The study, conducted by Tak Kyu Oh and In-Ae Song of Seoul National University Bundang Hospital, exploited one of the world’s most complete health data resources, the National Health Insurance Service database, which covers virtually the entire South Korean population. The researchers asked a deceptively simple question: among adults aged 18 and older, did having at least one documented diagnosis of a selected psychiatric disorder in 2021 predict greater healthcare utilization and higher medical costs in the following year? By anchoring the exposure and all baseline characteristics in 2021 and tracking outcomes in 2022, the design established a clear temporal sequence, with the psychiatric diagnosis recorded before the utilization and cost outcomes were measured.

The psychiatric conditions under scrutiny spanned ten diagnoses, a deliberately broad selection intended to capture the major categories of mental illness seen in routine clinical practice. Because the analysis relied on the International Classification of Diseases, 10th Revision codes recorded in insurance claims, a documented diagnosis meant that a clinician had formally identified the disorder and that the encounter had entered the reimbursement system. This administrative definition has both strengths and limitations: it captures conditions serious enough to prompt care and coding, but it may miss individuals who never seek treatment, a well-known blind spot in psychiatric epidemiology.

To isolate the effect of psychiatric diagnosis from the many other factors that drive healthcare use, the investigators turned to propensity score matching, a statistical technique that pairs each person with a psychiatric disorder to a person without one who shares an essentially identical profile of baseline characteristics. Age, sex, income level, comorbid physical illnesses measured with the Charlson Comorbidity Index, and other covariates all assessed in 2021 fed into the matching algorithm. After 1:1 matching, the analytic cohort comprised 567,884 individuals, split evenly into 283,942 adults with documented psychiatric disorders and 283,942 without. The balance achieved by matching was checked using absolute standardized mean differences, the standard diagnostic for confirming that the two groups are comparable on observed variables.

The results were striking across every outcome the team examined. In the matched cohort, 11.9 percent of adults with psychiatric disorders made at least one emergency department visit in 2022, compared with 8.6 percent of those without such diagnoses, a risk ratio of 1.38 with a 95 percent confidence interval of 1.35 to 1.41. Hospitalization told an even starker story: 8.8 percent of the psychiatric disorder group were admitted to hospital during the year, versus 4.4 percent of the comparison group, a risk ratio of 2.01. In other words, having a documented psychiatric disorder roughly doubled the odds of an inpatient stay.

The most dramatic disparities emerged for psychiatric-related care. Emergency department visits tied to psychiatric conditions occurred at a risk ratio of 43.85 between the two groups, and psychiatric-related hospitalizations at a risk ratio of 40.92, both with confidence intervals that excluded any possibility of chance explaining the gap. While the absolute numbers remained small, because psychiatric-related events are rare even in the exposed group, the magnitude of the relative differences underscores how concentrated acute psychiatric crises are among people already carrying a formal diagnosis. When a crisis does land someone in an emergency department or a hospital bed, it is overwhelmingly likely to be someone the system already knows has a mental illness.

Costs followed the same pattern. Mean annual medical expenditures reached 620.6 United States dollars among adults with documented psychiatric disorders, against 296.8 dollars for the matched comparison group, a cost ratio of 2.09. Every one of these differences was statistically significant at P less than 0.001. The cost analysis employed a gamma generalized linear model with a log link, a choice that reflects the deeply skewed distribution of medical spending, where a small fraction of patients accounts for a large share of total expenditure. Modeling costs directly on this skewed scale avoids the distortions that arise from transforming spending data or comparing raw means between unequal groups.

The statistical machinery behind the study deserves attention because it shapes how the findings should be read. For the binary outcomes, the authors used modified Poisson generalized linear models with robust standard errors, an approach that has become the method of choice for cohort studies with common outcomes. Traditional logistic regression yields odds ratios that can wildly overstate relative risks when an outcome is not rare, but the modified Poisson estimator produces risk ratios directly, making the results far more interpretable for clinicians and policymakers. A risk ratio of 2.01 genuinely means twice the risk, not an inflated odds-based approximation of it.

Still, the study’s observational design imposes important caveats. Propensity score matching can only balance the characteristics that are measured and recorded; unmeasured differences between the groups, from smoking and diet to social support and housing instability, could contribute to the observed gaps in utilization and cost. The documented-diagnosis definition also raises the possibility of detection bias, since people in contact with the health system for mental health reasons may be more likely to have physical complaints noticed and coded as well. And the findings are specific to South Korea’s single-payer insurance system, where financial barriers to care are relatively low; in systems with more fragmented coverage, the relationships could look different in both directions.

What the study loses in causal certainty, however, it gains in scale and policy relevance. A cohort of more than half a million matched individuals, followed through a national claims database, provides a precision that no collection of smaller studies can match, and the consistency of the signal across emergency visits, hospitalizations and costs suggests a robust phenomenon rather than a statistical artifact. The authors conclude that their findings support coordinated care models that address psychiatric and physical health needs together, an argument that aligns with a growing international movement toward integrated mental and physical healthcare. As health systems worldwide grapple with rising demand and constrained budgets, the arithmetic emerging from this study is hard to ignore: mental illness is not a separate ledger of costs but a multiplier running through the entire healthcare system, doubling hospital admissions and roughly doubling what each affected adult costs the system every year.

Subject of Research: Healthcare utilization and medical costs among adults with psychiatric disorders

Article Title: Healthcare utilization and medical costs among adults with selected psychiatric disorders: a nationwide retrospective cohort study

Article References: Healthcare utilization and medical costs among adults with selected psychiatric disorders: a nationwide retrospective cohort study. (n.d.). https://doi.org/10.1186/s12889-026-29907-4

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29907-4

Keywords: psychiatric disorders, healthcare utilization, emergency department visits, hospitalization, medical costs, South Korea, National Health Insurance Service, retrospective cohort study, propensity score matching, public health, health economics, mental health