Healthcare burden and clinical outcomes of polypharmacy in older adults: a population-based cohort study in South Korea | Archives of Public Health

Data source

We utilized data from the Korea National Health and Nutrition Examination Survey (KNHANES) from 2012 to 2016, which were linked to administrative claims data from the Health Insurance Review and Assessment Service (HIRA), as well as from the National Health Insurance Service (NHIS) from 2012 to 2017. (Supplementary Fig. 1).

KNHANES [9] is a national survey conducted by the Korean Agency for Disease Control and Prevention that assesses the health and nutritional status of the Korean population through interviews and physical examination. All participants were selected based on demographic data such as gender, age, region, and type of residence in a stratified sample representative of the total Korean population [10]. The data collected include health behaviors, socioeconomic factors, clinical health indicators, and dietary habits [10].

The NHIS dataset includes beneficiary eligibility information, such as disability and mortality. Korea has a single health insurance claims database that is representative of the entire country, and claims data are generated when healthcare providers make insurance benefit claims to HIRA after providing medical care [11]. Therefore, HIRA provides representative health insurance data for Korea [11]. The combination of these datasets provides a solid foundation for assessing the health outcomes and healthcare utilization of the Korean population.

The study was approved by Inha University Hospital (2022-09-039-001). Informed consent was waived due to the study’s retrospective design.

Study population

A total of 24,900 individuals participated in the KNHANES from 2012 to 2016. We excluded 18,457 respondents under 65 years of age, 1,397 who died or were hospitalized during the baseline survey, and 980 who did not have any outpatient medications or were prescribed for less than 30 days. After excluding another 769 individuals with missing data, 3,297 older adults were included in the final analysis. Claims data (2012–2017) were linked for 3,279 individuals to define polypharmacy and outcomes, with no missing values.

Definitions and measurement

Polypharmacy

Polypharmacy was defined as more than 30 prescription days with five or more distinct prescriptions in six months (July–December, 2012–2016) [2, 12, 13]. We counted the number of drugs per respondent using the Korean national drug code from the WHO-Anatomical Therapeutic Chemical Classification System [14].

Medical costs and clinical outcomes

For medical expenses, we considered total annual per capita medical, outpatient visit, hospitalization, and medication cost. We also observed clinical outcomes, including hospitalization and mortality. The above results were observed in the year following polypharmacy measurement.

Covariates

This study included several covariates. Age was classified as 65–79 years and ≥ 80 years. To reflect the healthcare vulnerability of older adults living alone, we stratified the population into individuals living alone and those residing with others. Health lifestyles included not drinking excessively, walking more than 5 days per week, not smoking, and a body mass index (BMI; kg/m²) < 25. The Charlson Comorbidity Index (CCI) [15, 16] was classified as 0, 1, or ≥ 2, and area of residence was divided into metropolitan, urban, and rural depending on population density. Household income was divided into four quartiles: low, lower-middle, upper-middle, and high. Education level was divided into below high school and high school or higher. Health insurance was classified as either national health insurance or medical aid [2]. We also controlled disability, private health insurance and calendar year (survey year).

Statistical analysis

We performed a multivariable linear regression analysis to examine the association of polypharmacy with medication cost, and total expenses. We used a two-part model to adjust the probability of hospitalization in predicting the hospitalization expenditure: firstly, a probability model was used to estimate the probability of being hospitalized in a one-year period; secondly, the same covariates with gamma distribution and log link were added to the model to estimate the cost of hospitalization conditional on incurring positive costs [17]. In addition, we performed a multivariable logistic regression analysis to examine whether polypharmacy is a risk factor for hospitalization or death. All analyses were conducted using SAS 9.4, with statistical significance defined as a two-sided p-value of less than 0.05.