Association between polypharmacy and the risk of requiring long‐term care among community‐dwelling older residents: A retrospective cohort study

Abstract

Background

Polypharmacy has been increasing attention as it is associated with a number of health problems, especially adverse outcomes in older adults. However, there is insufficient evidence regarding the risk of polypharmacy and long‐term care.

Methods

We analyzed a community‐based retrospective cohort of residents of Goto City by combining data from health checkups, prescription information, and long‐term care needs certification database. The study sample included residents aged 65–79 years in 2015 who were followed up for 4 years. The number of medications used was categorized as 0, 1–5, 6–9, and ≥10. The outcome was initiation of long‐term care. Cox regression models were used to calculate hazard ratios (HRs) with 95% confidence intervals (CIs) adjusted for potential confounders.

Results

Among 1083 participants, 58 used long‐term care for 4 years. Compared with participants taking no medication, the incidence of long‐term care initiation was approximately double in those taking 1–5 medications, four times higher in people taking 6–9 medications, and 13 times higher in people taking ≥10 medications. After adjusting for potential baseline confounders, the number of medications was significantly associated with the initiation of long‐term care services (1–5 drugs: adjusted HR 2.38, 95% CI 1.06–5.34; 6–9 drugs: adjusted HR 2.97, 95% CI 1.23–7.15; and ≥10 drugs: adjusted HR 5.54, 95% CI 1.89–16.23).

Conclusions

Among community‐dwelling residents aged 65–79 years, the risk of requiring long‐term care had a dose–response relationship with the number of prescribed medications.

1. INTRODUCTION

Global trends show a steady rise in the proportion of the adult population aged 65 years and above, which is likely to continue in the coming decades. Extending years of life without disabilities is a health policy priority in aging societies. ¹ Many countries around the world have established social insurance systems in response to the increase in the older population accompanied by an increase in the number of people with disabilities requiring assistance. ² However, many countries will certainly face challenges in the future due to the sharp increases in costs related to functional disabilities and long‐term care (LTC). In 2000, the Japanese government introduced a public LTC insurance system to extend healthcare support for older adults. ³ This insurance program covers daily life assistance and support (e.g., eating and bathing assistance and housework support), rehabilitation at home, and visiting nursing care, which are not covered by conventional medical insurance. However, to accommodate the increases in overall expenditure, insurance premiums for people aged 65 or older doubled between 2000 and 2021. ⁴ Controlling this increase in the number of LTC insurance certifications is an urgent healthcare and financial challenge. One solution is to extend healthy life expectancy through preventive health interventions for people at a high risk of requiring LTC. Such solutions require the identification of appropriate and sensitive risk factors for identifying individuals at higher risk.

Polypharmacy refers to the prescription of multiple medications to a single individual, and can be defined either numerically or qualitatively. ⁵ , ⁶ , ⁷ Although there is no universal definition of polypharmacy, it has been associated with many health problems, including mortality. ⁸ Polypharmacy may be associated with adverse outcomes, particularly in older adults. ⁷ , ⁸

A systematic review and meta‐analysis reported a strong association between polypharmacy and frailty syndromes. ⁹ However, as there have been a limited number of longitudinal studies of the relations between polypharmacy and frailty, it is difficult to establish any potential causal relationships. ⁹ Polypharmacy was most prevalent in individuals with the mildest LTC requirements. ¹⁰ As part of community‐based prevention programs, Nagasaki University has conducted many field surveys addressing health issues on remote islands, which makes it easier to follow participants and register their disabilities. We examined the association between polypharmacy and the risk of requiring LTC in the early stages of functional disability development by combining a community‐based cohort with associated healthcare‐related data.

2. METHODS

2.1. Data source and participants

We conducted a community‐based retrospective cohort study to explore the risk indicators for certified LTC needs among older adults. We analyzed combined data from three data sets obtained from Goto City, which encompasses a group of remote islands in Nagasaki Prefecture, Western Japan. The first data set consisted of health checkup data for 2015. The second data set was comprised of prescription information collected in June 2015. The third data set consisted of the LTC cost information from 2015 to 2019. These combined data cover people who were alive and living in Goto City as of September 1, 2020.

Participants were residents of Goto City aged 65–79 years who underwent a Japanese national medical checkup in 2015. Participants included people insured by National Health Insurance (public medical insurance for unemployed or self‐employed people and retirees aged < 75 years) or Late Elders' Health Insurance (public medical insurance for everybody aged ≥ 75 years and people aged 65–74 years with certified disability). None of the participants had used LTC between January 1, 2015 and December 31, 2015.

The requirement for informed consent was waived as data were anonymized before being provided to the researchers. The study was approved by the Institutional Review Board of Nagasaki University Institutional Ethics Committee (approval number 21012904).

