Analysis of Determinants of Medication Adherence among Older Adults with Hypertension and Diabetes using the WHO Multidimensional Model: A Study Based on Korean Medical Panel Data (KHP)
Article information
Abstract
Purpose
This study aimed to comprehensively analyze the multidimensional determinants of medication adherence among older adults in Korea using the World Health Organization (WHO) multidimensional model, which incorporates socioeconomic, health care system, condition-related, treatment-related, and patient-related dimensions, based on data from the Korea Medical Panel (KHP).
Methods
A total of 2,888 community-dwelling older adults aged 65 years and older were analyzed using the 2021 KHP dataset. Medication adherence was classified into adherence and non-adherence groups. Block-entry logistic regression analysis was conducted according to the WHO multidimensional model, and model fit was evaluated using the Hosmer-Lemeshow test and Nagelkerke R2.
Results
Among the participants, 69.4% were classified as adherent and 30.6% as non-adherent. Of the five dimensions, patient-related factors contributed most strongly to the final model (Nagelkerke R2=.045), including perceived stress, health literacy, and health-related quality of life. Condition-related factors, such as the number of chronic diseases, bedridden status, and need for care, also showed significant associations with medication adherence. Treatment-related factors, including medication side effects, and health system factors, such as unmet medical needs, were identified as additional predictors. Socioeconomic factors, including living alone and male sex, contributed modestly. Overall, the block-entry model demonstrated incremental improvements in explanatory power as each dimension was sequentially added.
Conclusion
Medication adherence among older adults is influenced by multidimensional determinants across the five WHO domains. Patient-related and condition-related factors exerted the strongest effects, followed by treatment-related, health care system, and socioeconomic factors. These findings underscore the need for comprehensive, multidimensional nursing interventions tailored to older adults to effectively improve medication adherence.
INTRODUCTION
With rapid aging, the number of older adults with chronic diseases is also rapidly increasing. According to a recent survey, 84.0% of older adults aged 65 years and over in Korea have chronic diseases, and 54.9% have three or more chronic diseases [1]. As of 2023, 63.5% of older adults in Korea have high blood pressure, and 27.2% have diabetes [2]. Hypertension and diabetes are often treated with long-term medication along with lifestyle changes. Adherence to prescribed medications is important in the management of these chronic diseases [3].
Medication non-compliance occurs when a patient does not take medicines according to the method prescribed according to the prescription of a medical institution; conversely, medication adherence occurs when a patient takes medicine appropriately according to the medication instructions [4]. An increase in the older adult population with chronic diseases may reduce patient adherence to medication due to factors such as polypharmacy and the absence of specific symptoms [5]. Medication non-adherence in patients with chronic diseases in South Korea is a serious problem that worsens clinical outcomes such as death, hospitalization, and falls, with a significant impact on healthcare costs and patient prognosis [6,7]. Medication is essential for the management of hypertension and diabetes, and non-adherence to medication can lead to serious treatment failure. Studies have shown that medication adherence in patients with diabetes is associated with cardiovascular and dyslipidemic outcomes. In addition, poor medication adherence promotes physiological aging in older adults, resulting in higher mortality rates than those in younger individuals [8].
When considering medication adherence in older adults, a comprehensive review of various factors, including cognitive function, personal beliefs, and health literacy, is essential. Medication factors, such as multiple medications, complexity of prescriptions, and experience of side effects, should be evaluated [7]. Moreover, socioeconomic factors and the quality of patient-provider interactions also play an important role in medication adherence [9,10]. The medication adherence rate of patients with chronic diseases in Korea ranges from 69.9% in patients with complex chronic diseases [11] to 93.1% in non-disabled patients with diabetes [12]. It has been reported to be as diverse as 67.9% in middle-aged and older adults [3]. When considering medication adherence in older adults, a comprehensive review of various factors, such as cognitive function, personal beliefs, health literacy, taking multiple medications, the complexity of prescriptions, and the experience of side effects, is essential. Moreover, socioeconomic factors, such as patient-provider interaction, complexity of medication prescriptions, patient-provider relationships, and socioeconomic status, play a significant role in medication adherence [5].
