PATIENT IDENTIFICATION FOR RARE DISEASES WITH RECENT ICD-10-CM DESIGNATION USING CLAIMS DATA
Author(s)
Queeny Ip, PharmD, PhD1, Pam Kumparatana, MPH2;
1Komodo Health, New York City, NY, USA, 2Komodo Health Inc., New York City, NY, USA
1Komodo Health, New York City, NY, USA, 2Komodo Health Inc., New York City, NY, USA
Presentation Documents
OBJECTIVES: Identifying sufficient rare disease patients in claims data is challenging, particularly for conditions with recently assigned codes. Using warm autoimmune hemolytic anemia (wAIHA) as a case study (ICD-10-CM code D59.11, introduced October 2022), we evaluated the performance of a commonly used claims-based algorithm by calculating key diagnostic metrics: sensitivity, specificity, Positive Predictive Value (PPV), and Negative Predictive Value (NPV).
METHODS: From an initial cohort of 178,428 Autoimmune Hemolytic Anemia (AIHA) patients using the Komodo Research Dataset January 1, 2016 to November 30, 2025, the commonly used algorithm (≥ 2 D59.11 codes ≥ 30 days apart) identified 7,731 potential wAIHA cases. Because direct antiglobulin test results were unavailable, two approaches (APP1 and APP2) estimated "true" cases. APP1 is pattern-based, leveraging diagnosing patterns pre/post wAIHA ICD introduction (e.g. D59.11 on or after October 1, 2022) and APP2 is treatment-based, including treatment data (corticosteroids, splenectomy, or immunosuppressants). Performance metrics were derived from the resulting true/false positives and negatives. AI tools used were validated by the quality check in the research process.
RESULTS: The algorithm demonstrated critically low sensitivity (APP1: 1.72%; APP2: 1.77%), identifying only 2,858 true positives while failing to capture the vast majority (False Negatives: APP1: 158,531; APP2: 162,829). Specificity was moderate (APP1: 61.75%; APP2: 71.40%). The PPV was 36.97% for both approaches, indicating that only about 37% of flagged patients were likely true cases. NPV was also low (APP1: 4.75%; APP2: 7.54%).
CONCLUSIONS: Requiring ≥ 2 D59.11 codes ≥ 30 days apart is insufficient as a standalone rule for identifying wAIHA patients. The low sensitivity suggests a mismatch with clinical coding practices, where many true cases are recorded with single or clustered codes. To achieve an acceptable capture rate, researchers may consider relaxing temporal requirements or augmenting definitions with broader clinical context.
METHODS: From an initial cohort of 178,428 Autoimmune Hemolytic Anemia (AIHA) patients using the Komodo Research Dataset January 1, 2016 to November 30, 2025, the commonly used algorithm (≥ 2 D59.11 codes ≥ 30 days apart) identified 7,731 potential wAIHA cases. Because direct antiglobulin test results were unavailable, two approaches (APP1 and APP2) estimated "true" cases. APP1 is pattern-based, leveraging diagnosing patterns pre/post wAIHA ICD introduction (e.g. D59.11 on or after October 1, 2022) and APP2 is treatment-based, including treatment data (corticosteroids, splenectomy, or immunosuppressants). Performance metrics were derived from the resulting true/false positives and negatives. AI tools used were validated by the quality check in the research process.
RESULTS: The algorithm demonstrated critically low sensitivity (APP1: 1.72%; APP2: 1.77%), identifying only 2,858 true positives while failing to capture the vast majority (False Negatives: APP1: 158,531; APP2: 162,829). Specificity was moderate (APP1: 61.75%; APP2: 71.40%). The PPV was 36.97% for both approaches, indicating that only about 37% of flagged patients were likely true cases. NPV was also low (APP1: 4.75%; APP2: 7.54%).
CONCLUSIONS: Requiring ≥ 2 D59.11 codes ≥ 30 days apart is insufficient as a standalone rule for identifying wAIHA patients. The low sensitivity suggests a mismatch with clinical coding practices, where many true cases are recorded with single or clustered codes. To achieve an acceptable capture rate, researchers may consider relaxing temporal requirements or augmenting definitions with broader clinical context.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
SA18
Topic
Study Approaches
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, SDC: Rare & Orphan Diseases