Status
Conditions
Treatments
Study type
Funder types
Identifiers
About
The goal of this clinical trial is to determine if artificial intelligence and machine learning (AI/ML) models can help address social needs in Medicaid enrollees. The main questions it aims to answer are:
Can AI/ML models accurately identify social needs from administrative healthcare data?
Can AI/ML models accurately predict which people will engage with social supports?
Researchers will compare individuals who live in different regions to see if AI/ML models perform better than the status quo.
Full description
Social drivers of health (SDoH) are the largest factors affecting our health and wellbeing but are difficult for healthcare systems to address. Despite new models that provide incentives for health plans and providers to reach beyond clinical care to improve patient health outcomes, existing data infrastructures lack relevant information to support such interventions. The first problem is one of identification; providers undercode social needs in existing schemas and ancillary data collection methods such as social screens are not common, standardized, or easily shared. The second problem is a lack of engagement between individuals and social services, which is especially frustrating since there are many evidence-based practices that community-based organizations (CBOs) use to address social needs. Without precise information on who needs social support and how to maximize their engagement with CBOs, providers and insurers have limited ability to deploy interventions that remove barriers to care and equalize health outcomes across vulnerable populations.
Our project will apply a precision medicine approach to the identification of, and engagement with, Medicaid recipients with social needs. The investigators have partnered with a managed care organization that coordinates benefits for over 250,000 Maryland Medicaid members. They have launched a population-wide social screening program to add member-reported social needs to their existing clinical data. The investigators will enhance their health information technology (IT) infrastructure with a set of machine learning models for risk identification, an engagement support system to maximize member's use of social supports, and a continuous qualitative and quantitative improvement process to establish a learning health system. We will accomplish this work through the following aims:
Aim 1: Develop and deploy a set of machine learning models that use multiple individual- and community-level data sources to predict which members use the emergency department to fulfill social or non-urgent needs as opposed to treatment for urgent medical conditions. These models will identify individuals whose social needs are driving inappropriate utilization so that high-risk individuals will be given enhanced outreach services to facilitate completion of a comprehensive social needs assessment. The investigators will analyze these assessments to determine if our models lead to the assessment of individuals with a higher social need profile.
Aim 2: Develop and deploy an engagement support system that identifies and displays the characteristics of members that prevent them from engaging with a CBO. This system will use artificial intelligence techniques to identify characteristics of individuals who have historically disengaged from the social service pipeline before receiving social services and suggest potential strategies for increasing engagement. The investigators will apply the models to newly assessed members and present predicted high risk individuals to the plan's community health workers through their existing IT platform, allowing them to proactively address members' barriers to accessing services. The investigators will analyze engagement success (i.e., whether a member who was referred to services received assistance from a CBO) to determine if our support system increased the likelihood of success.
Aim 3: Implement a continuous qualitative and quantitative improvement process that identifies recurring themes and disengagement points in cases where members were not able to complete their relevant social intervention. These findings will be analyzed by the research team to identify potential tactics to address engagement barriers, and resulting recommendations for increasing engagement will be propagated through the system either by updates to the health IT infrastructure or staff training sessions. Through this Aim the investigators will build a learning health system, with the team constantly refining engagement methods throughout the project.
The study team is well positioned to develop a social needs intervention protocol and will include rigorous evaluations to assess the effects of our intervention on the health and social outcomes of participating members by their demographic and geographic characteristics. Together, these aims will help inform the next generation of value-based care paradigms by identifying and addressing social needs and shrinking differences in health outcomes across a large, high-risk population.
Enrollment
Sex
Ages
Volunteers
Inclusion criteria
Exclusion criteria
Primary purpose
Allocation
Interventional model
Masking
249,660 participants in 2 patient groups
Loading...
Central trial contact
Ian Stockwell, PhD
Data sourced from clinicaltrials.gov
Clinical trials
Research sites
Resources
Legal