How Decades of Lifestyle Patterns Shape Healthcare Costs in Korea

What if the biggest driver of healthcare spending is not a single bad habit, but the long-term pattern formed by several habits interacting over decades? A new study drawing on nearly two decades of national insurance data from South Korea suggests exactly that. Researchers analyzed 167,343 Korean adults aged 40 to 79 years, tracking smoking, alcohol consumption, physical activity, and body mass index from 2002 to 2009, and then following their healthcare utilization and expenditure through 2019. The findings, published in the International Journal of Public Health, reveal that the way these four behaviors evolve together over time is a powerful and sex-specific predictor of how much healthcare people use and how much it costs.

The study exploited a uniquely rich data resource: the National Health Insurance Service Health Screening Cohort, known as NHIS-HEALS. Korea operates a universal single-payer system that covered approximately 97 percent of the population as early as 1989, and enrollment in the system mandates biennial health screenings that record anthropometry, laboratory markers, and self-reported behaviors. Because the cohort is built from administrative claims rather than surveys, it avoids the recall bias and measurement error that plague self-reported data. The researchers selected a 10 percent random sample of adults aged 40 to 79 who underwent national health screening in 2002 and 2003, then applied strict exclusions to remove former smokers, people with baseline chronic diseases such as hypertension, diabetes, stroke, or cancer, and anyone with missing behavioral data across the trajectory estimation waves.

To capture the joint evolution of the four behaviors, the team used group-based multi-trajectory modeling, a statistical technique designed to identify qualitatively distinct subgroups within a population rather than describing average change around a single mean. Smoking was categorized by cigarettes per day, alcohol intake by grams per day, physical activity by weekly frequency of moderate-to-vigorous exercise, and BMI by Korean obesity guidelines. Model fit was assessed using the Akaike Information Criterion, the Bayesian Information Criterion, entropy, and average posterior probability, and a six-group solution emerged as optimal for both males and females. This approach matters because single-behavior analyses fail to capture the clustering and interactive effects of co-occurring risk behaviors, which is precisely where much of the health burden hides.

The six trajectories told a striking story. Among the 95,226 men, roughly 30 percent maintained a stable healthy pattern of non-smoking, low alcohol intake, increasing physical activity, and normal BMI, while 28 percent showed progressive improvement across all behaviors. Two moderate-risk groups, together comprising about 40 percent, sustained smoking, moderate drinking, and low physical activity, differing mainly in whether they carried overweight or normal BMI. Two small severe-risk groups, each around 1 percent, combined light smoking, minimal activity, and either persistent underweight or severe obesity. Among the 72,117 women, the picture differed: nearly 35 percent were stably healthy, and the moderate-risk groups were defined primarily by overweight or obese BMI rather than by smoking or drinking, since most women were nonsmokers and nondrinkers.

The associations with healthcare outcomes were assessed using generalized estimating equations stratified by sex, with adjustments for age, region, insurance type, income, disability status, and the Charlson Comorbidity Index. For inpatient expenditure and length of stay, which showed substantial zero-inflation with roughly 80 percent of observations at zero, the researchers applied a two-part modeling framework: a logistic model for the probability of any hospitalization, and a gamma model with a log link for the magnitude among those hospitalized. Outpatient expenditure and visit counts, with fewer than 0.3 percent zero observations, required only single-part models. An autoregressive order 1 correlation structure was selected because healthcare spending in a given year correlates most strongly with the immediately preceding year.

The results were dramatic and sex-specific. Among men, the higher severe-risk trajectory was associated with outpatient expenditure 12.5 percent higher and inpatient expenditure 60.6 percent higher than the stable healthy group, along with 4 percent more outpatient visits and 22 percent longer hospital stays. But the most intriguing finding concerned the higher moderate-risk group: these men, despite normal BMI, younger age, and lower comorbidity, showed 3.7 percent fewer outpatient visits yet 24.4 percent longer hospital stays. This pattern echoes what epidemiologists call the lean paradox, in which normal-weight individuals with poor smoking, drinking, and activity profiles visit outpatient care less but hospitalize more, possibly because they engage in less preventive care and present with more advanced disease when they finally do.

