The reporting and handling of missing data in genetic epidemiological studies of mental health in childhood and adolescence: A systematic review

Abstract Background Genetic epidemiological analyses of child and adolescent mental health often use data from prospective longitudinal cohorts. Missingness due to selective attrition is therefore an important potential source of bias in such analyses. Informatively reporting on missingness and taking appropriate steps to handle it in analyses can mitigate this potential bias. Here, we aim to systematically assess how researchers report and address missingness in genetic epidemiological studies of child and adolescent mental health‐related outcomes using cohort data. Methods We systematically searched the Ovid Medline database for studies published between August 2012 and August 2025, reporting polygenic score, genome‐wide association, or Mendelian randomization analyses, of data on children or adolescents participating in cohort studies. We extracted information from eligible studies based on criteria adapted from the strengthening and reporting of observational studies in epidemiology (STROBE) guidelines. Results A total of 133 eligible studies were included, of which 125 (93.98%) reported the number of complete cases in all waves, while 84 (63.16%) detailed the amount of missingness on all key variables. Most studies used complete case analysis, while 39 studies explicitly reported applying other methods to handle missingness, with multiple imputation ( n = 20, 15.04%) being the most common, followed by full information maximum likelihood 10 (8.1%). Only 18 studies (13.53%) reported an assumed missing mechanism along with the method used to address missingness. Full reporting of both the extent and handling of missingness at the item level was rare, occurring in only 5 (3.76%) and 15 (11.28%) studies, respectively, among the 123 studies that used multi‐item instruments. Conclusion Best practice recommendations for reporting on missing data handling emphasize the importance of detailing the proportion of missingness, types of mechanisms underpinning missingness, and details of approaches used. Based on this review, these recommendations for proper reporting of missing data are rarely followed in full.