Although the World Health Organization (WHO) has emphasized that medication adherence to long-term care should be understood through a multidimensional framework that encompasses patients, treatments, conditions, healthcare systems, and socioeconomic factors [13], empirical studies applying this framework to the elderly population are still limited. Previous studies have relied on systematic literature reviews focusing primarily on hypertensive patients, mainly classifying adherence-related factors at a conceptual level [7,14,15]. In addition, there is a lack of empirical evidence on how multiple domains of the WHO multidimensional adherence model simultaneously affect medication adherence among older adults with chronic conditions. In order to effectively manage medication in elderly patients with chronic diseases, it is necessary to go beyond simply identifying individual variables that affect adherence, and to clarify the relative influence and contribution of each dimension proposed by WHO on medication implementation.
Therefore, this study aims to use the data of the Korean Medical Panel in 2021 to understand the actual status of medication adherence among older adults with hypertension and diabetes, and to comprehensively analyze the multidimensional determinants of medication implementation based on the WHO multidimensional model. Specifically, we will determine how the explanatory power of the model changes when socioeconomic, healthcare system, disease, treatment, and patient factors are phased in through block-entry logistic regression, and identify key factors in each dimension that increase the risk of nonadherence. Through this, the purpose is to provide basic data necessary for the preparation of customized intervention plans to improve health outcomes for the older adults.
METHODS
1. Study Design
This study is a secondary data analysis study using cross-sectional data extracted from the 2021 Korean Medical Panel Survey to identify medication adherence and multidimensional factors of prescription drugs for hypertension and diabetes in the elderly.
2. Data Collection and Study Participants
This study used the 2021 annual data (ver. 2.2) of the 2nd Korean Medical Panel, a nationally approved statistic jointly investigated by the Korea Institute of Health and Social Affairs and the National Health Insurance Service. The Korea Medical Panel has been conducting surveys since 2008 to calculate indicators related to the use of health care by the Korean people and to provide basic information on the implementation of health care policies. The second Korean Medical Panel recruited a new panel through stratified probability proportional system sampling based on the 2016 registration census and began surveying in 2020. Specifically, households living in 17 provinces and cities are classified according to stratification variables, and the sample survey area (colony) is extracted based on the number of households, and then the final sample household is systematically extracted according to the order of the household list in the sample survey area [16]. The Korea Medical Panel survey is based on general statistics approved by the National Statistical Office, and the researcher complied with the data utilization regulations and obtained consent before using it. Prior to the study, the review exemption was approved by the ‘I University Institutional Bioethics Board (IRB)’(approval number: 2025-ICCU-IRB-02).
This study analyzed drug adherence and related factors as the final analysis of 2,888 people aged 65 years or older out of a total of 14,741 household members who participated in the survey in 2021, who were taking at least one of the prescription drugs for hypertension or diabetes at the time of the survey, and who did not respond to the main variables of the study (Figure 1).
Flow chart of the study population.
3. Research Variables
In this study, the variables were divided into independent and dependent categories.
1) Dependent variables
The dependent variable of this study, medication adherence, was calculated using a question asking whether or not to take medication for hypertension prescription drugs and diabetes prescription drugs. The Korea Medical Panel survey asked subjects who responded that they were taking prescription drugs for hypertension or diabetes as of the survey date about four questions related to their medication behavior in the past year for all medications prescribed by doctors, including medications and injections. For each of the drugs for hypertension and diabetes, (1) whether the dosage was well followed, (2) whether the number of doses per day was well observed, (3) whether the dosage was kept well, and (4) whether the prescription drug was stopped arbitrarily without consulting a doctor. If the answer to the (1)∼(3) question was "I abide by it", it was considered medication adherence and was given 0 points, and if the answer was "Mostly observed", "I almost never obey", or "I do not follow at all", it was considered as medication non-adherence and 1 point was given. In addition, if the answer was "I don't know exactly (dosage/frequency/time)", it was considered that I did not follow the exact dosage method and gave 1 point. (4) Those who answered ‘none’ to the question were considered medication adherence (0 points), and those who answered ‘yes’ were considered medication non-adherence (1 point). If there was one or more non-adherence responses among the four questions related to medication adherence, and the total number of responses was 1 or more, it was classified as medication non-adherence, and if it was 0 points, it was classified as medication adherence. Medication adherence was judged based on 4 questions when taking only one type of prescription drug for hypertension and diabetes, and 8 questions were used to determine medication adherence if both hypertension and diabetes prescription drugs were taken, and medication non-adherence was classified as medication adherence if one or more questions were answered, and medication adherence was classified as medication adherence if all questions were answered as medication adherence [3]. Finally, it was coded as adherence=1 and non-adherence=0.
