Hearsay: Is patient self-report of health care utilization accurate?

Presenter: Teresa Longobardi, University of Manitoba

Abstract

Background
In the absence of centralized data repository for health care contacts, studies concerned with the utilization of health care services, and attendant costs, often depend on clinical charts. When it is prohibitively costly to obtain a complete record of clinical data, self-report by patients is an alternative. This study will test the accuracy of self-report by a population-based sample of patients with Inflammatory Bowel Disease (IBD).

Method
The Manitoba IBD Cohort Study is a 5-year longitudinal cohort study of subjects diagnosed within 7 years of enrollment. Participants undergo annual in-person interviews. During the 12-month in-person interview, 352 patients were asked to report the number of overnight hospitalizations and physician contacts they had in the prior 12 months and their prescription drug utilization in the past month. They were asked to indicate whether the utilization was for IBD-specific reasons or not. Manitoba Health, the single comprehensive provincial health insurer, maintains an administrative database that employs the universal classification system to classify diagnoses and procedures associated with health care utilization. All records for Cohort Study cases were extracted from this database for contrast against self-report of health care utilization. The calculations of specificity and sensitivity of self-report to administrative data were used to summarize the accuracy of self-report.

Results
Administrative health care records were extracted for the 352 Cohort Study participants, up to 12 months from the date of the interview. 46% of cases were identified as having contacted a physician for an IBD reason. These findings were collated with interview data to determine that subjects reported with 79% sensitivity and 72% specificity. Hospitalizations for IBD reasons were reported by 11% of cases. They reported with 75% sensitivity and 97% specificity. Non-IBD hospitalizations were reported by 17% of cases with 91% sensitivity and 97% specificity. These measures did not change significantly after allowing for a 6-month leeway. There was no detectable pattern in the demographics of cases that incorrectly reported utilization of health care services.

Conclusions: The validity of self-report, as it informs the estimation of health care costs, is critical. This study has drawn attention to the potential inaccuracy of recall of IBD patients over a time period of 12 months and cautions investigators. Interview tools and interviewing techniques may be devised to ward off inaccuracies. However, administrative data is a superior source of data and is better used than self-report to determine impact on utilization of any interventions.