Intrinsic Capacity Impairment Patterns and their Associations with Unfavorable Medication Utilization: A Nationwide Population-Based Study of 37,993 Community-Dwelling Older Adults

Abstract

Objectives

Our aim was to explore the patterns of intrinsic capacity (IC) impairments among community-dwelling older adults and the associations of these different patterns with excessive polypharmacy, potentially inappropriate medications, and adverse drug reactions in a nationwide population-based study.

Design

A cross-sectional study included older adults from the Taiwan Integrated Care for Older People (ICOPE) program in 2020.

Setting and Participants

The study subjects comprised 38,308 adults aged 65 years and older who participated in the ICOPE Step 1 screening and assessed six domains of IC following the World Health Organization (WHO) ICOPE approach.

Methods

Latent class analysis was adopted to identify distinct subgroups with different IC impairments patterns. The associations between different IC impairments patterns and unfavorable medication utilization, including excess polypharmacy (EPP), potentially inappropriate medications (PIMs), and adverse drug reactions (ADRs), were assessed by multivariate logistic regression models.

Results

Latent class analysis identified five distinct subgroups with different IC impairment patterns: robust (latent class prevalence: 59.4%), visual impairment (17.7%), physio-cognitive decline (PCD) with sensory impairment (12.3%), depression with cognitive impairment (7.7%), and impairments in all domains (2.9%). Compared to the robust group, all other groups were at higher odds for unfavorable medication utilization. The “depression with cognitive impairment” group (EPP: aOR=4.35, 95% CI 3.52–5.39, p<0.01; PIMs: aOR=2.73, 95% CI 2.46–3.02, p<0.01) and the “impairment in all domains” group (EPP: aOR=9.02, 95% CI 7.16–11.37, p<0.01; PIMs: aOR=3.75, 95% CI 3.24–4.34, p<0.01) remained at higher odds for EPP and PIMs after adjustment.

Conclusions

We identified five distinct impairment patterns of IC, and each impairment pattern, particularly the “depression with cognitive impairment” and “impairment in all domains”, was associated with higher odds of EPP and PIMs. Further longitudinal and intervention studies are needed to explore long-term outcomes of different impairment pattern and their reversibility.

Introduction

The World Health Organization (WHO) published the World Report on Aging and Health, which defined “healthy aging” as a process of developing and maintaining functional ability that enables wellbeing in older age (Rudnicka et al., 2020, Beard et al., 2016, Kino et al., 2021). In particular, the World Report on Aging and Health proposed using “functional ability” and “intrinsic capacity” (IC) as measures to promote the life-course approach to disability/dementia prevention. The constructs of IC include five domains: cognition, locomotion, psychosocial, sensory (vision and hearing), and vitality (4, 5). These domains, which were constructed based on scientific evidence, may individually represent different underlying pathophysiology; however, their clustering may reveal common aging pathways underlying the IC impairment phenotype. The definition of IC shifted the disease-centric paradigm to a function-centric one used to characterize the health of older adults (6), and it transformed the model of medicine from a reactive one to a preventive one (4). Under the umbrella of healthy aging, health care professionals and health care systems should consider all contributing factors of functional declines (e.g., aging, diseases, social determinants, and other factors) to formulate holistic care plans instead of only responding to certain clinical manifestations, symptoms, or diseases (7, 8). The emphasis on long-term functional outcomes is compatible with modern value-based health care schemes to construct future health care systems with person-centeredness and holisticity (9, 10).

Based on the WHO document “Integrated Care for Older People (ICOPE): guidance for person-centred assessment and pathways in primary care” (11), WHO introduced a simple assessment instrument for older people or other people in the communities to primarily screen deficits of intrinsic capacity, which was defined as Step 1. Step 1 was designed to facilitate early identification of impaired intrinsic capacity, so it was designed as conveniently as possible to be utilized in the community settings. For those with deficits of intrinsic capacity, they should visit their primary care physicians with the results and the primary care physicians would conduct further comprehensive assessments focused on the deficits of intrinsic capacity. At this stage, it was defined as Step 2 of ICOPE process.

