OBJECTIVES: Medical records (MR) are an important source of retrospective bleed data in hemophilia studies, captured as individual bleed events and bleed rates (e.g. “5 bleeds since last visit”) in narrative text. Studies without access to MR narrative text are restricted to individual ICD-coded bleed events. A methodology to calculate annualized bleed rate (ABR) by incorporating bleed rates from narrative text is needed to prevent underestimating ABR in real-world studies.

METHODS: PicnicHealth is a data platform working directly with individuals who contribute their MR data to research. This observational cohort study includes individuals with hemophilia A (HA) followed from January 1, 2018 to December 31, 2022. ABRs were calculated within observation periods including ≥1 hematology visit every 18 months. Wilcoxon rank sum tests evaluated differences in median ABRs calculated from bleed events with and without rates. Analyses were also stratified by hemophilia severity (severe <1%, moderate 1-5%, mild 6-50% baseline factor VIII) and restricted to time on prophylactic medication.

RESULTS: This study included 367 individuals with HA. Mean age was 27 years and 72% had severe HA. Among all patients, median (Q1-Q3) ABR increased when calculated with bleed rates [0.88 (0.00-1.90)] versus without [0.72 (0.00-1.53)] (p=0.086). During time on prophylactic medication (N=291), differences were statistically significant with rates [0.80 (0.00-1.84)] versus without [0.58 (0.00-1.51)] (p=0.047). Within severity subgroups, ABR persistently increased when incorporating bleed rates.

CONCLUSIONS: Access to comprehensive bleed data in MRs is important for quantifying the burden of bleeding in the hemophilia community. This study found that incorporating bleed rates increased estimated ABRs, suggesting that access to MR narrative text is important to prevent underestimating ABR. Real-world hemophilia studies should consider leveraging MR narrative text in ABR estimations.