Director, RWE Rare Disease Scientist Pfizer, United States
Background: Real-world evidence (RWE) studies and applications are increasingly being recognized to improve healthcare delivery and increase the speed of patient access to new drugs. Studies involving real-world data (RWD) must include understanding of the transparency of data collection and limitations before drawing any conclusions.
Objectives: We assessed the completeness of key data elements from two linked datasets in a subset of patients with hemophilia B (PwHB): Komodo Health (KH) claims database and PicnicHealth’s (PH) Hemophilia B RWD Cohort derived from provider medical records (MR) across a patient’s healthcare journey.
Methods: We linked 93 PwHB in the US from PH’s dataset to KH’s (January 2015 – April 2022). Both datasets had tokens created using first name, last name, date of birth and gender. The PH platform retrieves MRs directly from providers, digitizes MRs for patients across providers, and produces longitudinal RWD. We assessed the match rate to KH claims of the following metrics found in PH’s dataset: hemophilia factor medication and select labs routinely utilized in the management of PwHB. Additionally, we compared the overall completeness of select variables in PH and KH.
Results: Of the 90 patients found in PH’s platform with a hemophilia factor medication, 11 (12.2%) were found in KH with all their medications matched with PH’s platform, 18 (20.0%) were found with some of their medications in KH, 50 (55.6%) did not have any medications listed in Komodo. Additionally, 12 (13.3%) patients in PH were missing listed medications found in Komodo.
Of the 86 patients found in PH’s platform with a lab of interest claim, 2 (2.3%) were found in KH with all labs of interest, 23 (26.7%) were found with ≥ 1 of labs of interest in KH, 19 (22.1%) did not have any labs of interest listed in KH. Additionally, 42 (48.8%) patients in PH were missing ≥ 1 lab of interest found in KH.
Other variables found in both PH and KH were also assessed for general missingness (n = 93). In PH, 0 (0%) patients were missing ZIP codes for sites of healthcare encounters, 3 (3.2%) were missing all factor medication, 7 (7.5%) were missing all select labs, 10 (10.8%) were missing severity. In KH, 0 (0%) patients were missing ZIP codes, 52 (55.9%) were missing all factor medications, 25 (26.9%) were missing all select labs and 93 (100%) were missing severity.
Conclusions: While there will unlikely ever be a perfect source of information, claims linked with MR data present opportunities to create a complete patient journey. The ideal composite dataset should include MR data which provide more detailed information, such as severity and lab result values, together with claims which help to identify critical gaps.
