In a European cohort of older patients with multimorbidity and polypharmacy, the prevalence of potentially clinically significant DDIs was high (54%) at hospital admission. Patients with certain medical conditions, such as depression, coronary artery disease, heart failure, atrial fibrillation and COPD, and those with polypharmacy were more likely to have at least one DDI at baseline. The top five most frequent potentially clinically significant DDIs (drugs that reduce potassium [diuretics, inhaled beta2 agonists, systemic corticosteroids], centrally-acting drugs [psychotropics, antidepressants, opioids, antiepileptics], potassium-sparing drugs [angiotensin-converting enzyme, angiotensin II type 1 receptor blockers, spironolactone] and antithrombotics] comprised 80% of all potentially clinically significant DDIs. These five potentially clinically significant DDIs were all classified as pharmacodynamic interactions mainly causing cardiovascular adverse events. The prevalence of potentially clinically significant DDIs increased significantly at discharge and then remained stable over the subsequent 12 months (i.e., did not return to the baseline prevalence level).
Comparisons with previous studies
The high prevalence of DDIs in our cohort aligns with the in-patient and out-patient prevalences reported in the literature for comparable populations (geriatric outpatient cohort: 44.5%  and 58.3% ; geriatric inpatient cohort: 60.5% ). The drug classes most frequently involved in DDIs in our cohort were also the same as those previously reported [1, 4, 16, 17, 26,27,28,29]. These drugs are often prescribed to older patients to treat common age-related conditions, particularly cardiovascular and neurological conditions. Consistent with our results, Vonbach et al. reported that more than 70% of all major and moderately severe potential DDIs were pharmacodynamic interactions .
In line with our results, two studies reported that hospitalization was associated with an increase in DDIs in older patients, [26, 30] one of which reported that almost half of the DDIs at hospital discharge were incurred during hospitalization . However, and this is the originality of our study, none of the published studies assessed DDIs over time to describe trends after discharge in older patients. Changes in the prevalence of DDIs over time, with a higher prevalence at discharge than at baseline, and a decrease after hospital discharge but not to baseline levels, were noted particularly for DDI 65 (concomitant prescription of ≥2 drugs that reduce potassium), which includes cardiovascular and respiratory drugs (high ceiling diuretics and inhaled beta2 agonists). This trend has already been described in patients with coronary heart disease and COPD,  conditions for which these drugs are indicated. In addition, factors associated with an increase in DDIs at 2 months were similar to those previously reported in the literature, especially polypharmacy [2,3,4, 7, 32]. The fact that the patients included in Bern had more comorbidities, were more likely to have polypharmacy, and were generally hospitalised for non-elective reasons, may explain the differences in trends across trial sites.
DDIs are often classified according to their potentially harmful effects, but no study has reported the prevalence of DDIs according to whether or not they may be appropriate. Evaluation of appropriateness of a DDI often requires an individual, case-by-case assessment, which is not possible on a large database. However, from the list of 66 DDIs, some can be considered inappropriate in many cases (“avoid concurrent use”), such as DDI 32, “beta-blocker and verapamil or diltiazem” (potentially serious cardiovascular adverse effects in particular in patients predisposed to heart failure); DDI 36, “concomitant use of ≥3 centrally-acting drugs” (increased risk of falls, fracture, impaired cognition); and DDI 63, “tamoxifen and paroxetine or fluoxetine or bupropion” (risk of sudden death – ventricular arrhythmias, torsade de pointes). Most drugs in these DDI categories were rarely prescribed in our cohort with the exception of DDI 36. Other DDIs may be appropriate, or even intentional, in some cases if preventive measures are taken (e.g., dose adaptation, close monitoring, patient education), for example, for DDI 12, “oral anticoagulant and antiplatelet drug”; DDI 65, “concomitant prescription of ≥ 2 drugs that reduce potassium”; and DDI 66: “selective serotonin reuptake inhibitor and loop or thiazide diuretic”. DDI 65 prescribed in acute clinical situations, such as cardiac failure or COPD exacerbation, may, for example, be appropriate if serum potassium concentrations are monitored closely. Similarly, DDI 12 is only appropriate in certain specific clinical situations (e.g., atrial fibrillation and recent acute coronary syndrome) and for a fixed duration . The fact that 57% of patients in the experimental arm (i.e., patients for whom medication review was performed by a physician and pharmacist) had at least one DDI at hospital discharge suggests that most of these DDIs were considered appropriate in these patients and thus did not require drug discontinuation or modification after the medication review. Only the patients in the control arm had an increase in the prevalence of DDI 36 from baseline to discharge, which may explain why there was a significant increase in the prevalence of DDIs during hospitalization only in patients in the control arm. These data suggest that frequent medication reviews for those admitted to hospital, not only during hospital admission, but also after discharge, can contribute to reducing the risk of inappropriate DDIs.
Strength and limitations
To our knowledge, this is the first international, multicenter study to report data on the prevalence of DDIs at hospital admission and at different time points over a one-year follow-up period in an older population with multimorbidity. To identify the DDIs, we used for the first time a new list of 66 potentially clinically significant DDIs developed by an international consensus panel .
This study has some limitations. First, we were unable to distinguish between appropriate and inappropriate DDIs (which would require detailed information on, for example, dose adaptations, indications, monitoring, patient specifics). Second, because of a lack of statistical power, we did not assess ADEs and drug-related hospital admissions associated with DDIs. This information would have enabled us to evaluate actual DDIs resulting from potential DDIs. As reported in the literature, it is therefore likely that we have overestimated the true clinical significance of the DDIs . Indeed, studies that focused on actual DDIs reported lower prevalences in ambulatory settings (9.5 and 25.5%) [24, 34] as well as for in-patients (8.8%) . Nevertheless, the presence of a potentially clinically significant DDI remains a strong signal that should alert the prescriber to the need for greater patient monitoring and education.
It seems important to help clinicians identify potentially harmful DDIs more effectively, in particular the five most prevalent, to improve their assessment of risk-benefit ratios in individual patients. Risk minimization measures for prescription of psychotropic drugs, antithrombotics and drugs that reduce or increase potassium concentrations should be deployed.
The next logical step would be to assess the clinical impact of these DDIs, especially of the most common ones. To this end, a large scale cohort study from healthcare databases could be carried out. Better identification of the impact of DDIs is important for clinicians, researchers and health policy decision-makers to plan for safer healthcare in ageing societies.