Survey methods for seriously ill hospitalized adults: practical lessons from SUPPORT. The Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments.

OBJECTIVE: To describe issues confronted in collecting questionnaire-based data in a large study of seriously ill hospitalized adults, discuss methods used to achieve acceptable results, summarize the reliability of selected measures, and explore challenges for patient-based research in this population.
DESIGN: A prospective cohort study.
SETTING: Five teaching hospitals in the US. PARTICIPANTS: 9105 patients, with a subset of 80 patients and 10 proxy respondents for reliability testing. MEASUREMENTS AND MAIN OUTCOMES: Response rates; percent agreement and Kappa statistics to assess test-retest reliability.
RESULTS: A pretest and pilot test were used to help develop the assessment strategy, establish study procedures, and identify staff characteristics likely to facilitate the project. Achieving adequate response rates required great perseverance and flexibility by interview staff as well as close monitoring and central supervision. Standardized quality assurance with timely feedback, and social support provided by peer and supervising staff were both important. Older women proved to be the best interviewers. Reliability was acceptable for existing and newly developed scales to measure subjective health status, preferences, and satisfaction.
CONCLUSIONS: It was possible to collect reliable and useful data on preferences and subjective health outcomes from a cohort of seriously ill hospitalized patients and their proxies. Achieving acceptable data quality required a large interview team, mature staff, flexibility about modes of administration, and centralized, structured quality assurance. Successful outcomes research in this population is likely to require similar planning and efforts.