The development of IC features new conceptual modifications for promoting the health of older people based on longitudinal evidence across countries (4). Also, IC is proposed to serve as a potential indicator of functional reserve in the aging process. Several studies have shown the association between impairments in different IC domains and numerous adverse outcomes (12, 13). Although IC captures a composite state of healthy aging, it remains unclear whether these construct domains are of the same prognostic significance or independent from each other in the construct of IC. For example, previous studies have clearly identified subtypes of physical frailty using the data-driven approach and found that frailty subtypes carry different underlying pathophysiologies, natural courses, and prognoses (14, 15). In our previous study, we have assessed the association between the number of impairments in different IC domains and adverse outcomes by a weighted-IC scoring system (12). Extending our question from “numbers” to “clusters” of IC impairments would provide further insights regarding the interplays of the different domains of IC impairments.

In addition to interplays of the different domains of IC impairments, linking the potential cluster effect of the IC impairments with clinical outcomes, such as unfavorable medication uses, is of great interest. Several indicators of unfavorable medication use, including excess polypharmacy (EPP), defined as the concomitant use of ten or more medicines; and potentially inappropriate medications (PIMs), defined as drugs for which use among older adults should be avoided, have been reported to be associated with an increased risk of geriatric syndromes, including cognitive impairment and poor nutritional status (16, 17). Suboptimal prescribing and the occurrence of adverse drug reactions (ADRs), which is a harmful result caused by the use of drug(s), may also negatively affect functional status in older people (16). On the other hand, decreased physical and mental health status are also associated with unfavorable medication use (18). Therefore, IC assessment with the composite consideration of all physical and mental capacities of an individual, may thus provide good prediction ability to unfavorable medication use and extend the clinical implications of IC assessment. Hence, this cross-sectional study adopts a data-driven approach, latent class analysis, to examine IC impairment patterns and their associations with unfavorable medication utilization, including EPP, PIMs, and ADRs, in a nationwide population-based sample in Taiwan.

Methods

Data source and study population

The Taiwan Integrated Care for Older People (ICOPE) program, a prospective population-based study, was conducted by the Health Promotion Administration (HPA) of the Ministry of Health and Welfare in 2020. Community-dwelling people aged 65 years and older were invited to participate in the study by hospital outpatient departments, primary care physicians, and community public health centers. Primary care physicians are those who run primary care services in the communities, either public or private services, that provide mainly health care instead of public health services. Public health centers in Taiwan are locality-based institutions that provide preventive services, infectious disease controls, and care management (such as pulmonary tuberculosis).

The assessment for IC in Taiwan followed the World Health Organization (WHO) ICOPE approach (11, 19); that is, a screening tool for IC impairment was implemented for all participants (ICOPE Step 1), and then an in-depth assessment (ICOPE Step 2) was applied to confirm IC impairments. Through face-to-face interviews, six domains of IC were evaluated by clinicians: cognition, locomotion, vitality/nutrition, vision, hearing, and psychology. Herein, sensory capacity was split into the domains of vision and hearing following the WHO ICOPE operationalization. Other covariates, including demographic characteristics, medication use, and institutions of ICOPE assessment, were also collected. The HPA collected all data and authorized the research team for data analysis, so the Institutional Review Board requirement was waived for this study. All participants in the program provided consent agreements for the use of their data for research.

This cross-sectional study comprised 38,308 eligible individuals who participated in the ICOPE Step 1 screening in 2020. After excluding people with invalid age (n=2), incomplete information about intrinsic capacity (n=36), and duplicated assessments (n=277), a total of 37,993 participants were included in the present study.

