Background Prescribing safety indicators for use in UK general practice have been developed through expert consensus but have not been validated in a large UK-wide dataset. We aimed to quantify indicators describing hazardous prescribing patterns that could put patients at risk of harm (prescribing indicators) or repeat prescriptions without the appropriate monitoring tests (monitoring indicators) using the Clinical Practice Research Database (CPRD). We further investigated the relationship between patient and practice characteristics and the prevalence of the indicators and their usefulness for potentially distinguishing between practices.

Methods The prevalence of individual and composite indicators addressing prescribing of anticoagulants, anti-platelets, NSAIDs, β blockers, glitazones, metformin, digoxin, antipsychotics, combined hormonal contraceptives (CHC) and oestrogens to patients with contra-indicating conditions or prescribing of angiotensin-converting-enzyme inhibitors and loop diuretics, amiodarone, methotrexate, lithium and warfarin without appropriate monitoring was measured. Multilevel logistic regression models with random effects at the practice level were used to quantify the impact of patient and practice level variables on the indicator prevalence and the variation between practices. The variation between practices and the number of patients at risk were used to provide estimates of reliability for individual practices.

Results 57,479 of 1,255 637 patients at risk triggered at least one prescribing indicator (4.58%; 95% CI: 4.54% to 4.61%) and 21,501 of 182,721 (11.8%; 11.2% to 11.9%) triggered at least one monitoring indicator. For individual prescribing indicators prevalence rates ranged from almost zero to 10.2%, and for monitoring indicators from 10.4% to 41.9%. Older male patients and those prescribed multiple repeat medications had significantly higher risk of triggering a prescribing indicator whereas younger female patients with fewer repeat prescriptions had significantly higher risk of triggering a monitoring indicator. A low prevalence of prescribing indicators did not predict a low prevalence of monitoring indicators or vice versa.

Conclusion These findings highlight the hazards of prescribing multiple medications. Variation in the prevalence of prescribing and monitoring indicators between practices was high, even after adjusting for patient and practice level variables, presenting important targets to improve patient safety in primary care. These indicators can be used by primary care practitioners in their own practice and establish a baseline for investigating changes in prescribing safety over time.