Very early in the COVID-19 outbreak it became obvious that the virus’ incidence and associated mortality were disproportionately impacting Black and Hispanic people in the United States. Even as the pandemic shifted geographically, COVID-related deaths among Black and Hispanic patients far exceeded their proportion of the population. Addressing social determinants of health has come into stark focus during COVID-19, but health disparities pre-date the pandemic.
My colleagues Shaifali Ray, MHA, Vizient senior networks director, and Karyl Kopaskie, PhD, Sg2 associate principal, intelligence, recently shared our analysis of inpatient data from more than 550 Vizient-member hospitals across the nation from March–September 2020 that identified racial and ethnic disparities in patients’ risk of COVID-19 diagnosis and admission. How we leveraged that data can be applied to other health equity problems and to develop patient-centered and population-specific strategies for addressing disparities.
The patient trajectory
Health disparities are rarely confined to an individual encounter. Patients in an acute health episode—such as an infection or hospitalization—bring with them a history of prior comorbidities, exposure risks and obstacles that contributed to their current episode.
As hospitals and health systems analyze the course of an acute disease, it’s important to also consider the path that has brought the patient to their current point. Consider each step of the patient trajectory—not only what happens in the hospital, but which patients are diagnosed and admitted. And even prior to who is diagnosed, who is exposed to the virus. Each of these steps identifies a different population of interest, and each step may have its own distinct risk factors. Analyzing each step provides a clearer picture of what factors can alter a patient’s trajectory from exposure to outcome and where disparities appear.
The analysis found that Black and Hispanic patients are a far larger proportion of COVID-19 diagnoses than the population as a whole. In fact, there is more risk associated with race and ethnicity than there is associated with age or with comorbidities. Once a patient is admitted, however, the risk of needing ICU care and of mortality is far less variable by race and ethnicity. This shift suggests that the racial and ethnic disparities we see reflect differences in exposure risks and access to health care, and that the data that can further clarify this situation may come from factors outside of the COVID encounter and outside of the medical record.
Specific risks in a broader view
When hospitals expand their view to include data beyond a single acute episode, they gain the context of the patient and their environment. In our analysis, we extended our scope for risk factors in three directions: longitudinal factors from a patient’s prior care, environmental variables from the patient’s neighborhood, and intersectional factors to identify specific subsets of larger groups. In each case the additional factors and the broader view enabled deeper insights into how risk profiles are different for different groups.
Longitudinal data—Although hospitals have a wealth of claims data supplying diagnosis and procedure codes for an acute episode, it misses patient history that would show contributing factors. A longitudinal record—assembling both present and past data on each patient—provides this context. In our analysis we assembled data from all encounters since January 2019 to provide each established patient with at least a year of data prior to the start of the outbreak in 2020.
We found that prior comorbidities like diabetes, hypertension and obesity increased a patient’s risk of COVID diagnosis, and that those comorbidities were already more prevalent among Black and Hispanic patients.
As we noticed that prior comorbidities increased risk, we also saw that patients whose diabetes had been under control were at lower risk than those whose diabetes were uncontrolled. This suggests that the work hospitals did in past years to help patients manage chronic conditions is one of the most important interventions for the current crisis, and that chronic disease management in general will be a vital tool in reducing health disparities.
One interesting finding was that patients with new comorbidities were much more likely to be admitted and need ICU care. These new diagnoses might reflect the effects of COVID and not its causes, and it is only in the comparison of the current episode to the past data that this distinction becomes clear.
Environmental context—Publicly available data, including census data, population health statistics or unemployment data can describe a patient’s neighborhood and provide some context to potential obstacles to care or risks of exposure. In our analysis, we used patient zip codes mapped to census data and current unemployment rates.
We found that patients who live in a neighborhood with high poverty rates were at a higher risk for a COVID diagnosis, but those who live in areas with higher unemployment were at lower risk. This was a particularly strong correlation during the summer of 2020 when some areas shut down more than others and suggests that there may be risks associated with the workplace.
Awareness of the circumstances around patients’ lives may provide some hints as to some of the differences in their health risks, not just in a pandemic, but also in their vulnerability to chronic diseases and in obstacles to accessing care. These contextual details can highlight the factors where intervention could have real benefits.
Intersectional risks—In our analysis, when we looked at the risk factors one by one, it suggested that older patients and Black and Hispanic patients were at higher risk of getting COVID. A closer look, however, shows a lot of variability within these groups. Younger Black and Hispanic patients—those of working age especially—have extra risks that do not seem to affect patients over 60 to the same extent. And in those specific risks, Black and Hispanic women were at the highest risk around the 20–40 age range.
Without patient-specific employment data we can only hypothesize, but the fact that Black and Hispanic women of working age were experiencing higher rates of COVID when we had also found higher risks related to employment in our environmental context is suggestive. Data from the Bureau of Labor Statistics confirm that Black and Hispanic women are overrepresented in many of the essential worker roles and industries that were more likely to have continued work during the pandemic. These three findings taken together suggest a very specific area that needs a further look, and which might provide an actionable strategy for reducing disparities by addressing the mechanisms by which they are operating.
While this example is specific to COVID, the strategy is adaptable to many situations and provides a view of risk that might suggest concrete interventions. The risk of a too-broad analysis of risk leads to support for a too-broad intervention, which could entirely miss the specific needs of the most vulnerable populations.
In our analysis, we took advantage of several data sources outside the acute episode. We used the context to better interpret the specific risks to specific patients at specific points in their trajectory, and the relationship that each of those risks had to the overall view of a pandemic. From there the data suggests areas for further study that might provide concrete interventions to address those disparities.
As hospitals continue to identify and address issues of health equity, it will be essential that they consider information from outside of the single acute episode. Both longer timelines and data sources that provide socioeconomic context outside the medical record have much to contribute to our understanding of inequity and the points where it can be effectively addressed in the most specific way possible. If we take care to consider intersectional risks among race/ethnicity, gender, age and comorbidities, along with the context in which those variations exist, we can begin to address the specific risks that affect each vulnerable subset of patients.
About the author: Heather Blonsky works closely with Vizient’s Equity Steering Committee to pursue data-driven insights into issues of health equity and social determinants of health. As a data scientist in Vizient’s Center for Advanced Analytics and Informatics she contributes to the development of metrics that members can use in evaluating their own performance and benchmark against their peers.