Screening for IC impairments — ICOPE Step 1

The IC assessment was introduced based on the WHO ICOPE guidelines (Supplementary Table 1) (18, 19); it allows a rapid and feasible assessment of six domains of IC. The cognition domain was evaluated by a 3-item recall test, a quick screening test for short-term memory, and time orientation, which is theorized as an unconscious yet fundamental cognitive process. Failing in either the item recall test or time orientation was defined as an impairment in the cognition domain. The locomotion domain was assessed by the Five Times Sit to Stand test. According to the recommendation of the Asian Working Group for Sarcopenia (AWGS) (20), participants who were unable to complete the test within 12 seconds were regarded as having limited mobility, i.e., an impairment in locomotion domain. Unintentional weight loss and appetite loss were assessed in the vitality domain; participants who had an unintentional weight loss of more than 3 kg or had experienced a loss of appetite in the past three months were defined as malnutrition. Visual impairment was determined by asking participants “Do you have any problems in seeing, reading, or having any eye diseases limiting the ability to see (e.g. diabetes, hypertension)?” Participants answering “yes” were considered to have an impairment in the vision domain. Hearing ability was assessed by the whisper test. Whisper test was performed once on each ear. Those who reported difficulties for hearing on either ear would be classified as impaired hearing. For those with hearing aid, they were asked to remove the hearing aid for whisper tests. In that test, participants were asked to repeat a series of three numbers whispered by the examiners. In the psychological domain, two questions were asked to evaluate psychological status: “Have you been bothered by feeling down, depressed or hopeless in the past two weeks?” and “Have you lost interest in your usual activities in the past two weeks?”. Participants answering “yes” to either question were considered to have an impairment in the psychosocial domain.

Outcome measures — unfavorable medication utilization

In this study, information about medication utilization was collected from a self-reported medication history through face-to-face interviews and further verified by the Pharma Cloud System (21) of the National Health Insurance Administration (NHIA). Taiwan's National Health Insurance (NHI) is a single-payer and government-run healthcare system that covers more than 99% of Taiwan's population (>23 million people). Accessed by the NHI IC Card (issued by the NHIA to all beneficiaries), the Pharma Cloud System recorded all prescriptions received by the beneficiaries under the NHI, which allows clinicians to retrieve records of prescriptions (drug names, dosage, frequencies and days of supply) in the past 3 months for all beneficiaries (including data in the same day of the IC assessment). All participants in the program have granted their consent to use their data on the Pharma Cloud system for research. Excess polypharmacy (EPP), potentially inappropriate medications (PIMs), and adverse drug reactions (ADRs) were assessed accordingly. Based on previous studies (22, 23) participants who regularly took ten or more drugs were considered to have EPP. Study subjects using pain medications, including nonsteroidal anti-inflammatory drugs (NSAIDs) and opioids, or hypnotic drugs, such as benzodiazepines and Z-drugs, were determined to use PIM accordingly (24). To detect ADRs, one question was included in the Taiwan ICOPE questionnaire: “Have you ever suffered from balance problems, fatigue, dizziness, hypotension, or dry mouth after taking medicines?” Participants responding “yes” to the question were classified as experiencing ADRs.

Other variables

The following demographic characteristics for each study participant were collected in the Taiwan ICOPE program: age, sex, institutions and settings of ICOPE assessment (e.g., hospital outpatient departments, primary care physicians, and community public health centers).

Statistical analysis

Descriptive statistics of the demographic and clinical characteristics for all participants were analyzed. Continuous variables in the text and tables are expressed as the mean ± standard deviation, and categorical variables are expressed as percentages.

Phase 1: Latent class analysis (LCA) to identify impairment patterns of IC

In the first phase of the study, LCA was used to identify distinct participant subgroups with different patterns of IC impairments. LCA, a subset of structural equation modeling, is performed to classify heterogeneous individuals into clusters of similar latent classes based on their patterns of associations in the set of categorical variables (21). We used this method to assess whether the population of older adults could aggregate into subgroups characterized by the cooccurrence of impairments in the six domains of IC. To determine the optimal number of IC impairment patterns, we compared the goodness-of-fit of different models using the Akaike Information Criteria (AIC), Bayesian Information Criteria (BIC), and sample-size adjusted BIC (aBIC); lower AIC/BIC/aBIC values indicate a better model fit (25, 26). In addition, we also conducted the bootstrap log-likelihood ratio (BLR) test to select the final class. The fit statistics of the model were evaluated from two to six classes, and the best model was identified based on the results of both model fit and clinical insights. Comparisons of baseline characteristics across different IC impairment patterns were performed by analysis of variance (ANOVA) for continuous variables and chi-square tests or Fisher's exact test for categorical variables.

