The influence of physicians on medication adherence

Background:\n\nMedication adherence is defined as the extent to which patients take medications according to agreed recommendations from a health care provider. Correspondingly, medication non-adherence is the failure to take medications according to a prescribed medication regimen. Considered one of the greatest challenges to the successful management of people with chronic conditions in the community setting, patients who are non-adherent to medications have higher risks for hospitalization and even death compared to patients who take medications as prescribed. According to a report in 2008, the costs due to non-adherence in the United States (US) was estimated to be between US$100 billion to US$310 billion per year; however, these numbers are based on general assumptions and rigorous estimates are not available in the US or Canada. Despite years of research, major gaps remain in our understanding of the causes of non-adherence. Studies often focus on patient characteristics and patient behaviour. Although some of these factors are influential, they typically only explain a small fraction of the variance in models predicting non-adherence. Prescribing physicians have been identified as having a strong influence on their patient’s adherence to medications; however, their impact has never been comprehensively incorporated into population-based models to help explain the residual variance. \nPurpose and research approach:\nThe purpose of this research was to examine the impact of physicians on population-based models of medication adherence. Three retrospective cohort studies were conducted using population-based, administrative databases from Saskatchewan, Canada. The study population consisted of new statin users (no statin claims in the previous five year) between 2012 and 2017. Statin medication was the focus in these studies because they are prescribed for chronic treatment only, they had no therapeutic equivalent during the period of study, they are prescribed to a large percentage of the population, and they are associated with reduced morbidity and mortality from atherosclerotic cardiovascular disease. Each study focused on different aspects of the physician’s potential impact on the outcome of optimal medication adherence to statins defined as proportion of days covered (PDC) of at least 80%. Study 1 measured the impact of continuity of care (COC) provided by physician prescribers on optimal adherence; study 2 focused on the impact of demographic characteristics of physicians on optimal adherence; and study 3 measured the overall effect of physicians on the outcome of optimal adherence. \nStudy 1 – The impact of physician continuity of care on medication adherence\nThe first study investigated continuity of care (COC), a factor related to physician practice that is associated with medication adherence and is commonly used as a baseline explanatory variable in population-based studies. COC is typically represented by the usual provider continuity index (UPCI), which is calculated exclusively from the number of outpatient physician visits. However, the number of outpatient visits only represents one aspect of COC. Our aim was to improve the measurement of COC by integrating information on physician services and pharmacy claims (i.e., medication dispensing) data. Our new “integrated COC” definition required patients to have one physician who satisfied all three criteria: a) the most frequently visited general practitioner physician (i.e., usual care provider); b) the statin prescriber; and c) provider of a complete medical examination within the past year. Logistic regression models were constructed with each measure of COC (high UPCI index or integrated COC) on the outcome of optimal statin adherence (PDC ≥80%). Predictive performance of the two models was compared using the DeLong test. In a cohort of 55,144 new statin users, the integrated COC measure had a stronger association with optimal adherence [adjusted odds ratio (aOR) =1.56, 95% confidence interval (CI) 1.50 to 1.63] than UPCI (aOR = 1.23, 95% CI 1.19 to 1.28), and produced greater prediction accuracy of the multivariable model (DeLong test, p<0.0001). The results suggest that physician service and pharmacy claim data should be adopted in COC measures for population-based adherence models because of greater predictive performance in models predicting optimal adherence to statin.\nStudy 2 – Physician demographic factors and medication adherence\nThe second study examined the impact of age or sex concordance (i.e., same age range or same sex) between physicians and patients on optimal adherence to statin medications. We hypothesized that age or sex concordance between physicians and patients would result in higher medication adherence through improved communication and trust compared to non-concordant pairs. Multivariable logistic regression models by generalized estimating equations were applied to examine odds of optimal adherence associated with age and/or sex concordance. Among 51,874 pairs of new statin users and 1,562 prescribers, no influence of age concordance on the odds of optimal adherence could be detected (aOR = 1.02, 95%CI 0.97 to 1.07). The association between sex concordance and optimal statin adherence was stronger but failed to reach statistical significance by a very small margin (aOR=1.05, 95%CI 1.00 to 1.11). It suggested that the potential for an important influence of sex concordance remains and should be investigated in other health care settings. \nStudy 3 – The overall impact of physicians on medication non-adherence\nThe third study aimed to quantify the overall impact of physicians on optimal statin adherence. We identified the prescriber for each new statin user and measured each patient’s adherence one-year after the initial dispensation. The overall physician impact on optimal medication adherence (i.e., PDC >= 80%) was estimated from the intraclass correlation coefficient (ICC) derived from a random intercept model controlled by numerous patient-level variables (e.g., sex, residence, income, etc.). We also examined the impact of unmeasured physician factors or latent effects based on the ICC of a random intercept model controlled by both patient variables and physician-level factors (e.g., country of medical training, remuneration type, statin patient count, etc.). Finally, we estimated the impact of specific physician-level factors [sex, country of medical training, years in practice, remuneration type, number of patients, and number of patients taking a statin (statin patient count)]. Unadjusted odds ratios (uOR) for each factor were generated from logistic regression models; adjusted odds ratios (aORs) were obtained from non-linear mixed-effects logistic regression models adjusted by patient-level variables. Our results were derived from 51,874 new statin users. Addition of the physician effect to a model consisting of multiple patient-level factors only explained an additional 6.4% of the observed variance in adherence between patients, of which physician-level factors had a minimal contribution. The vast majority of the overall physician impact (5.2% out of a possible 6.4%) was attributed to a “latent effect” of the prescriber. The results suggest that the overall impact of prescribers on optimal statin adherence appears to be very limited. \nFuture research\nResearch on the influence of physicians should continue with different types of medications and conditions. Also, specific factors such as COC, type of physician remuneration, sex concordance, and country of medical education require further study to help understand the complex role of physicians and potential new targets for improving medication adherence.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.