Billing code algorithms to identify cases of peripheral artery disease from administrative data

Abstract

ObjectiveTo construct and validate billing code algorithms for identifying patients with peripheral arterial disease (PAD).MethodsWe extracted all encounters and line item details including PAD-related billing codes at Mayo Clinic Rochester, Minnesota, between July 1, 1997 and June 30, 2008; 22 712 patients evaluated in the vascular laboratory were divided into training and validation sets. Multiple logistic regression analysis was used to create an integer code score from the training dataset, and this was tested in the validation set. We applied a model-based code algorithm to patients evaluated in the vascular laboratory and compared this with a simpler algorithm (presence of at least one of the ICD-9 PAD codes 440.20–440.29). We also applied both algorithms to a community-based sample (n=4420), followed by a manual review.ResultsThe logistic regression model performed well in both training and validation datasets (c statistic=0.91). In patients evaluated in the vascular laboratory, the model-based code algorithm provided better negative predictive value. The simpler algorithm was reasonably accurate for identification of PAD status, with lesser sensitivity and greater specificity. In the community-based sample, the sensitivity (38.7% vs 68.0%) of the simpler algorithm was much lower, whereas the specificity (92.0% vs 87.6%) was higher than the model-based algorithm.ConclusionsA model-based billing code algorithm had reasonable accuracy in identifying PAD cases from the community, and in patients referred to the non-invasive vascular laboratory. The simpler algorithm had reasonable accuracy for identification of PAD in patients referred to the vascular laboratory but was significantly less sensitive in a community-based sample.

Corpus domain effects on distributional semantic modeling of medical terms

“…Using structured billing codes (a.k.a. claims data) for case ascertainment has been shown to have limitations in terms of accuracy and completeness, particularly in community-based samples (Bazarian et al, 2006;Fan et al, 2013;Pakhomov et al, 2007). Furthermore, a number of conditions (e.g.…”

Patient Stratification Using Electronic Health Records from a Chronic Disease Management Program

“…Table 1 shows that one may not be able to segregate patients effectively by using obvious risk factors for hypertension such as demographic features (sex, age, race), or the presence of common comorbidities, further supporting the fact that an automated method for identifying risk factors is needed in order to more effectively predict differences in patient outcomes. The nature of the disease may explain why the AUC of 0.71 found in the risk factor identification analysis is lower than in the application of logistic regression and feature selection towards more straightforward use cases such as the identification of rheumatoid arthritis or peripheral artery disease [32], [33], [53]. More importantly, this finding underscores the importance of identifying disease subtypes, each of which may exhibit hypertension with varying degrees of severity.…”

Novel screening metric for the identification of at-risk peripheral artery disease patients using administrative claims data

“…This algorithm was based on a thorough review of the literature. 4,10,[22][23][24] These codes indicate reasonable diagnostic accuracy in identifying PAD cases using administrative claims data. 22 The diagnosis and procedure codes used to identify PAD and non-PAD cases are listed in Appendix A.…”

“…4,10,[22][23][24] These codes indicate reasonable diagnostic accuracy in identifying PAD cases using administrative claims data. 22 The diagnosis and procedure codes used to identify PAD and non-PAD cases are listed in Appendix A. Patients were identified as PAD cases if they had at least one inpatient, outpatient, or professional claim with a primary or secondary diagnosis or procedure code listed for PAD in Supplementary Appendix A.…”

“…However, a thorough review of the literature led to the development of our claims-based algorithm to identify PAD, 4,10,22,23 which had been shown to have reasonable diagnostic accuracy. 22 Further studies to confirm the predictive value of this screening tool using an ABI would be useful. Our analyses were limited to the information available in the administrative claims data, which can be incomplete.…”