2) Independent variables
In this study, independent variables were divided into five dimensions according to the WHO's multidimensional adherence model to identify the influencing factors of medication adherence in older adults with hypertension or diabetes. The WHO model emphasizes that an indivi-dual's health behavior is not determined by a single factor, but is influenced by interaction between personal, social, and health care systems [15]. Therefore, in order to analyze the factors affecting medication adherence more comprehensively and systematically, the factors according to each dimension were structured as follows, taking into account their theoretical significance among previous studies and panel data items [14,17].
First, in this study, the social and economic factors included in the panel data consist of gender, age, living alone, education level (below elementary school/middle school/high school/college or higher), household equalized income, economic activity status, and private insurance status. Household income was calculated by calculating the equalized household income divided by the number of household members squared, and then classified into the fourth quartile based on the entire survey subjects of the Korean Medical Panel [3].
Second, the health care system factors consisted of the type of national health insurance, whether there were commercial treatment centers, and whether there was unmet medical care.
Third, disease-related factors were classified into the number of chronic diseases (1, 2∼3, 4 or more), pre-existing conditions, care needs, disability, and subjective health status (good, moderate, poor).
Fourth, treatment-related factors were divided into whether or not they had experienced adverse drug reactions, body mass index, and time of first diagnosis (10∼20s, 30∼ 40s, 50∼65 years old, and 65 years or older).
Fifth, patient-related factors consisted of regular exercise, current smoking status, current alcohol consumption, perceived stress, mental health, health-related quality of life, and health literacy. Perceived stress is the question "How much stress do you feel in your daily life?" It was measured using a single question. Responses were surveyed on a 4-point scale from ‘very much’ to ‘almost nothing’. Health-related Quality of Life (HRQoL) is an EQ-5D (EuroQoL-5 Dimension) tool developed by the EuroQoL group, which is a 3-point scale (no hindrance, somewhat disturbing, and severely disturbing) for each of the five domains of exercise capacity, self-management, daily activities, pain/discomfort, and anxiety/depression, and the value of the EQ-5D Index score was converted into the EQ-5D Index score by applying the Korean weighted model (Time Trade-Off method) presented by the Korea Centers for Disease Control and Prevention. The closer the score to 1, the higher the quality of life, which was put into the analysis. Health Literacy is a shortened three-question (The Brief Health Literacy Screen, BHLS) used by Korean medical panels, and includes questions such as "Is it difficult to read and understand printed materials or notices received from hospitals or pharmacies on your own?" and "Do you have difficulty filling out hospital documents (e.g., medical appointments, consent forms, etc.)?" Each item was measured on a 5-point Likert scale from ‘always (1 point)’ to ‘not at all (5 points)’, and was categorized into three groups based on the sum score of the three items (3∼15 points): Inadequate (3∼9 points), Marginal (10∼11 points), and Appropriate (Adequate, 12∼ 15 points).
4. Data Analysis
The data collected in this study were analyzed using IBM SPSS/WINdows 26.0 Program, and the specific data analysis method is as follows.
Frequencies and percentages were calculated to identify the general characteristics of the subjects.
The difference in medication adherence according to general characteristics was analyzed using the chi-square test (x2).
In order to reflect the multidimensional framework in the model as it is, the fit of the model was evaluated by the Hosmer-Lemeshow test and Nagelkerke R2.
According to the WHO multidimensional factor model, block entry logistic regression analysis was performed in which variables were injected step by block. The dependent variable of this study, drug adherence, was treated as a dummy variable with ‘1’ for the adherence group and ‘0’ for the non-adherent group. In logistic regression analysis, if the odds ratio (OR) is less than 1 (OR < 1), it means that there is a high probability of lower drug adherence compared to the reference group, and if it is greater than 1 (OR >1), it means that there is a high probability of higher drug adherence.
Although the primary analyses focused on estimating conditional associations between medication adherence and related factors using unweighted logistic regression, the Korea Health Panel Study (KHP) employs a complex survey design incorporating stratification, clustering, and sampling weights to ensure national representativeness. To assess the robustness of the main findings and address potential bias arising from ignoring the survey design, a sensitivity analysis was conducted using weighted logistic regression.
Specifically, the individual sampling weight provided by the KHP (I_WGC) was applied to re-estimate the final multivariable model. The same set of covariates and blockwise modeling strategy used in the primary analysis was retained to facilitate direct comparison. Odds ratios (ORs), 95% confidence intervals (CIs), and p-values obtained from the weighted model are presented in Supplementary Table 1.
