Medication Management Difficulty, Medication Nonadherence, and Risk of Hospitalization Among Cognitively Impaired Older Americans: A Nationally Representative Study

Abstract

Effective disease management in older adults relies on medication adherence to prevent adverse outcomes like hospitalization, particularly among those with cognitive impairment. In this study, we examined the impact of cognitive impairment on medication management, adherence, and hospitalization risk across levels of cognitive function. Analyzing data from 28,558 community-dwelling older adults, we found that those with dementia had the most difficulty managing medications (13.12%), followed by cognitive impairment without dementia (5.80%), and intact cognition (1.96%). Only persons with dementia showed a significant association between medication management difficulty and hospitalization risk (Odds Ratio [OR] = 1.71; 95% Confidence Intervals: 1.08, 2.70; p = .02). Cost-related medication nonadherence was associated with hospitalization risk solely among those with intact cognition (OR = 1.25; 95% CI: 1.07, 1.45; p = .004). Dementia was associated with higher odds of medication management difficulty and subsequently hospitalization risk, underscoring the need for resources to support medication use for this population.

Introduction

Nearly 95% of older adults in the U.S. have at least one chronic condition that requires ongoing medication management (National Council on Aging, 2022). Over 65% of older Americans regularly take two or more prescription medications, and over 40% of older Americans take five or more medications (Young et al., 2021).

One key to effective disease management is adherence to the prescribed medication regimen, requiring the patient and caregiver(s) to manage and administer the medications. Poor medication management and medication nonadherence, caused by health barriers (e.g., functional limitations) or socioeconomic barriers (e.g., difficulty affording medications), put one at risk for adverse outcomes, for example, accelerated functional decline (Lau et al., 2011) and hospitalization (Kuzuya et al., 2008; Sokol et al., 2005).

Cognitive impairment, such as due to Alzheimer’s disease and related dementia (ADRD), affects one’s ability to manage medications and can lead to poor medication adherence (El-Saifi et al., 2018; Smith et al., 2017). Memory and executive functioning may be the cognitive domains most relevant to medication management (Smith et al., 2017). In a sample of 38 older adults without ADRD, lower cognitive function was associated with reduced regularity in medication-taking behaviors (Austin et al., 2017). Among persons with ADRD, higher costs of dementia medications may hinder adherence to these medications (Borah et al., 2010). Yet, scarce evidence is available, especially from nationally representative studies, on how overall medication cost may affect adherence and how changes in medication management are manifested among people with different levels of cognitive levels, such as cognitive impairment before having an ADRD diagnosis. Such knowledge is critical to developing targeted interventions most responsive to the drivers of poor medication management and adherence among persons with cognitive impairment, preferably at a stage of cognitive function when such an intervention may be most influential. Moreover, gaining a deeper understanding of how costs act as a barrier to medication adherence and management could facilitate systematic efforts to enhance the affordability of medications. This, in turn, could improve adherence to treatment regimens among older adults with cognitive impairment, who often experience multimorbidity, polypharmacy, and medication burden.

This study has two objectives: 1) to examine the association between different levels of cognitive function and self-reported difficulty in medication management and medication non-adherence; and 2) to examine the association between difficulty in medication management, medication non-adherence, and hospitalization among patients with different levels of cognitive function. We hypothesized that 1) greater cognitive impairment is linked to increased difficulty in medication management and non-adherence; and 2) the effects of medication management difficulty and medication non-adherence on hospitalization risk will be greater with more cognitive impairment.

Methods

Study Design and Data Source

This study involved a secondary analysis of data from the Health and Retirement Study (HRS; conducted by University of Michigan [NIA U01AG009740]) (Sonnega et al., 2014). HRS investigators utilized multi-stage area probability design including geographical stratification, clustering, and oversampling of certain demographic groups (e.g., African Americans and Hispanics) (Sonnega et al., 2014). The HRS includes a nationally representative sample of more than 37,000 individuals ≥50 years of age from 23,000 households in the U.S. (Sonnega et al., 2014). HRS surveys, having been conducted every 2 years since 1992, collect rich information about changes in individual health and economic circumstances associated with aging (Sonnega et al., 2014). Data in this study were obtained from the 2016, 2018, and 2020 HRS public surveys (HRS, 2018). The study protocol was approved by the Institutional Review Board of the University of Rochester (# MOD00009269).

Study Sample.

