Predictive models for low birth weight: a comparative analysis of algorithmic fairness-improving approaches.
Uložené v:
| Názov: | Predictive models for low birth weight: a comparative analysis of algorithmic fairness-improving approaches. |
|---|---|
| Autori: | Brown CC; University of Arkansas for Medical Sciences, 4301 W Markham St, Slot #820-12, Little Rock, AR 72205. Email: cbrown3@uams.edu., Gomez-Acevedo H, Amick BC 3rd, Tilford JM, Bryant-Moore K, Thomsen M |
| Zdroj: | The American journal of managed care [Am J Manag Care] 2025 May 01; Vol. 31 (5), pp. e132-e137. Date of Electronic Publication: 2025 May 01. |
| Spôsob vydávania: | Journal Article; Comparative Study |
| Jazyk: | English |
| Informácie o časopise: | Publisher: Clinical Care Targeted Communications Group, LLC Country of Publication: United States NLM ID: 9613960 Publication Model: Electronic Cited Medium: Internet ISSN: 1936-2692 (Electronic) Linking ISSN: 10880224 NLM ISO Abbreviation: Am J Manag Care Subsets: MEDLINE |
| Imprint Name(s): | Publication: Cranbury, NJ : Clinical Care Targeted Communications Group, LLC Original Publication: Old Bridge, NJ : American Medical Pub., c1995- |
| Výrazy zo slovníka MeSH: | Algorithms* , Infant, Low Birth Weight*, Humans ; Retrospective Studies ; Infant, Newborn ; Cross-Sectional Studies ; Female ; Male ; Birth Certificates ; Arkansas ; Adult |
| Abstrakt: | Objective: Evaluating whether common algorithmic fairness-improving approaches can improve low-birth-weight predictive model performance can provide important implications for population health management and health equity. This study aimed to evaluate alternative approaches for improving algorithmic fairness for low-birth-weight predictive models. Study Design: Retrospective, cross-sectional study of birth certificates linked with medical insurance claims. Methods: Birth certificates (n = 191,943; 2014-2022) were linked with insurance claims (2013-2021) from the Arkansas All-Payer Claims Database to assess alternative approaches for algorithmic fairness in predictive models for low birth weight (< 2500 g). We fit an original model and compared 6 fairness-improving approaches using elastic net models trained and tested with 70/30 balanced random split samples and 10-fold cross validation. Results: The original model had lower accuracy (percent predicted correctly) in predicting low birth weight among Black, Native Hawaiian/Other Pacific Islander, Asian, and unknown racial/ethnic populations relative to White individuals. For Black individuals, accuracy increased with all 6 fairness-improving approaches relative to the original model; however, sensitivity (true-positives correctly predicted as low birth weight) significantly declined, as much as 31% (from 0.824 to 0.565), in 5 of 6 approaches. Conclusions: When developing and implementing decision-making algorithms, it is critical that model performance metrics align with management goals for the predictive tool. In our study, fairness-improving models improved accuracy and area under the curve scores for Black individuals but decreased sensitivity and negative predictive value, suggesting that the original model, although unfair, was not improved. Implementation of unfair models for allocating preventive services could perpetuate racial/ethnic inequities by failing to identify individuals most at risk for a low-birth-weight delivery. |