The results of the weighted analysis were compared with those of the unweighted primary model in terms of the direction, magnitude, and statistical significance of associations. Overall, the direction and relative effect sizes of key predictors were largely consistent across weighted and unweighted models, supporting the robustness of the main findings. Differences in statistical significance observed in the weighted analysis were interpreted as reflecting reduced standard errors due to the application of sampling weights rather than substantive changes in the underlying associations.
RESULTS
1. Participants' Characteristics
A total of 2,888 subjects were included in this study, with 43.0%(n=1,241) males and 57.0%(n=1,647) females. The drug adherence group of the study subjects was 69.4% (n=2,003) and the non-adherent group was 30.6%(n=885). The most common age group was 65∼74 years old at 49.7% (n=1,436). ‘living alone’ was 25.0%(n=723) and 74.7% (n= 2,157) were living together, and the education level was ‘elementary school graduated or no school education’ was the most common at 48.5%(n=1,400). In terms of "annual household income", the "lowest group (Q1)" accounted for the most at 93.0% (n=2,686), and 47.1% (n=1,360) of the respondents were engaged in economic activities. In the National Health Security type, 93.5% (n=2,699) of the respondents had National Health Insurance (NHI) subscribers, and 12.7%(n=366) of the respondents said they had a usual source of care. The group with 2∼3 chronic diseases was the most common at 56.7% (n=1,638), and the subjects were bedridden at 18.5% (n=535). 6.1% (n=175) of the respondents responded that they needed care from others, and 14.1% (n=408) had disability. The most common respondents had a subjective health status of ‘Moderate’ at 41.8% (n=1,209). 2.7% (n=78) of the subjects responded that they had experienced drug side effects. When classified according to BMI, the normal weight group was 32.2% (n=931), the overweight group was 29.1% (n=839), and the obesity group was 36.1% (n=1,044). In the past year, 56.6% (n=1,634) of the subjects did regular exercise, 71.6% (n=2,066) of smokers, and 58.2% (n=1,679) of abstainers. As for the stress level, ‘a lot’ was the most common at 48.4% (n=1,398), and 8.4% (n=243) responded that they had psychological distress. The largest number of respondents said that health literacy was inadequate, with 57.0% (n=1,645).
2. Differences between Medication Adherence and Non-adherence Groups according to General Characteristics
The variables that showed statistically significant differences in medication adherence were status of living alone (p=.010), presence or absence of usual source of care (p=.002), number of chronic diseases (p=.001), bedridden status (p<.001), need care from others (p=.001), subjective health status (p=.028), experience of drug side effects (p= .001), regular exercise (p=.044), stress perception level (p= .001), and health literacy (p=.003).
Among socioeconomic factors, there was a significant difference between living alone and medication adherence (p =.010). The non-adherence rate of the elderly living alone was 34.6%, which was higher than that of the elderly living together (29.4%). Other socioeconomic factors, such as gender, age, education level, and household income, were found to have no significant differences with medication adherence.
As for factors related to health status and disease, the non-adherence rate was 36.5% in the group with 4 or more chronic diseases, which was higher than that of the group with a small number of diseases (1: 28.3%) (p=.001). In addition, the non-adherent group had a rate of 38.5% and 41.7%, respectively, which was significantly higher than that of the adherent group (28.9% and 29.9%) (p<.001). At the same time, the percentage of non-adherent groups who experienced adverse drug reactions was 47.4%, which was higher than that of the group that did not experience adverse reactions (30.2%) (p=.001). The non-adherent group had a lower rate of non-adherence (p=.044) than the group that did not exercise regularly (29.1%) (32.6%). When the perceived stress level was ‘almost none (26.0%)’ group, the proportion of non-adherence was higher in the group that responded ‘Very much (33.8%)’ or ‘A lot (36.1%)’ (p=.001). The proportion of non-adherence in the ‘adequate’ group was the lowest at 24.8%, and the proportion of non-adherence in the ‘inadequate’ group was the highest at 32.5% (p=.003).
General Characteristics and Differences between Medication Adherence and Non-adherence Groups (N=2,888)
3. Analyze the Fit of the Model
In order to evaluate the relative contribution of factors step by step and reflect the multidimensional framework explaining medication adherence in the model, block entry logistic regression was carried out reflecting the WHO multifaceted model. As a result of the diagnosis of the regression model, the VIF of each independent variable was in the range of 1.02∼2.45, and the tolerance was above 0.43, which met the generally recommended criteria (VIF<10, Tolerance >0.1), and it was confirmed that there was no problem of multicollinearity in this model.
