Abstract Title

Improving Risk-Adjustment Methods for Cardiac Patients by Using Present-On-Admission(POA) Data

Presenter Name

Saehwan Park

Abstract

The purpose of this study is to examine whether the predictive power of risk-adjustment models can be improved by incorporating Present-On-Admission (POA) codes for patients with coronary artery disease (CAD).

We studied POA-coded data from 174,909 CAD discharges of the Texas Hospital Inpatient Discharge Public Use Data File for 2012. Conditions could either be complications or comorbidities, depending on whether they were present on admission. Major complications were identified based on both frequency of occurrence and in-hospital mortality. We compared the performance of three logistic regression models in terms of mortality prediction and explanatory power.

We found these: Risk models including POA had greater explanatory power, by approximately 10% (pseudo-R2: 9.2% vs. 10.4%), compared with the baseline model which did not include POA information. Separation of complications and comorbidities revealed additional information for some conditions. When present-on-admission, Peptic Ulcer Disease was not a significant predictor of mortality (OR=1.05; p-value>0.9), but was highly significant when it occurred as a complication (OR=20.55; p-value<0.01).

Overall, we concluded that POA information can greatly improve the utility of administrative data for risk-adjustment for patients with CAD. Report cards of hospitals' risk-adjusted mortality rates for cardiac patients can increase their validity and end-user acceptance by incorporating POA information. However, more standardization and uniformity in POA coding is needed. Some hospitals may still be reluctant to report complications or, alternatively, chronic comorbidities (e.g., depression, alcoholism) may be misreported as complications.

Presentation Type

Poster

This document is currently not available here.

Share

COinS
 

Improving Risk-Adjustment Methods for Cardiac Patients by Using Present-On-Admission(POA) Data

The purpose of this study is to examine whether the predictive power of risk-adjustment models can be improved by incorporating Present-On-Admission (POA) codes for patients with coronary artery disease (CAD).

We studied POA-coded data from 174,909 CAD discharges of the Texas Hospital Inpatient Discharge Public Use Data File for 2012. Conditions could either be complications or comorbidities, depending on whether they were present on admission. Major complications were identified based on both frequency of occurrence and in-hospital mortality. We compared the performance of three logistic regression models in terms of mortality prediction and explanatory power.

We found these: Risk models including POA had greater explanatory power, by approximately 10% (pseudo-R2: 9.2% vs. 10.4%), compared with the baseline model which did not include POA information. Separation of complications and comorbidities revealed additional information for some conditions. When present-on-admission, Peptic Ulcer Disease was not a significant predictor of mortality (OR=1.05; p-value>0.9), but was highly significant when it occurred as a complication (OR=20.55; p-value<0.01).

Overall, we concluded that POA information can greatly improve the utility of administrative data for risk-adjustment for patients with CAD. Report cards of hospitals' risk-adjusted mortality rates for cardiac patients can increase their validity and end-user acceptance by incorporating POA information. However, more standardization and uniformity in POA coding is needed. Some hospitals may still be reluctant to report complications or, alternatively, chronic comorbidities (e.g., depression, alcoholism) may be misreported as complications.