2.2. Outcome measurement based on LTC

Use of the LTC insurance system in Japan requires application for LTC (support) certification and an assessment to determine the level of care required. ³ Trained investigators visit applicants to assess their activities of daily living, and primary care physicians send an opinion describing the applicant's LTC needs from a medical perspective. Based on this information, LTC needs certification is approved by a municipal review board of health and welfare professionals. Many studies used LTC certification as an index of functional disability. ¹¹ In this study, the outcome was the initiation of LTC, and follow‐up was terminated if the participant received LTC services. The outcome variables were obtained as time‐to‐event data. As annual LTC cost information was obtained from Goto City, the last day of the year was used as the date of the initiation of LTC. Residents were eligible for follow‐up from January 1, 2016 to December 31, 2019.

2.3. Definition of polypharmacy

In Goto City, a system is in operation that connects all dispensing pharmacies via Information and Communication Technology (ICT) and shares dispensing information between pharmacies. In this study, we used the dispensing data stored on the cloud server of this system. According to “Guidelines for Medical Treatment and its Safety in the Elderly” and studies mentioning “hyperpolypharmacy,” individuals were divided into four groups depending on the number of oral medications taken monthly ⁹ , ¹² : no medication, a few medications, polypharmacy (≥6 medications), and hyperpolypharmacy (≥10 medications).

2.4. Covariates

To assess the relation between number of medications and risk of requiring LTC, we identified six potential confounders that could be obtained from health checkup data. The health checkup data included basic attributes from health checkups and the participants' questionnaire responses. Basic characteristics included age and gender. Anthropometric information included height, weight, and body mass index (BMI). Information on smoking status, drinking status, exercise habit, and disease history was obtained using a questionnaire.

With regard to alcohol consumption, the participants were asked, “How often do you drink? (sake, shochu, beer, liquor, etc.)” The participants who answered “every day” or “sometimes” were defined as positive for alcohol consumption. With regard to smoking, we asked, “Do you currently smoke cigarettes?” The participants who answered “yes” were defined as current smokers. With regard to exercise, the participants were asked, “Have you exercised for 30 minutes or more 2 days a week for at least 1 year?” The participants who answered “yes” were defined as positive for exercise habit. With regard to stroke, the participants were asked, “Have you ever been told by a doctor you have had a stroke (cerebral hemorrhage, brain infarction, etc.) and received treatment?” The participants who answered “yes” were defined as having a history of stroke. With regard to a heart disease, the participants were asked, “Have you ever been told by a doctor you have a heart disease (angina pectoris, myocardial infarction, etc.) and received treatment?” The participants who answered “yes” were defined as having a history of heart disease.

2.5. Statistical analysis

Statistical analyses included comparisons based on the number of medications used and multivariable analyses. For univariate analysis, the chi‐square test was used to compare categorical variables, and the Kruskal–Wallis test was used to compare continuous variables. The Kaplan–Meier curve was used to compare the cumulative time to requiring LTC according to the number of medications. The chance of survival (“to not receiving LTC services”) is shown at different time points throughout the observation period separately for the four groups. The log‐rank test was used to check for statistical significance of differences between the four groups in the cumulative time to receiving LTC. Incidences of LTC services per group were calculated, and their 95% confidence intervals (CIs) were calculated using the Clopper–Pearson method. For multivariable analysis, we used Cox regression analysis with receiving LTC services as the dependent variable and calculated the adjusted hazard ratio (HR) after adjusting for age (as continuous), gender, BMI (as continuous), drinking status, current smoking status, exercise habit, history of stroke, and history of heart disease.

Statistical analyses were performed using IBM SPSS Statistics version 29 for Windows (IBM, Tokyo, Japan). In all analyses, p < 0.05 was taken to indicate statistical significance.

3. RESULTS

3.1. Characteristics of the participants

The study population consisted of 1083 participants (432 men and 651 women), and 58 required LTC during the 4‐year follow‐up period. Table 1 shows the participant characteristics classified by the number of medications used. There were significant differences in age, history of stroke, and history of heart disease according to the number of medications. However, there were no significant differences in gender, BMI, drinking status, current smoking status, or exercise habit between the groups.

TABLE 1.

Participant characteristics classified by the number of medications used.

