Organisational variation and reliability of national polypharmacy performance indicators

Talk Code: 
Rupert Payne
Gary Abel, Rachel Denholm, Deborah McCahon
Author institutions: 
University of Bristol, University of Exeter


The national ePACT2 polypharmacy prescribing indicators were introduced in 2017, and use patient-level primary care electronic dispensing data to identify patients taking multiple medicines or certain potentially high-risk combinations of medicines. The indicators can be used to facilitate the identification of patients in individual GP practices, but it is unclear whether they might have a role in comparing prescribing across practices.


We used practice-level ePACT2 data from all GP surgeries in England from June 2019. Five key indicators were studied, applied to patients prescribed at least one drug during the study month: mean number of drugs per patient, and percentage of patients per practice receiving ≥10 drugs, exposed to an anticholinergic burden score (ABS) ≥9, receiving ≥3 anticoagulant or antiplatelet drugs, and receiving ≥2 unique medications likely to cause kidney injury. Indicators were based on prescribing for all ages, and for patients ≥75 years. Basic descriptive statistics were calculated describing variation (mean and standard deviation) across all practices for each indicator and age group. We calculated inter-unit (Spearman-Brown) reliability to determine how reliably practices can be distinguished (classified) from each other for a particular indicator, using mixed-effects Poisson (mean drug number) or logistic (other indicators) regression.


Across practices (n=6974) and all age groups, the mean (±SD) number of drugs prescribed was 3.5±0.5, with 5.2±2.5% of patients receiving ≥10 medicines, 0.08±0.10% with ABS≥9, 0.09±0.24% taking multiple anticoagulants/antiplatelets, and 31±5.8% taking drugs likely to cause kidney injury. Polypharmacy was considerably higher in patients ≥75 years (mean number drugs 5.1±0.8, 10.7±5.8% on ≥10 medicines), but other indicators had similar prevalence. For prescribing for people of any age, reliability was excellent for indicators looking at mean number of drugs (0.98±0.08) and percentage patients taking ≥10 medicines (0.93±0.08). Reliability of the kidney injury prescribing indicator was lower although still acceptable (0.82±0.13). Reliability of indicators measuring high ABS or high risk of bleeding was poor (0.39±0.16, 0.17±0.06 respectively). For patients ≥75 years, reliability remained high for mean number of drugs (0.95±0.10), but was notably lower for percentage patients taking ≥10 drugs (0.85±0.12) and kidney injury prescribing (0.61±0.18).


Polypharmacy is common in primary care, and use of multiple drugs likely to cause kidney injury is widespread. However other measures of potentially inappropriate prescribing are relatively uncommon. Further research is required to determine to what extent prescribing variation reflects differences in practice demographics. Indicators may have utility for individual practices for identifying patients receiving potentially problematic medication regimens. Although the basic polypharmacy indicators may have utility in distinguishing practices with varying levels of multiple drug use, including potentially in high-stakes applications, the other indicators cannot be reliably used to assess differences in potentially inappropriate prescribing between practices and their use as quality indicators is inadvisable.

Submitted by: 
Rupert Payne
Funding acknowledgement: 
This study is funded by the National Institute for Health Research (NIHR) School for Primary Care Research (project reference 409). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care.