In this cross-sectional analysis, hyperpolypharmacy was independently associated with female sex, more comorbidities, higher BMI, nursing home residency, and increasing healthcare utilization, when compared with polypharmacy participants. However, patients with a higher education level, older age, and only GP contact (as compared with specialist/ED visits or any hospitalizations) had a lower risk of hyperpolypharmacy. Hyperpolypharmacy was not significantly associated with medication adherence in our study. Among adults with hyperpolypharmacy, high medication adherence (i.e., 8 points on MMAS-8©) was more likely in those aged ≥ 80 years, individuals receiving community nurse visits at home, and in those with better self-reported health-related quality of life.

4.1 Comparison with Prior Evidence4.1.1 Factors Associated with Hyperpolypharmacy

Our finding of an increasing risk of hyperpolypharmacy with increasing number of healthcare utilization is in keeping with results from a US cross-sectional study in a selected population of primary care patients with heart failure [39], as well as a nationwide South Korean cohort study in older individuals [40]. This is not surprising, since increasing healthcare utilization provides more opportunity for physicians to prescribe medications leading to hyperpolypharmacy. Given that our analyses were adjusted for the number of comorbidities and predicted mortality (as measured by the Charlson Comorbidity Index), differences in severity of illness are unlikely to fully explain the association of healthcare utilization and hyperpolypharmacy. Both the US and Korean studies showed that ambulatory healthcare visits carried higher risk of hyperpolypharmacy than hospitalization [39, 40]; whereas we observed the opposite in our study. One possible explanation could be different definitions used for “ambulatory/outpatient” visits. For example, we combined specialist and ED visits together into one category, and this was recorded separately to GP visits. In the other studies, although ED visits were determined, it is not clear whether GP visits were included in “ambulatory healthcare services.” In addition, in our unadjusted sensitivity analysis, (Supplementary Fig. 1) we noted that fewer hospitalizations and more GP-only contacts were associated with fewer medications. However, this could be explained by patients being less sick or having fewer comorbidities, therefore naturally requiring fewer medications and being less likely to be hospitalized.

Regarding age, we noted that age ≥ 80 years appeared to be associated with lower rates of hyperpolypharmacy. This could be because of deprescribing in the oldest old, in whom the risks of medication side effects increase (e.g., intensive glycemic control among older adults can be harmful [41]), and the benefits of preventive medication decrease (e.g., statin deprescribing in primary prevention for patients with limited life expectancy [42, 43]). Our findings concur with two other European multicenter studies conducted in home care-dependent patients with high levels of complexity [4] and nursing home patients with severe cognitive impairment [44], which found an inverse association between hyperpolypharmacy and age. Conversely, data from the English Longitudinal Study of Ageing showed that individuals aged ≥ 80 years carried a fivefold greater risk of hyperpolypharmacy [3]. A possible explanation for this is that in this study, data were collected in 2012, which was before deprescribing health initiatives and incentives such as medication review, patient and physician educational materials, cost savings analysis, and improved quality of life were promoted [45]. There may also have been less awareness at the time among physicians regarding potentially inappropriate medication (PIM) prescribing and medication overuse.

4.1.2 Association Between Hyperpolypharmacy and Low/Medium Medication Adherence

In a sensitivity analysis we observed a U-shaped pattern between number of medications and adherence, with a trend toward low/medium medication adherence in participants with either a low or high number of regular medications. Though this finding did not reach statistical significance, this U-shaped pattern has been described before when exploring the impact of number of medications on adherence to antihypertensive medications [46]. It is to be expected that medication adherence decreases as the number of medications increase, e.g., because of regimen complexity, the challenges of remembering too many medications, drug–drug interactions, and side effects promoting nonadherance. Surprisingly, adherence worsened with a low number of regular medications [47]. This phenomenon has been observed before, and can be explained by the health belief model [46, 48], whereby someone taking less medications believes they are not as sick and therefore will not put much effort into maintaining their health and taking medication prescribed by physicians.