Phase 2: Association between IC impairment patterns and medication utilization

In the second phase, multivariate logistic regression models were used to examine the association between IC impairment patterns and medication utilization, including EPP, PIMs, and ADRs. These models were further adjusted by age, sex, and institutions of ICOPE assessment. Crude and adjusted odds ratios (aORs) with 95% confidence intervals (CIs) were reported.

All analyses were performed using SAS, version 9.4 (SAS Institute Inc., Cary, NC) (Lanza et al., 2007, PROC LCA & PROC LTA (Version 1.3.2) [Software], 2015, LCABootstrap SAS Macro (Version 4.0) [Software], 2016). A two-tailed p value of <0.05 was considered to indicate statistical significance.

Results

Demographic characteristics

Table 1 summarizes the demographics of the study participants, and a total of 37,993 older adults (mean age: 73.2±6.6 years; 60.9% females) were identified. Approximately half of the participants received the ICOPE assessment in hospital outpatient settings, and one-third of participants were screened in community public health centers. Among all study participants, impairments in cognitive ability were most prevalent (20.3%), followed by vision (17.6%) and locomotion (11.2%) impairments. However, only 4.6% of the study participants had impairments in vitality.

Table 1.

Characteristics of total participants included in the study

Data values show Mean ± SD, or number (%)	Total participants
Numbers	37993
Age (years)	73.2 ±6.6
65–69	13955 (36.7)
70–74	10255 (27.0)
75–79	6903 (18.2)
80–84	4269 (11.2)
±85	2611 (6.9)
Sex
Male	14869 (39.1)
Female	23124 (60.9)
Institution receiving IC assessment
Outpatient	18479 (48.7)
Health examination	2506 (6.6)
Community-based screening	12364 (32.5)
Others	4644 (12.2)
Assessment of IC impairment
Cognitive impairment	7697 (20.3)
Limited mobility	4263 (11.2)
Malnutrition	1761 (4.6)
Visual impairment	6684(17.6)
Hearing loss	4093 (10.8)
Depressive symptoms	2871 (7.6)

*IC, intrinsic capacity

Impairment patterns of intrinsic capacity

We presented the fit statistics from the 2- to 6-class models to identify appropriate IC impairment patterns in Table 2. As the 6-class model became unstable, AIC/BIC/aBIC values favored the 5-class model; the result of BLR test also showed that the 5-class model was statistically better than the 4-class model with an improvement of goodness-of-fit (p=0.01). In addition, the IC impairment patterns in the 5-class model were clinically meaningful according to the expert opinions (Supplementary Table 2). With these considerations, we selected the 5-class model to define IC impairment patterns in this study.

Table 2.

Latent class analysis model fit of IC impairment

No. of Classes	Degrees of Freedom	AIC	BIC	aBIC	BLR Test
2	50	765.75	876.83	835.52	-
3	43	388.15	559.06	495.50	0.01
4	36	186.19	416.91	331.11	0.01
5	29	125.68	416.21	308.16	0.01

*The model became unstable with the 6-class model; AIC = Akaike information criterion; BIC = Bayesian information criterion; aBIC = Adjusted Bayesian information criterion; BLR = bootstrapped log-likelihood ratio

Qualitative labels were assigned to the five impairment patterns based on the conditional probabilities: “robust”, “visual impairment”, “physio-cognitive decline (PCD) with sensory impairment”, “depression with cognitive impairment”, and “impairments in all domains” (Figure 1; Supplementary Table 2). The class prevalence of the “robust” group was 59.4%, indicating that our study participants were relatively healthy. Approximately 18% of the study participants were categorized into the “visual impairment” group, and 12.3% of study participants had co-occurrent physical and cognitive declines with sensory impairment and were categorized into the “PCD with sensory impairment” group. The “depression with cognitive impairment” group included those suffering from depressive symptoms and cognitive impairments, and 7.7% of the study population displayed this impairment pattern. Only a few participants (2.9%) were categorized as having “impairments in all domains”, indicating high probabilities of impairments across all domains of IC.

