From 3M Health Information Systems
Data challenges for the Social ACO
A relatively recent construct in Value-Based Care programs is the merger of the Accountable Care Organization (ACO) with programs designed to address specific Social Determinants of Health (SDoH). One such program, the Commonwealth Care Alliance, has been called the nation’s first Social ACO. The Social ACO seeks to integrate not only the health and social service programs targeted at special populations such as Medicaid beneficiaries, but the program dollars and risk-sharing arrangements as well.
In a recent 3M Inside Angle blog, Dr. Gordon Moore cites one of the enablers of a “fully integrated system of [health] care” as Methodologies to identify medical and non-medical risk factors. These non-medical risk factors are the multiple determinants of health “beyond diagnosis, severity, age and sex.” They include characteristics such as socioeconomic status (income and education), housing characteristics, food security, health care access, social support, and exposure to crime and violence.
Ideally, we would like individual assessments of non-medical risk factors in the same way that we prefer individual clinical measurements to assess a person’s risk for disease. Your healthcare provider reviews your individual physical and laboratory measures—BMI, blood pressure, cholesterol, HbA1c, etc.—not the average measures of persons in your neighborhood when diagnosing, monitoring, and treating your disease. How does a care provider acquire access to important non-clinical measure of individual risk when assessing a patient and devising an optimal treatment strategy?
As discussed in a previous blog, the advent of ICD-10 has provided health providers with a place to code on a claim—within the supplemental Z-Codes—selected social factors affecting health status. Through this mechanism, your provider can note relevant non-medical factors in your health profile. If your provider or health plan administers a questionnaire collecting Patient Reported Outcomes (PRO), your provider may have some information on social factors impacting your health such as the strength of your support network or your ability to get to clinic appointments and pick up prescription medications.
At this point in time, however, these data are not very prevalent. In our early analysis of Z-Codes in ICD-10 claims, only a small proportion of patients have codes relevant to adverse social situations. In addition, PRO surveys that include assessment of social factors impacting health status and interaction with the healthcare system are insufficient.
What other options exist for obtaining non-medical data points to provide better information on a patient’s social environment? We can think of these data as impacting the environment in which patients live, work, and receive their health care. Although not all environmental factors impacting health are collected in databases that can be linked to administrative claims and electronic medical record (EMR) data, an ever increasing amount of information on small geographic areas is being collected and made available. Sources include databases developed and maintained by the U.S. Census, the Centers for Disease Control, the National Institutes of Health, various state health and local departments, federal, state and local social service agencies, and population surveys on various health and sociodemographic questions.
How can you make sense of such numerous and diverse data elements? Fortunately, there are statistical methods like principal component analysis (PCA) and structural equation modeling (SEM) that aid us in the compilation and understanding of numerous and seemingly disparate data points, and the computing technology to employ these methods in very large databases. We can compile numerous data points in small areas such as census tracts and use these techniques to help us construct composite scores for SDoH for an entire state by small area. These scores can be linked to persons in a healthcare claims database by geocoding patient addresses and merging with small area SDoH scores.
Merging information on a person’s social environment with their healthcare utilization and cost experience will allow us to better understand the relationship between clinical risk and social-environmental risk. This understanding will provide valuable insight for those creating integrated health programs like the Social ACO. We in the 3M HIS Clinical and Economic Research team are undertaking this work and will have more to report in future blogs.
Paul LaBrec is research director for Populations and Payment Solutions with 3M Health Information Systems.