Self-Report of Healthcare Utilization among Community-Dwelling Older Persons: A Prospective Cohort Study

BACKGROUND Self-reported data are often used for estimates on healthcare utilization in cost-effectiveness studies. OBJECTIVE To analyze older adults' self-report of healthcare utilization compared to data obtained from the general practitioners' (GP) electronic medical record (EMR) and to study the differences in healthcare utilization between those who completed the study, those who did not respond, and those lost to follow-up. METHODS A prospective cohort study was conducted among community-dwelling persons aged 70 years and above, without dementia and not living in a nursing home. Self-reporting questionnaires were compared to healthcare utilization data extracted from the EMR at the GP-office. RESULTS Overall, 790 persons completed questionnaires at baseline, median age 75 years (IQR 72-80), 55.8% had no disabilities in (instrumental) activities of daily living. Correlations between self-report data and EMR data on healthcare utilization were substantial for 'hospitalizations' and 'GP home visits' at 12 months intraclass correlation coefficient 0.63 (95% CI; 0.58-0.68). Compared to the EMR, self-reported healthcare utilization was generally slightly over-reported. Non-respondents received more GP home visits (p<0.05). Of the participants who died or were institutionalized 62.2% received 2 or more home visits (p<0.001) and 18.9% had 2 or more hospital admissions (p<0.001) versus respectively 18.6% and 3.9% of the participants who completed the study. Of the participants lost to follow-up for other reasons 33.0% received 2 or more home visits (p<0.01) versus 18.6 of the participants who completed the study. CONCLUSIONS Self-report of hospitalizations and GP home visits in a broadly 'healthy' community-dwelling older population seems adequate and efficient. However, as people become older and more functionally impaired, collecting healthcare utilization data from the EMR should be considered to avoid measurement bias, particularly if the data will be used to support economic evaluation.