Figure 1.

Latent profiles of IC impairment for 5-class model

Yellow line represents the latent profile of robust group; red line represents the latent profile of visual impairment group; blue line represents the latent profile of PCD with sensory impairment group; green line represents the latent profile of depression with cognitive impairment group; black line represents the latent profile of impairments in all domains group; IC, intrinsic capacity; PCD, physio-cognitive decline.

Table 3 shows comparisons between different IC impairment patterns based on LCA, including robust, visual impairment, PCD with sensory impairment, depression with cognitive impairment, and impairments in all domains. Those in the “PCD with sensory impairment” group and “impairments in all domains” group were significantly older (p<0.01) and had more impairments in domains of IC (p<0.01). Compared to those with other impairment patterns, participants in the “depression with cognitive impairment” group were more likely to be women (p<0.01). The distribution of institutions of ICOPE assessment was different across the five distinct subgroups (p<0.01), although most study participants received their ICOPE assessment in the hospital outpatient departments.

Table 3.

Comparisons between different subgroups based on LCA

Data values show M ± SD, or number (%)	Robust	Visual impairment	PCD with sensory impairment	Depression with cognitive impairment	Impairments in all domains	p-value*
Numbers	26883	5395	2957	1932	826
Age (years)	72.5 ± 6.1	72.9 ± 6.4	78.5 ± 7.7	73.7 ± 6.5	78.8 ± 7.7	<0.01
65–69	10689 (39.8)	2055 (38.1)	462 (15.6)	631 (32.7)	118 (14.3)	<0.01
70–74	7599 (28.3)	1468 (27.2)	526 (17.8)	520 (26.9)	142 (17.2)
75–79	4755 (17.7)	990 (18.4)	590 (20.0)	385 (19.9)	183 (22.2)
80–84	2566 (9.6)	570 (10.6)	685 (23.2)	274 (14.2)	174 (21.1)
±85	1274 (4.7)	312 (5.8)	694 (23.5)	122 (6.3)	209 (25.3)
Sex
Male	10637 (39.6)	2057 (38.1)	1243 (42.0)	637 (33.0)	295 (35.7)	<0.01
Female	16246 (60.4)	3338 (61.9)	1714 (58.0)	1295 (67.0)	531 (64.3)
Institution receiving IC assessment
Outpatient	12761 (47.5)	2937 (54.5)	1405 (47.5)	946 (49.0)	430 (52.1)	<0.01
Health examination	1402 (5.2)	593 (11.0)	285 (9.6)	146 (7.6)	80 (9.7)
Community-based screening	8930 (33.2)	1431 (26.5)	1026 (34.7)	704 (36.4)	273 (33.0)
Others	3790 (14.1)	434 (8.0)	241 (8.2)	136 (7.0)	43 (5.2)
Assessment of IC impairment
Cognitive impairment	3091 (11.5)	969 (18.0)	2209 (74.7)	724 (37.5)	704 (85.2)	<0.01
Limited mobility	1136 (4.2)	0 (0.0)	2166 (73.3)	282 (14.6)	679 (82.2)	<0.01
Malnutrition	0 (0.0)	784 (14.5)	0 (0.0)	535 (27.7)	442 (53.5)	<0.01
Visual impairment	0 (0.0)	4835 (89.6)	1099 (37.2)	289 (15.0)	461 (53.5)	<0.01
Hearing loss	1345 (5.0)	738 (13.7)	1665 (56.3)	0 (0.0)	345 (41.8)	<0.01
Depressive symptoms	0 (0.0)	453 (8.4)	86 (2.9)	1710 (88.5)	622 (75.3)	<0.01

*ANOVA test for continuous variables and Chi-square test or Fisher exact test for categorical variable; IC, intrinsic capacity; PCD, physio-cognitive decline

