Does computerised decision support reduce hazardous prescribing in general practice?

Problem

Prescribing errors are an important and expensive preventable cause of safety incidents, illness, hospitalisations and deaths. In England, they are estimated to contribute to 1,708 deaths annually at a cost of over £98 million, and a third of clinically significant errors occur in primary care.A number of studies (including systematic reviews) have indicated that well-designed computerised decision support can reduce prescribing errors, but there is a lack of evidence in primary care.As part of a NIHR Programme Grant for Applied Research we evaluated the rollout of a new computerised decision support system (OptimiseRx) to general practices in England to determine if this reduced hazardous prescribing.

Approach

We used an incomplete (not formally randomised) stepped-wedge design with control groups. We used data from all general practices that contributed data to ResearchOne database between 1st January 2011 and 31st December 2019, irrespective of whether they have ever implemented OptimiseRx. Practices that implemented OptimiseRx between 2014 and 2018 were counted as intervention practices, all other practices were counted as non-intervention practices. The inclusion of the latter enabled secular trends to be estimated and adjusted for in the analyses. We derived a composite outcome 48 prescribing safety indicators . Outcomes were analysed using binomial mixed regression analysis with a random intercept for each practice and adjusting for both secular trends over time and seasonal effects. Assessment points were 6 months, 12 months, and 24 months.

Findings

Of the 409 practices participating in the study, 227 (56%) were intervention practices. At baseline 560,035 patients were at risk of being exposed to hazardous prescribing, and of these, 18,552 (3.31%) were actually exposed. Exposure to hazardous prescribing decreased over time, and this trend was accelerated after deployment of OptimiseRx. Deployment of OptimizeRx was associated with an additional reduction of hazardous prescribing of 10% two years post-intervention (odds ratio [OR] 0.90, 95% confidence interval [CI] 0.88 to 0.92); 7% at one year post-intervention (OR 0.93 (0.91 to 0.94)) and 2% at six-months post-intervention (OR 0.98 (0.96 to 0.99)).

Consequences

The introduction of OptimiseRx was associated with a significant reduction in hazardous prescribing and these reductions were more pronounced the longer software had been in use. This suggests a ‘learning’ effect where repeated exposure of safety alerts effects clinician prescribing behaviour over time. Given the harm associated with hazardous prescribing this study supports continued use of clinical decision support in general practice.