We included individuals who participated in at least one HRS survey from 2016 to 2020 (n = 53,781). We excluded nursing home residents (n = 1002), participants <50 years old (n = 1543), and those whose interviews were completed by a proxy, that is, those who were not self-respondents (n = 1757), because of the large amount of missing data in important covariates among proxy respondents (e.g., depressive symptoms). Moreover, because the focus of this study is on the ability to manage medications and medication adherence, we limited the sample to persons who regularly took prescription medications, resulting in a final sample of 28,558 subjects.

Key Variables

Two HRS variables were used to assess inadequacy in medication management—medication management difficulty (due to health) and medication nonadherence (due to cost). Medication management difficulty was identified if the response was “Yes” to the question “Because of a health or memory problem, do you have any difficulty taking medications?” Cost-related medication nonadherence was identified if the response was “Yes” to the question “Have you ever taken less medication than was prescribed because of the cost in the last 2 years?”

Cognitive Function.

Cognitive function was assessed with tests adapted from the Telephone Interview for Cognitive Status (TICSm), including an immediate and delayed 10-noun free recall test, a serial 7 subtraction test, and a backward count from 20 test. The TICS is a validated instrument for screening cognitive impairment, including dementia, in older adults (Breitner et al., 1995). Based on the total TICSm score (0–27), cognitive function was categorized as “dementia” (0–6), “cognitive impairment, no dementia (CIND)” (7–11), or “intact cognition” (12–27). This approach, including the TICSm cutoff points, was validated against gold-standard ADRD diagnosis among older adults including comprehensive in-home neuropsychological and clinical assessments and expert clinician adjudication, as shown in multiple studies (Crimmins et al., 2011; Langa et al., 2005, 2017; Ofstedal et al., 2005).

Hospitalization.

Specific to each wave of HRS survey, inpatient hospitalization was identified if the patient responded “Yes” to the question “In the last two years, have you been a patient in a hospital overnight?”

Covariates.

Covariates that are potentially related to cognitive function, medication management, adherence, or hospitalization were controlled for in regression analyses, including age, gender, marital status, race/ethnicity, educational level, employment status, number of co-residing children, insurance coverage, self-rated health status, ADL limitations, depressive symptoms (Center for Epidemiologic Studies Depression score ≥4), and number of chronic conditions.

Data Analysis

Descriptive statistics were used to summarize sample characteristics as means (standard deviations [SDs]) for continuous variables and frequency (%) for categorical variables. To test hypothesis 1, we performed separate logistic regression to examine the relationship between cognitive function and medication management difficulty as well as cost-related medication nonadherence. To test hypothesis 2, we conducted logistic regression to examine the relationship between medication management difficulty, cost-related medication nonadherence, and hospitalization in the subgroups of individuals with intact cognition, CIND, and dementia. All regression models adjusted for respondent covariates, and all analyses used HRS sampling weights to adjust for nonresponse and the complex sampling design of the HRS survey; standard errors of estimates were adjusted using the Huber–White sandwich estimator (Neilsen & Seay, 2012). Statistical analyses were conducted using Stata 16.1 (College Station, TX).

Results

The sample of 28,558 HRS participants who regularly took medications had an average age of 67.4 years and were primarily white, female, married, retired Medicare beneficiaries who had a high school level of education with “good” or “very good” self-rated health. On average, each participant had 2.5 chronic conditions with hypertension, osteoarthritis, diabetes, heart disease, and psychiatric disorders being the most common. Some participants had ADL limitations and depressive symptoms (Table 1).

Table 1.

Weighted Sample Characteristics Among Participants Who Regularly Took Medications in 2020.

