Simsek, Irmak

Background: Patient reported measures are fundamental to patient centered research and are often collected by questionnaires. Low participation can compromise the validity of the results; transparent reporting and comparing characteristics of participants to nonparticipants is important to assess the generalizability of survey results. We analyzed participation in a survey of social determinants of health among a population-based sample of patients with heart failure (HF). Methods: A total of 7920 residents from 11 Minnesota counties were identified with a first-ever code for HF (ICD-9 428 and ICD-10 I50) between 1/1/2013 and 12/31/2016. Participants completed a battery of 12 questionnaires by mail or telephone. Characteristics and outcomes data were extracted from medical records and compared between participants and nonparticipants. Response rate (RR) was calculated using guidelines of the American Association for Public Opinion Research (AAPOR). The association between nonparticipation and outcomes was examined by Cox proportional hazards regression for death and Andersen-Gill modeling for hospitalizations. Results: Among 7920 patients, 3438 responded to the survey (AAPOR RR#2= 43%). Numerous clinical and demographic differences between participants and non-participants were noted (Table), particularly for education, marital status, and neuropsychiatric conditions. After a mean (±SD) of 1.5±1 years after survey administration, 1,577 deaths and 5,873 hospitalizations had occurred. Nonparticipation was associated with a two-fold increased risk of death (HR: 2.30; 95% CI: 2.06-2.57; p<0.001) and a 10% increased risk of hospitalizations (HR: 1.10; 95% CI: 1.01-1.21; p=0.03) after adjusting for age, sex, marital status, coronary disease, arrhythmia, hyperlipidemia, cancer, chronic kidney disease, depression and anxiety. Conclusions: In a large survey of patients with HF, participation was associated with large differences in clinical and demographic characteristics and outcomes. Examining the impact of participation is critical to draw inference from studies of patient reported measures.