| References: | J Midwifery Womens Health. 2016 Nov;61(6):707-720. (PMID: 27737504) Health Aff (Millwood). 2022 Feb;41(2):212-218. (PMID: 35130064) Med Care. 2007 Nov;45(11):1034-42. (PMID: 18049343) Matern Child Health J. 2013 Feb;17(2):299-306. (PMID: 22392602) Health Aff (Millwood). 2023 Oct;42(10):1359-1368. (PMID: 37782868) Obstet Gynecol. 2020 Sep;136(3):440-449. (PMID: 32769656) Am J Obstet Gynecol. 2022 Dec;227(6):893.e1-893.e15. (PMID: 36113576) NPJ Digit Med. 2020 Jul 30;3:99. (PMID: 32821854) J Womens Health (Larchmt). 2019 Jul;28(7):919-928. (PMID: 31259671) JAMA Intern Med. 2018 Nov 1;178(11):1544-1547. (PMID: 30128552) JAMA Netw Open. 2023 Dec 1;6(12):e2345050. (PMID: 38100101) BMC Pregnancy Childbirth. 2023 Aug 31;23(1):628. (PMID: 37653383) J Perinatol. 2018 Aug;38(8):997-1008. (PMID: 29593355) JMIR Med Inform. 2022 May 31;10(5):e36388. (PMID: 35639450) MMWR Surveill Summ. 2018 Jan 19;67(1):1-16. (PMID: 29346340) JAMA Health Forum. 2022 Jun 10;3(6):e221398. (PMID: 35977238) Anesthesiology. 2011 Nov;115(5):963-72. (PMID: 21934482) Natl Vital Stat Rep. 2023 Jan;72(1):1-53. (PMID: 36723449) BMC Pregnancy Childbirth. 2023 Aug 22;23(1):600. (PMID: 37608358) PLoS One. 2021 Oct 27;16(10):e0259258. (PMID: 34705872) J Community Health. 2003 Oct;28(5):335-46. (PMID: 14535599) J Obstet Gynecol Neonatal Nurs. 2006 Jan-Feb;35(1):3-12. (PMID: 16466348) Matern Child Health J. 2015 Mar;19(3):643-50. (PMID: 25138628) Science. 2019 Oct 25;366(6464):447-453. (PMID: 31649194) Med Care. 1998 Jan;36(1):8-27. (PMID: 9431328) |
| Grant Information: | K01 MD018072 United States MD NIMHD NIH HHS; R01 DK125641 United States DK NIDDK NIH HHS; R01 MH133857 United States MH NIMH NIH HHS; U54 TR001629 United States TR NCATS NIH HHS |
| Entry Date(s): | Date Created: 20250519 Date Completed: 20250519 Latest Revision: 20250531 |
| Update Code: | 20250531 |
| PubMed Central ID: | PMC12109546 |
| DOI: | 10.37765/ajmc.2025.89737 |
| PMID: | 40387721 |
| Databáza: | MEDLINE |
Full Text Finder 
Nájsť tento článok vo Web of Science | Abstrakt: | Objective: Evaluating whether common algorithmic fairness-improving approaches can improve low-birth-weight predictive model performance can provide important implications for population health management and health equity. This study aimed to evaluate alternative approaches for improving algorithmic fairness for low-birth-weight predictive models.<br />Study Design: Retrospective, cross-sectional study of birth certificates linked with medical insurance claims.<br />Methods: Birth certificates (n = 191,943; 2014-2022) were linked with insurance claims (2013-2021) from the Arkansas All-Payer Claims Database to assess alternative approaches for algorithmic fairness in predictive models for low birth weight (< 2500 g). We fit an original model and compared 6 fairness-improving approaches using elastic net models trained and tested with 70/30 balanced random split samples and 10-fold cross validation.<br />Results: The original model had lower accuracy (percent predicted correctly) in predicting low birth weight among Black, Native Hawaiian/Other Pacific Islander, Asian, and unknown racial/ethnic populations relative to White individuals. For Black individuals, accuracy increased with all 6 fairness-improving approaches relative to the original model; however, sensitivity (true-positives correctly predicted as low birth weight) significantly declined, as much as 31% (from 0.824 to 0.565), in 5 of 6 approaches.<br />Conclusions: When developing and implementing decision-making algorithms, it is critical that model performance metrics align with management goals for the predictive tool. In our study, fairness-improving models improved accuracy and area under the curve scores for Black individuals but decreased sensitivity and negative predictive value, suggesting that the original model, although unfair, was not improved. Implementation of unfair models for allocating preventive services could perpetuate racial/ethnic inequities by failing to identify individuals most at risk for a low-birth-weight delivery. |
|---|---|
| ISSN: | 1936-2692 |
| DOI: | 10.37765/ajmc.2025.89737 |