Social Determinants as a Tool for Improving Preventative Care Rates Among the Most Vulnerable Populations

Written by Raafia Ahmed on 12 Aug 2021

Co-authors: Corinne Stroum, Dr. Steve Overman and Muhammad Aurangzeb Ahmad

We teach children to look both ways before they cross. We are given life jackets when we fly. We are told to wear a mask. In all aspects of our lives, we are reminded to care for and protect ourselves. At times it is because we may be accountable to someone or something, such as the airline who we trust to fly with, or to a loved one we don’t want to spread a disease to. Most of the time however, we simply want to protect ourselves. 

 

We are quick to take caution (and to be patrolled) in situations involving immediate possible danger; in our cars we wear our seatbelts. Why do we not take caution when possible dangers to our wellness may exist over time? Is it because of time, and not having enough of it because we work two jobs, and by the time we can visit the doctor, their office has closed? Or is it due to a knowledge gap, and not receiving proper education on our diet restrictions as a pre-diabetic patient? There are an endless number of personal and social reasons why individuals do not access preventative care services.  

 

Health care systems, therefore, must take shared accountability over our overall health and wellness, not only for the sick or ill, but for the seemingly healthy as well. As more health care systems switch over to the value-based care model, they take on the financial risk as they become responsible for a patient’s overall wellness. A first step to minimizing risk is to keep the healthy, healthy, and not to not exacerbate a condition through the continuum of care (figure 1). By wholly understanding social factors at a macro level and how they may affect individuals’ access to preventative care, health care systems can shelter themselves from preventable (and costly) health care utilization. 

 

Figure 1  

 

Social factors, such as education, income level, and individual demographics, grouped together are known as Social Determinant of Health (SDoH). Social Determinants correlate with health inequities and can directly influence a person’s health and wellbeing. SDoH therefore can impact chronic condition status, count of preventable inpatient and ED visits, or even the mortality rate of individuals. These health outcomes (and others) are impacted more so by SDoH factors than the actual medical care received.   

 

Hospitals are investing in collecting SDoH within their EMR by conducting programs to capture patient self-assessments and documenting social determinant ICD Codes. Even with these efforts, EMR systems contain limited, and sometimes unreliable SDoH variables. With limited SDoH data available at the patient level, health systems can use publicly available datasets to conduct patient outreach for preventative care. Publicly available data sources are generally available at an aggregate level, e.g. by County, State, or more granular, at a Census Tract Level. Of course, data attained at the population level will still have some gaps, but a study by Soy Chen et al. demonstrates that it is possible to understand healthcare utilization by simply using publicly available SDoH datasets, behavioral indexes, and basic patient information (such as age, gender and address). Indicating that understanding population level social factors may provide us with rich insights, regardless of how well the EMR populates SDoH data. 

 

A KenSci project focused on predicting which patients are likely to defer their well care visit, confirmed the validity of this approach.  After joining the CDC’s Social Vulnerability Index and AHRQ’s Social Determinants of Health Aggregate Dataset to a customer’s EMR data on the patient’s County, our Machine Learning algorithm performed significantly better at proactively identifying patients likely to defer care. Precision and Recall jumped from about .60 to .90 and .70 to .80 respectively, after adding in features like: percent of population that does not speak English well, percent of the population that is disabled, percent of housing units with no vehicle available, etc.  

 

Using SDoH variables as input features is the first step in using publicly available data to understand what perpetuates health inequities within a community. As next steps, we imported the model outputs into PowerBI and analyzed patients with predictions geospatially to visualize the communities in which they reside. We pulled in SDoH, location, demographic data, and predictions into a key influencers chart to learn about variables that increase the likelihood of a care deferral prediction. By discovering the social factors behind care deferral, ACO’s can take strategic action to reduce quality measure gaps and consequently reduce financial risk. 

 

Access Our Latest Thinking

Research, blogs, and other whitepapers on AI Led healthcare transformation

Case Study

See how Advocate Aurora Health is working on fighting the opioid battle

May 13, 2020

Blog

How to accelerate the movement of data onto Azure in FHIR format

March 17, 2020

Webinar

SLUHN is fighting COVID-19 by staying ahead with a real-time command center

September 4, 2020

PodCast

HIT Like a girl~ Health from Corinne Stroum on Healthcare Informatics

August 12, 2020

Press Release

SCAN Health and KenSci are leveraging AI to help the senior members

August 11, 2020