Effects of Neighborhood-level Data on Performance and Algorithmic Equity of a Model That Predicts 30-day Heart Failure Readmissions at an Urban Academic Medical Center.
Algorithmic equity
congestive heart failure
hospital readmission
Journal
Journal of cardiac failure
ISSN: 1532-8414
Titre abrégé: J Card Fail
Pays: United States
ID NLM: 9442138
Informations de publication
Date de publication:
09 2021
09 2021
Historique:
received:
25
01
2021
revised:
24
03
2021
accepted:
26
04
2021
pubmed:
29
5
2021
medline:
20
11
2021
entrez:
28
5
2021
Statut:
ppublish
Résumé
Socioeconomic data may improve predictions of clinical events. However, owing to structural racism, algorithms may not perform equitably across racial subgroups. Therefore, we sought to compare the predictive performance overall, and by racial subgroup, of commonly used predictor variables for heart failure readmission with and without the area deprivation index (ADI), a neighborhood-level socioeconomic measure. We conducted a retrospective cohort study of 1316 Philadelphia residents discharged with a primary diagnosis of congestive heart failure from the University of Pennsylvania Health System between April 1, 2015, and March 31, 2017. We trained a regression model to predict the probability of a 30-day readmission using clinical and demographic variables. A second model also included the ADI as a predictor variable. We measured predictive performance with the Brier Score (BS) in a held-out test set. The baseline model had moderate performance overall (BS 0.13, 95% CI 0.13-0.14), and among White (BS 0.12, 95% CI 0.12-0.13) and non-White (BS 0.13, 95% CI 0.13-0.14) patients. Neither performance nor algorithmic equity were significantly changed with the addition of the ADI. The inclusion of neighborhood-level data may not reliably improve performance or algorithmic equity.
Sections du résumé
BACKGROUND
Socioeconomic data may improve predictions of clinical events. However, owing to structural racism, algorithms may not perform equitably across racial subgroups. Therefore, we sought to compare the predictive performance overall, and by racial subgroup, of commonly used predictor variables for heart failure readmission with and without the area deprivation index (ADI), a neighborhood-level socioeconomic measure.
METHODS AND RESULTS
We conducted a retrospective cohort study of 1316 Philadelphia residents discharged with a primary diagnosis of congestive heart failure from the University of Pennsylvania Health System between April 1, 2015, and March 31, 2017. We trained a regression model to predict the probability of a 30-day readmission using clinical and demographic variables. A second model also included the ADI as a predictor variable. We measured predictive performance with the Brier Score (BS) in a held-out test set. The baseline model had moderate performance overall (BS 0.13, 95% CI 0.13-0.14), and among White (BS 0.12, 95% CI 0.12-0.13) and non-White (BS 0.13, 95% CI 0.13-0.14) patients. Neither performance nor algorithmic equity were significantly changed with the addition of the ADI.
CONCLUSIONS
The inclusion of neighborhood-level data may not reliably improve performance or algorithmic equity.
Identifiants
pubmed: 34048918
pii: S1071-9164(21)00193-7
doi: 10.1016/j.cardfail.2021.04.021
pmc: PMC8434976
mid: NIHMS1708647
pii:
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
965-973Subventions
Organisme : NLM NIH HHS
ID : F31 LM013403
Pays : United States
Organisme : NHLBI NIH HHS
ID : K23 HL128837
Pays : United States
Organisme : NHLBI NIH HHS
ID : K23 HL141639
Pays : United States
Informations de copyright
Copyright © 2021 Elsevier Inc. All rights reserved.
Références
Am J Cardiol. 2004 Jul 1;94(1):20-4
pubmed: 15219502
Ann Intern Med. 2018 Dec 18;169(12):866-872
pubmed: 30508424
J Stat Softw. 2010;33(1):1-22
pubmed: 20808728
Am J Med Qual. 2018 Sep/Oct;33(5):493-501
pubmed: 29357679
Hastings Cent Rep. 2020 May;50(3):46-49
pubmed: 32468631
AMA J Ethics. 2019 Feb 1;21(2):E167-179
pubmed: 30794127
Bioinformatics. 2012 Jan 1;28(1):112-8
pubmed: 22039212
Ann Intern Med. 2014 Dec 2;161(11):765-74
pubmed: 25437404
Proc Natl Acad Sci U S A. 2020 Jun 9;117(23):12592-12594
pubmed: 32457147
Lancet Digit Health. 2020 May;2(5):e221-e223
pubmed: 33328054
J Gen Intern Med. 2015 Jun;30(6):719-23
pubmed: 25527336
Int J Cardiol. 2019 Mar 1;278:186-191
pubmed: 30579719
Health Serv Res. 2019 Apr;54(2):327-336
pubmed: 30848491
Ann Epidemiol. 2020 Mar;43:37-43.e4
pubmed: 32151518
Epidemiology. 2010 Jan;21(1):128-38
pubmed: 20010215
J Card Fail. 2021 May;27(5):522-532
pubmed: 33358958
JAMA. 2019 Jul 9;322(2):113-114
pubmed: 31169890
J Biomed Inform. 2016 Dec;64:10-19
pubmed: 27658885
Med Care. 2010 Nov;48(11):981-8
pubmed: 20940649
J Gen Intern Med. 2017 Jan;32(1):71-80
pubmed: 27848189
Stroke. 2014 Jun;45(6):1716-20
pubmed: 24736237
Nat Methods. 2018 Apr;15(4):233-234
pubmed: 30100822
Am J Kidney Dis. 2020 Dec;76(6):749-751
pubmed: 33077316
Med Care. 2007 Aug;45(8):789-805
pubmed: 17667314
J Am Heart Assoc. 2020 May 5;9(9):e014836
pubmed: 32316807
Science. 2019 Oct 25;366(6464):421-422
pubmed: 31649182
Crit Care. 2015 Aug 13;19:285
pubmed: 26268570
Heart. 2014 Jan;100(1):60-7
pubmed: 24186564
J Clin Epidemiol. 2016 Jan;69:245-7
pubmed: 25981519
J Health Soc Behav. 1995;Spec No:80-94
pubmed: 7560851
BMC Psychiatry. 2014 Oct 17;14:289
pubmed: 25322690
Int J Cardiol. 2019 Mar 1;278:167-172
pubmed: 30587417
J Womens Health (Larchmt). 2007 Mar;16(2):245-55
pubmed: 17388741
Circ Cardiovasc Qual Outcomes. 2010 Mar;3(2):181-7
pubmed: 20124526
Circulation. 2020 Jun 2;141(22):e841-e863
pubmed: 32349541
JAMA Cardiol. 2019 Dec 1;4(12):1189-1190
pubmed: 31596433
Science. 2019 Oct 25;366(6464):447-453
pubmed: 31649194
Crit Care Med. 2020 Jun;48(6):808-814
pubmed: 32271185