As a result of logistic regression analysis, it was found that the explanatory power of the model gradually improved as the steps were added. First, the model with demographic sociological factors (block 1) was not statistically significant (Δ-2LL=14.66, p =.261), and the Nagelkerke R2=.008 was less explanatory. The model with the addition of health system and service factors (block 2) was marginally significant at the significance level (Δ-2LL= 10.45, p=.048), and the Nagelkerke R2 increased slightly to .013. The model with disease-related factors (block 3) improved significantly (Δ-2LL=25.93, p <.001), and the explanatory power also increased to Nagelkerke R2=.026. Even when treatment-related factors (block 4) were added, the model showed significant changes (Δ-2LL=7.82, p=.002), and the explanatory power increased to .030. The final model, including patient-related factors (block 5), showed a statistically significant improvement (Δ-2LL= 30.65, p=.001), and the explanatory power was the highest at Nagelkerke R2=.045. In addition, the results of the Hosmer-Lemeshow test showed that the significance probability was more than .05 at all stages (Block 5: x2=6.11, p=.635), and statistically confirmed that the discrepancy between the model and the observational data was not significant. However, since this test can be sensitive in a large sample, it was used only as a supplementary indicator to support model suitability (Table 2).
Logistic Regression Model Fit Indicators by Block
4. Influence and Contribution by Variables
Logistic regression analysis showed that the factor that contributed most to medication adherence was "patient-related factors (block 5)". This block showed the greatest improvement in model explanatory power (R2), and that variables such as stress perception, health-related quality of life, and health literacy level had an important impact on medication adherence. At the perceived stress level, the perceived ‘Very much stress’ group had lower medication adherence compared to the ‘almost none’ group (OR=0.67, 95% CI=0.51∼0.86, p=.002), and the ‘A lot’ group was also low (OR=0.74, 95% CI=0.61∼0.90, p=.003). Medication adherence increased with higher health-related quality of life (OR=1.44, 95% CI=1.01∼2.04, p=.042). The group with inadequate health literacy had lower medication adherence compared to the adequate group (OR=0.71, 95% CI=0.55∼ 0.92, p=.010).
The second important factor was ‘disease-related factors (block 3)’, which was analyzed that the number of chronic diseases, whether or not they had underlying diseases, and whether they needed care from others had a significant impact on medication adherence. In terms of the number of chronic diseases, the group with 2∼3 had higher medication adherence than the group with 4 or more (OR=1.26, 95% CI=1.02∼1.56, p=.029). In the case of bedridden, medication adherence was lower than that of non-bedridden patients (OR=0.73, 95% CI=0.59∼0.92, p= .007). Medication adherence was lower with or without care needs compared to those without it (OR=0.65, 95% CI=0.44∼0.95, p=.027).
The third is ‘health care system & team-related factors (block 2)’, in which unmet medical need (OR=0.77, 95% CI=0.59∼0.99, p=.039) was identified as an important variable that negatively affects medication adherence.
Next, ‘treatment-related factors (block 4)’ showed that the experience of drug side effects was a factor that reduced medication adherence. Medication adherence was lower in those who had experienced drug side effects than in those who had not experienced side effects (OR=0.60, 95% CI=0.37∼0.98, p=.042).
Finally, ‘socioeconomic factors (block 1)’ did not show a statistically significant model improvement, but some variables were shown to influence medication adherence. For gender, men had lower medication adherence compared to women (OR=0.76, 95% CI=0.59∼0.98, p =.035). Older adults living alone had lower medication adherence compared to cohabiting elderly (OR=0.77, 95% CI=0.62∼0.95, p=.015) (Table 3).
Logistic Regression of Factors Associated with Medication Adherence (N=2,888)
DISCUSSION
This study aimed to explore the factors affecting medication adherence in a multidimensional manner, based on the multifaceted medication adherence model proposed by the WHO [15]. Since medication adherence is not determined by a single factor but by a combination of socioeconomic conditions, healthcare systems, disease characteristics, treatment factors, and individual patient factors, it is necessary to analyze them separately and step by step. Therefore, in this study, a block entry method was used, in which each of the five-factor dimensions was set as a block and sequentially inserted by reflecting the WHO's multifaceted evaluation framework to construct a logistic regression model. This approach leads to an understanding of medication adherence as a multidimensional outcome rather than piecemeal behavior. It further provides a theoretical and empirical basis for formulating strategies for managing chronic diseases in older adults. This model construction method has the advantage of, first, allowing the relative influence of each factor dimension on medication adherence to be independently identified. Second, the contribution of each factor can be evaluated through the change in explanatory power (R2) with each added block. Furthermore, by confirming the cumulative and interactive effects of the factors explaining medication adherence, we aimed to provide a basis for the development of multidimensional intervention strategies.
