INTRODUCTION

Adjustment for confounding variables is a fundamental process in observational research for identifying valid associations between exposures and health outcomes [1]. With the increasing use of administrative databases such as the National Health Insurance Claim Data (NHICD) of Korea, a variety of analytical methods, including multivariable regression and propensity score approaches, have been applied to control for confounders. However, because the NHICD are primarily collected for reimbursement purposes, they lack many important confounders, such as laboratory results or measures of clinical severity.

Comorbidity is defined as the presence of a condition other than the primary disease under study that may influence a patient’s prognosis and potentially alter treatment plans [2]. In health outcome studies using the NHICD, comorbidities are frequently used as surrogate variables for disease severity. For example, Hong et al. [3] estimated adjusted hazard ratios for suicide death after discharge among individuals with psychiatric illness by adjusting for confounders, including comorbidity. As numerous outcome studies have applied comorbidity as a confounding variable, review articles have also been conducted to recommend appropriate methods for comorbidity adjustment in the NHICD. Kim [4] proposed that key factors to consider in comorbidity adjustment include the choice of comorbidity measurement, determination of the data and diagnostic code scope, and specification of an appropriate look-back period. In addition, simulation studies that integrated these factors and focused on patients undergoing percutaneous coronary intervention or diagnosed with myocardial infarction have recommended specific methodologies for measuring comorbidities [5,6].

Because comorbidities are generally identified based on diagnostic codes, limitations related to diagnostic inaccuracy inherent in the NHICD are unavoidable. In particular, because diagnostic codes serve as key criteria for defining study populations, the Reporting of Studies Conducted using Observational Routinely Collected Health Data (RECORD) statement recommends that researchers provide detailed descriptions of the codes or algorithms used to identify study populations, along with explicit references to any validation studies [7]. Similarly, to ensure the reproducibility and transparency of research findings, the methods used for comorbidity adjustment should be reported in sufficient detail. Nevertheless, the reporting quality regarding comorbidity adjustment has not yet been evaluated. Therefore, this study aimed to assess the reporting quality of comorbidity adjustment in studies utilizing the NHICD.

METHODS

We reviewed health outcome studies that used the NHICD. A literature search was conducted in PubMed in April 2025 using the following search strategy: (Korea OR Korean) AND (“health insurance claim*” OR claims data OR NHIS OR HIRA) AND (“2024”). Two reviewers (SHL and DKH) independently screened eligible studies and extracted data. Discrepancies were resolved through discussion, and if consensus was not reached, a reviewer (author) adjudicated the decision. Of the 239 articles identified, we included only health outcome studies that exclusively utilized the NHICD. After excluding review articles (n=9), studies using non-Korean data (n=11), and other studies including international data (n=12), a total of 207 articles remained. Among these, we further excluded studies that linked claim data with other sources, such as health screening or hospital data (n=58), epidemiologic studies (n=60), and cost-effectiveness analyses (n=2). In addition, 5 studies that did not use comorbidities as confounding variables were excluded, resulting in a final sample of 82 studies included for review.

Indicators for evaluating the reporting quality of comorbidity adjustment were selected based on a review of previous studies that recommended methods for comorbidity adjustment [7,8].

Ethics Statement

As this study is a scoping review based on previously published literature, ethical approval and patient consent were not required.

RESULTS

Comorbidity Measurement

Methods for measuring comorbidities can generally be categorized into 2 approaches: one that uses standardized comorbidity measurement tools, and another in which researchers manually select specific comorbidities for adjustment (ad hoc selection). The latter approach involves identifying comorbidities to be included as confounding variables based on clinical expertise or evidence derived from previous studies.

