Our objective was to evaluate the effect of an ED-based pharmacist-led medication review intervention on health services utilization using an interrupted time-series design. Our results indicate that the intervention did not modify long-term trends of total physician, family physician or ED visits, even when we stratified our results by age, discharge status, or by hospital site.
The lack of observed differences in GP visits between the intervention and control groups is consistent with prior literature in the field, with three prior hospital-based studies indicating no effect, and one indicating a modest increase in GP and urgent care visits following the intervention [
25‐
28]. While Okere et al. observed an increase in GP visits among patients who received ED-based medication review, their study enrolled a much younger patient population than others [
25‐
28]. The patients included in the present study were older, with a median age of 70 years. As older patients are more likely to return to residential care or be admitted to hospital after an ED visits, they may be less likely to follow up on the result of a medication review with a GP in the community [
29,
30]. As a result, outpatient health services utilization outcomes may be less discerning in this older patient population. In addition, while ADEs are common in older adults with high levels of morbidity, their subsequent health services utilization may be driven by factors other than more rapid resolution of medication-related problems or preventative interventions, such as their frailty, lack of social supports and loss of independence [
31].
This may lead to smaller discernible differences between groups in a controlled study attempting to measure the effect of medication review in older ADE patients on health services utilization.
We observed no differences in level or trend of ED revisits per 1000 patients following medication review. In light of the primary study, which indicated a clinically important reduction in the number of hospital days in the intervention group compared to control, our finding of no difference in repeat ED visits is reassuring, and indicates that patients in the intervention group were not inappropriately discharged earlier [
14].
Strengths and limitations
Our study was a non-randomized, non-blinded controlled clinical trial. In order to mitigate the inherent risk of bias in non-randomized designs we used an interrupted time series approach, allowing us to confirm that there were no pre-intervention differences in the level or trend of any of the health outcomes we measured. This indicates that the systematic allocation algorithm was successful in creating comparable groups of patient. We enrolled a large sample size of patients under real-world circumstances and completed our evaluation using exclusively administrative health data. This provided us the opportunity to evaluate a medication review as implemented outside of the highly controlled environment of randomized trials. In addition, we compared medication review to medication reconciliation with as needed pharmacist consultation, improving the external validity of our findings relative to other studies, which used no intervention as a control. It is unlikely in acute care settings today, that a control group of patients would not receive any medication management interventions, as medication reconciliation is considered standard care in acute care [
12]. Therefore, our results are likely generalizable to other provinces in Canada, and other countries with similar healthcare systems.
Future studies evaluating clinical pharmacy interventions when randomized trials are not ethically feasible or too costly, may consider using a similar design, particularly when health outcome trends over time are relevant. Unlike previous studies, the measurement of trends using this approach controlled for outliers, which may not have been representative of a true change in the outcome [
32].
This study is not without limitations. Several studies have previously measured the appropriateness of pharmacist recommendations, and have found that pharmacists’ recommendations are generally appropriate and clinically relevant [
33,
34]. We did not implement the use of any implicit or explicit tools to identify inappropriate medications (e.g. MAI or STOPP/ START checklist), as the focus of our ED-based intervention was the identification of ADEs, and many ADEs occur in appropriately prescribed and administered medications (e.g., hypoglycemia in a patient with diabetes who was on insulin). As few tools have been developed and validated to standardize ED-based medication review in patients presenting with ADEs, we were unable to standardize the intervention, aside from emphasizing the clinical focus on identifying and treating ADEs bringing patients into hospital. Research has shown that explicit tools may not be effective at identifying all types of ADEs in ED patients, and could restrict the medication review and limit its potential impact [
35‐
38]. Including both implicit and explicit review criteria may capture the benefits of highly standardized approaches while ensuring that clinicians are empowered to make clinical decisions and provide individualized care [
39].
Recently, an international core outcome set released seven recommended outcomes to measure medication review [
40]. These outcomes suggested that researchers examine adverse events using drug-related hospitalizations. While drug-related ED visits and hospitalizations would have been ideal outcomes for our study drug-related causality is consistently underreported in administrative data, and therefore not useful [
41].
This study was unable to assess whether recommendations made by clinical pharmacists in the ED were successfully communicated to and adopted by GPs. If recommendations were not adopted, our finding of no difference across all measures of subsequent outpatient health services utilization may indicate ineffective communication across health sectors. Communication and acceptance of pharmacist recommendations have been identified as a threat to the success of medication review interventions [
25,
39,
42]. A recent study by Santolaya-Perrín et al. trialed four different communication techniques on the uptake of physician acceptance of pharmacists’ recommendations [
39]. The technique most similar to the approach we used in our study led to an acceptance of only 27% of recommendations. In contrast, a site that used electronic clinical record systems observed an acceptance rate of 52%. Recently, a separate survey found that 96.7% of GPs stated electronic communication of medication recommendations as their preferred method of receiving prescribing information [
43].
The effectiveness of technological solutions for communicating between care providers may continue to improve with growing use and could provide a more cost-effective pathway for information exchange than face-to-face interactions [
44]. Future research is needed to develop and evaluate communication strategies to ensure the uptake of pharmacist recommendations, while evaluating their impact on health outcomes.