Fix Population Health Equity Issues Inherent in All Data

Three Interventions to Fix Population Health Equity Problems Inherent in All Data

Written By: Corinne Stroum & Dalila Zelkanovic

Summary: Traditional approaches to using data, such as relying on a single source, will leave behind some of the most critical information about healthcare and your members. Utilizing advanced analytics helps you spot things that aggregations hide. By incorporating additional datasets, you can better understand your membership and trends in new services. Together, these ensure the best utilization for each member.

This summer, Advata made its first public debut at America’s Health Insurance Professionals (AHIP) in Las Vegas. Familiar focus areas usually associated with this conference, including medical and pharmacy needs coverage, were not the only vendors present.

The exhibition hall also had supplemental benefits: vision, dental, multiple fertility and family planning products, meal delivery, and behavioral or mental health services. The discussions at AHIP demonstrated that it took a global pandemic for us to break down as a nation before we broke open about our disconnect between physical and behavioral health.

Whole-person care was a meaningful focus of the exhibitors, looking beyond only one type of need. Similarly, an analytics strategy focusing on medical and pharmacy claims or purely on clinical services does not look at the complete view, leading to representation bias. That’s because analytics and modeling can only study those patients and members engaging with the healthcare system, ignoring those who do not.

Therefore, it perpetuates the gap between those who can and cannot access care: those who leave a strong clinical or claims data trail is more likely to be picked up by predictive models. Further, traditional datasets leave behind actionable and high-value data. The gaps in how we diagnose, treat, and monitor behavioral health are evident in the health disparities we see across the spectrum. Disparities are the results we can measure when we measure inequities. So, the onus is on analytics vendors to act on health equity.

We believe there are three enhancements to improve your organization’s usage of non-traditional data that isn’t captured in claims:

1. Develop a strong data governance methodology to ensure guidance on when and where information is meaningful. Determine whether your data can stand alone or whether it is more meaningful as part of a composite score. Perform analyses to see the impact of treating a field as a bucketized value, ensuring better comparisons or analyses across demographics and geography. Record the source of the data. Document how often it is updated and what assumptions or inclusion criteria are applied.

2. Utilize advanced analytics strategies to ensure you can see areas worthy of further review. In an example: it is common to aggregate metrics in a population, such as in finding the median or average. However, this often leaves behind those meaningful differences across the cohorts of your population. Advata frequently uses boxplots and Bayes’ Confidence Intervals to highlight ranges and outliers and to make comparisons across cohorts. We also use statistical methods to seek meaningful change in metrics over time, ensuring that a large increase in one group’s utilization, for example, is not canceled out by the collective decreases in the remainder of the population.

3. Identify what is actionable if not predictive. Data for homelessness and housing instability, or social risk factors, is inconsistently recorded. This sparsity means they are not suited as a predictor in a machine learning model and could lead to overfitting. They are, however, a big influence on how a user will act on a risk score. Consider surfacing data like this, which can inform member intervention strategies in your user’s tool as a decision aid.

Advata is proudly committed to health equity through our demonstrated use of meaningful predictive guidance and decision-making.