Many countries are now super-aged societies where more than 20% of the population is older than 65 years [1]. The demographic changes occurring in the past decades are modifying the prevalence of multiple chronic diseases and the related prescription of multiple medications [1–3], putting a strain on social and healthcare policy planning. Multimorbidity and polypharmacy are the most common conditions managed in geriatric clinical practice. Additionally, older individuals with multimorbidity and related polypharmacy prescriptions, who are at increased risk for adverse events, are the most common users of healthcare and generate high costs [2, 4–6].

Medication consumption among older adults is rising, with rates ranging from 42.5% to 77.0%, estimated to be three times higher than their proportion in the population [6, 7]. The majority of clinical practice guidelines focus on the management of single disease states and do not adequately consider multimorbidity, resulting in long-term treatment with multiple medications [6, 8, 9]. Therefore, prescriptions based on clinical protocols aimed at patient benefit may result in not only favorable but also adverse health outcomes due to unanticipated polypharmacy therapy [10, 11]. Older adults taking multiple medications have been reported to have worse health status compared with those taking fewer medications, and appear to be a vulnerable population [12].

Polypharmacy therapy may sometimes be necessary to manage multiple chronic conditions, but it also poses several risks to the health of older adults and has been found to be associated with various adverse outcomes [13]. In particular, polypharmacy increases the likelihood of inappropriate prescribing and reduces adherence to complex regimens, thereby increasing the risk of morbidity, cognitive and functional impairment, falls and fractures, hospitalizations, and mortality [13–15].

Currently, there is no universally accepted definition of polypharmacy. The most commonly referenced threshold is 5 medications, while higher levels of polypharmacy are often defined by the use of 10 or more medications [16]. Available evidence regarding the efficacy of polypharmacy therapy in the growing elderly population is scarce and rarely derived from real-life conditions, such as those outside hospital settings, but rather from specific clinical contexts [2, 7, 17, 18].

In the framework of the Moli-sani study [19, 20], a large cohort of Italian adults, we prospectively evaluated the impact of polypharmacy therapy on health (hospitalization, length of hospital stay, mortality) in a general population of community-dwelling elderly. In particular, polypharmacy was also considered as a time-varying variable in further survival analyses and we examined whether potentially inappropriate prescriptions (PIP) could potentially mediate the association between polytherapy and poor health.

The study included 5,631 older adults (50.9% men), with a mean baseline age of 72.4 ± 5.4 years. Of those, 1,059 (18.8%) were not taking long-term medications, 1,875 (33.3%) were taking 1–4 medications a day, 1,496 (26.6%) were taking 5–9 medications (polypharmacy), and 1,201 (21.3%) were taking 10 or more medications (heightened polypharmacy).

In total, 29% of the older adults were on chronic polytherapy [C-PT; median number of drugs 10 (IQR: 8–13)], 18.9% were on non-chronic polytherapy [NC-PT; 7 (6–9)], while 52.1% were classified as not on polytherapy [No PT; 1 (0–2)].

Supplementary Tables S2, S3 show the distribution at baseline of the main characteristics, the most common chronic degenerative diseases in older adults according to not being on polypharmacy or being on polypharmacy therapy. Individuals on polypharmacy (both NC-PT and C-PT) were older than those not on polypharmacy (P-value < 0.0001), had lower levels of education and social status in childhood, and were more frequently retired (P-value = 0.0001, 0.0019 and 0.0013, respectively; Supplementary Table S2). Moreover, they had lower levels of physical activity, ate fewer calories, and drank less alcohol, being more frequently abstainers or ex-drinkers. However, they did not differ in their adherence to the Mediterranean diet. Women were more represented in the NC-PT group in comparison to the other groups (P-value <0.0001). There were no significant sociodemographic differences in terms of marital status, place of residence and income between individuals not on polypharmacy therapy and those on C-PT. In general, individuals on polypharmacy therapy showed a higher prevalence of multimorbidity, chronic-degenerative diseases and obesity, but a lower prevalence of liver disease, Parkinson’s disease and Alzheimer’s disease (Supplementary Table S3).

During a median follow-up of 12.6 years (IQR 10.6–13.8 years; 64,716 person-years), a total of 2,001 deaths (41.1% for CVD and 28.7% for cancer) were ascertained; additionally, 4,342 hospital admissions (56.6% for CVD, 14.2% for IHD and 16.2% for cerebrovascular disease) occurred.

Figure 1 shows the adjusted survival curves (a: for all-cause mortality; b: for all-cause hospitalization) according to polypharmacy. Both figures show that individuals on polypharmacy (both NC-PT and C-PT) had a lower probability of survival or being free from hospitalization for any cause, during follow-up.

Tables 1, 2 report the hazard for all-cause, CVD and cancer mortality according to polypharmacy. Additionally, the same analyses were reported for all-cause and CVD, IHD and cerebrovascular hospitalizations.

Considering the individuals not on polypharmacy as a reference and after adjusting for possible confounders (model 2, Table 1), the individuals in NC-PT and C-PT showed a higher hazard of mortality [21% (95% CI 7%–37%) and 30% (16%–46%), respectively, P-value < 0.0001) and all-cause hospitalization [39% (28%–51%) and 61% (49%–75%), respectively, P-value < 0.0001)]. Similar trends were found for all secondary outcomes (CVD and cancer mortality, and CVD, IHD, and cerebrovascular hospitalizations), except for cancer mortality in C-PT elderly and cerebrovascular hospitalization in NC-PT elderly (Table 2).

Additionally, a propensity score approach was used to control for residual confounding by comorbidities. The unadjusted and propensity score-adjusted differences in NC-PT vs. No PT and C-PT vs. No PT for each variable included in the propensity score are shown in Supplementary Figure S1; the c-statistic of the propensity score models was 0.68 and 0.83, respectively. All the pre-treatment differences disappeared after adjustment by propensity score weighting. Using the propensity score method (Table 3), the association between the studied health outcomes and polypharmacy did not differ from the previous results, mainly for the primary outcomes and for C-PT elderly overall (Tables 1, 2).

Supplementary Table S4 shows the distribution of the total number of hospitalizations and the total number of hospital days accrued during follow-up for all-cause, CVD, IHD and cerebrovascular hospitalizations. The 15,161 total multiple hospital admissions had a median duration of 20 days (IQR: 9–41). Compared to individuals not on polytherapy, those on NC-PT and C-PT accumulated significantly more total hospitalizations (22%, 95% CI: 17%–28% and 50%, 44%–57%, respectively; Supplementary Table S5) and total hospital days (14%, 12%–15% and 15%, 13%–16%, respectively). Similar trends were observed when considering the total number of hospitalizations for CVD and IHD and the total number of hospital days for cardiovascular hospitalizations. The observed inverse trend relative to the total number of hospital days for cerebrovascular hospitalizations should be further investigated (Supplementary Table S5).

The prevalence of PIP was 11.1%, 29.1% and 43.3% in the No PT, NC-PT and C-PT groups, respectively. Figure 2 and Supplementary Table S6 show that individuals with PIP had a higher risk of all-cause mortality and hospitalizations. Additionally, Table 4 shows that potentially inappropriate prescriptions are important mediators of the association between polypharmacy therapy and all-cause mortality, with PTE being 11.6% and 13.6% for NC-PT and C-PT, respectively. Similar results were found for cardiovascular mortality and hospitalizations (Table 4). The PIP mediation effects for all-cause hospitalization were slightly inferior (5.9% and 6.0% for NC-PT and C-PT, respectively).

Trends in polypharmacy between 2005 and 2020 among older Moli-sani participants using data from the regional drug register are shown in Supplementary Table S7. Considering the variation in this exposure during the follow-up period, we repeated survival analyses, which showed that the hazard increased for all outcomes studied (Supplementary Table S8).

Stratified analyses by sex, age groups (65–75 years, ≥75 years) and education (low, high) showed no difference in the association between polypharmacy and the primary outcomes (Supplementary Table S9). On the other hand, stratification for a history of CVD showed that the higher hazard was mostly evident in individuals without a previous event of CVD (P-value for interaction 0.011 for all-cause mortality). Supplementary Table S10 reports the same analyses with CVD mortality and hospitalizations as outcomes, showing that older adults aged 65–75 years at baseline had a higher hazard than those ≥75 years (P-value for interaction = 0.039 and 0.022, respectively). It should be noted that individuals without multimorbidity at baseline showed the same hazards for all outcomes compared to those with two or more comorbidities (Supplementary Tables S9, S10).

In a general elderly population in Southern Italy, we observed that polypharmacy therapy was associated with a higher hazard of all-cause and specific (mainly cardiovascular) mortality and hospitalization. The results remained consistent after adjusting for a number of variables, including lifestyle habits, socioeconomic status, and various comorbidities, and after employing several statistical approaches and sensitivity analyses to minimize potential confounding factors and biases. However, we remain aware of the inherent difficulty in establishing a clear causal relationship between the exposure of interest and the outcome in observational studies.

Additionally, when the association between the polypharmacy regimen and the total number of hospitalizations and the total hospital days accrued during follow-up was investigated, an increase in the all-cause hospitalization “burden” on the National Health Service was observed in those on polypharmacy. When the variation in therapy regimen during follow-up was considered, the hazard of polytherapy increased for all outcomes.

Numerous studies have indicated that the use of multiple medications is associated with a wide range of adverse clinical events [13–15, 29–31]. However, it is difficult to understand whether the increased risk is due to the poor health status that required the prescription of medications or to the prescription of multiple medications per se. Moreover, despite rigorous adjustment for comorbidities, confounding by disease may persist in observational studies. We used propensity score analysis as an additional approach to control for residual confounding by comorbidities, which showed similar results. Moreover, stratification for CVD history showed that the higher hazard was mostly evident in individuals without a previous CVD event, strongly suggesting that polypharmacy may represent a health hazard beyond the confounding effect of comorbidities.

Interestingly, older adults without multimorbidity showed similar harms from polypharmacy compared to those with multimorbidity. This highlights that the risks associated with the use of multiple medications are not limited to those with multiple health conditions.

Our results are in line with previous studies [29–32]. The English Longitudinal Study of Ageing (6,295 individuals, aged ≥50 years), observed that, over a 6-year follow-up period, both polypharmacy (5–9 medications) and heightened polypharmacy (10+) were associated with a higher risk of all-cause mortality and CVD mortality, whereas cancer mortality was only related to heightened polypharmacy [30]. Chang TI et al., analyzing a large cohort of Korean older community-indwelling individuals (more than 3 million individuals, aged ≥65 years), found a graded association between the number of medications and the risk of adverse clinical outcomes and in particular, polypharmacy was associated with a significantly higher risk of hospitalization and mortality [29].

Several mechanisms may explain the relationship between polypharmacy and poor health status or increased mortality. Older individuals are more susceptible to serious adverse drug events due to age-related physiological changes that heighten the body’s sensitivity to drug effects. The harmful effect of polypharmacy on health and survival could be explained by the cumulative effects of multiple medications on the renal or hepatic system, which trigger a cascade of interactions in elderly individuals already suffering from multiple comorbidities [14]. Systematic reviews have reported that reducing specific classes of medications may reduce adverse events and improve quality of life [33–35].

Finally, polypharmacy is potentially harmful because it increases the possibility of inappropriate prescriptions. Therefore, we specifically evaluated whether PIP could potentially mediate the association between polypharmacy and poor health. Our longitudinal analyses showed that PIP, which are quite common among older adults receiving multiple drug prescriptions, are important mediators of the association between polypharmacy and key health outcomes. These results are supported by the finding that among individuals on polypharmacy, those without severe multimorbidity had the same polypharmacy burden as those with multimorbidity. As the number of individuals taking multiple medications on a regular basis increases, strategies to address the challenges related to polypharmacy burden are needed [35, 36].

The results of our study do not call into question the efficacy of any individual medication (among those examined in the study). This is because our study design, particularly the reference group used, was not intended to assess the efficacy of a single medication. Such efficacy evaluations should be conducted for each medication in its appropriate setting, such as hypertensive patients with or without the specific antihypertensive drug being tested, etc. However, this scenario does not align with the scope of our study. Clinical guidelines typically rely on evidence derived from randomized clinical trials or meta-analyses. However, these sources frequently exhibit bias due to the exclusion or inadequate representation of older individuals, particularly those with multiple health conditions and undergoing multiple medication treatments [37, 38].

Improving drug precription for older adults is a global priority for all healthcare systems. The majority of older individuals are cared for by general practitioners who may have inadequate expertise in geriatrics and multimorbidity particularly in polypharmacy management. Our findings strongly support the need for public action to improve the culture of appropriate prescribing and specific research and clinical pharmacological knowledge for the elderly population.