Variable	Entire sample	Sub-samples				p-Value*
	(N = 28,558; Weighted N = 155,891,560)	Intact cognition (75.76%)	CIND (11.62%)	Dementia (2.58%)	Missing (10.04%)
Age, mean (S.D.)	67.44 (0.19)	66.50 (0.19)	71.31 (0.30)	74.17 (0.52)	68.29 (28.70)	<0.001
Female, %	56	55.94	56.33	58.66	55.39	0.59
Race/ethnicity, %
Non-Hispanic White	75.39	77.95	56.08	45.83	86.00	<0.001
African American	10.83	9.17	21.68	26.64	6.73
Hispanic	9.20	8.37	16.28	22.71	3.82
Other	4.58	4.51	5.96	4.82	3.45
Highest educational degree, %
No degree	11.49	7.74	31.70	48.79	6.91	<0.001
GED	5.11	4.96	7.69	6.27	2.91
High school diploma	43.56	44.12	44.41	35.67	40.41
Two-year college	7.41	7.76	4.79	2.55	9.02
Four-year college	18.79	20.57	7.44	3.85	22.34
Master	10.58	11.64	3.27	2.01	13.32
Doctorate	3.06	3.22	0.70	0.86	5.09
Employment, %
Unemployed	2.67	2.56	2.50	1.98	3.88	<0.001
Employed	33.13	37.11	14.89	4.80	31.44
Retired	46.84	44.80	52.99	55.59	52.84
Homemaker	4.62	4.44	5.95	7.59	3.62
Other	12.75	11.09	23.67	30.04	8.22
Insurance coverage (Yes/No for each), %
Medicare	60.53	56.73	76.54	84.38	64.57	<0.001
Medicaid	12.11	9.64	25.97	37.21	8.27	<0.001
Private insurance	54.92	59.51	30.87	20.01	56.57	<0.001
Other	6.72	6.91	6.76	5.33	5.64	0.20
Legally married, %	60.32	62.74	46.38	37.27	64.1	<0.001
Number of children living together, mean (S.D.)	0.35 (0.01)	0.34 (0.01)	0.40 (0.02)	0.39 (0.03)	0.29 (0.02)	<0.001
Self-rated health status, %
Excellent	6.32	6.53	4.20	3.99	7.75	<0.001
Very good	30.92	32.59	17.24	15.17	38.26
Good	34.65	35.64	31.06	24.18	34.02
Fair	21.42	19.74	34.00	36.25	15.77
Poor	6.68	5.50	13.50	20.41	4.21
Having ≥1 ADL limitation, %	17.37	15.19	30.45	40.96	12.64	<0.001
Having depressive symptoms, %	15.2	13.24	25.35	31.19	14.18	<0.001
Number of pre-existing conditions, mean (S.D.)	2.45 (0.02)	2.37 (0.02)	3.00 (0.03)	3.19 (0.06)	2.28 (0.04)	<0.001
Medication difficulty, %	2.66	1.96	5.80	13.12	1.63	<0.001
Medication non-adherence, %	9.25	8.92	13.15	12.96	6.31	<0.001
Hospitalization, %	24.6	23.67	32.80	34.08	19.71	<0.001

Note. CIND, cognitive impairment no dementia.

^*p-Values based on ANOVA (continuous variable) or Chi-square (categorical variables) analyses.

The majority of the sample had intact cognition (75.8%), 11.6% had CIND, and 2.6% had dementia. Participants with CIND and dementia were older, less likely to be married, more likely to be retired Medicare beneficiaries of racial/ethnic minorities who had lower educational levels, poorer self-rated health, ADL limitations, depressive symptoms, and more chronic conditions (Table 1).

The rate of medication management difficulty was 13.12% among participants with dementia, 5.8% among those with CIND, and 1.96% among those with intact cognition (Table 1). Compared with participants with intact cognition and adjusting for covariates, those with CIND (Odds Ratio [OR] = 1.48; 95% Confidence Interval [CI]: 1.17, 1.88; p = .001) and dementia (OR = 2.91; 95% CI: 2.14, 3.97; p < .001) were more likely to have difficulty managing medications (Table 2). Other factors associated with greater odds of medication management difficulty included being Hispanic, having Medicaid insurance coverage, poorer self-rated health status, having at least one ADL limitation, more chronic conditions, and depressive symptoms (Supplementary Table S1).

Table 2.

Multilevel Mixed-Effects Logistic Regression Results Showing Odds Ratios of Having Difficulty in Medication Management and Cost-Related Medication Nonadherence, With Sampling Weights.^a

Outcome: Medication difficulty	Odds ratio	95% CI	p-Value
Dementia status (ref. = No cognitive impairment)
Dementia	2.91	(2.14, 3.97)	<0.001
CIND	1.48	(1.17, 1.88)	0.001
Missing	1.02	(0.72, 1.45)	0.89
Outcome: Medication non-adherence
Dementia status (ref. = No cognitive impairment)
Dementia	1.02	(0.77, 1.34)	0.90
CIND	1.06	(0.91, 1.24)	0.44
Missing	0.91	(0.72, 1.15)	0.43

Note. CIND, cognitive impairment no dementia.

^aBoth models adjusted for covariates.