			No medication (n = 366)	1–5 medications (n = 517)	6–9 medications (n = 167)	≥10 medications (n = 33)	p‐value
			Median (Q1–Q3)
Age (years)			69 (67–74)	72 (68–75)	74 (70–76)	72 (71–76)	<0.001 a
BMI (kg/m2)			22.9 (21.1–25.2)	23.2 (21.2–25.5)	23.8 (21.9–25.9)	23.5 (21.3–26.1)	0.073 a

	Category	Reference	n (%)
Gender	Women	Men	216 (59.0)	312 (60.3)	102 (61.1)	21 (63.6)	0.93 b
Alcohol drinking	Yesc	Noc	146 (39.9)	194 (37.5)	50 (29.9)	10 (30.3)	0.13 b
Current smoking	Yes	No	43 (11.7)	38 (7.4)	13 (7.8)	3 (9.1)	0.14 b
Exercise habit	Yes	No	200 (54.6)	257 (49.7)	82 (49.1)	11 (33.3)	0.085 b
History of stroke	Yes	No	21 (5.7)	16 (3.1)	29 (17.4)	9 (27.3)	<0.001 b
History of heart disease	Yes	No	30 (8.2)	32 (6.2)	34 (20.4)	14 (42.4)	<0.001 b

Note: Yes^c: every day or sometimes, No^c: rarely or nondrinker. Q1, Quartile 1; Q3, Quartile 3.

^a Kruskal–Wallis test.

^b Chi–square test.

3.2. Proportion of LTC‐free survival

The proportion of LTC‐free survival over 4 years is shown as a Kaplan–Meier curve in Figure 1. LTC services were required for 8 of 366 participants (2.2%) in the no medication group, 26 of 517 (5.0%) in the 1–5 medications group, 16 of 167 (9.6%) in the 6–9 medications group, and 8 of 33 (24.2%) in the ≥10 medications group. The log‐rank test indicated that these differences between groups were statistically significant (no medication vs. 1–5 medications, p = 0.030; no medication vs. 6–9 medications, p < 0.001; no medication vs. 10–medications, p < 0.001). The median survival could not be displayed because the cumulative survival did not drop below 50% in any of the four groups.

FIGURE 1.

Proportion of long‐term care (LTC)‐free survival over 4 years (Kaplan–Meier survival curves). LTC services were required for 8 of 366 participants (2.2%) in the no medication group, 26 of 517 (5.0%) in the 1–5 medications group, 16 of 167 (9.6%) in the 6–9 medications group, and 8 of 33 (24.2%) in the ≥10 medications group.

3.3. Association between medication use and initiation of LTC

Table 2 shows the association between medication use and the initiation of LTC. With the no medication group as the reference, the incidence of LTC services was approximately double in the 1–5 medications group, 4 times higher in the 6–9 medications group, and 13 times higher in the ≥10 medications group. The fully adjusted HRs (aHRs) were calculated by Cox regression analysis adjusted for gender, age, BMI, multiple lifestyle factors, and disease history. The number of medications was significantly associated with initiation of LTC (1–5 medications: aHR 2.38, 95% CI 1.06–5.34; 6–9 medications: aHR 2.97, 95% CI 1.23–7.15; and ≥10 medications: aHR 5.54, 95% CI 1.89–16.23).

TABLE 2.

Association between medication use and initiation of long‐term care.

	Cumulative incidence (%)	Incidence/1000 person‐years (95% CI)	Age‐ and gender‐adjusted HR (95% CI)	p‐value	Fully adjusted HR (95% CI)	p‐value
No medication	8/366 (2.2)	5.6 (2.4–10.9)	1 [reference]		1 [reference]
1–5 medications	26/517 (5.0)	12.9 (8.4–18.8)	2.04 (0.92–4.53)	0.081	2.38 (1.06–5.34)	0.035
6–9 medications	16/167 (9.6)	25.0 (14.4–40.3)	3.60 (1.52–8.57)	0.004	2.97 (1.23–7.15)	0.015
≥10 medications	8/33 (24.2)	70.8 (31.1–134.7)	10.51 (3.91–28.28)	<0.001	5.54 (1.89–16.23)	0.002

Note: Data were subjected to Cox regression analysis. The fully adjusted model included as covariates: age (as continuous); gender; BMI (as continuous); drinking status; smoking habits; exercise habit; history of stroke; history of heart disease.

Abbreviations: CI, confidence interval; HR, hazard ratio.

4. DISCUSSION

This study demonstrated an association between polypharmacy and initiation of LTC for up to 4 years among community‐dwelling residents aged 65–79 years. According to previous studies, polypharmacy has been reported to show associations with frailty and mortality. Gnjidic et al. reported that the use of more than five medications was associated with incident frailty across a population of 1662 men aged ≥ 70 years with a follow‐up period of 2 years. ¹³ Jamsen et al. found that a higher number of medications was associated with a greater risk of mortality and with a higher risk of transitioning from robust to the prefrail state. ¹⁴ However, polypharmacy represents a multifactorial issue with both modifiable and nonmodifiable components. The diseases leading to polypharmacy contribute to frailty and vice versa. In several previous studies, the association between polypharmacy and incident frailty was statistically significant even after adjusting for the type and number of medical conditions at baseline. ¹⁵ , ¹⁶ In contrast, Kuroda et al. reported a relation between LTC needs certification and polypharmacy. They performed a population‐based study to combine medical claims data and LTC needs certification records via a case–control study. ¹⁷ Their study did not adjust for potential confounders, such as BMI, drinking, smoking, and physical activity, which were adjusted in most other studies. ¹¹ , ¹⁷ To our knowledge, our retrospective cohort study is the first to estimate the risk of polypharmacy for requiring LTC among community‐dwelling older residents who underwent health checkups. Our study showed that the risk of requiring LTC has a dose–response relationship with the number of medications prescribed. Our community‐based retrospective cohort study could guarantee the order of exposure and outcome. Japanese LTC insurance claims have been used in many studies, and their usefulness has been demonstrated. ¹⁸ We found that it is possible to investigate the association between medications and LTC progression by integrating administrative data sets.

