Social care, encompassing behavior/mental and built-environment health, is integral to healthcare. NLP algorithms captured behavior/mental health from EHR notes; ArcGIS geocode algorithms captured built-environment health. We combined medical and social care factors from 104,038 diabetes patients (2012-2023). Our machine learning models predicting cardiovascular diseases uncover implications for personalized "whole person" intervention strategies.

Using ArcGIS, environmental factors including groceries, restaurants, parks, and trails were extracted using EHR data. Prior literature suggests where you live impacts well-being. Proximity algorithms were generated to determine whether environmental factors influence health outcomes.

The Impact of Accelerated Digitization on Patient Portal Use by Underprivileged Racial Minority Groups During COVID-19: Longitudinal Study

J Med Internet Res 2023;25:e44981

URL: https://www.jmir.org/2023/1/e44981 

DOI: 10.2196/44981 

Diabetes affects 37.3 million Americans, accounts for an estimated $24.6B in wasteful and avoidable spending, and its incidence is growing. The disease is a key cause of many comorbidities such as end-stage renal disease and complications such as neuropathy. Exposure to health-impeding social care, or the social and behavioral health, contributes to poor diabetes outcomes. Our goal is to develop a predictive model that identifies patients with type 2 diabetes at risk of developing complications. 

Capturing built environment factors associated with patients is a powerful approach to gaining additional insights about obstacles that may prevent individuals from living a more fulfilling and healthy lives. Easy access to groceries (vs eateries), health centers, pharmacies, or public transportation are examples that geocoding can help further explain the quality of health care and outcomes. 

Diet is one of the most challenging lifestyle factors impacting patients with diabetes. There is no good way to log nutrition intake. Patients are asked to recall food consumed, describe how they were prepared, and estimate portions consumed. Evaluating food purchases (for the family) offers an objective approach to calculating nutrition intake and move away from subjective estimations. As a first step, we developed a tool that captures Kroger food purchases and an algorithm that extracts and calculates macro-nutrients.