Nemours

Earlier in 2022, I shared a story about my team’s experience uncovering racial disparities in our patient’s influenza vaccination rate that would not have been discovered had we not stratified our data by race and ethnicity. The article goes on to discuss the role that IT leaders can play in promoting the analysis of health equity and disparities within a health system by setting expectations for the appropriate use and analysis of outcomes data.

2022 was a year of great progress in this area for Nemours in a number of ways – from leadership to team structure to our Health Equity Pledge and finally, the launch of our standard REaL data reporting model. All of this work is anchored to our goal of building the healthiest generation of children, supported by our Health, Value, and Equity strategic pillar.

The premise of all of this work is as follows: To truly improve the well-being of children, we need to define health beyond just the treatment of illness. The well-being of the whole child, from physical health to mental health, to educational outcomes and community well-being needs to be considered. By changing our reimbursement strategy to focus more on value-based care payment models, we can align our revenue with our mission in a way that sustains us as we make investments in whole child health. Finally, we realize we cannot truly support a healthier generation of children, or even succeed in alternative payment models if we leave populations behind by failing to act on inequities both in access to care and in outcomes.

Our Analytics and Technology team is taking several key steps to support this work in 2023 and beyond, including partnering on staffing with our Office of Health Equity and Inclusion (OHEI), developing new standard data assets, and supporting enterprise communications/ awareness-raising about disparities analysis.

Staffing Partnership

In 2023, my department will be adding a dedicated data analyst and an evaluation scientist who focus on supporting OHEI-related work. If we truly lived by the idea that addressing inequities is part of everything we are doing in terms of quality, workflow optimization, clinical standardization, etc, and not just a niche function, then you may ask why we need dedicated analytics staff rather than just leveraging existing teams. One way to think of it is that we want our data analysts enterprise-wide to focus on identifying disparities during the work. For example, different prevention rates, surgical outcomes, or even wait times across populations.

While we will encourage all of our quality improvement experts, data analysts, process improvement experts, etc. to support issue the identification and escalation of any disparities that emerge during their work, our newly dedicated equity data analyst and evaluation scientist will also focus on measuring the effectiveness of interventions: How will we know we have been successful at closing a gap? What interventions should be scaled? What types of community feedback have we received about new programs? They will work with both qualitative and quantitative data, framed using our new standard REaL categories and reporting models and informed by healthcare and population health best practices.

Standard Data Assets

One issue we face as a large enterprise with many decentralized data teams is multiple models and methods for reporting and stratifying patient demographic information. This makes it challenging, despite common best intentions, to leverage outcomes analysis done by different enterprise teams. For example, some analysts might treat race and ethnicity as separate independent dimensions (which is how they are captured during patient registration), while others combine them into a single category list. Likewise, some analysts create distinct multi-race/multi-ethnic categories, others use the first race or ethnicity selected by a family, while yet others assign a patient to every category they indicated during registration. Even though the various reporting approaches all have pros and cons, we felt that a single unified approach for REaL data stratification was an essential first step in our efforts to look at population-level outcomes and potential disparities.

 Multiple models and methods for reporting and stratifying patient demographic information make it challenging, despite common best intentions, to leverage outcomes analysis done by different enterprise teams 

Another area we tackled was aggregating the many distinct, specific, race/ethnic categories into roll-up groups that would provide sufficient numbers for analysis. For example, various Hispanic ethnicities such as Puerto Rican, Mexican, etc. While we respect the need for families to have specific classifications that they identify with, the comparatively low N-size of pediatric clinical populations dictates that we start with a higher level of aggregation, particularly when exploring disease-specific or procedure-specific outcomes that impact even smaller populations.Our new model combines race and ethnicity into a single classification and establishes Hispanic identity as a primary characteristic when present as an ethnic group choice. It also handles multi-racial/multi-ethnic patients by creating a category of Two or More Races.

Analysts engaging in different types of demographic stratification can use more granular or specific categories based on their specific needs, with the understanding that this new enterprise model is to serve as both the starting point and required standard for our system-wide equity and disparity measures.

To implement our new analytics infrastructure, we started by inventorying the various data sources used to pull demographic data across the organization. These included: EHR-integrated reporting tools such as Workbench and Radar, visualization tool data marts (Qlik QVDs), and SQL raw data extracts (Clarity & Caboodle). To validate that our new mapping logic factored in all of the various unique combinations of multi-race/multi-ethnic/partial refusal/”other”, etc. we wrote an elaborate error-checking formula in SQL and tested the results against several million patient records in our database. We believe that our end-product data asset is now ready for prime time and are rolling it out first for our Qlik data visualization tools and ad-hoc SQL queries. Within a few months, we will load it into our EHR for use in integrated reporting tools.

Moving Forward

We have our whole child health strategic imperative, we have our standard enterprise data assets for starting to look into equity and disparities. The next steps involve persuasive communication, education, and outreach. Fortunately, we can build upon very relatable studies that provide great examples of the types of analysis we plan to perform; such as the study on physician-patient gender concordance and heart attack mortality published in PNAS in 2018, and the analysis of racial gaps in stroke treatment published in the journal Stroke in 2019. Between our new analytics capabilities and readily explainable examples of disparities in outcomes, Nemours is well positioned with both the “why” and the “how” surrounding this important work. The only part left is the “when”, and that starts now.