Beyond Diagnostic Codes: A New Multiple Myeloma Algorithm
Posted: Tuesday, March 26, 2019
Administrative databases have certain research advantages over clinical trials and tumor registries. Clinical trials include relatively small numbers of patients, following them over a short period, and often do not reflect real-world practice, whereas a tumor registry typically includes information from only a patient’s first therapy course, to name one drawback. On the other hand, “large, population-based data sets are particularly valuable when studying rare diseases such as multiple myeloma, because they may include enough individuals to identify a sufficient number of cases and to assess generalizability of the results,” wrote Kimberly J. Woodcroft, PhD, of Henry Ford Health System, Detroit, and colleagues. “However, it is critical to accurately identify cancer cases when using such data sources”—so the team set out to validate at least one code-based algorithm to identify multiple myeloma cases.
Pharmacoepidemiology & Drug Safety published the successful results. In several rounds of analyses, the team tested each algorithm candidate by applying it to a large cohort (a Henry Ford Health System tumor registry, which contributes to the Surveillance, Epidemiology, and End Results program) to generate a list of potential cases. Then, they compared that list with a list of cases from the Optum Research Database. The best algorithm generated in the study’s first stage had the “combination of the highest acceptable positive predictive value (PPV; 0.81) and sensitivity (0.73),” they noted. When it was validated in stage 2, its PPV was 0.86.
Neither diagnostic nor International Classification of Diseases, 9th Revision, Clinical Modification (ICD‐9‐CM) codes alone should be used to identify multiple myeloma in administrative data, Dr. Woodcroft and her co-investigators concluded. Rather, “an algorithm that includes ICD‐9‐CM codes preceding procedure codes for diagnostic tests for multiple myeloma followed by ICD‐9‐CM codes for multiple myeloma within a specific time window can identify valid cases of multiple myeloma in administrative data and achieve a reasonable PPV.” Its importance goes beyond the academic, the authors noted, with use possible “in future epidemiologic studies to investigate outcomes, survival, and treatment patterns.”
Disclosure: The study authors’ disclosure information can be found at onlinelibrary.wiley.com.
