Reducing Injuries From Medication-Related Falls Using Computerized Alerts for High Risk Patients

Background: Fall-related injuries account for significant morbidity and mortality, particularly in the elderly where multiple comorbidities and age-related changes in bone density increase the risk of fall-related fractures Indeed use of psychotropic medications in elderly persons is associated with a 2 to 29 fold increase in the risk of falls and a 2 to 5 fold increase in the risk of hip fracture. At particular risk are individuals over the age of 70, those with a prior history of falls, cognitive impairment, stroke, Parkinson's disease, or other conditions that would impair balance or gait. In our particular study population, 67.5% of persons with a psychotropic drug prescribing problem had at least one additional risk factor for fall-related injuries. This was particularly true for women who not only were more likely to have a psychotropic drug prescribing alerts than men but were also more likely to have other risk factors. 70.3% of women who had a psychotropic prescribing alert had other risk factors in comparison to 62.1% of men, particularly as it related to older age and a history of a fall-related fracture or soft-tissue injury in the past 12 months. A recent in-hospital study showed that providing physicians with patient-specific recommendations for changes in high risk psychotropic therapy through a computerized order-entry system reduced the prescription of non-recommended drugs and doses by 10%, which in turn was associated with a significant two-fold reduction in the in-hospital fall rate{5007}. If even a 5% reduction (annual prevalence 16.1% to 11.1%) could be achieved in primary care through targeted recommendations for high risk patients with psychotropic drug prescribing alerts, we would expect that it could conservatively reduce the number of falls among Canadian elderly (assuming the lowest risk of RR=1.66) from 116,064 to 82,212 and the number of fall-related injuries from 11,606 to 8,221. Based on the average costs of treating fall-related injuries of $20,000/injury{5006}, a reduction in adverse events of this magnitude would be associated with an annual savings of $67,708,000 in direct care costs. The research question is the following: Can medication-related fall injuries be reduced by using computerized electronic prescribing and drug management systems to identify high risk patients and provide physicians with patient-specific recommendations for modifying psychotropic medication use to reduce this risk?

Objective: To determine the extent to which a targeted psychotropic drug alert and recommendation system will reduce

a) the rate of potentially inappropriate psychotropic medication for patients at risk of fall-related injuries, and b) fall-related injury risk, fall-related injuries and hospitalizations.

Research Plan : A single blind, cluster randomized controlled trial will be conducted to test the hypothesized benefits of the targeted psychotropic drug alert and recommendation system versus the standard automated generic drug alert system within a fixed cohort of primary care physicians and an open cohort of patients seen by study physicians in the 16 month follow-up period for the assessment of reductions in potentially inappropriate psychotropic prescriptions and fall-related injuries. A single blind trial was planned because intervention status cannot be blinded for physicians in the study. However, study participants are blinded to the outcomes assessed, because the data required to assess these outcomes can be predominantly collected and assessed using data sources that are independent of the intervention status. Patients, clustered within physicians, is the unit of analysis because patient level information provides the most precise, non-ecological, method of the study outcomes as well as potential confounders, and because hierarchical multivariate analytic methods are now available to model clustering in the assessment of treatment effect{Chuang, 2000 4339 /id}. The benefit of the intervention will be assessed by comparing patients of physicians who received the psychotropic drug alert and recommendation system and patients of physicians who received automated drug decision support. This approach minimizes Hawthorne effects, arising from the intensive nature of practice intervention required to support computer-based systems in primary care that would likely result in over-estimation of benefit if computer-based decision support for drug management were compared to physicians with no computerized intervention. Further, it provides a means by which information on prescriptions, drug and disease profile can be assessed in an equivalent way between patients of physicians with automated control or targeted alert experimental decision-support, reducing biases related to differences in measurement sources.