Accuracy of Internet-Based Patient Self-Report of Postdischarge Health Care Utilization and Complications Following Orthopedic Procedures: Observational Cohort Study

BACKGROUND: The accuracy of patient self-report of health care utilization and complications has yet to be determined. If patients are accurate and engaged self-reporters, collecting this information in a manner that is temporally proximate to the health care utilization events themselves may prove valuable to health care organizations undertaking quality improvement initiatives for which such data are often unavailable. OBJECTIVE: The objective of this study was to measure the accuracy of patient self-report of health care utilization and complications in the 90 days following orthopedic procedures using an automated digital patient engagement platform. METHODS: We conducted a multicenter real-world observational cohort study across 10 orthopedic practices in California and Nevada. A total of 371 Anthem members with claims data meeting inclusion criteria who had undergone orthopedic procedures between March 1, 2015, and July 1, 2016, at participating practices already routinely using an automated digital patient engagement platform for asynchronous remote guidance and telemonitoring were sent surveys through the platform (in addition to the other materials being provided to them through the platform) regarding 90-day postencounter health care utilization and complications. Their self-reports to structured survey questions of health care utilization and complications were compared to claims data as a reference. RESULTS: The mean age of the 371 survey recipients was 56.5 (SD 15.7) years, 48.8% (181/371) of whom were female; 285 individuals who responded to 1 or more survey questions had a mean age of 56.9 (SD 15.4) years and a 49.5% (141/285) female distribution. There were no significant differences in demographics or event prevalence rates between responders and nonresponders. With an overall survey completion rate of 76.8% (285/371), patients were found to have accuracy of self-report characterized by a kappa of 0.80 and agreement of 0.99 and a kappa of 1.00 and agreement of 1.00 for 90-day hospital admissions and pulmonary embolism, respectively. Accuracy of self-report of 90-day emergency room/urgent care visits and of surgical site infection were characterized by a kappa of 0.45 and agreement of 0.96 and a kappa of 0.53 and agreement of 0.97, respectively. Accuracy for other complications such as deep vein thrombosis, hemorrhage, severe constipation, and fracture/dislocation was lower, influenced by low event prevalence rates within our sample. CONCLUSIONS: In this multicenter observational cohort study using an automated internet-based digital patient engagement platform, we found that patients were most accurate self-reporters of 90-day hospital admissions and pulmonary embolism, followed by 90-day surgical site infection and emergency room/urgent care visits. They were less accurate for deep vein thrombosis and least accurate for hemorrhage, severe constipation, and fracture/dislocation. A total of 76.8% (285/371) of patients completed surveys without the need for clinical staff to collect responses, suggesting the acceptability to patients of internet-based survey dissemination from and collection by clinical teams. While our methods enabled detection of events outside of index institutions, assessment of accuracy of self-report for presence and absence of events and nonresponse bias analysis, low event prevalence rates, particularly for several of the complications, limit the conclusions that may be drawn for some of the findings. Nevertheless, this investigation suggests the potential that engaging patients in self-report through such survey modalities may offer for the timely and accurate measurement of matters germane to health care organizations engaged in quality improvement efforts post discharge.