Comorbidity measurements, on the other hand, define sets of disease groups that are known to be associated with health outcomes such as mortality and medical costs. These measurements can be developed using either prescription codes or diagnostic codes. Among prescription-based measures, the chronic disease score is a representative example [9]. Although this method offers high sensitivity for detecting comorbid conditions, it has practical limitations, as a single drug may be prescribed for multiple conditions and medication lists must be continuously updated when new drugs are reimbursed. In contrast, measurements based on diagnostic codes include the Charlson comorbidity index (CCI) and Elixhauser comorbidity measure (ECM). The CCI assigns weighted scores to 17 comorbid conditions that significantly affect in-hospital mortality, based on their relative risk estimates, and the total score represents the overall comorbidity burden (Table 1) [10]. Quan et al. [11] highlighted the need to update the weights to reflect contemporary medical practice and subsequently recalibrated them using data from 6 countries. They demonstrated that the updated weights showed good-to-excellent discrimination in predicting in-hospital mortality. In Korea, a modified version was developed using an Asian nationwide database (mCCI-A), and it was reported that the c-statistic for predicting mortality was significantly improved compared with the original CCI [12]. The ECM, meanwhile, consists of 30 diagnoses that have been shown to be significantly associated with length of hospital stay, medical costs, and in-hospital mortality [13]. Although several studies have reported that the ECM is statistically superior to the CCI, it carries a risk of overfitting in small datasets and is limited by computational requirements in large datasets [14,15]. The ECM can also be simplified by counting the number of comorbid conditions to mitigate these limitations [16]. Furthermore, van Walraven et al. [17] derived a composite index by estimating disease-specific weights through mortality modeling, and Thompson et al. [14] reported a similar approach (Table 2).

Indicators for Evaluating the Reporting Quality for Comorbidity Adjustment

When measuring comorbidities using the NHICD, it is essential to determine the scope of data, the range of diagnostic codes, and the length of the look-back period [4,8]. First, the NHICD is classified into inpatient and outpatient records, and the choice of data scope should be carefully considered. Comorbidities can be identified from both inpatient and outpatient records, and restricting analyses to only 1 source may lead to underestimation. For example, hypertension is primarily managed in outpatient settings; therefore, relying solely on inpatient data may result in an underestimated prevalence of this comorbidity.

Second, the range of diagnostic codes must be clearly defined. Diagnoses in the NHICD are categorized as primary diagnoses, secondary diagnoses, or rule-out diagnoses. Rule-out diagnoses represent conditions that were suspected but not confirmed after diagnostic testing; therefore, only primary and secondary diagnoses should be included when identifying comorbidities. In addition, primary and first secondary diagnoses are stored in the general information table, whereas all diagnoses, including these, are stored in the diagnosis table. Consequently, the estimated prevalence of comorbidities may differ depending on whether the general information table or the diagnosis table is used.

Third, it is necessary to determine the look-back period, defined as the time window for screening prior medical records before the index date. The index date serves as the reference point for assessing either health outcomes or patients’ medical histories. Most previous studies selected relatively short look-back periods (<1 year), prioritizing data accessibility and practical convenience over clinical judgment [18]. Studies comparing model performance across different observation periods have shown that although longer look-back periods tend to increase the estimated prevalence of comorbidities, they may also reduce overall model efficiency [5,6].

Reporting Quality for Comorbidity Adjustment

Among the 82 studies, 33 studies (40.2%) applied either ad hoc selection or the CCI alone, whereas 13 studies (15.9%) applied both methods concurrently (Table 3). The ECM was used in 3 studies, of which only 1 applied the ECM exclusively, while the other 2 applied the ECM in combination with either the CCI or ad hoc selection. Among the 47 studies that applied ad hoc selection, a wide range of comorbidities was used to adjust for associations between study exposures and health outcomes. Among these comorbidities, diabetes mellitus (80.9%) and hypertension (72.3%) were the most frequently included (Table 4).

Only 3 studies explicitly reported the data sources used to identify comorbidities. For example, Kim et al. [19] screened comorbidities if they were diagnosed in either inpatient or outpatient settings. Other studies similarly utilized both outpatient and inpatient records [20,21]. Six studies reported the range of diagnostic codes used, with 3 studies considering primary diagnoses only and 3 studies including both primary and secondary diagnoses when identifying comorbidities.

The look-back period was reported in 26 studies (31.7%). Sixteen studies (19.5%) examined comorbidities during the 1 year preceding the index date, whereas 5 studies referred only to “prior to the index date” without specifying the duration. The observation period varied across studies, ranging from 6 months to 5 years before the index date.

DISCUSSION

Comorbidities serve as an essential surrogate variable for patient severity in observational studies using administrative data. In this scoping review, we evaluated the reporting quality of comorbidity adjustment in health outcome studies using the NHICD. Our review demonstrated that reporting transparency remains suboptimal, with substantial variability in the selection and operationalization of comorbidity measures. Taken together, these findings highlight the need for methodological guidance specifically tailored to NHICD-based research.

