Karin E, Dear BF, Heller GZ, Crane MF, Titov N. (2018). “Wish you were here”: Examining characteristics, outcomes, and statistical solutions for missing cases in web-based psychotherapeutic trials. JMIR Mental Health. 5(2): e22. doi: 10.2196/mental.8363
Researchers used data from 3 randomized controlled trials of web-based cognitive behavioral therapy (N=820) to evaluate characteristics of participants missing at post-treatment assessment, but who responded to 3-month follow-up assessments (n=55; re-contacted cases) to identify how to best address missing cases in web-based psychotherapy trials. Researchers identified predictors of being missing at post-treatment, compared baseline characteristics of re-contacted cases to cases missing at post-treatment and follow-up (n=83; completely missing cases), and evaluated statistical methods for replacing clinical outcomes for missing cases. Outcomes included depression symptoms, comorbid anxiety symptoms, and treatment adherence (completion of treatment modules). Higher baseline depression and lower treatment adherence significantly predicted missing post-treatment assessments, accounting for 40.3% of variance in probability of missing the post-treatment assessment. Participants who completed post-treatment and follow-up assessments (n=682; retained cases) had significantly lower baseline depression symptoms than re-contacted and completely missing cases, but there were no significant differences between re-contacted and completely missing cases. These results indicate that retained cases were clinically distinct from re-contacted cases, but re-contacted and completely missing cases may be similar based on baseline depression symptoms. Statistical methods that did not adjust for depression or treatment adherence underestimated depression symptoms of re-contacted cases by up to 30% and last observation carried forward and baseline observation carried forward replacement methods significantly overestimated depression symptoms by up to 24% and 69%, respectively. Since results indicated that missing cases do not occur at random, researchers suggest statistical methods of replacement should adjust for baseline depression symptoms and treatment adherence.