The rate of cost-related medication nonadherence was the highest among participants with CIND (13.15%), followed by dementia (12.96%) and intact cognition (8.92%) (see Table 1). Adjusting for covariates, CIND and dementia were not significantly associated with cost-related medication nonadherence as compared to intact cognition (Table 2). Rather, being female, African Americans, having Medicare insurance coverage, being unemployed, poorer self-rated health status, having at least one ADL limitation, more chronic conditions, and depressive symptoms were associated with greater odds of cost-related medication non-adherence (Supplementary Table S2). Factors associated with lower odds of cost-related medication nonadherence included higher educational levels, having Medicaid or other insurance coverage, being married, and participating in the HRS 2018/2020 wave (vs. 2016).

Among all participants in the sample, there is evidence suggesting that medication management difficulty is associated with a higher risk of hospitalization (OR = 1.18, 95% CI: 0.97–1.44; p = .09), and that cost-related medication nonadherence was significantly associated with a higher risk of hospitalization (OR = 1.21. 95% CI: 1.07–1.37; p = .003 [Table 3]).

Table 3.

Multilevel Mixed-Effects Logistic Regression Results Showing Odds Ratios of Hospitalization, With Sampling Weights.

Hospitalization				Subgroups
	Entire sample (N = 28,558; weighted N = 155,891,560)			Intact cognition (75.76%)			CIND (11.62%)			Dementia (2.58%)
	Odds ratio	95% CI	p-Value	Odds ratio	95% CI	p-Value	Odds ratio	95% CI	p-Value	Odds ratio	95% CI	p-Value
Medication difficulty	1.18	(0.97, 1.44)	0.09	1.18	(0.90, 1.55)	0.23	1.09	(0.75, 1.58)	0.65	1.71	(1.08, 2.70)	0.02
Medication nonadherence	1.21	(1.07, 1.37)	0.003	1.25	(1.07, 1.45)	0.004	1.08	(0.82, 1.43)	0.59	1.20	(0.73, 1.97)	0.46

Subgroup analysis showed that the association between medication management difficulty, cost-related medication nonadherence, and risk of hospitalization varied across different levels of cognitive function (Table 3). Among participants with intact cognition, medication management difficulty was not significantly associated with the risk of hospitalization, yet cost-related medication nonadherence was associated with a higher hospitalization risk (OR = 1.25, 95% CI: 1.07–1.45; p = .004). Among participants with CIND, neither medication management difficulty nor cost-related medication nonadherence was significantly associated with hospitalization risk. Among participants with dementia, medication management difficulty was associated with a higher hospitalization risk (OR = 1.71, 95% CI: 1.08, 2.70, p = .02), whereas cost-related medication nonadherence was not significantly associated with hospitalization risk. Full models of hospitalization risk are shown in Supplementary Tables S3–S6.

Discussion

In this study, we used three waves of HRS data spanning six years to examine the association between different levels of cognitive function and medication management difficulty as well as cost-related medication nonadherence, and the association between difficulty in medication management, cost-related medication nonadherence, and risk of hospitalization, among individuals with different levels of cognitive function in a nationally representative sample of older persons in the U.S. This study has two principal findings. First, cognitive impairment, including CIND and dementia, was associated with greater odds of medication management difficulty, whereas cognitive impairment was not significantly associated with cost-related medication nonadherence; this partially supported our first hypothesis. Second, the association between medication management difficulty, cost-related medication nonadherence, and risk of hospitalization differed at varying levels of cognitive function. Namely, cost-related medication nonadherence was associated with higher hospitalization risk among persons with intact cognition, medication management difficulty was associated with higher hospitalization risk among persons with dementia, and neither medication management nor nonadherence was associated with hospitalization among persons with CIND. This finding partially supported our second hypothesis.

Consistent with prior evidence (Kuzuya et al., 2008), in this study, cognitive impairment was associated with increased odds of medication management difficulty, which we found to increase as the degree of cognitive impairment worsened from CIND to dementia. However, we did not find cognitive impairment, either at the level of CIND or dementia, to be associated with cost-related medication nonadherence. This may be because although dementia significantly increases medical and personal care costs (Cantarero-Prieto et al., 2020), such increases are more pronounced with moderate or severe cognitive impairment due to the need for extensive long-term care (Schaller et al., 2015). Since our sample was limited to community-dwelling HRS participants who were self-respondents (i.e., no need for a proxy respondent), the level of their cognitive impairment may be lower than what would necessitate extensive long-term care and thus may have limited bearing on socioeconomic measures, such as whether medication costs would cause one to stop taking certain medications.

