Improving the Quality of Patient Reported Outcomes by Minimizing Missing Data
Patient-Reported Outcomes (PROs) are often included as an endpoint in prospective studies. Clinicians use the study results to help individuals with cancer make informed decisions about treatment. Missing PRO data threaten the validity of randomized clinical trials (RCTs) and can undermine scientific credibility of PRO-specific conclusions. The thesis goal was to use quantitative and qualitative approaches to identify causes of missing PRO data. A systematic review of the literature identified 46 unique categories of factors reported to have an association with missing PRO data. These factors were organized into a framework having five main components: Instrument, Participant, Centre, Staff, and Study. Cox regression examined the association of 15 factors with time to the first missing instrument in data from three RCTs: MA17 (breast cancer), PR7 (prostate cancer), and LY12 (non-Hodgkin’s lymphoma). Most factors did not have significant independent association with outcome, but a few did, depending on the RCT. Centre presence of a post-graduate residency training program (MA17, PR7) and geographic location (PR7, LY12) had a stronger association. Participant age (PR7) and baseline level of well-being (LY12), and centre level of activity (LY12) had a weaker association. Interviews were conducted with 11 Clinical Research Associates (CRAs). Ten themes were identified in analysis based on descriptive phenomenology. The CRAs described their experience with different patterns of missing PRO data. PROs for routine clinical care compete with PROs for RCTs. Several important influences of missing PRO data were inherent to the instruments (quantity, repetition, wording of questions, and format), patients (characteristics, attributes of the cancer, and toxicity from the treatment), centres (deficiencies in research infrastructure), staff (developing relationships with patients), and RCTs (data capture in electronic versus paper format). Adapting the assessment schedule and providing sufficient time for the patient to complete the instrument were challenging for the CRAs, particularly at disease progression. There is a need for field guidance, and a motivation and knowledge base among CRAs to develop such guidance. This dissertation provides evidence to facilitate understanding of which factors influence missing PRO data, so that steps can be taken to improve the quality of PROs by minimizing missing data.