Polypharmacy, patient or practice? A multilevel analysis of the association between patient activation and polypharmacy

Talk Code: 
Robert Williamson
Robert Williamson, Isaac Barker
Author institutions: 
The Health Foundation


Providing high quality care, including safe prescribing, for patients with multiple long-term conditions is a growing challenge in general practice. Interventions to improve prescribing and reduce unnecessary polypharmacy have targeted both general practitioners and patients. Despite the likely association between the success of such interventions and patient ability to self manage (which can be assessed by the patient activation measure), there has yet to be analysis of the association between the two. We analyse the association between patient activation and polypharmacy, controlling for relevant patient demographic and clinical characteristics, and practice characteristics and quantify the variance in prescribing quality that is attributable to a patient’s practice compared to patient level variation.


This was an observational, cross-sectional study of 9,597 patients registered with Islington CCG who had one or more long-term condition and returned a patient activation measure postal questionnaire. We analysed a linked pseudonymised dataset of electronic health records from 1st January 2013 until 30th April 2016 (40 months), as well as active clinical characteristics. Polypharmacy was defined as concurrent prescription of ten or more medications, taken from the electronic health record of the patient. The association between polypharmacy and a patient’s level of activation was analysed using multivariate logistic regression, controlling for relevant observable patient and practice characteristics and accounted clustering at practice level. To evaluate the proportion of variance explained by the practice clustering we calculated the intraclass correlation coefficient using the latent variable approach


In our cohort 15% of patients had 10 or more concurrent prescriptions and were therefore defined as having polypharmacy. Adjusting for demographic and clinical patient characteristics and practice characteristics higher activation (PAM levels 2, 3 and 4) were significantly associated with a lower odds of a patient having polypharmacy compared to lowest level of activation (PAM level 1) - odds ratios for PAM levels 2, 3 and 4 of 0.77 (95% CI 0.63 to 0.93), 0.79 (95% CI 0.67 to 0.93) and 0.72 (95% CI 0.57 to 0.91) respectively The median odds ratio, which quantifies the median change in likelihood of polypharmacy for a patient moving between two random practices, was 1.31 and is of a comparable magnitude to being in PAM level 4 compared to level 1 (odds ratio 0.72).


Highly activated patients are less likely to have polypharmacy, and the magnitude of the effect is similar to that of the between practice variation. Interventions targeting prescribing quality have previously focused on between practice variation, however this study encourages a renewed focus on the variation resulting from patient ability to self-manage and engage in shared decision making.

Submitted by: 
Sarah Deeny
Funding acknowledgement: