Frailty is broadly defined a “clinical state characterized by a decrease of an individual's homeostatic reserves and is responsible for enhanced vulnerability to endogenous and/or exogenous stressors” (1). Multiple operational definitions of frailty are available in literature (x). The most frequently used definition (2) focuses on the evaluation of five domains (nutritional status, energy, physical activity, mobility, and strength) to identify the frail phenotype. Independently of the adopted definition, frailty has been repeatedly associated with several negative health outcomes such as fracture risk (3), hospitalization (4), disability (5), multimorbidity (6), and medication harm (7). A recent meta-analysis of studies using the Frailty Index, reported that frailty is a significant predictor of mortality (8). It has been also suggested that polypharmacy may be an explanatory factor of the association between frailty and mortality in older individuals (9). Polypharmacy is common in people with frailty (10). Both frailty and prefrailty significantly predict nursing home (NH) placement (11) and a meta-analysis reported a very high prevalence of both conditions (40.2 and 52.3%, respectively) in NH patients (12), especially among women (13). More than 90% of patients awaiting NH placement have at least one potentially inappropriate medication (PIM) and 65% are eligible for the application of STOPPFrail criteria (14) (Screening Tool of Older Persons Prescriptions in Frail adults with limited life expectancy), which is designed to help physicians in deprescribing medications in older patients with frailty and limited life expectancy. The interplay between polypharmacy, multimorbidity, and frailty is potentially complex (15). For example, number of drugs can be higher than number of comorbidities, they both can be determinants of frailty, and disease-drug and drug-drug interactions are frequent. Given the high prevalence of these conditions in NH patients, it is of interest to see how these common phenomena affect mortality in this setting of care.

The aims of the current paper are to (i) assess whether nursing home residents with polypharmacy (5–9 medications) or hyperpolypharmacy (≥10 drugs), have an increased risk of death and (ii) whether any association is modified according to the presence of frailty or disability and multimorbidity.

At baseline, from the initial sample of 4,156 participants, 133 (3.2%) were excluded due to missing data, leading to a final sample of 4,023 participants. The sociodemographic and clinical characteristics of participants have been described for the overall study population and according to the number of medications used (Table 1). The mean age in the overall study population was 83.6 years and almost two-thirds were female. Almost half (47%) of the population were classified as frail and 40.6% had multimorbidity. A quarter (n = 1,042, 25.9%) of participants were not on polypharmacy regimen, while a half (n = 2,002, 49.8%) were on polypharmacy and a quarter (n = 979, 24.3%) were on hyperpolypharmacy. Laxatives, antiulcer medications, psychotropic drugs (such as benzodiazepines, antidepressants, and antipsychotics), and diuretics were the most common classes prescribed in the study population.

Comorbidities such as Parkinson's disease, stroke, heart failure, and cancer were more common among patients on polypharmacy or hyperpolypharmacy compared to those who were not. Multimorbidity was more common in participant on polypharmacy (42.8%) or hyperpolypharmacy (52.9%) than people taking 0–4 drugs (27.9%) (p < 0.001). Dementia and poor cognitive status were inversely associated with being on polypharmacy or hyperpolypharmacy (p < 0.001). The proportion of residents with disability (ADL score 5–6) was comparable among polypharmacy groups (p = 0.147). On the contrary, frailty was less common in participants on polypharmacy (46.2%) or hyperpolypharmacy (44.1%) compared to the group of those taking <5 drugs (51.6%) (p = 0.002).

Overall, 761 (18.9%) participants died during follow-up and 262 (6.5%) moved out of the nursing home and were censored. Overall the incidence rate of death in the study population was 0.21 persons/year. Table 2 shows factors associated with death during 1-year follow-up. After adjusting for sociodemographics and potential confounders (facilities, cognitive status), frailty and disability were associated with the highest risk of death (frailty: HR = 1.85, 95%CI 1.49–2.28; disability: HR = 2.10, 95%CI 1.68–2.47), while both polypharmacy and hyperpolypharmacy and cognitive status showed no effect on mortality.

Table 3 shows the results of the Cox proportional hazards regression after stratification for frailty status (see also Figures 1, 2 for the Kaplan-Meyer survival curves for polypharmacy). Among non-frail participants, multimorbidity (adjusted HR 1.34, 95%CI = 1.01–1.82) and hyperpolypharmacy (adjusted HR 1.61, 95%CI = 1.09–2.40) were associated with a higher risk of death during the follow-up. Among frail participants, no other factors were significantly associated with death. Conversely, both polypharmacy and multimorbidity were not associated with a different risk of death after stratification for disability.