Studies on polypharmacy‐related interventions have increased substantially. ¹⁹ However, there is a lack of high‐quality evidence to support the broad implementation of polypharmacy‐related interventions. ¹⁹ , ²⁰ Since 2016, clinics and hospitals in Japan have received incentives for each case where they were successful in reducing two or more medications in patients previously receiving six or more drugs, and these incentives were extended to pharmacies in 2018. After implementing these incentive‐based policies, polypharmacy has decreased nationwide. ²¹ , ²² However, it remains unclear whether reducing medications through these policies will actually lower LTC costs.

This study had some limitations. First, this was a cohort study in a single region, so further studies will be necessary to replicate our results in other regional populations. Certification of the level of care required is a program unique to Japan, and these results are not directly applicable to other countries. Furthermore, using LTC certification as a proxy for disability may lead to differential outcome misclassifications because of the application‐based system of LTC certification. On the other hand, from a medical economic perspective of controlling LTC costs, there is great significance in conducting analysis on an application basis. Second, this study focused on community‐dwelling residents using data from the Japanese national health checkup program; therefore, it was not possible to adjust for some potential confounders. Specifically, this study did not account for musculoskeletal diseases, disease severity, or social factors, all of which may be important risk factors for requiring LTC. Future studies are warranted to explore the association between polypharmacy and the risk of requiring LTC while adjusting for a broader range of potential covariates. Third, participation in health checkups was not mandatory. It is possible that those who participated in health checkups were more health‐conscious, which may have led to selection bias. Finally, we were unable to track people who died or emigrated from the island during the observation period.

In conclusion, among community‐dwelling residents aged 65–79 years, the risk of requiring LTC had a dose–response relationship with the number of prescribed medications. This was the first study to estimate the risk of polypharmacy for requiring LTC among community‐dwelling older residents undergoing health checkups. Further studies are necessary to explore whether reducing medications will actually lower LTC costs.

AUTHOR CONTRIBUTIONS

Kengo Maeda: Methodology; visualization; investigation; writing – original draft; conceptualization; formal analysis. Shin‐Ya Kawashiri: Conceptualization; methodology; writing – review and editing. Kazuhiko Arima: Formal analysis; methodology; writing – review and editing. Tetsuro Niri: Writing – review and editing. Yukiko Honda: Conceptualization; data curation; resources; writing – review and editing. Jun Miyata: Writing – review and editing. Fumiaki Nonaka: Conceptualization; data curation; resources; writing – review and editing. Hirokazu Kumazaki: Writing – review and editing. Takahiro Maeda: Conceptualization; data curation; methodology; writing – review and editing. Yasuhiro Nagata: Conceptualization; data curation; funding acquisition; methodology; resources; project administration; supervision; writing – original draft.

FUNDING INFORMATION

This work was supported by JSPS KAKENHI (grant number JP21K06690).

CONFLICT OF INTEREST STATEMENT

Jun Miyata, Fumiaki Nonaka, and Takahiro Maeda are affiliated with an endowed department (Department of Island and Community Medicine) at Nagasaki University Graduate School of Biomedical Science, which is supported by Nagasaki Prefecture and Goto City.

ETHICS STATEMENT

Ethics approval statement: The study was approved by the Institutional Review Board of Nagasaki University Institutional Ethics Committee (approval number 21012904).

Patient consent statement: The requirement for informed consent was waived as data were anonymized before being provided to the researchers.

Clinical trial registration: Not applicable.

PERMISSION TO REPRODUCE MATERIAL FROM OTHER SOURCES

Not applicable.

Maeda K, Kawashiri S‐Y, Arima K, Niri T, Honda Y, Miyata J, et al. Association between polypharmacy and the risk of requiring long‐term care among community‐dwelling older residents: A retrospective cohort study. J Gen Fam Med. 2025;26:402–407. 10.1002/jgf2.70041

DATA AVAILABILITY STATEMENT

Data cannot be shared for privacy and ethical reasons.