Among women, the higher severe-risk trajectory, affecting less than 1 percent of the cohort, was linked to outpatient expenditure 14.8 percent higher, inpatient expenditure 74.7 percent higher, 6 percent more outpatient visits, and 45.4 percent longer hospital stays. Smoking lowers estrogen levels, and women show greater physiological susceptibility to alcohol-related harm, which may partly explain why adverse behaviors carry such heavy costs in this group. More paradoxically, the female decreasing-risk trajectory, representing women who improved their behaviors over time, showed higher expenditure and utilization than the stable healthy reference group, unlike their male counterparts. The authors propose several mechanisms: health-conscious women may simultaneously adopt broader health-promoting behaviors that elevate short-term utilization, the benefits of behavioral improvement may require a longer time horizon than the 11-year follow-up to manifest as cost savings, and estrogen-mediated protection in younger women may attenuate observable benefits while its withdrawal after menopause amplifies residual damage from prior unhealthy behaviors.

Age-stratified analyses added another layer of nuance. Among men under 65, severe-risk trajectories showed the strongest associations with inpatient expenditure and length of stay, likely reflecting early disease onset and intensive treatment needs. Among men 65 and older, moderate-risk trajectories dominated, consistent with the cumulative damage of persistent smoking and heavy drinking in later life worsening comorbidities and hospitalization rates. Among women, the higher severe-risk trajectory was strongly associated with higher inpatient expenditure and longer stays in both age groups, with coefficients of 0.624 in the younger group and 0.477 in the older group. These patterns suggest that the behavioral drivers of healthcare demand shift across the life course, and that prevention strategies must be tailored accordingly.

The stakes are enormous. Non-communicable diseases account for approximately 71 trillion Korean won annually in Korea, nearly 85 percent of total healthcare costs, and Korean healthcare expenditure has been growing at the fastest rate among OECD countries. The socioeconomic losses attributable to smoking, high-risk drinking, physical inactivity, and obesity reached approximately 42 trillion won in 2019. Out-of-pocket spending remains above the OECD average, and the long-term financial sustainability of the health system is under pressure. The authors argue that community-level health policies in Korea have focused primarily on interventions after disease onset, and that adopting prevention-oriented approaches, as demonstrated in Japan where community-based health promotion programs reduced healthcare expenditure, could enhance the effectiveness of preventive policy.

The study has limitations worth noting. Substantial sample exclusions, including 258,970 individuals removed for missing data or death and 83,918 for baseline chronic conditions, likely produced a healthier-than-average analytic sample, and excluded participants differed significantly in age, income, and disability status. Unmeasured confounders such as diet, sleep, education, occupation, and health literacy may have influenced both behaviors and utilization. Physical activity was measured only by weekly frequency, without duration or intensity, and pharmacy costs were excluded because Korean claims data do not disaggregate pharmaceutical expenditure by care setting, which may underestimate the total financial burden of unhealthy trajectories. Nevertheless, the core message stands: healthcare demand is shaped by long-term, multidimensional behavioral patterns rather than by single risk factors, and preventive interventions should address not only high-risk groups but also individuals with normal BMI who continue to smoke, drink, or remain inactive. Strengthening community-based prevention may be the key to reducing the healthcare burden and sustaining Korea’s health system for decades to come.

Subject of Research: Health

Article Title: Joint trajectories of health behaviors and their associations with healthcare utilization and expenditure

Article References: Joint trajectories of health behaviors and their associations with healthcare utilization and expenditure. (n.d.). https://doi.org/10.3389/ijph.2026.1609686

Image Credits: AI Generated

DOI: 10.3389/ijph.2026.1609686

Keywords: health behavior, healthcare expenditure, South Korea, NHIS-HEALS, trajectory modeling, public health, preventive medicine, Joint, trajectories, health, behaviors, associations