Abstract 5752: Identifying monoclonal gammopathy of undetermined significance from electronic health records in a large, integrated healthcare network

Abstract

Abstract Monoclonal gammopathy of undetermined significance (MGUS) is an obligate precursor to multiple myeloma (MM) and occurs among approximately 3% of US adults age ≥50 years. Use of data programmatically retrieved from electronic health records (EHR) may facilitate research to better understand risk factors for MGUS and the progression to MM. The aim of this study was to evaluate the accuracy of two EHR-based approaches for identifying MGUS cases. The study population included members of Kaiser Permanente Southern California (KPSC), an integrated healthcare network with more than 4.6 million members. Data were retrieved from KPSC's Surveillance, Epidemiology and End Results (SEER)-affiliated cancer registry as well as from comprehensive EHR, which contain chart notes from medical encounters, lab data, and diagnosis codes. Using KPSC's EHR, we identified potential MGUS cases programmatically using two approaches: serum monoclonal (M)-protein lab values and ICD-9 codes. For both approaches, ≥1 year of KPSC membership prior to the date of the qualifying lab test or diagnosis (“index date”) was required. We excluded members with a diagnosis of MM within 6 months after the index date using cancer registry data. For the M-protein approach, we included cases who met the following criteria: 1) a first recorded serum M-protein (identified using CPT code 84165) occurring after 2008 (when M-protein data became fully accessible electronically) and 2) an M-protein lab value of 0-3 g/dL from 2008-2014. For the ICD approach, we included cases with a first recorded ICD-9 diagnosis code for MGUS (273.1) from 2008-2014. For each approach, we randomly selected 100 cases to be confirmed via manual chart review of documentation +/- 6 months from the index date. A physician diagnosis of MGUS in the chart notes was considered the gold standard for confirmation. Positive predictive values (PPV) were calculated. The manual chart review for the M-protein approach indicated a low PPV (61.2%), and reviews were halted after 67 cases. To improve the accuracy, we expanded our exclusion criteria to any hematopoietic/lymphoid malignancy from 2007-2014. Among the remaining 54 chart-reviewed cases, 40 (74.1%) were confirmed as incident or prevalent MGUS cases, and 14 (25.9%) were classified as “other.” From the cases identified with ICD-9 codes, 92 remained after we applied the expanded exclusion criteria. Among those, 90 (97.8%) were confirmed with chart notes and 2 (2.2%) were not confirmed. We established an algorithmic approach to efficiently and accurately identify MGUS cases for population-based research. Our findings suggest that the ICD code approach has excellent PPV. The poor performance of the M-protein approach appears to be due both to a lack of physician documentation and the use of M-protein to monitor patients previously diagnosed with lymphoproliferative disorders. Further investigation is warranted to evaluate the generalizability of the ICD approach outside KPSC. Citation Format: Hilary C. Tanenbaum, Brenda Birmann, Kimberly Bertrand, Lauren Teras, Amrita Krishnan, Sophia Wang, Chun Chao. Identifying monoclonal gammopathy of undetermined significance from electronic health records in a large, integrated healthcare network [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5752.