TY - JOUR
T1 - Development and Validation of the Summary Elixhauser Comorbidity Score for Use With ICD-10-CM-Coded Data Among Older Adults
AU - Mehta, Hemalkumar B.
AU - Li, Shuang
AU - An, Huijun
AU - Goodwin, James S.
AU - Alexander, G. Caleb
AU - Segal, Jodi B.
N1 - Publisher Copyright:
© 2022 American College of Physicians. All rights reserved.
PY - 2022/10/1
Y1 - 2022/10/1
N2 - Background: Older adults have many comorbidities contributing to mortality. Objective: To develop a summary Elixhauser (S-Elixhauser) comorbidity score to predict 30-day, in-hospital, and 1-year mortality in older adults using the 38 comorbidities operationalized by the Agency for Healthcare Research and Quality (AHRQ). Design: Retrospective cohort study. Setting: Medicare beneficiaries from 2017 to 2019. Patients: Persons hospitalized in 2018 (n = 899 844) and 3 disease-specific hospitalized cohorts. Measurements: Weights were derived for 38 comorbidities to predict 30-day, in-hospital, and 1-year mortality. The S-Elixhauser score was internally validated and calibrated. Individual Elixhauser comorbidity indicators (38 comorbidities), the modified application of the AHRQ-derived Elixhauser summary score, the Charlson comorbidity indicators (17 comorbidities), and the Charlson summary score were externally validated. The c-statistic was used to evaluate discrimination of a comorbidity score model. Results: The S-Elixhauser score was well calibrated and internally validated, with a c-statistic of 0.705 (95% CI, 0.703 to 0.707) in predicting 30-day mortality, 0.654 (CI, 0.651 to 0.657) for in-hospital mortality, and 0.743 (CI, 0.741 to 0.744) for 1-year mortality. In external validation of other comorbidity indices for 30-day mortality, the c-statistic was 0.711 (CI, 0.709 to 0.713) for the individual Elixhauser comorbidity indicators, 0.688 (CI, 0.686 to 0.690) for the AHRQ Elixhauser score, 0.696 (CI, 0.694 to 0.698) for the Charlson comorbidity indicators, and 0.690 (CI, 0.688 to 0.693) for the Charlson summary score. In 3 disease-specific populations, the discrimination of the S-Elixhauser score in predicting 30-day mortality ranged from 0.657 to 0.732. Limitation: Validation of the S-Elixhauser comorbidity score and head-to-head comparison with other comorbidity scores in an external population are needed to evaluate comparative performance. Conclusion: The S-Elixhauser comorbidity score is well calibrated and internally validated but its advantage over the AHRQ Elixhauser and Charlson summary scores is unclear.
AB - Background: Older adults have many comorbidities contributing to mortality. Objective: To develop a summary Elixhauser (S-Elixhauser) comorbidity score to predict 30-day, in-hospital, and 1-year mortality in older adults using the 38 comorbidities operationalized by the Agency for Healthcare Research and Quality (AHRQ). Design: Retrospective cohort study. Setting: Medicare beneficiaries from 2017 to 2019. Patients: Persons hospitalized in 2018 (n = 899 844) and 3 disease-specific hospitalized cohorts. Measurements: Weights were derived for 38 comorbidities to predict 30-day, in-hospital, and 1-year mortality. The S-Elixhauser score was internally validated and calibrated. Individual Elixhauser comorbidity indicators (38 comorbidities), the modified application of the AHRQ-derived Elixhauser summary score, the Charlson comorbidity indicators (17 comorbidities), and the Charlson summary score were externally validated. The c-statistic was used to evaluate discrimination of a comorbidity score model. Results: The S-Elixhauser score was well calibrated and internally validated, with a c-statistic of 0.705 (95% CI, 0.703 to 0.707) in predicting 30-day mortality, 0.654 (CI, 0.651 to 0.657) for in-hospital mortality, and 0.743 (CI, 0.741 to 0.744) for 1-year mortality. In external validation of other comorbidity indices for 30-day mortality, the c-statistic was 0.711 (CI, 0.709 to 0.713) for the individual Elixhauser comorbidity indicators, 0.688 (CI, 0.686 to 0.690) for the AHRQ Elixhauser score, 0.696 (CI, 0.694 to 0.698) for the Charlson comorbidity indicators, and 0.690 (CI, 0.688 to 0.693) for the Charlson summary score. In 3 disease-specific populations, the discrimination of the S-Elixhauser score in predicting 30-day mortality ranged from 0.657 to 0.732. Limitation: Validation of the S-Elixhauser comorbidity score and head-to-head comparison with other comorbidity scores in an external population are needed to evaluate comparative performance. Conclusion: The S-Elixhauser comorbidity score is well calibrated and internally validated but its advantage over the AHRQ Elixhauser and Charlson summary scores is unclear.
UR - http://www.scopus.com/inward/record.url?scp=85140272599&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85140272599&partnerID=8YFLogxK
U2 - 10.7326/M21-4204
DO - 10.7326/M21-4204
M3 - Article
C2 - 36095314
AN - SCOPUS:85140272599
SN - 0003-4819
VL - 175
SP - 1423
EP - 1430
JO - Annals of internal medicine
JF - Annals of internal medicine
IS - 10
ER -