Although no consistent rule was identified for selecting comorbidity measurements according to study population and health outcomes, the CCI was the most frequently applied method among the reviewed studies, and its application varied substantially depending on how the CCI score was distributed. Among the 47 studies that applied the CCI, 20 categorized it using the conventional classification of 0 points, 1 point, 2 points, and ≥3 points. Ten studies treated the CCI as a continuous variable, while the remaining 17 categorized it based on its distribution (e.g., 0/1/≥2, 0–1/≥2, or 0–2/3–5/≥6 points). While the CCI provides a standardized approach for measuring comorbidities, it also presents several limitations that warrant careful consideration. First, the CCI does not adequately capture psychiatric conditions that are increasingly relevant in NHICD-based studies, as it includes only dementia among mental health disorders. This limitation has led researchers to supplement the CCI by adding selected psychiatric conditions through ad hoc inclusion, an approach that enhances clinical relevance but reduces comparability across studies. For example, Kim et al. [22] applied the CCI and additionally incorporated psychiatric disorders such as personality disorders, psychotic symptoms, and recurrent depression. Second, although updated weighting systems such as Quan et al. [11]’s revisions and the Korean-modified mCCI-A [12] have demonstrated improved predictive validity, none of the included studies applied these updated weights, and only a few adopted population-specific weighting [19]. One possible reason for their limited uptake is that these updated weighting schemes are not yet widely disseminated or validated across diverse study populations. To facilitate broader use in NHICD-based research, further studies are needed to assess the performance and generalizability of these updated or locally derived weights across different clinical and demographic subgroups.

The ECM was developed specifically to measure comorbidities using administrative data. Although the ECM is conceptually advantageous in its breadth and inclusion of mental health conditions, its complexity and computational demands likely contributed to its underutilization [14,15]. Simplified ECM-based indices, such as the van Walraven score [17], may offer a pragmatic alternative that balances conceptual rigor with feasibility. In contrast, the ad hoc selection method allows for condition-specific tailoring but introduces considerable heterogeneity across studies, thereby limiting cross-study comparability [4]. For example, in a study examining the association between insomnia and major adverse cardiac and cerebrovascular events among patients with end-stage kidney disease, comorbidities included diabetes mellitus, congestive heart failure, cerebrovascular disease, myocardial infarction, peripheral vascular disease, chronic obstructive pulmonary disease, obstructive sleep apnea, and restless leg syndrome [23]. Similarly, a study analyzing the association between periodontal disease treatment and mortality in patients with dementia adjusted for diabetes mellitus, ischemic heart disease, cerebrovascular disease, and kidney disease as covariates [24]. Although hypertension and diabetes mellitus were the most frequently applied comorbidities, a wide range of conditions was used depending on the study population and the health outcomes investigated.

Our findings demonstrate significant deficiencies in the reporting of methodological details related to comorbidity measurement. Only a small proportion of studies explicitly described the scope of data used, the diagnostic code range, or the look-back period. None of the studies reported all of these factors, even though they critically influence comorbidity identification and, consequently, the validity of risk adjustment. Only 3 studies explicitly reported the scope of data used. Restricting analyses to inpatient claims may underestimate the prevalence of outpatient-managed conditions such as hypertension. Consequently, comorbidities should be screened in both inpatient and outpatient records to ensure comprehensive measurement. Similarly, the estimated prevalence of comorbidities may vary depending on the range of diagnostic codes used. The diagnostic code range is generally determined by the validity of diagnostic coding, which remains a key limitation of the NHICD. Some studies applied operational definitions to enhance the validity of diagnostic codes. Yoon et al. [25] identified comorbidities when patients had 2 or more visits to medical institutions with the same comorbidity code. Similarly, Kim et al. [19] defined comorbidities such as diabetes, hypertension, dyslipidemia, osteoporosis, Parkinson’s disease, and dementia using diagnostic codes combined with medical claims for at least 1 principal or secondary diagnosis and prescription medication use.