The association between IC impairment patterns and unfavorable medication utilization

Excess polypharmacy (EPP)

The results of adjusted multivariate logistic regression are presented in Table 4. Compared to the “robust” group, the “visual impairment” group (aOR=2.42; 95% CI, 2.03–2.88; p<0.01), “PCD with sensory impairment” group (aOR=3.97; 95% CI, 3.29–4.78; p<0.01), and “depression with cognitive impairment” group (aOR=4.35; 95% CI, 3.52–5.39; p<0.01) were at greater odds for EPP after adjustment for age, sex, and institution receiving IC assessment. For those in the “impairments in all domains” group, the odds of EPP were nine times higher than for those in the “robust” group (aOR=9.02; 95% CI, 7.16–11.37; p<0.01).

Table 4.

The association between IC impairment subgroups and medication use

	OR	95% CI	p-value	aOR	95% CI	p-value
Excess polypharmacy
Robust	ref	-	-	ref	-	-
Visual impairment	2.53	2.13–3.01	<0.01	2.42	2.03–2.88	<0.01
PCD with sensory impairment	4.87	4.09–5.79	<0.01	3.97	3.29–4.78	<0.01
Depression with cognitive impairment	4.36	3.53–5.39	<0.01	4.35	3.52–5.39	<0.01
Impairments in all domains	10.96	8.80–13.66	<0.01	9.02	7.16–11.37	<0.01
Potentially inappropriate medications
Robust	ref	-	-	ref	-	-
Visual impairment	1.91	1.78–2.04	<0.01	1.79	1.67–1.92	<0.01
PCD with sensory impairment	2.04	1.87–2.23	<0.01	1.94	1.77–2.12	<0.01
Depression with cognitive impairment	2.82	2.55–3.11	<0.01	2.73	2.46–3.02	<0.01
Impairments in all domains	4.08	3.54–4.70	<0.01	3.75	3.24–4.34	<0.01
Adverse drug reactions
Robust	ref	-	-	ref	-	-
Visual impairment	2.91	2.56–3.32	<0.01	2.69	2.36–3.06	<0.01
PCD with sensory impairment	3.22	2.77–3.78	<0.01	2.96	2.52–3.49	<0.01
Depression with cognitive impairment	6.26	5.39–7.28	<0.01	6.00	5.16–6.98	<0.01
Impairments in all domains	10.53	8.77–12.65	<0.01	9.45	7.81–11.44	<0.01

aOR= adjusted odds ratio from the multivariate logistic regression with adjustment for age, sex, and institution receiving IC assessment; IC, intrinsic capacity; PCD, physio-cognitive impairment

Potentially inappropriate medications (PIMs)

After adjusting for age, sex, and institutions of ICOPE assessment, we found that study participants in the “visual impairment” group were associated with a 79% increase in the odds of PIM use compared with the “robust” group (aOR=1.79; 95% CI, 1.67–1.92; p<0.01). Moreover, both the “PCD with sensory impairment” group (aOR=1.94; 95% CI, 1.77–2.12; p<0.01) and the “depression with cognitive impairment” group (aOR=2.73; 95% CI, 2.46–3.02; p<0.01) were associated with a statistically higher odds of PIM use than the “robust” group. Participants in the “impairments in all domains” group were significantly more likely to be prescribed PIMs than the “robust” group (aOR=3.75; 95% CI, 3.24–4.34; p<0.01).

Adverse drug reactions (ADRs)

Compared to the “robust” group, participants in other IC impairment pattern groups were more likely to experience ADRs after adjustment for age, sex, and institution receiving IC assessment, particularly those in the “depression with cognitive impairment” group (aOR=6.00; 95% CI, 5.16–6.98; p<0.01) and the “impairments in all domains” group (aOR=9.45; 95% CI, 7.81–11.44; p<0.01).

