The COVID-19 pandemic had a profound effect on American life, and disproportionately affected the lives and livelihoods of some of the country's most vulnerable residents.1 While demographic trends about populations who have been particularly vulnerable to COVID-19-related illness and death have been documented,1 there has been less certain ability to pinpoint why these trends occurred. Though many have speculated that housing conditions, access to and trust in good healthcare, high risk jobs, and other factors have played roles, these factors have been hard to document empirically for many vulnerable populations, including people with intellectual and developmental disabilities (IDD), who are the focal population of this study.

Research suggests that people with IDD may have been particularly vulnerable to serious outcomes from COVID-19, including an increased chance of being hospitalized or dying from the virus compared to the general population. A cross-sectional study of electronic health data from more than 500 health care organizations in the United States found that patients with IDD had higher odds of hospitalization and death following a COVID-19 diagnosis compared to patients without IDD.2 Furthermore, the odds of mortality for patients with IDD was significantly higher than for other identified high-risk conditions.2 Similar disparities were also found in large samples in the United Kingdom.3

Despite the strong evidence for IDD as a risk factor for poor outcomes related to COVID-19, relatively few studies have examined the associations between IDD and being diagnosed with the virus. Evidence is mixed about whether people with IDD are at an increased risk of contracting COVID-19, with some studies reporting a higher prevalence of infection among people with IDD2 (and others reporting lower infection rates compared to people without disabilities.4

What literature does exist has identified potential associations with being diagnosed with COVID-19 at the individual- and systems-levels. The same chronic health conditions and comorbidities that put a person at risk for severe outcomes from COVID-19 may also be associated with their risk of infection, though the direction of these relationships is not always clear. For example, a cross-sectional analysis of electronic medical records identified diabetes, chronic kidney disease, and immunosuppression to be significantly associated with COVID-19 infection risk.4 However, having a higher number of chronic health conditions was associated with a lower risk of infection, suggesting a possible interplay between personal health and risk reduction behaviors.5 A cohort analysis of people with IDD who had been diagnosed with COVID-19 identified age, chronic kidney disease, and Down Syndrome as factors associated with increased odds of being diagnosed with COVID-19.6

For people with IDD, understanding the ways that they live in and interact with the community may be particularly important in identifying factors associated with COVID-19 diagnosis. For instance, while an analysis of data on COVID-19 outcomes for Californians receiving IDD services found an overall lower case rate compared to people not receiving services, case rates varied considerably based on where a person lived, with higher rates in settings with more residents.7 Similarly, other research suggests people with IDD may be in regular contact with multiple support personnel, use shared transportation, and live in congregate care settings, all of which may increase their risk of exposure to COVID-19.2 Notably, research has also identified higher case fatality rates for people with IDD living in congregate care settings compared to those living independently or in a family home6 (Landes et al., 2021), highlighting the importance of understanding the relationship between a person's services and their risk of COVID-19 infection.

Given the evidence for a heightened risk of poor outcomes from COVID-19 for people with IDD and the potential for future public health emergencies, it is imperative to understand the factors that put people at risk of contracting COVID-19 in the first place. COVID-19 testing and surveillance tracking were challenging in the general population, especially in the early stages of the pandemic,8,9 and these challenges may have been exacerbated for people with IDD. For example, a lack of disability identifiers in testing and surveillance data made it impossible to make meaningful comparisons of case rates and outcomes between people with and without disabilities.10 As the pandemic progressed, the dearth of disability identifiers in testing data made it more difficult to link to other administrative datasets, including hospitalization and mortality data, to understand the ongoing impact of COVID-19 on people with disabilities.8

Understanding the factors that put people with IDD at risk of contracting COVID-19 has implications for public health and for promoting health equity. The COVID-19 pandemic highlighted how a lack of attention to the specific experiences of people with IDD can extend to state level policy that exacerbates existing inequities. Guidelines for reporting COVID-19 cases in nursing homes and other long term care facilities were not issued until April 2020, well after the virus had been present in the United States.8 Even when guidelines were available, some states did not prioritize congregate care settings for people with IDD at the same degree as nursing homes.5 This lack of guidelines may have contributed to a delay in policies to protect people living in congregate care settings; the federal government did not issue guidance for COVID-19 mitigation strategies in long term care facilities until May of 2020,5 months after it was apparent that such facilities appeared to have high rates of infection,11 leaving many people with IDD particularly vulnerable.

Despite the evidence that people with IDD may be at increased risk for poor outcomes from COVID-19 infection, relatively few studies have examined the factors that put people at risk for COVID-19 diagnosis in the first place. Understanding the factors that were most predictive of COVID-19 diagnosis among people with IDD could be useful in future public health planning to inform strategies for preventing disease infection for people with IDD, which is an important step toward addressing inequitable outcomes once infection occurs.

To do so, we aimed to approach major IDD datasets inductively, without allowing existing theory or literature to bias our approach to analysis. To accomplish this, we used machine learning to inductively analyze a merged dataset specific to people with IDD to determine the factors that were most predictive of a singular outcome: having been diagnosed with COVID-19.

Using a representative, random sample of Medicaid Home and Community Based Services (HCBS; 1915(c)) users with IDD from one U.S. state, our study sought to explore these research questions.

To what extent can machine learning models accurately predict a documented COVID-19 diagnosis for people with IDD who use HCBS?

What factors contribute most to the prediction of a documented COVID-19 diagnosis for people with IDD who use HCBS?