A version of this article was published on HCP Live.
A recent study showcased the feasibility of utilizing an algorithm to accurately detect pregnant patients with specific inherited bleeding disorders in an electronic health record (EHR). The findings were unveiled at the 2023 American Society of Hematology (ASH) Annual Meeting and Exposition held in San Diego, California.
Takeaways
- The study demonstrated the effectiveness of an algorithm in accurately identifying pregnant patients with specific inherited bleeding disorders within electronic health records (EHR), showcasing potential advancements in healthcare data analysis.
- Prior research highlighted challenges in relying solely on diagnosis codes for EHR searches, leading to a high number of false positives. This limitation prompted the need for a more accurate and efficient approach to identification.
- Traditional methods of ensuring data integrity through in-depth chart reviews often limited researchers to local institutional levels. The study emphasized the importance of overcoming these limitations to leverage national research infrastructure like the National Patient-Centered Clinical Research Network (PCORnet).
- The research team refined the initial algorithm based on retrospective queries, manual chart reviews, and a local registry. This refinement process aimed to enhance accuracy and address issues such as contamination in certain ICD-diagnosis codes.
- The revised algorithm demonstrated a high level of accuracy in identifying pregnant women with inherited bleeding disorders, with a sensitivity of 97.0% and a positive predicted value (PPV) of 91.4%. This suggests the potential for improved diagnostic capabilities in identifying specific health conditions during pregnancy.
The lead investigator, Ming Y Lim, MBBCh, an associate professor in the Division of Hematology and Hematologic Malignancies at the University of Utah, along with her colleagues, highlighted that relying solely on diagnosis codes for EHR searches led to a significant number of false positive identifications based on earlier research.
Lim emphasized, “To enhance data integrity, comprehensive chart reviews are often necessary, typically limited to a local institutional level. This restriction impedes the utilization of national research infrastructure like the National Patient-Centered Clinical Research Network (PCORnet), which draws data from millions of EHRs across healthcare institutions in the United States.”
In response, Lim and her team sought to assess the diagnostic accuracy of an algorithm designed to utilize common data definitions along with multiple data elements in the EHR to identify affected patients. The study focused on data from pregnant women with inherited blood disorders who delivered at a University of Utah-affiliated hospital between January 2016 and December 2020. The inherited blood disorders included hemophilia and hemophilia carriers, von Willebrand disease, and other rare bleeding disorders. Data elements aligned with the PCORnet Common Data Model (CDM) to ensure consistency in data definitions and formats across multiple sites.
The initial algorithm version incorporated ICD-9/10 codes for inherited bleeding disorders, medications for managing these disorders, and coagulation factor test results. Confirmation of inherited bleeding disorders required meeting the criteria for either codes and medications, or codes and coagulation factor test results.
However, a revision was introduced based on retrospective queries, confirmed through manual chart reviews and the local registry, containing demographic and laboratory data of patients with an inherited bleeding disorder diagnosis treated at a federally funded adult hemophilia treatment center.
The original algorithm (query 1.0) identified 301 pregnant women meeting criterion 1 with 1 or more live birth or fetal death during the study period. Among them, 25 patients fulfilled criteria 2 and/or 3, with the remaining 276 cases verified using the registry or manual chart review.
Results indicated that certain ICD-diagnosis codes caused contamination and were inadequate for diagnosing carriers of bleeding disorders or those with rarer bleeding disorders. Consequently, these codes were removed, and a fourth criterion was added in the revised version. This criterion stated that for cases fulfilling criterion 1 but not 2 or 3, a diagnosis was confirmed if a patient had 2 or more identical ICD diagnosis codes for an inherited bleeding disorder at 2 or more separate visit types.
In the revised algorithm (query 1.1), 35 pregnant women were identified with inherited bleeding disorders, with 32 confirmed as true positives. The 3 initially misdiagnosed individuals underwent further laboratory testing and were subsequently ruled out. The revised algorithm exhibited a sensitivity of 97.0% and a positive predicted value (PPV) of 91.4%.
This article was written with help from ChatGPT
Reference
Lim MY, Sivaloganathan V, and Simonsen SE. Developing an Algorithm to Better Identify Pregnant Women with Inherited Bleeding Disorders within Electronic Health Records. Presented at: ASH Annual Meeting and Exposition. San Diego, CA. December 9-12, 2023.