Discussion

This study identified 5 distinct IC impairment patterns (robust, visual impairment, PCD with sensory impairment, depression with cognitive impairment, and impairments in all dimensions), and the odds for unfavorable medication utilization was significantly higher in all IC impairment groups than in the robust group. To the best of our knowledge, this is the first study using a data-driven approach to explore IC impairment patterns and to examine their associations with unfavorable medication utilization using a nationwide population-based sample. The results supported the possibilities of different pathogenic processes of IC impairments and their associations with pharmaceutical management in the clinical course. Notably, in this study, nearly 60% of all participants were robust, i.e., they had no impairments in any IC domain, which suggested the good health status of the study participants. The prevalence of IC impairments in this study was similar to that in one previous study from China (29), but was it lower than those in other studies (30, 31). Since most participants were enrolled from outpatient settings, healthy and active older people were more likely to be enrolled in the outpatient visits.

A large international study confirmed the association between IC impairments and the risk of incident disability (31). Social frailty has also been reported to be associated with IC impairments (32). Interestingly, from this study, cognitive impairment and limited mobility tended to cluster with other domains of IC instead forming independent categories. Although the cross-sectional design limited the potential to define the pathogenic process of each IC impairment pattern, this study still disclosed the important finding that different IC impairments may tend to present as clusters. Also, this study clearly demonstrated the importance of the cognition domain in the constructs of IC because it was present in multiple impairment patterns, such as PCD with sensory impairment, depression with cognitive impairment, and impairments in all domains.

These findings echo the identification of the physiocognitive decline syndrome (PCDS, defined as the co-occurrence of impairments in physical and cognitive function) phenotype in our previous studies conducted in community-dwelling older people (33). Longitudinal studies have clearly shown that PCDS substantially increases the risk of incident disability, incident dementia, all-cause mortality and specific neurocircuit impairments in older people (Chung et al., 2021, Liang et al., 2021, Liu et al., 2020), but no special information has been reported regarding sensory impairments. The results of randomized controlled trials also supported the potential to improve mobility, cognition, nutritional status, depressive symptoms and others by multidomain interventions (34, 36, 37) which lends support to the preventive nature of IC and its reversibility to promote healthy aging.

Compared to the robust group, all IC impairment pattern groups were of higher odds for EPP, PIMs, and ADRs, which may be related to the complexity of the clinical conditions. Although IC addresses the potential for healthy aging, diseases and multimorbidity eventually occur alongside aging. Impairments in IC domains are eventually associated with complex care needs and medication utilization.

After adjustment, depression with cognitive impairment and impairments in all domains were the two IC impairment pattern groups with higher odds of unfavorable medication utilization, which suggested the highest care complexity. One possible reason for depression with cognitive impairment group being of greater odds for PIM and ADR is that they may undergo an additional physiopathological process that exposes them to higher odds of PIMs and ADRs. We did perform a series of sensitivity analyses to examine the association between number of impairments and odds of unfavorable medication use (Supplementary Figure 1 and Supplementary Table 3) as well as each domain of impairment and odds of unfavorable medication use (Supplementary Table 4 and 5). We found potential additive effects regarding the number of impairments and odds of unfavorable medication use. In addition, the OR of PIMs was 4.35 for the cluster of psychological + cognitive (Table 4) and psychological was 2.31 (Supplementary Table 5), which also suggested a potential additive effect.

Nevertheless, we can only hypothesize there might be a superposed pathological process. To the best of our knowledge, there are only a few studies which investigated the relationship between depression and PIM use. One study from Japan reported a significant association between the presence of depression and PIM use according to the modified Beers criteria in older adult patients (38) while another study of 16,568 community living older adults reported that those with depression had higher odds of PIM use when compared with those without depression (39). Although major depression is less prevalent, depressive disorders are frequently encountered in the geriatric population. So does cognitive impairment. It is not surprising that people with psychological domain were more likely to be prescribed PIMs as several psychotropics such as benzodiazepines are included in the STOPP and 2015 Beers criteria. In addition, the health care system may play certain roles since older adults with cognitive impairment may experience more psychotropic drug use for behavioral & psychiatric symptoms (40, 41). Furthermore, previous studies have shown that the odds of ADRs may go along with PIMs (23). Hence, it is reasonable that people with a higher odds of PIMs were more likely to experience ADRs. A person-centered approach for deprescribing could thus been proposed and implemented to optimize the pharmacotherapy for older people (42). Therefore, promoting healthy aging should also address the importance of associated clinical care needs and provide appropriate deprescribing interventions.

