Gathering clinical events in pragmatic clinical trials may be aided by patient-reported health data. However, there needs to be more information about the viability of using patient-reported data in population-based studies of health. The purpose of this pragmatic clinical experiment was to assess the degree of agreement between patient-reported and EHR-based characteristics. A concordance substudy of a comparative effectiveness research trial utilized data from the pragmatic clinical trial ADAPTABLE (Aspirin Dosing: A Patient-Centric Trial Assessing Benefits and Long-Term Effectiveness). Between April 2016 and June 30, 2019, the study randomly assigned 15,076 participants with preexisting atherosclerotic cardiovascular disease to receive either low or high-dose aspirin. Patient-reported data gathered through an online portal and data from electronic sources (electronic health record data) were compared across 4 categories (demographic characteristics, encounters, diagnoses, and procedures).
The overall agreement, sensitivity, specificity, positive predictive value, negative predictive value, and κ statistics with 95% CIs were calculated using the patient report as the criterion standard for demographic characteristics and the electronic health record as the criterion standard for clinical outcomes. Out of the 15,076 patients for whom data was available, the median age was 67.6 (range, 21-99), and 68.7% were male. Using self-reported information as the gold standard, agreement (κ) was high for both black and white ancestry and current smoking status but was lower for the former. Hispanic ethnicity data from electronic health records were highly specific (99.6%) but less sensitive (82.5%). Patient reported clinical end points demonstrated low sensitivity for myocardial infarction (33%), stroke (34.2%), and significant bleeding (36.6%) compared to data from electronic health records. Myocardial infarction (40.7%), stroke (33.8%), and significant bleeding (21.9%) all had dismal levels of positive predictive value.
Positive predictive value and moderate agreement (κ=0.54) were found for coronary revascularization, the most concordant event by the data source. In all demographic parameters and in numerous clinical events. The sensitivity and chance-corrected agreement of patient-reported data obtained via an online portal for cardiovascular events were low to moderate in a concordance substudy of a large, pragmatic comparative effectiveness research trial. Findings imply more effort is required to perfect the incorporation of patient-reported health data into pragmatic research projects.