In this study, medication adherence in patients with hypertension and diabetes was 69.4%, which was higher than a previous study of 67.9% in middle-aged and older adults with chronic diseases using the same medical panel data [3], but similar to or lower than 69.9∼93.1% based on medical panel data in 2018 [12,20]. Although it is difficult to directly compare these results due to differences in the types of chronic diseases and measurement methods, this study tracked a total of eight dosing methods related to the two prescription drugs and classified them into medication adherence groups; therefore, the criteria for medication adherence may have been higher than those in previous studies. In addition, since whether the subjects responded on their own became the criterion for judging medication adherence, it is possible that the results may have been overestimated or underestimated according to the subjects’ judgment criteria. Longitudinal studies are needed to compare outcomes by adding new medical panel data to be collected in the future.
The analysis revealed that the socioeconomic, healthcare system, and disease-, treatment-, and patient-related factors of older adults with chronic diseases gradually improved the explanatory power of the model, especially disease- and patient-related factors Although the Nagelkerke R2 in the final model was relatively low, this is common in behavioral outcomes such as medication adherence, where multiple psychosocial factors interact and individual differences are high. Previous studies using large survey data on adherence have also reported similarly low pseudo-R2 values [6], Thus, effect directions, theoretical coherence, and incremental improvements across blocks were also considered in evaluating model adequacy. In conclusion, the psychological and cognitive characteristics of patients, direct health status, and disease-related factors had the greatest impact on medication adherence. It is necessary to prepare customized intervention plans that consider these factors.
First, in terms of socioeconomic factors, men and older adults living alone showed low medication adherence. This is partly consistent with the results of domestic and international studies showing that living alone and male sex are associated with low medication adherence in adults, including older adults [3]. However, the independent explanatory power of socioeconomic factors in this study was limited, which is in line with the WHO [15]'s point that medication adherence is difficult to explain as a single factor and must be understood in interactions with other factors. Nevertheless, in this study, the non-compliance rate of older adults living alone was 34.6%. This was higher than that of non-single older adults (29.4%), and the effects of solitary living were continuously observed even after controlling for other factors. This means that if one lives alone, they may have difficulty managing their medication because there is no one to take care of it, suggesting that strengthening social support networks and companion-based medication support programs are important strategies. Unlike this study, which showed that men were associated with lower adherence in terms of sex, some previous studies reported lower adherence among women, indicating that sex effects vary depending on the context of the study and measurement indicators [18,19]. These differences are likely due to structural differences in the measurement methods, medication complexity, and social support networks. Therefore, future studies need to evaluate consistency through verification of the interaction between sex and related factors and the expansion of other big data-based analyses, such as claims data.
In the healthcare system, unmet medical needs are significantly associated with poor medication adherence. This is consistent with previous studies that showed that when medical services are not provided on time due to lack of access to medical care, patients’ willingness and persistence to take medication decreases, resulting in lower medication adherence [20]. Although access to healthcare is relatively high among older Koreans, unmet needs due to social and economic constraints negatively impact medication adherence [21]. Unmet medical experiences, such as accessibility to the healthcare system and lack of communication with medical staff, have been reported as major factors that reduce medication adherence [9]. This study showed that the impact of this unmet experience was significant even after considering other dimensions, which supports the idea that strengthening the patient-centered continuum of the care system, the system of primary care physicians, and the policy of supporting medication costs can contribute to improving medication adherence.
