) and structured chart data that included information on the patient’s statin use, SASE, ASCVD history, and statin fill history. Those in the control sites, usual care, had access to a patient dashboard displaying compliance with statin therapy..
- Centrally processed individualized statin specific reminders, when compared to usual care, lead to a modest yet statistically significant increase in HIS use and statin adherence in VA patients with ASCVD.
The rates of non-guideline concordant statin use among patients with established ASCVD remain high, and may be driven in part by both TI and concerns related to SASE. It is unknown whether personalized electronic reminders can help increase appropriate statin prescriptions among high risk VA patients.
On March 5, 2023 Dr. Salim Virani from Aga Khan University in Karachi, Pakistan presented the results of A Randomized Trial Of A Personalized Clinical Decision Support Intervention To Improve Statin Prescribing In Patients With Atherosclerotic Cardiovascular Disease (PCDS Statin) during the Late Breaking Clinical Trials Section of ACC.23/WCC, with simultaneous publication in Circulation.1
The creation of the intervention in PCDS Statin was described in previous studies: natural language processing and machine learning algorithms were used to identify details of SASE from patient charts and qualitative interviews were done to determine patients and provider perspectives on SASEs and what data clinicians find helpful when making statin prescription decisions.2–4 This information was used to create a centrally processed, patient specific, electronic reminder meant to promote statin prescription.
In this cluster randomized control trial, randomization occurred at the clinic level. Primary care providers of patients in an intervention clinic would receive the electronic reminder 2-7 days prior to the patients visit (synchronous) and at a different timepoint outside of the visit (asynchronous). Reminders included date and type of ASCD diagnosis, current statin and dose, date of last fill, date and type of SASE, and guidelines resources on HIS use and SASE management. The control arm received usual care, which included a patient dashboard that provides information on patient compliance with statin therapy. Clinicians in the intervention sites could opt out of electronic reminders, and reminders were not sent if a clinician did not acknowledge the receipt of three previous alerts. The primary outcome was pre-post change in HIS use. The secondary outcome was pre-post change in overall statin use. An exploratory outcome was pre-post statin adherence using the metric of proportion of days covered.
Overall, 14 clinics (18427 patients) were randomized to the intervention and 13 clinics to the control (18214 patients). The mean age was 71.1 years and the most common ASCVD phenotype was ischemic heart disease; 41.6% of patients had an SASE identified, 52.4% had diabetes, 96.5% were male, 67.9% were White, and 28.6% were Black. Baseline HIS/overall statin use was 52.6%/81.0% at intervention sites and 55.6%/82.2% at usual care sites. Nearly 5000 reminders were sent on 4532 unique patients, with 73% being synchronous reminders and 27% being asynchronous. A total of 37 clinicians (31.6%) in the intervention arm opted out of the reminders.
This study found a modest but statistically significant positive effect for the primary outcome of HIS use. At the end of the study HIS use among intervention patients was 55.2%, an increase of 1.6%, compared to 53.7% for usual care patients, a decrease in 2.2%. The odds ratio for HIS use with the intervention was 1.06 (95% CI 1.02-1.11). Within the intervention group, this effect was driven by those clinicians who received a reminder, as not all eligible patients had reminders sent to their clinicians because of the algorithms to limit alert fatigue: the absolute change was +10.1% in HIS for those who received a reminder versus a -0.18% change among those who did not receive a reminder. Therefore, there was a “number needed to remind” of 10 for each HIS prescription. For the secondary outcome, overall statin use actually decreased in both groups, but it decreased less in intervention group (-2.4%) compared to the control group (-5.2%), with an odds ratio for any statin use with the intervention of 1.12 (95% CI 1.06-1.18). Gains in overall statin adherence was higher in the intervention group (+6.4%) compared to the usual care group (+3.6%), odds ratio 1.38 (95% CI 1.32-1.45).
In conclusion, sending centrally processed, electronic reminders tailored with patient-specific information to primary care providers of VA patients with established ASCVD led to an increase is HIS use, an increase in statin adherence, and lower decrease in overall statin use when compared to usual care. Because of algorithms to avoid alarm fatigue and clinician drop out, only 53% of eligible patients actually had a reminder sent, and the effect size in this “per-protocol” group was much higher as compared to the “intention to treat” group. This trial was unique in its use of using both structured data as well as natural language processing to tailor the reminders to the specific patient with the goal of increasing HIS uptake. These results “inform how informatic driven interventions can improve evidence-based care delivery in large healthcare systems,” according to Dr. Virani. “It will be important to see what the long term outcomes of these patients are as we follow them for another year.”
- Virani SS, Ramsey DJ, Westerman D, et al. Cluster Randomized Trial of a Personalized Clinical Decision Support Intervention to Improve Statin Prescribing in Patients With Atherosclerotic Cardiovascular Disease (PCDS Statin). Circ [Published Ahead Print] 2023;
- Virani SS, Akeroyd JM, Ahmed ST, et al. The use of structured data elements to identify ASCVD patients with statin-associated side effects: Insights from the Department of Veterans Affairs. J Clin Lipidol [Internet] 2019;13(5):797-803.e1. Available from: https://linkinghub.elsevier.com/retrieve/pii/S1933287419302661
- Ahmed ST, Akeroyd JM, Mahtta D, et al. Shared decisions: A qualitative study on clinician and patient perspectives on statin therapy and statin-associated side effects. J Am Heart Assoc 2020;9(22).
- Gobbel GT, Matheny ME, Reeves RR, et al. Leveraging structured and unstructured electronic health record data to detect reasons for suboptimal statin therapy use in patients with atherosclerotic cardiovascular disease. Am J Prev Cardiol [Internet] 2022;9(August 2021):100300. Available from: https://doi.org/10.1016/j.ajpc.2021.100300