Impact of community health workers on diabetes management in an urban United States Community with high diabetes burden through the COVID pandemic

Abstract

Objective: Community Health Worker (CHW) interventions are promising approaches to increasing access to health care, garnering better health outcomes, and decreasing health inequities for historically marginalized populations. This study examines the impact of a health system-based CHW program embedded in the Diabetes Impact Project - Indianapolis Neighborhoods (DIP-IN), a large, place-based, multi-year intervention to reduce diabetes burden. We assessed the CHW program's effectiveness in managing glucose control and reducing diabetes-associated complications across the COVID timeline.

Methods: We examined the association between the CHW intervention and diabetes management in 454 CHW patients and 1,020 propensity score-matched comparison patients. Using electronic medical records for encounters between January 1, 2017, and March 31, 2022, we estimated the CHW program effect using a difference-in-difference approach through generalized linear mixed models.

Results: Participation was associated with a significant reduction (-0.54-unit (95 % CI: -0.73, -0.35) in glycosylated hemoglobin (A1C) on average over time that was beyond the change observed among comparison patients, higher odds of having ≥ 2 A1C measures in a year (OR = 2.32, 95 % CI: 1.79, 3.00), lower odds of ED visits (OR: 0.88; 95 % CI: 0.73, 1.05), and lower odds of hospital admission (OR: 0.81; 95 % CI: 0.60,1.09). When analyses were restricted to a pre-pandemic timeframe, the pattern of results were similar.

Conclusion: This program was effective in improving diabetes management among patients living in diabetes-burdened communities, and the effects were persistent throughout the pandemic timeline. CHW programs offer crucial reinforcement for diabetes management during periods when routine healthcare access is constrained.

Keywords: COVID-19; Chronic Disease; Community Health Workers; Diabetes Mellitus, Type 2; Glycated Hemoglobin; Health Equity; Health Promotion; Place-Based Intervention; Public Health Practice; Tertiary Prevention.

Fig. 1

Indianapolis health system patient analytic flowchart for A1C and hospital outcomes (2017–2022; Indiana, United States).

Fig. 2

Mean A1C by year (2017–2022) and analysis group (Comparison, DIP-IN pre-enrollment, DIP-IN post-enrollment) (Indianapolis, Indiana, United States).

Fig. 3

Of Indianapolis health system patients included in study with at least 1 A1C^a, proportion with ≥ 2 A1C measures by year (2017–2022) and analysis group (Comparison, DIP-IN pre-enrollment, DIP-IN post-enrollment) (Indiana, United States) ^a Required to be in the A1C analytic sample.

Fig. 4

Percent of Indianapolis health system patients included in study with hospital ED visit in six-month period (2017–2021) by analysis group (Comparison, DIP-IN pre-enrollment, DIP-IN post-enrollment) (Indiana, United States) ^a April 2019: Start of DIP-IN CHW program ^b March 2020: First COVID-19 case in Marion County, Indiana ^c April 2021: COVID-19 vaccine available for all adults in Indiana.

Fig. 5

Percent of Indianapolis health system patients included in study with hospital admission in six-month period (2017–2021) by analysis group (Comparison, DIP-IN pre-enrollment, DIP-IN post-enrollment) (Indiana, United States) ^a April 2019: Start of DIP-IN CHW program ^b March 2020: First COVID-19 case in Marion County, Indiana ^c April 2021: COVID-19 vaccine available for all adults in Indiana.

Fig. A1

Directed acyclic graph^a identifying confounders for A1C and hospital outcome analytic models (Indianapolis, Indiana, United States). Abbreviations: A1C, Glycated Hemoglobin; DIP-IN, Diabetes Impact Project – Indianapolis Neighborhoods. ^a Directed acyclic graph created using: Johannes Textor, Benito van der Zander, Mark K. Gilthorpe, Maciej Liskiewicz, George T.H. Ellison. Robust causal inference using directed acyclic graphs: the R package 'dagitty'. International Journal of Epidemiology 45(6):1887–1894, 2016. ^b The exposure is a green oval with an arrow inside, and the outcome is a blue oval with a vertical line inside. Adjusted variables are depicted with white ovals. Ancestors of the exposure and outcome are depicted with a salmon oval – in this study, these are unobserved variables. A causal path is indicated with a green line, and a biasing path is indicated with a pink line. Race and ethnicity are a proxy for racism and marginalization. While we did include payor type in our models as an indicator of socioeconomic status, we assume too much residual confounding remains.