Effectiveness of the Nationwide Depression Screening: A Systematic Review and A Target Trial Emulation Study Using National Health Insurance Service Database

Abstract

INTRODUCTION Over the past two decades, previous studies about the effectiveness of organized depression screening at the primary care or national level have reached mixed conclusions. Despite its early age of onset and substantial disease burden, depression often remains undetected and untreated in the general population. Although screening for depression has been proposed to prevent the delay of treatment, the public health effectiveness of such screening remains inconclusive. Therefore, the first study used a systematic review and meta-analysis to comprehensively evaluate the effectiveness of depression screening based on both randomized controlled trials (RCTs) and non-randomized observational studies. Following this, a target trial emulation study analyzed the effectiveness of the 2019 nationwide depression screening program in South Korea, which used the Patient Health Questionnaire-9 (PHQ-9), leveraging data from the National Health Insurance Service (NHIS) database to support more informed decision-making. METHODS In the first study, a systematic review and meta-analysis was conducted to identify and synthesize evidence evaluating the effectiveness of depression screening published from June 1959 to August 2024. Databases including PubMed, EMBASE, PsychINFO, and Web of Science were searched, pooling 4,039 articles initially. Two independent reviewers performed the study search, data extraction, and study evaluation. Each included study was summarized based on the predefined population, intervention, comparison, outcome, and study design (PICOS) framework. The quality of the studies was assessed using Version 2 of the Cochrane Risk-of-Bias Tool for Randomized Trials and the Risk of Bias in Non-Randomized Studies - of Interventions. The evaluation of the evidence was based on six criteria: (1) targeting specific population, (2) exclusion of prevalent cases of depression, (3) eligibility determined before follow-up, (4) inclusion of additional components besides screening, (5) use of a validated screening tool, and (6) analysis of outcomes using individual-level data. Random effects models were used to conduct the meta-analysis, applying restricted maximum likelihood estimation method. Heterogeneity across studies was assessed using the I2 statistic and prediction intervals (PI). Publication bias was assessed using visual evaluation of Begg’s funnel plot and Egger’s test. Subgroup analyses were performed based on outcome definition, sex, and geographic region, while sensitivity analysis excluded studies with a high risk of bias. In the second study, an emulation of a hypothesized target trial was conducted using the Korea NHIS database (2019 – 2022). The estimands were derived using the traditional intent-to-treat (ITT) and modified ITT approaches. As the depression screening for individuals in their 20s and 30s was introduced in 2019, the follow-up period started on January 1, 2019, and ended on December 31, 2022. The primary outcome was the incidence of hospitalization for mood disorders, with secondary outcomes including (1) initial use of antidepressants, (2) emergency department visits for mood disorders, and (3) suicide and suicidal behaviors. Individuals with a psychiatric diagnosis prior to 2019 or those who had participated in the depression screening before 2019 were excluded. Coarsened exact matching (1:1 ratio) based on age, sex, and subscriber types was used to match the screened and non-screened individuals. The multivariate cause-specific Cox regression models were used to estimate the hazard ratios (HRs), which adjusted for urbanicity, prior health check-ups, income, and the Charlson Comorbidity Index. Subgroup analyses were conducted based on sex, age groups (20–50s and 60–70s), and income levels. To address unmeasured confounding, a negative control outcome (hospitalization for cancer) and the E-value were applied. Exploratory analyses were also performed to examine the distribution of depression-related outcomes, the diagnostic performance of the PHQ-9, and the impact of post-screening interventions using a dataset from a cohort of screeners. RESULTS The systematic review and meta-analysis found significant but weak evidence that depression screening for the general adult population in primary care or at the national level is effective (pooled OR 0.74, 95% CI: 0.62–0.87), particularly among women and Asian populations. However, the included studies were mostly of poor methodological quality, with significant bias and considerable heterogeneity across studies. None of the studies met all six criteria required to provide valid evidence for the effectiveness of depression screening in the general adult population. The target trial emulation study aimed to estimate the effectiveness of depression screening based on the modified ITT approach (n = 53,688), and the traditional ITT (n = 276,214) as secondary. After adjusting for confounders, the screened group showed a 44% lower risk of hospitalization for mood disorders compared to the non-screened group (adjusted hazard ratio [aHR] for modified ITT 0.56, 95% confidence intervals [CI] 0.41–0.76), especially in women and in older adults. The initiation of antidepressant use was significantly increased in screened individuals than in non-screened (aHR for modified ITT 1.21, 95% CI 1.10–1.33). The hazards of emergency department visits were significantly lower in the screened group than in non-screened group (aHR for modified ITT 0.71, 95% CI 0.51–0.98). No significant difference was observed in suicide and suicidal behaviors between the screened and non-screened groups (aHR for modified ITT 0.62, 95% CI 0.32–1.21). Hospitalization for cancer, used as a negative control outcome, showed no difference between these groups. CONCLUSION The systematic review and meta-analysis found statistically significant but limited evidence supporting effectiveness of depression screening in general adult populations. The target trial emulation study using NHIS data in South Korea showed that the 2019 nationwide depression screening program reduced hospitalization rates for mood disorders, increased antidepressant use, and decreased emergency department visits, particularly among women and older adults. However, the observed differences between screened and non-screened individuals may result from non-specific effects of the screening program, highlighting the need for ongoing evaluation to support evidence-based decisions on nationwide depression screening.

