A study of the agreement between patient self‐assessment and study personnel assessment of bleeding symptoms

BACKGROUND: Clinical trials investigating new platelet transfusion therapies frequently require the assessment of bleeding for the study outcome. These assessments are commonly performed by study personnel and can be time-consuming. The purpose of this study was to assess whether patients were able to reliably assess their bleeding status on a daily basis. STUDY DESIGN AND METHODS: Patients admitted to hospital to receive chemotherapy for acute leukemia or to undergo allogeneic peripheral blood progenitor cell transplant were included. Patients were given an introduction to a form for documenting the occurrence of 16 bleeding symptoms. Patients completed this form and were examined daily by a study assessor. A weekly health record review was also performed by a study assessor. The agreement between raters was determined by calculating the raw agreement, chance-corrected agreement, and chance-independent agreement. RESULTS: Thirty-five patients completed 458 assessment forms that were paired with 559 forms completed by a study assessor with 450 matched forms available for analysis (mean, 12.86 per patient). Agreement for most individual bleeding symptoms was high. Thirteen items had agreement greater than 90 percent and all items had agreement greater than 77 percent. The lowest agreement was seen for skin symptoms: petechiae (89.2%), purpura (80.9%), and ecchymosis (77.6%). The negative predictive value of patient self-assessment was high (range, 71.1%-100%) whereas the positive predictive value was lower (range, 0%-86.5%). CONCLUSION: The reliability was very good between patients and study assessors with the patients reporting excellent negative predictive value and variable positive predictive value.