Building on prior studies that reported a higher risk of hospitalization associated with medication management difficulty and nonadherence (Kuzuya et al., 2008; Sokol et al., 2005), we found in this study that these associations manifested differently across various levels of cognitive impairment. Specifically, medication management difficulty was associated with a pronounced increase in hospitalization risk among individuals with dementia. As an individual’s ADRD progresses and cognitive and physical function deteriorates, they may gradually lose the ability to self-manage medications and become prone to the deterioration of chronic conditions. Considering that individuals with ADRD are more likely to have a complex medication regimen due to their higher rates of multiple chronic conditions (Snowden et al., 2017; Zhao et al., 2008), they may be particularly at risk for hospitalization resulting from poorly managed chronic conditions.

It is surprising that cost-related medication nonadherence was only associated with hospitalization risk among persons with intact cognition and not among those with CIND or dementia. The reason for this finding is unclear, though it may be related to the fact that among persons with CIND and dementia, there are competing factors that may exert a greater impact on hospitalization than cost-related medication nonadherence, such as depressive symptoms (measured in this study) and frailty (Fried et al., 2001) (not measured in this study). It is also possible that the impact of cost-related medication nonadherence may be more pronounced among financially challenged older adults. Approximately 1 in every 5 older Americans reported cost-related medication nonadherence (Dusetzina et al., 2023; Nekui et al., 2021). Older adults with limited financial resources reported having only $16 left each week after paying for health care and basic living needs, and many even skipped basic needs to pay for medications (Briesacher et al., 2009; Nekui et al., 2021); these individuals may be subject to a higher risk of hospitalization due to cost-related medication nonadherence.

For individuals with ADRD who experience moderate and severe cognitive impairment, medication management often falls in the hands of their caregivers. However, caregivers of persons with ADRD frequently express low confidence and high stress levels when it comes to assisting with medications (Reinhard, 2019). Hence, it’s essential to provide support for both community-dwelling individuals with ADRD and their caregivers, offering them appropriate education, resources, and services to effectively manage prescribed medications. For instance, clinicians in the community, such as primary care providers and pharmacists, could incorporate regular assessments of patients’ and caregivers’ medication knowledge and awareness of social determinants of health, such as cost, during office visits. Additionally, health care professionals conducting home visits, like home health care nurses, can serve as valuable resources for medication education.

This study has limitations. First, HRS data did not include information about the diagnosis or severity of ADRD. However, the method of identifying patients with dementia and CIND in HRS has robust reliability and has been validated against gold-standard assessments and diagnosis of ADRD (Langa et al., 2017; Ofstedal et al., 2005). Second, though we used longitudinal data spanning six years, we couldn’t ascertain the causality of identified associations, such as how changes in cognitive function between two waves cause medication management difficulty and further an increase in hospitalization risk. Future studies can employ a sample of community-dwelling older adults who are free from cognitive impairment and problems with medication management at baseline, and examine if changes in cognitive impairment precede changes in medication management. Third, the measure of medication adherence in this study was cost-related. Comprehensive assessments of medication adherence, including barriers, may be useful.

Conclusions

In a nationally representative sample of Americans aged 50 and older, we investigated the impact of cognitive function on medication management and adherence, as well as the relationship between medication management, adherence, and the risk of hospitalization. Our analysis revealed that cognitive impairment was associated with an increased likelihood of experiencing difficulties in medication management, which, in turn, heightened the risk of hospitalization, particularly among individuals with dementia. These findings emphasize the critical need to provide support for community-dwelling older adults with ADRD and their caregivers who may encounter challenges with medication management and adherence. It is imperative to implement systematic efforts aimed at providing resources and training to caregivers regarding medication management, as well as ensuring access to home-based services and resources for both patients and caregivers. While our study has its limitations, including reliance on survey-based observational data and the focus on cost-related adherence, future research should aim to directly explore longitudinal changes in cognitive function and medication management while employing comprehensive adherence assessment methodologies.

What this paper adds

1. In a nationally representative sample of older adults in the U.S., cognitive impairment was found to be associated with higher odds of having difficulty managing medications, which subsequently increased the risk of hospitalization.

2. Differences were noted in the association between cognitive impairment and medication management difficulty: As the degree of cognitive impairment increased from cognitive impairment without dementia (CIND) to dementia, the odds of having medication management difficulty continued to rise.

3. Medication nonadherence was associated with a higher risk of hospitalization among older adults, and this association was only found among older persons with intact cognition, not among those with CIND or dementia.

Application of study findings

1. This study showed that cognitive impairment increases one’s risk for medication management difficulty, which subsequently was associated with a pronounced increase in the risk of hospitalization.

2. More support regarding medication management should be provided for older persons with cognitive impairment, particularly dementia, and their caregivers.