[연구배경] 지난 20년간 1차 진료나 국가 차원에서 조직적으로 시행된 우울증 선별검사의 효과에 대한 이전 연구들은 서로 다른 결론을 내었다. 따라서 본 연구에서는 체계적 문헌 고찰 및 메타분석과 국민건강보험공단 데이터베이스를 이용한 표적 시험 에뮬레이션 연구를 수행하여, 전국민 대상 우울증 선별검사의 효과성을 평가하였다. [방법] 체계적 문헌 검토 및 메타분석은 일반 성인 인구를 대상으로 시행된 우울증 선별검사의 효과성을 평가하는 무작위 임상 시험(RCT)과 관찰 연구들의 결과들을 종합하였다. PubMed, EMBASE, PsychINFO 및 Web of Science와 같은 데이터베이스를 검색하여 2024년 8월까지 발표된 관련 논문들을 추출했다. 두 명의 독립적인 연구자가 연구 검색, 적합성 검토, 데이터 추출, 편향 위험 평가 및 증거 수준의 질적 평가를 수행했다. 메타 분석은 제한된 최대 가능도 추정법을 적용해 랜덤 효과 모델을 사용하여 수행했다. 연구 간 이질성은 I² 통계와 예측 구간을, 출판 비뚤림은 깔때기 도표와 Egger’s test를 이용해 평가했다. 국내에 도입된 2019년 우울증 선별검사의 효과성을 평가하기 위해 국민건강보험공단 데이터베이스를 사용하여 표적 시험 에뮬레이션 연구를 수행했다. 추적 관찰 기간은 2019년 1월 1일에 시작하여 2022년 12월 31일에 종료되었다. 주요 결과는 기분 장애로 인한 입원 발생이고, 부차적 결과는 (1) 첫 항우울제 사용, (2) 기분 장애로 인한 응급실 방문의 발생, (3) 자살 또는 자살 관련 행동의 발생이다. 2019년에 생존한 18세 이상 성인 중 4%를 랜덤 추출한 후, 이전에 정신과 진단을 받거나 우울증 선별검사에 참여했던 사람들은 제외했고, 수검자격인 특정 연령을 기준으로 유자격자와 무자격자를 분류했다. 그 다음, 연령(±2세), 성별, 보험 자격에 따라 1:1 비율로 유자격자와 무자격자를 대략화된 일치 매칭(coarsened exact matching)하였다. 그 다음 중대한 프로토콜 위반 사례를 제외한 수검자와 비수검자의 데이터를 이용해 수정된 치료의향분석(intent-to-treat, ITT) 효과를 추정하였다(n=53,688). 선별검사 효과는 매칭된 쌍에 따라 층화된 다변량 원인-특정 Cox 회귀 모델로 추정했다. 보정된 모델에서는 거주지의 도시화 수준, 과거 건강 검진 횟수, 소득 수준, Charlson 동반 질환 지수와 같은 교란 변수를 보정했다. 하위 그룹 분석은 성별, 연령군, 소득 수준으로 층화해 분석했다. 음성 대조군 분석과 E-value로 측정되지 않은 교란 효과를 평가했다. [결과] 체계적 검토 및 메타분석에서는 일반 성인 인구를 대상으로 한 우울증 선별검사가 우울 관련 결과를 예방할 수 있다는 증거를 발견했으나(통합된 odds ratio 0.74, 95% CI: 0.62–0.87), 포함된 연구 대부분이 낮은 품질을 보였으며 연구 간 이질성이 컸다. 수정된 ITT 접근법을 기반으로 한 표적 시험 에뮬레이션 연구 결과, 2019년 우울증 선별검사 수검자는 비수검자에 비해 기분 장애로 인한 입원 위험이 44% 낮았고(보정된 위험 비[adjusted hazard ratio, aHR]: 0.56, 95% CI: 0.41–0.76), 이는 여성과 고령층에서만 나타났다. 또한 수검자는 비수검자에 비해 항우울제 치료 시작이 많고(aHR: 1.21, 95% CI: 1.10–1.33), 응급실 방문 위험은 낮았으나(aHR: 0.71, 95% CI: 0.51–0.98), 자살 및 자살 행동에서는 유의한 차이가 없었다(aHR: 0.62, 95% CI: 0.32–1.21). [결론] 국내에 도입된 전국민 대상 우울증 선별검사는 기분장애로 인한 입원과 응급실 방문의 감소 및 항우울제 치료의 활성화와 연관이 있다는 증거를 발견하였다. 그러나 이러한 수검자와 비수검자의 차이는 선별검사의 비특이적 효과로 인한 것으로 보이며, 향후 지속적인 효과성 평가를 통해 불필요한 비용을 감소할 필요가 있다.