Validity of Identification Methods of Lower Extremity Amputation in the Veterans Health Administration Electronic Medical Records

Abstract

It is not currently known what the best method to identify individuals with lower extremity amputation (LEA) is. The purpose of this study is to determine the positive predictive value (PPV) of algorithms used to identify patients with LEA using Veterans Health Administration (VHA) electronic medical records (EMR) and to determine if PPV varies by age, gender, and race. 685 patients identified as having at least one diagnosis or procedure code for LEA and being alive were mailed a survey and asked to provide self-reported LEA status. We received 441 (64%) responses. We calculated PPV estimates and false negative percentages for nine algorithms. Algorithm 1, allowing any procedure or diagnosis code for LEA, consistently had the lowest PPV estimate across both the entire sample and all subgroups. Algorithms requiring at least one procedure code or two or more diagnosis codes generally had high PPVs, with the algorithm with the highest PPV estimate varying among subgroups and for the entire sample. In general, increasing restriction in an algorithm resulted in a higher PPV and proportion of false negatives. Ultimately, the best choice of algorithm depends on specific study needs.