In addition to identifying impairment patterns of IC, another merit of our study was to further link these patterns to unfavorable medication utilization, including EPP, PIMs and ADRs. Although the World Report on Ageing and Health and ICOPE guidelines have advocated that IC assessment could prevent unnecessary treatments (43, 44), polypharmacy and side effects, only a few studies have described the baseline status of polypharmacy when conducting IC assessment (45, 46). To the best of our knowledge, our study may be the first to examine whether an association exists between IC impairment patterns and unfavorable medication utilization. Excessive polypharmacy and PIMs are commonly used quality indicators of medication utilization in older adults, and many existing studies have proven the association between these quality indicators and adverse clinical outcomes, such as all-cause hospitalization and fracture-specific hospitalizations in the elderly (23, 46, Huang et al., 2019, Hsu et al., 2017, Tseng et al., 2020, Lin et al., 2017). Our findings thus could serve as a good screening tool to identify people who may suffer adverse clinical outcomes associated with unfavorable medication utilization by assessing impairment patterns of IC.

Despite our efforts in this study, there were still some limitations. First, the cross-sectional design limited the possibility of examining the clinical outcomes and progression processes of different IC impairment patterns, which needs a longitudinal study for confirmation. Second, convenient sampling strategies from health care providers and public health centers may result in inevitable selection bias. Disadvantaged older people with limited access to these settings were less likely to be enrolled. Third, we evaluated the ADRs based on a relatively nonspecific question. Although this is the only one question we can acquire from Taiwan's ICOPE questionnaire as the ICOPE 1 did not intend to capture unfavorable medication uses and may not fully capture (underestimate) the prevalence of ADRs, we still find good links between IC impairments and odds of ADRs. More researches are warranted to further verify our findings. Fourth, due to data restriction, our study didn't include the number of comorbidities as the covariate. Based on the principles of ICOPE Step 1 assessment, comorbidity is not the required component and senior respondents may not 100% clear for their underlying medical conditions. The Taiwan government followed the WHO principles to design Step 1 questionnaire, and information of comorbidity should be collected at Step 2. Future research should consider adjusting the number of comorbidities as a confounding factor if the information is available. Fifth, unfavorable medication utilization may be the result of the fragmentation of clinical care, and we did not have data for further evaluation. Nevertheless, these limitations did not weaken the strength of this study, and a large nationwide population-based study assessing unfavorable medication utilization using a data-driven approach has not been reported before.

Conclusion

In conclusion, five distinct IC impairment patterns were identified, and each impairment pattern, particularly the “depression with cognitive impairment” and “impairment in all domains”, was associated with higher odds of EPP and PIMs. Further longitudinal and interventional studies are needed to further verify our findings and their clinical implications.

Ethical standards

Institutional Review Board requirement was waived for this study.

Contributors

All authors meet the criteria for authorship stated in the Uniform Requirements for Manuscripts Submitted to Biomedical Journals. All authors drafted the article, revised it it critically for important intellectual content, and approved the final version to be published. Meng LC, Hsiao FY, Huang ST, Lu WH, Chen LK, Peng LN designed the research. Meng LC, Hsiao FY, Peng LN, and Chen LK drafted and prepared the manuscript. Meng LC analyzed the data. Hsiao FY, Peng LN, and Chen LK provided critical methodological input. Meng LC, Huang ST, Lu WH, Hsiao FY provided methodological and statistical input. Peng LN and Chen LK contributed to the clinical interpretation.

Conflict of interest

All authors declare no conflicts of interest.

Electronic Supplementary Material

Supplementary material is available for this article at https://doi.org/10.1007/s12603-022-1847-z and is accessible for authorized users.

Supplementary material, approximately 43.8 KB.