Talk to the Veterans Crisis Line now
U.S. flag
An official website of the United States government

Health Services Research & Development

Go to the ORD website
Go to the QUERI website
HSRD Conference Logo



2023 HSR&D/QUERI National Conference Abstract

Printable View

4068 — Who was left out? Identifying the Post COVID-19 Digital Divide in Chronic Pain Care Using Ensembled Machine Learning.

Lead/Presenter: Evan Carey,  COIN - Seattle/Denver
All Authors: Carey EP (Seattle/Denver COIN, VA Eastern Colorado Health Care System), Gunzburger E (Seattle/Denver COIN, VA Eastern Colorado Health Care System) Blankenhorn R (Seattle/Denver COIN, VA Eastern Colorado Health Care System) Sayre G (Seattle/Denver COIN, VA Puget Sound Health Care System ) Mog A (Seattle/Denver COIN, VA Puget Sound Health Care System) Firestone C (VA Northeast Ohio Healthcare System) Stevenson L (VA Northeast Ohio Healthcare System)

Objectives:
The impact of COVID on specialty chronic pain care is unknown. We investigated (1) the impact of COVID on access to pain specialty care in VA and community settings, then (2) identified traits of Veterans experiencing larger than average losses in access to care.

Methods:
We identified a national cohort of Veterans diagnosed with Chronic pain and in primary care as of 2020-03-01 (n = 1,751,969), referred to as the ‘post COVID onset’ cohort. We implemented a machine learning ensemble modeling framework to identify Veteran level predicted probabilities of receiving pain specialty care between 2020-03-01 to 2020-11-01 in this cohort. We then used an identically defined cohort and follow-up from one year prior to build a model estimating pre-COVID care probabilities. Using these model predictions as the counterfactual, we identified Veteran level predicted loss of specialty pain care utilization for the ‘post COVID onset’ cohort for both VA and community care settings. Veterans with at least a 20% pre-COVID probability of care were examined and stratified into groups based on their change in probability of care: large (>0.25 absolute decrease), small (0 to 0.25 absolute decrease), and no decrease. Models were repeated for Seattle and Denver catchment areas, with Denver results presented here.

Results:
Denver results: Following COVID, any pain specialty care use declined by 29% compared to the year prior, translating to 534 Veterans not utilizing care. Among the 1,278 Veterans receiving pain care, 41% used VA, 63.5% used community care, and 3.8% used both. Among the 23,534 Veterans, we estimated 4.8% (1,132) were likely to receive pain specialty care in VA settings had they been seen pre-COVID. Among them, the predicted loss of access to care was large for 25% (n = 284), small for 65.8% (n = 745), and none for 9% (n = 103). Veterans with a higher predicted loss of access to VA pain care had lower mental health and other pain comorbidities and lower rates of substance use disorders. Among the 23,534 Veterans, we estimated 6.7% (1,586) were likely to receive pain specialty care in the community had they been seen pre-COVID. Among them, the predicted loss of access to care was large (>0.25 absolute decrease) for 4.4% (n = 70), small (0 to 0.25 absolute decrease) for 67.3% (n = 1,067), and none for 28% (n = 449). Veterans with a higher predicted loss of access to community pain care had higher mental health and other pain comorbidities and higher rates of substance use disorders. For both VA and community care, higher recent pain scores, recent prior pain care utilization and higher rates of long-term opioid use were associated with a lower predicted loss of access to care.

Implications:
Although VA purchased community pain care was more resilient to COVID interruptions on average, higher complexity Veterans were most likely to lose pain care if seen in the community, and least likely to lose pain care if seen in the VA.

Impacts:
Among community care reliant Veterans with complex chronic pain, ongoing access to pain care should monitored with VA provided pain care options targeted as access disparities develop.