Connect

Member Research and Reports

Member Research and Reports

UAB Estimates Predictive Value of Claims-based Algorithms in Older Women with Osteoporosis

“Validation of claims-based algorithms to identify outcomes of interest is becoming more important with the increasing use of administrative claims data,” says Dr. Nicole Wright, assistant professor in the department of epidemiology at the University of Alabama at Birmingham. “The objective of our recent study was to estimate the validity, using positive predictive value, of the claims-based algorithms to identify two adverse events of interest in older women with osteoporosis enrolled in Medicare. This work will inform the findings of the 10-year long-term safety of osteoporosis medications project being conducted at UAB.” Co-investigators in the project are department colleagues Mr. Tarun Arora, statistician, and Dr. Elizabeth S. Delzell, professor emeritus; Mr. Wilson K. Smith, statistical analyst in UAB’s department of medicine; Dr. Jeffrey R. Curtis, assistant professor, and Dr. Kenneth G. Saag, professor, in the division of clinical immunology and rheumatology; Dr. Meredith L. Kilgore, professor and chair in the department of health care organization and policy; and Dr. Monika M. Safford, professor in the division of preventive medicine.

WrightN_UAB_ASPPH
[Photo: Dr. Nicole Wright]

Using Medicare’s national 5 percent sample data from 2006 to 2008, Dr. Wright and her team used ICD-9 (International Classification of Diseases, revision nine) diagnosis codes to identify possible hypersensitivity (ICD-9 from inpatient or ER claims: 995.0, 995.2, or 995.3 ) and osteonecrosis of the jaw (ONJ) (ICD-9 from any claim: 522.7, 526.4, 526.5, or 733.45). Medical records for the potential cases were requested, and all retrieved records were reviewed by experts to determine case status. The positive predictive value (PPV) was calculated as the number of confirmed cases divided by the total number of retrieved records with sufficient information.

A total of six of the 84 potential ONJ cases were confirmed, resulting in a PPV (95 percent confidence interval [CI]) of 7.1 percent (2.7, 14.9); 95 of the 174 retrieved potential hypersensitivity records were corroborated, resulting in a PPV (95 percent CI) of 76.0 percent (67.5, 83.2). Dr. Wright indicated the team was happy to see that the hypersensitivity algorithm has moderately high validity but that more work is needed to optimize the algorithm to identify ONJ with high validity. As Dr. Wright said, “ONJ is a rare, yet serious adverse event associated with long-term use of certain osteoporosis medications. Having an algorithm that identifies true cases at high validity can help the field to identify additional risk factors that may be associated with the ONJ, so that therapy can be modified in high risk patients.”

“The Validity of Claims-Based Algorithms to Identify Serious Hypersensitivity Reactions and Osteonecrosis of the Jaw” was published in July in PLOS ONE.

Journal article: http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0131601