Disease-related factors contributed significantly to the improvement of the model, and the number of chronic diseases, disease status, and the need for care had a significant impact on medication adherence. The non-adherence rate in the older adult group with four or more chronic diseases was 36.5%, which was significantly higher than that in the group with a small number of diseases (28.3%), which can be interpreted as the number of medications to be taken increasing and management becoming more difficult. This is similar to the results of previous studies showing that medication adherence decreases due to functional decline (sick condition and need for care) [22]. However, regression analysis showed that medication adherence in the elderly with chronic diseases with the number of 4 or more chronic diseases was lower than that of the group with 2∼3 chronic diseases. This shows that when the number of diseases exceeds a certain level, the increase in the number of drugs to be taken and the complicated medication schedule can actually act as a burden on elderly patients and reduce adherence. In other words, it suggests that moderate disease burden may rather enhance the motivation of older adults to self-care [10]. But caution is required when interpreting this result. The low adherence rate in older adults with multimorbidity despite frequent outpatient visits and many opportunities for monitoring by medical staff may have been caused by psychological and physical fatigue caused by the severity of the disease or the composition of the complex disease (comorbidities other than hypertension/diabetes). In future studies, it is necessary to clarify this causal relationship through analysis that specifically controls the number and ingredients of prescribed drugs. In addition, functional decline, such as chronic diseases and care needs that are commonly experienced by older adults, limits their ability to perform medication, leading to poor medication adherence. These results indicated that simplified medication management and care support are essential for older adult patients undergoing polypharmacy in clinical practice [23]. Previous studies at home and abroad have also shown that older adults with chronic diseases with physical activity restrictions have difficulty taking medication on time, and their overall self-management ability for medication management is reduced, resulting in lower medication adherence; the results of this study are supported by reporting that functional status is important for medication adherence [22,24].
In terms of treatment-related factors, adverse drug reactions were identified as major factors in poor medication adherence. This finding reaffirms that drug reactions are representative determinants of poor medication adherence, as reported in a domestic study [20]. In addition, a previous study in middle-aged and older adults [3] reported that fear or actual experience of drug side effects in patients with chronic diseases led to discontinuation of medication or arbitrary dose adjustment, reducing adherence. In this context, the results of this study are consistent with those of previous studies in Korea, and the experience of side effects supports the idea that it acts as a barrier to medication adherence in older adult patients. However, a previous study [5] of older adult patients with diabetes showed that the experience of adverse drug side effects did not have a significant effect on medication adherence, which differed from the results of this study. This difference may be due to the characteristics and contextual factors of the study participants. In other words, in a study of a group of older adult diabetic patients, these groups may have accumulated experience in tolerance or coping with side effects through long-term medication treatment and continuous self-management education due to the nature of the disease. However, in this study, it is possible that the type of medication and complexity of taking were relatively high in older adults with complex chronic diseases, by expanding to older adults with hypertension, and that the complexity and vulnerability of drugs in older adults with multiple diseases may have acted as a factor that aggravated the decline in medication adherence compared to the single-disease population. This emphasizes the need for systematic interventions in adverse event management and the use of combination drugs rather than disease-specific drug education to promote medication adherence among older adult patients in clinical settings.
Finally, patient-related factors contributed the most to the improvement in the explanatory power of the final model. This showed a relatively large improvement in explanatory power compared to the addition of other factor groups, but the absolute size of the actual increase was not large. Therefore, it is clear that patient-related factors have a significant impact on medication implementation, but it is necessary to understand them within the combined action with other factors rather than interpreting them as isolated determinants. High stress levels and poor health literacy in older adults are associated with poor medication adherence, which increases in patients with higher health-related quality of life [25]. The negative association between stress levels and medication adherence identified in this study is consistent with the findings of previous studies. An Iranian population-based study [26] reported a significant reduction in the number of patients with diabetes and hypertension with high stress levels. In addition, a study of older adult patients with hypertension by Rajpura & Nayak [27] found that patients reported the most stress as the main cause of hypertension. It is recognized that the higher the disease burden and stress level, the lower the medication adherence. These results suggest that stress is a psychological risk factor that negatively affects self-management behaviors in patients with chronic diseases. For older adults, the burden of managing multiple chronic diseases simultaneously and the stress caused by the constraints of daily life can hinder the regular practice of taking medication.
The results of this study show that a lack of health literacy is associated with poor medication adherence, which is partially consistent with the findings of previous studies. However, a systematic review by Loke et al. [28] found that only one of seven studies reported a meaningful association between health literacy and medication re-prescription implementation, indicating that health literacy improvement interventions did not significantly increase medication adherence. In contrast, a study of older adult Korean diabetic patients [29] found that health literacy was associated with self-efficacy, which showed a static correlation with self-management and a negative correlation with glycated hemoglobin levels, confirming the positive effect of health literacy. These conflicting results can be attributed to the cultural backgrounds of the study participants, differences in the health system, and differences in health literacy measurement tools.