4.1.3 Factors Associated with Low/Medium Medication Adherence in Patients with Hyperpolypharmacy

In a systematic review of older patients receiving polypharmacy, the prevalence of nonadherance ranged from 6 to 55% [16]. By comparison, in our population of patients with hyperpolypharmacy, 53% were not taking their medication as prescribed. Thus, in keeping with results from the polypharmacy systematic review [16], in our study more medication (i.e., hyperpolypharmacy) did not appear to lead to more nonadherance.

We observed that receipt of community nursing care, better self-reported quality of life, and age ≥ 80 years favored medication adherence. We suspect these factors are interrelated. The prevalence of cognitive impairment and frailty increases with advancing age, making the requirement of community nurse visits more likely in the oldest old. Community nursing care in the four European countries included in this study is common and involved nurses visiting patients in their own homes to provide personal and social care such as administering medication, measuring blood pressure and glucose, or assisting with personal hygiene (community nurse visits did not involve care of patients living in nursing homes or long-term residential care). It is unsurprising that having a healthcare professional prepare, check, or administer the medications contributes to better medication adherence. While previous data showed that visiting the same physician regularly was associated with better medication adherence [49], we did not find an association between GP/type of physician contacts and medication adherence in our study.

4.2 Clinical Implications

This study identifies factors associated with hyperpolypharmacy, which may help healthcare professionals to detect a vulnerable, at-risk population and intervene with early medication reviews and drug optimization. Medication optimization trials in polypharmacy have demonstrated a reduction in inappropriate prescribing [18], as well as reductions in hospitalizations and healthcare costs [50]. In particular, we have shown that visits to specialists, EDs, or hospitals are associated with greater hyperpolypharmacy prevalence compared with contacts with the GP alone. These results reinforce the central role of the primary care physician as the coordinator of patient healthcare in the older population with multimorbidity and support the introduction of screening criteria for high-risk groups, i.e., all patients discharged from hospital, as well as the implementation of regular medication reviews, or medication adherence risk assessments every 6 months performed in the GP practice or by community pharmacists.

The consequences of nonadherance are far-reaching and amount to billions of euros in preventable healthcare costs, many thousands of avoidable hospitalizations, and premature deaths in Europe [13], representing a missed opportunity for healthcare providers and healthcare systems to intervene. Greater knowledge on medication nonadherance and its risk factors in older people with multimorbidity is important to prevent worsening of disease, re-hospitalization, and increased healthcare costs. Importantly, in this study, we have identified community nurse visits and self-reported quality of life as significant factors associated with better medication adherence. These are modifiable factors and could be used to develop personalized medication adherence interventions. Currently, medication adherence-enhancing interventions are used infrequently, and even more rarely reimbursed [51]; representing further aspects for improvement.

4.3 Strengths and Limitations of This Study

To our knowledge, this is the first study to explore the association between medication adherence in an older population experiencing hyperpolypharmacy. While there is an abundance of data on polypharmacy in this population, the outcomes do not automatically translate over to hyperpolypharmacy given that pharmacological side effects and interactions increase with increasing number of medications. The multinational design of this study included patients from four large European centers and enrolled many older medical inpatients with multimorbidity, increasing generalizability of results. The granular, prospectively collected data improve the quality of our findings.

However, this study has some limitations. First, while the cross-sectional design allowed us to explore associations, we cannot assess causal relationships. Second, healthcare utilization (e.g., number of physician visits, nursing home residency, or receipt of community nurse visits), as well as medication adherence, was self-reported from a patient questionnaire, meaning that we cannot exclude recall bias or social-desirability bias. As a disadvantage, the MMAS-8© and other adherence scales provide information on medication adherence in general, i.e., for all medications prescribed, but do not consider that adherence may be medication specific (e.g., only for statins or diuretics but not for analgesics). Third, unmeasured confounders may have affected our results, as is the case for all observational studies. For example, there was no information on pill burden (e.g., number of pills taken per day versus multiple drugs combined into one daily pill), or presence of a caregiver versus living alone, which could all act as confounders for the association between hyperpolypharmacy and medication adherence. Fourth, the generalizability of our results to a multiracial population outside Europe is limited by the very low number of non-white participants enrolled in the study. Fifth, we were not able to specifically differentiate ambulatory specialist visits from ED visits owing to the way the data were collected. Finally, we did not account for clustering by study site, and we cannot exclude that country differences in the organization and provision of healthcare may have affected our results.