In this longitudinal study of older nursing home residents, we found that hyperpolypharmacy and multimorbidity both increase the risk of death, but only in people without frailty, even after adjustment for demographics, facilities, comorbidities, and cognitive status. Person without frailty who had multimorbidity had 35% increased risk of death compared to those without multimorbidity. There was a 29% increased mortality risk in people with hyperpolypharmacy compared to those without polypharmacy, within the non-frail stratum.

Chart review of patients in the last year of life suggest that there is a continued prescribing of futile medications (25). However, our findings are in contrast to the SHARE study, which reported that people with both frailty and polypharmacy have an increased risk of death over a 30-month period (26), similar to a British study (27). Further, a Spanish study found opposite findings to ours; polypharmacy was associated with mortality in frail and prefrail older adults, but not in non-frail individuals (28). It should be noted that our population included only nursing home residents not the general population and this may partly explain our findings. Nursing home residents already have a high vulnerability and complex care needs and may have specific characteristics that affect the interplay between frailty and polypharmacy. A previous paper from the SHELTER study found that frailty is associated with less polypharmacy and with higher prevalence of symptomatic drugs use among residents (18). Further, drug patterns are different among frail and non-frail persons, with the latter showing higher prevalence of disease-modifying drugs, while frail patients show lower prevalence of drugs related to adverse health-outcomes (18).

Frailty is associated with short-term mortality also in older people without multimorbidity (29) and disability, rather than multimorbidity, has been shown to be predictive of death in older adults (30, 31). It is possible that, in our sample of complex older NH residents, frailty is one of the main determinant of negative outcomes including death while, in participants without frailty, the comorbidities and the related treatment burden have an impact on prognosis. The effect of these conditions could the be flattened by the presence of frailty and the association between polypharmacy and mortality only becomes evident only in absence of frailty.

Future research is needed to replicate our findings and identify mechanisms behind the results (25–28). Indeed, it is important to assess the specific vulnerability and characteristics of nursing home residents to verify why there is a different mortality risk associated with frailty and polypharmacy in this population. It may also be relevant to assess how the interplay between these variables differs in younger individuals; frailty is a predictor of mortality also in younger adults (aged 37–73) (32) and is associated with multimorbidity even at these ages. Further, it may be of interest to assess whether interventions such a deprescribing or geriatric cognitive assessment can modify frailty or polypharmacy and alter subsequent risk of death, both in frail and non-frail individuals. Within frail individuals, a systematic review concluded that deprescribing could be safe, feasible, well tolerated and can lead to important benefits (33). However, evidence is still conflicting on whether desprescribing will have any effect on mortality as well as several factors are involved into the deprescribing process (34). A STOPPFrail-guided deprescribing plan in older nursing home residents with frailty and polypharmacy was successful in reducing polypharmacy but did not affect mortality, although the study may not have been sufficiently powered to assess this (35). As our study found higher mortality associated with hyperpolypharmacy in non-frail individuals, the question is whether deprescribing should be prioritized in individual's without frailty. One study comparing nursing home and home-based settings reported that nursing homes provide a highly suitable scenario to carry out a periodic medication review as it is more feasible to apply the review recommendations (36). Polypharmacy is also related to other relevant outcomes in older adults including quality of life. Appropriate deprescribing interventions should be then promoted in the older population, regardless of the frailty status (37).

On the other hand, NHs resident could ondergo other interventions in order to ameliorate their health outcomes, besides deprescribing. That is why it is worth it to thoroughfully assess nursing home patients both with and without frailty: one randomized control study found that an outpatient intervention with Comprehensive Geriatric Assessment (CGA) may both delay the progression of frailty and may contribute to the improvement of frail patients in older persons with multimorbidity (38). A CGA should also assess emotional wellbeing that is strongly associated with frailty (39). Moreover, multidimensional interventions in geriatric settings are likely to be effective in the care of hospitalized frail elderly (40).

In conclusion, among nursing home residents in the SHELTER study multimorbidity and polypharmacy increased the risk of death only in people without frailty. This may be relevant for planning which vulnerable older people should be targeting for deprescription. However, more research is needed to verify whether these findings are replicated in other health care settings and to identify the mechanisms behind the difference in the risk of negative health outcomes and death according to frailty status.

All data generated or analyzed during this study are included in this article. Further enquiries can be directed to the corresponding author. Requests to access these datasets should be directed to graziano.onder@unicatt.it.

The studies involving human participants were reviewed and approved by Ethical Committee of the Catholic Unviersity of the Sacred Heart, approval number P/220/CE/2009. The patients/participants provided their written informed consent to participate in this study.

Study concept and design: EV and RL. Acquisition of data: GO. Analysis and interpretation of data: EV, RL, and KP. Preparation of the manuscript: MZ, EV, KP, and RL. Revision of the manuscript: LS, AC, DFi, MD, DFu, and MC. All authors contributed to the article and approved the submitted version.