The Patient Health Questionnaire-9 (PHQ-9) screening tool has been evaluated against various reference standards including semi-structured, fully structured, and the Mini International Neuropsychiatric (MINI) interviews that have been shown to have different accuracy for diagnosing major depression. This could influence estimates of PHQ-9 sensitivity and specificity and depression prevalence. This study proposed and validated Frequentist and Bayesian latent class models (LCM) that, in addition to depression prevalence, simultaneously estimate the accuracy of the PHQ-9 and different reference standards in the context of individual participant data meta-analysis. The study found that the diagnostic accuracy of the PHQ-9 estimated via the standard bivariate random-effects model depends on the reference standard used. Latent class models, which account for the imperfect nature of the reference standards, are alternative approaches to account for differences in reference standard accuracy.