Abstract
Introduction: Children, adolescents, and young adults (AYAs) with cancer are a special population who are subjected to a number of unique challenges, stressors, and barriers to high-quality psychological care. In a recent systematic review of measurement properties, we found that 5 of 18 identified patient-reported outcome measures (PROMs) had sufficient psychometric properties to justify their use. A next step is to analyze the reliability of these scale scores in a reliability-generalization meta-analysis. Methods: We conducted a systematic review of three databases for all studies reporting reliability data for previously identified PROMs. Included studies were further required to include patients with cancer, or survivors of cancer, ages 2-39. We next synthesized alpha and test-retest coefficients using best statistical practices, according to prespecified subgroups, where possible. We considered a threshold of 0.7 to represent sufficient evidence of reliability. Results: Seventy-one studies were included. Overall, reliability coefficients for scale and subscale scores exceeded 0.7. Subgroup analyses were limited by incomplete reporting and a lack of sufficient studies for each subgroup; however, where conducted, these subgroup analyses showed significant differences in the reliability of self-reports versus proxy reports and original versus adapted versions of PROMs. Discussion: We recommend better reporting of reliability data in future studies of children and AYAs with cancer. We discourage relying on historical reliability data in different samples and the reporting of only ranges of reliability coefficients for subscales. Our study suggests that significant differences in the reliability of PROMs may be associated with the PROM respondent and the version of the PROM, thus highlighting the need for further investigation.