> ## Content Index
> Fetch the complete content index at: https://www.coios.me/llms.txt
> Use this file to discover other available public pages before exploring further.

# Care management algorithms trained on past utilisation under-select disadvantaged adults
- URL: https://www.coios.me/care-management-algorithms-trained-on-past-utilisation/
- Published: 2026-11-03T13:08:10.000Z
- Updated: 2026-11-03T17:56:11.000Z
- Description: Among 201,216 US adults, utilisation-trained algorithms under-selected disadvantaged groups, and a health status target nearly closed the Black-White gap.
- Author: Daniel Ryan
- Tags: Data, models and systems, What linked data reveals, AI in medicine, United States, #preprint, #refines, #item, #source-2026-11, #score-68

This simulation study used 201,216 US adults from linked health record and survey data in All of Us, 2017-2019\. Logistic regression, random forest and XGBoost models were tested, varying threshold, feature set or training objective one at a time. Fair or poor health exceeded high utilisation among Black participants (33% vs 22%), but not White participants (19% vs 25%). Disparities persisted at every selection threshold; at baseline, uninsured people had the largest equal opportunity difference (-0.148). Outpatient visits dominated the Black-White gap; dropping them gave an EOD of 0.001 (AUC 0.77), a health status objective 0.008 (AUC 0.69). Need rested only on self-reported general health, and results come from simulation, not deployed programmes.

*Why it is interesting: Algorithms that allocate scarce care often learn who used care, not who needed it. This study isolates which design choice drives that gap, and finds the training target matters most.*

Source

[medRxiv, 1 November 2026](https://doi.org/10.64898/2026.10.30.26366732?ref=coios.me)

Paper

Algorithmic design choices as determinants of health equity: a causal simulation study of healthcare resource allocation

Authors

Reed C, Espinoza ME

DOI

10.64898/2026.10.30.26366732

Type

Preprint

Design

Causal simulation on a cohort of 201,216 adults from linked EHR and survey data, All of Us, 2017-2019; three prediction models with one design choice varied at a time

Verdict

Refines prior evidence

Driver

[What linked data reveals](https://www.coios.me/d-linked-data/)

Driver

[AI in medicine](https://www.coios.me/d-ai-medicine/)