In this study, the results showed that a higher health-related quality of life was associated with increased medication adherence, which is consistent with several previous studies. A study of patients with diabetes [25] confirmed that those with higher medication adherence had a significantly higher health-related quality of life, and a study of patients with diabetes and hypertension [30] reported a significant positive correlation between medication adherence and quality of life. This suggests a two-way relationship between medication adherence and quality of life, with patients with a high quality of life perceiving their health status positively and increasing their motivation for treatment. Conversely, patients with good medication adherence have better disease management and improved quality of life.
The effect of patient-related factors on medication adherence identified in this study suggests the need for a multidimensional approach to improve medication adherence in older adults with chronic diseases. Stress management, health literacy improvement, and quality of life improvement should be considered in an integrated manner, and customized intervention strategies that reflect the cultural characteristics and medical environments of Korean older adults are required. Future studies are needed to clarify the interaction effects between these factors and the changes over time for developing more effective medication adherence improvement programs.
This study had some limitations. First, the explanatory power (Nagelkerke R2) shown by the final model of this study (Block 5) was numerically low at .045. However, this does not imply the inadequacy of the model itself, but rather because the behavioral variable of medication adherence is compoundly influenced by various micro factors that are difficult to capture with secondary panel data, such as individual beliefs, momentary emotional states, and daily micro-habits, in addition to the demographic and clinical factors included in the study. The Hosmer-Lemeshow test, which statistically determines the fit of the model, showed a significance probability (p-value) of .635 (p>.05), confirming that the model in this study adequately reflects the observed data. Therefore, the value of this study lies in identifying key risk factors—such as living alone, stress, and health literacy—to inform policy priorities, rather than in maximizing predictive precision for individual behavior. Second, although sampling weights were available in the Korea Health Panel, they were not applied because the analysis prioritized examining conditional associations between variables over estimating population prevalence. While this approach is suitable for identifying risk factors, not accounting for the complex sample design may lead to underestimated standard errors and limited generalizability to the entire Korean older adult population. Thus, the statistical significance reported in this study should be interpreted with these methodological limitations in mind. Third, owing to the design of the cross-sectional survey study, it was difficult to determine the causal relationship between medication adherence and the influencing factors. Considering the possibility of a bidirectional relationship between HRQoL and medication adherence, it is necessary to confirm this temporal sequence through longitudinal studies. Forth, the strict binary classification, where a single negative response defined non-adherence, limits distinguishing between intentional and unintentional behaviors. This broad categorization likely contributed to the model's relatively low explanatory power. Fifth, this study integrates hypertension, diabetes alone, and complex morbidity groups, and has limitations in not accurately reflecting the complexity differences or interaction effects of drug therapy according to disease type. In the process of analysis, the number of chronic diseases was put into the control variable to correct the disease burden, but the possibility of potential classification bias cannot be ruled out because dummy variables for each disease group or stratified analysis were not performed. Therefore, future studies need to try to identify transitional factors specific to each disease group through subgroup analysis by disease type.
CONCLUSION
This study systematically analyzed the influencing factors of medication adherence in Korean elderly based on the WHO multidimensional model, and confirmed that patient-related factors (stress, health literacy, quality of life) and disease-related factors (number of chronic diseases, functional status) were the most important predictors. In particular, the risk of medication non-adherence was high in older adults with high stress levels and lack of health literacy, as well as in elderly people with chronic illnesses or care needs. These results suggest that a multidimensional approach is needed to improve medication adherence in the elderly, including psychological support, health literacy, and functional status improvement, beyond simple medication education. In addition, special attention and support are required for the elderly living alone and those with unmet medical experience. In future studies, it is necessary to clarify the interaction effect between these factors, clarify the causal relationship through longitudinal studies, and use objective medication adherence measurement methods. The results of this study are expected to provide important basic data for the development of customized intervention programs and policy formulation to improve medication adherence in elderly patients with chronic diseases.
Notes
CONFLICTS OF INTEREST
Dongwon Choi is an editorial board member of the Journal of Fundamental Nursing Science. She was not involved in the review process of this manuscript. Otherwise, there was no conflict of interest.
AUTHORSHIP
Study conception and design acquisition - Choi D; Data collection - Choi D; Data analysis & Interpretation - Choi D; Drafting & Revision of the manuscript - Choi D.
DATA AVAILABILITY
Please contact the corresponding author for data availability