Twenty-six studies reported the look-back period used for measuring comorbidity. Reporting the look-back period is important for enhancing the reproducibility of study findings and facilitating comparability across studies. Among the studies that reported this information, the observation periods ranged from 6 months to 5 years [26,27], with 1 year being the most commonly applied duration. Although no clear guidelines exist for determining an appropriate observation period, it should be selected in accordance with the study objectives. If the study aims to estimate comorbidity prevalence, the observation period can be determined based on clinical judgment to ensure accurate measurement [4]. Conversely, if the objective is to optimize predictive model performance, the observation period should be selected with efficiency in mind. Previous studies have compared look-back periods in terms of model performance. Kim and Ahn [6] assessed predictive performance for in-hospital mortality among patients undergoing percutaneous coronary intervention according to different observation periods. They reported that extending the observation period to 3 years increased the estimated prevalence of comorbidities but did not improve model performance compared with a 1-year period, consistent with findings in myocardial infarction [5]. Preen et al. [18] evaluated patients with medical and surgical conditions and compared observation periods of 1 year, 2 years, 3 years, and 5 years. They found that for mortality models, a 1-year observation period yielded the highest predictive performance, whereas for readmission models, a 5-year period was optimal.

To our knowledge, no prior study has specifically evaluated the reporting quality of comorbidity adjustment in research using administrative data. Nevertheless, 2 recent international evaluations based on the RECORD statement provide valuable insights into the overall reporting quality of real-world data studies. Zhao et al. [28] evaluated 187 cohort studies published between 2013 and 2021 according to the RECORD checklist. The mean proportion of adequately reported items was 44.7%, indicating that adherence to RECORD remains insufficient and that overall reporting quality has not improved in recent years. Similarly, Zhao et al. [29] assessed adherence to RECORD among studies published in 8 high-impact general medical journals between 2016 and 2023. The mean adherence rate was 70.7%, reflecting only moderate compliance, with incomplete reporting of several key items. These findings suggest that reporting quality remains suboptimal even in the international context. Therefore, although our analysis focused on domestic studies, the low reporting quality observed in our review is consistent with global evidence and underscores the ongoing need to strengthen reporting transparency in administrative data–based research. The lack of standardized reporting identified in this review is particularly concerning. Even when researchers appear to have considered essential methodological aspects, the absence of explicit documentation undermines reproducibility and limits the interpretability of comorbidity-adjusted estimates. To improve methodological transparency, future studies should clearly specify the data scope, diagnostic code range, and look-back period used for comorbidity assessment. In alignment with the RECORD statement, detailed and standardized reporting of these elements is essential to ensure methodological rigor and comparability across studies using the NHICD.

This review has several limitations. First, we restricted our search to PubMed-indexed articles written in English. This decision was made to enhance the feasibility and consistency of the scoping review. Furthermore, the literature search was confined to studies published in 2024 to capture the most recent trends in reporting quality. However, this approach may have introduced selection bias and limited the ability to assess longitudinal changes by excluding Korean-language studies that might provide complementary perspectives on reporting practices in NHICD-based research. Future studies should include Korean-language databases such as KoreaMed and extend the observation period to provide a more comprehensive and balanced overview of reporting practices. Second, the evaluation was based solely on information available in published reports; therefore, the actual methodological quality may have been underestimated if detailed descriptions were omitted. Finally, although this review focused primarily on the reporting quality of comorbidity adjustment, statistical validity and model performance are equally critical components of methodological soundness. Future investigations should therefore evaluate whether the applied comorbidity adjustment methods were statistically appropriate and examine the extent to which such adjustments influenced effect estimates and overall model fit. Despite these limitations, this review provides the first comprehensive assessment of comorbidity adjustment reporting in NHICD-based studies and offers an empirical foundation for improving transparency and methodological rigor in future administrative data research.

CONCLUSION

The findings indicate that the overall reporting quality of comorbidity adjustment in studies using the NHICD remains suboptimal. Transparent and standardized reporting of comorbidity adjustment is crucial to enhance the reproducibility, comparability, and validity of research findings. Establishing methodological guidelines and adopting reporting standards aligned with the RECORD statement would substantially improve the transparency and scientific rigor of future studies.

Notes

Conflict of Interest

The author has no conflicts of interest associated with the material presented in this paper.

Funding

None.

