Association between potentially inappropriate prescribing and adverse patient outcomes using codified STOPP-START and Beers criteria in large, routinely collected population health datasets: a retrospective cohort study

Background and objective — To assess the association between potentially inappropriate prescribing and adverse patient outcomes (mortality, hospitalization, and emergency department visits) in a large, population-based cohort of older adults, using previously codified STOPPSTART and Beers criteria.

Methods — We conducted a population-based, retrospective cohort study using linked health administrative data from Ontario, Canada. The cohort included all Ontario residents aged 65 years or older who received at least one prescription between April 1, 2003, and March 31, 2017 (N = 2,937,927). Potentially inappropriate prescribing was identified using subsets of the 2014 STOPP-START and 2015 Beers criteria applicable to health administrative data. Associations between potentially inappropriate prescribing and outcomes were examined using multivariable logistic regression.

Results — Of the 2,937,927 patients, 2,410,626 (82.1%) had the outcome of death, hospitalization or ED visit. Among those with an outcome, 2,002,651 (83.1%) and 1,412,278 (58.6%) had at least one potentially inappropriate prescription as identified by the STOPP-START and Beers criteria, respectively. After multivariable adjustment, the presence of any potentially inappropriate prescription identified by STOPP-START criteria was significantly associated with increased odds of mortality (adjusted odds ratio: 3.68, 95% confidence interval: 3.65-3.72), hospitalization (aOR: 4.86, 95% CI: 4.83-4.90), ED visits (aOR: 4.12, 95% CI: 4.09-4.14), and of the composite outcome (i.e. any one of emergency department visit, hospitalization and/or mortality) (aOR: 4.59, 95% CI: 4.55-4.62). For the Beers criteria, any potentially inappropriate prescription was similarly associated with increased odds of mortality (aOR: 1.74, 95% CI: 1.73-1.75), hospitalization (aOR: 2.93, 95% CI: 2.91-2.94), ED visits (aOR: 4.29, 95% CI: 4.26-4.32), and the composite outcome (aOR: 4.44, 95% CI: 4.40-4.47). A strong dose-response relationship was observed with the adjusted odds of all adverse outcomes increasing progressively with each additional PIP compared to one potentially inappropriate prescription. For example, five or more STOPP-START potentially inappropriate prescriptions were associated with a nearly 15-fold increase in the odds of death.

Conclusion — In this large, population-based study, potentially inappropriate prescribing was strongly and independently associated with an increased risk of emergency department visits, hospitalization, and mortality in older adults. The risk escalated significantly with an increasing number of potentially inappropriate prescriptions. These findings highlight the value of using codified medication appropriateness criteria to identify potentially inappropriate prescribing as a modifiable risk factor for adverse outcomes at a population level.