Acknowledgements

We thank Seo Hyun Lee and Da Kyoung Hwang for their support in searching and selecting the literature used in this review.

Author Contributions

All work was done by Kim KH.

REFERENCES

• 1. Han SJ, Kim KH. Adjusting for confounders in outcome studies using the Korea National Health Insurance Claim Database: a review of methods and applications. J Prev Med Public Health 2024;57(1):1-7. https://doi.org/10.3961/jpmph.23.250ArticlePubMedPMC
• 2. Charlson ME, Carrozzino D, Guidi J, Patierno C. Charlson comorbidity index: a critical review of clinimetric properties. Psychother Psychosom 2022;91(1):8-35. https://doi.org/10.1159/000521288ArticlePubMedPMC
• 3. Hong M, Lee SM, Han KM, Kim KH, Paik JW. Suicide death and other-cause mortality in psychiatric patients: a South Korean study using nationwide claims data. J Affect Disord 2024;352: 288-295. https://doi.org/10.1016/j.jad.2024.02.075ArticlePubMed
• 4. Kim KH. Comorbidity adjustment in health insurance claim database. Health Policy Manag 2016;26(1):71-78. (Korean)https://doi.org/10.4332/KJHPA.2016.26.1.71Article
• 5. Kim KH. Comparative study on three algorithms of the ICD-10 Charlson comorbidity index with myocardial infarction patients. J Prev Med Public Health 2010;43(1):42-49. (Korean)https://doi.org/10.3961/jpmph.2010.43.1.42ArticlePubMed
• 6. Kim KH, Ahn LS. A comparative study on comorbidity measurements with lookback period using health insurance database: focused on patients who underwent percutaneous coronary intervention. J Prev Med Public Health 2009;42(4):267-273. (Korean)https://doi.org/10.3961/jpmph.2009.42.4.267ArticlePubMed
• 7. Benchimol EI, Smeeth L, Guttmann A, Harron K, Moher D, Petersen I, et al. The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLoS Med 2015;12(10):e1001885. https://doi.org/10.1371/journal.pmed.1001885ArticlePubMedPMC
• 8. Shin DW, Han K. The use of Charlson comorbidity index for observational studies using administrative data in Korea. Precis Future Med 2025;9(1):2-14. https://doi.org/10.23838/pfm.2024.00191Article
• 9. Von Korff M, Wagner EH, Saunders K. A chronic disease score from automated pharmacy data. J Clin Epidemiol 1992;45(2):197-203. https://doi.org/10.1016/0895-4356(92)90016-gArticlePubMed
• 10. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40(5):373-383. https://doi.org/10.1016/0021-9681(87)90171-8ArticlePubMedPMC
• 11. Quan H, Li B, Couris CM, Fushimi K, Graham P, Hider P, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol 2011;173(6):676-682. https://doi.org/10.1093/aje/kwq433ArticlePubMed
• 12. Choi JS, Kim MH, Kim YC, Lim YH, Bae HJ, Kim DK, et al. Recalibration and validation of the Charlson Comorbidity Index in an Asian population: the National Health Insurance Service-National Sample Cohort study. Sci Rep 2020;10(1):13715. https://doi.org/10.1038/s41598-020-70624-8ArticlePubMedPMC
• 13. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care 1998;36(1):8-27. https://doi.org/10.1097/00005650-199801000-00004ArticlePubMed
• 14. Thompson NR, Fan Y, Dalton JE, Jehi L, Rosenbaum BP, Vadera S, et al. A new Elixhauser-based comorbidity summary measure to predict in-hospital mortality. Med Care 2015;53(4):374-379. https://doi.org/10.1097/MLR.0000000000000326ArticlePubMedPMC
• 15. Sharma N, Schwendimann R, Endrich O, Ausserhofer D, Simon M. Comparing Charlson and Elixhauser comorbidity indices with different weightings to predict in-hospital mortality: an analysis of national inpatient data. BMC Health Serv Res 2021;21(1):13. https://doi.org/10.1186/s12913-020-05999-5ArticlePubMedPMC
• 16. Dominick KL, Dudley TK, Coffman CJ, Bosworth HB. Comparison of three comorbidity measures for predicting health service use in patients with osteoarthritis. Arthritis Rheum 2005;53(5):666-672. https://doi.org/10.1002/art.21440ArticlePubMed
• 17. van Walraven C, Austin PC, Jennings A, Quan H, Forster AJ. A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Med Care 2009;47(6):626-633. https://doi.org/10.1097/MLR.0b013e31819432e5ArticlePubMed
• 18. Preen DB, Holman CD, Spilsbury K, Semmens JB, Brameld KJ. Length of comorbidity lookback period affected regression model performance of administrative health data. J Clin Epidemiol 2006;59(9):940-946. https://doi.org/10.1016/j.jclinepi.2005.12.013ArticlePubMedPMC
• 19. Kim JS, Park M, Park S, Chae J, Hong YH, Park KS, et al. Prognosis of amyotrophic lateral sclerosis patients after tracheostomy invasive ventilation in Korea. Amyotroph Lateral Scler Frontotemporal Degener 2024;25(3–4):271-281. https://doi.org/10.1080/21678421.2024.2314064ArticlePubMed
• 20. Choi E, Lee J, Choo E, Jang EJ, Lee IH. Continuity of care between dyslipidemia patients and multiple providers: a cohort study. PLoS One 2024;19(5):e0300745. https://doi.org/10.1371/journal.pone.0300745ArticlePubMedPMC
• 21. Cho A, Bae Y, Kim M, Kim DH, Lee YK, Park HC. Safety of high-dose intravenous iron in hemodialysis patients: results from the National Health Insurance Service (2019–2020) in South Korea. J Clin Med 2024;14(1):63. https://doi.org/10.3390/jcm14010063ArticlePubMedPMC
• 22. Kim H, Lee YB, Lee J, Kang D, Kim G, Jin SM, et al. Association between depression, antidepressant use, and the incidence of atherosclerotic cardiovascular diseases. J Affect Disord 2024;352: 214-221. https://doi.org/10.1016/j.jad.2024.02.034ArticlePubMed
• 23. Kim HW, Heo GY, Kim HJ, Kang SW, Park JT, Lee E. Insomnia in patients on incident maintenance dialysis and the risk of major acute cardio-cerebrovascular events and all-cause mortality. Nephrol Dial Transplant 2024;39(5):830-837. https://doi.org/10.1093/ndt/gfad231ArticlePubMed
• 24. Cho HA, Kim BR, Shin H. Association of periodontal disease treatment with mortality in patients with dementia: a population-based retrospective cohort study (2002–. 2018;Sci Rep 2024;14(1):5243. https://doi.org/10.1038/s41598-024-55272-6ArticlePubMedPMC
• 25. Yoon HY, Kim H, Bae Y, Song JW. Smoking status and clinical outcome in idiopathic pulmonary fibrosis: a nationwide study. Respir Res 2024;25(1):191. https://doi.org/10.1186/s12931-024-02819-wArticlePubMedPMC
• 26. Kim JY, Jeong Y, An H, Suh JW, Sohn JW, Yoon YK. Clinical outcomes of coronavirus disease 2019 in people living with human immunodeficiency virus in South Korea: a nationwide population-based cohort study. Influenza Other Respir Viruses 2024;18(6):e13337. https://doi.org/10.1111/irv.13337ArticlePubMedPMC
• 27. Kim DG, Kim SH, Lee JY, Lee JG. Antiplatelet agent for the prevention of late hepatic vascular complications in living donor-dominant liver transplant population. Asian J Surg 2024;47(9):3864-3869. https://doi.org/10.1016/j.asjsur.2024.04.002ArticlePubMed
• 28. Zhao R, Zhang W, Zhang Z, He C, Xu R, Tang X, et al. Evaluation of reporting quality of cohort studies using real-world data based on RECORD: systematic review. BMC Med Res Methodol 2023;23(1):152. https://doi.org/10.1186/s12874-023-01960-2ArticlePubMedPMC
• 29. Zhao HJ, Ushcatz I, Walwyn C, Lowe MS, Kim KS, Benchimol EI, et al. Adherence to RECORD reporting guidelines among observational studies using routinely collected health data published in general medical journals: a meta-epidemiologic study. J Clin Epidemiol 2025;185: 111876. https://doi.org/10.1016/j.jclinepi.2025.111876ArticlePubMed

Figure & Data

References

Citations

Citations to this article as recorded by