Automating the Analysis of Social Determinants of Health

0

Use social determinants to reduce costs and improve outcomes

Medical records rarely tell the whole story of a patient’s health. Some of the most important risk factors that contribute to preventable emergency room visits, 30-day readmissions, medication noncompliance and post-operative complications never even cross a care team’s radar.

A patient struggling financially, for example, may convince her doctor that she understands her medication plan but later decide not to fill a prescription because it costs more than she can afford. Another patient may fail to reveal that he has nobody to drive him to follow-up appointments and that he lives outside the reach of public transportation. How can a health system consistently uncover this type of information and use it to mitigate risk?

Social determinants of health, including race, gender, income, housing type and location, family size, education and many other factors not strictly related medical care, influence health and clinical outcomes much more than the information gathered during care episodes. Putting that information at the fingertips of care providers, delivering it to them directly through electronic medical records, should be a top priority as we attempt to achieve the aims of enhancing population health, reducing per capita costs and improving the patient experience of care.

Risk Modeling Without Social Determinants of Health

Clinical protocols, even those based on the latest evidence and best practices, tend to overlook vital clues to a patient’s true risk profile and health status — information the care team may not discover during the clinical encounter. Each patient presents unique extra-clinical challenges, defined by social and economic factors as well as health behaviors that often go undetected.

When risk assessments omit social determinants of health, unexpected and often costly consequences occur. For a country that currently spends $3 trillion per year on healthcare, about 32% of it on hospital care, preventable visits to emergency rooms are particularly problematic. Census data indicates that 428 of every 1,000 people in the United States made at least one emergency department visit in 2014. According to the American Institute for Preventive Medicine, about 55% of those visits (65.5 million) were unnecessary.

While preventable emergency room visits raise costs and put patients at undue risk, they also attenuate the limited resources of care management teams, making the mitigation of risk all the more challenging. Lacking the time and tools to evaluate social determinants of health efficiently, care teams make assessments of risk that place greater weight on clinical information. Inefficient resource management and suboptimal outcomes are almost inevitable in this state of affairs.

Meanwhile, hospitals are increasingly at the forefront, bearing the brunt of financial liability for a systemic problem of ineffective risk assessment. The Centers for Medicare & Medicaid Services (CMS) expects to withhold a total of $528 million in payments from hospitals with above-average 30-day readmission rates in the coming fiscal year.

Automating SDH Analysis to Identify Impactable Risk

Predictive analysis using social determinants of health holds tremendous promise for reducing costs and improving outcomes, but its success depends on the quality of the data used and the way health IT systems deliver its insights.

Care teams traditionally leverage human judgment to identify and mitigate risks stemming from social determinants of health. Interactions between care managers and patients will continue to be the best source of information about patient-specific risk. The objective of predictive analysis tools is not to displace these interactions, but to enhance them with insights that help caregivers hone in on critical information more quickly and efficiently.

By automating the analysis of social determinants of health and delivering the results to the point of care, predictive analytics solutions can help care teams identify patients that warrant special attention and give them a head start on the types of issues to expect. In some cases, these tools can even tie into care team workflows and automatically generate preliminary plans of care that specifically address non-medical risk factors.

The impact of the insights these solutions provide ultimately depends on the quality of the data they analyze: the more geographically precise and patient-specific the data, the more accurate the predictive model. Many solutions claim to analyze social determinants of health, but the data they use is highly aggregated and limited to a ZIP-code level of specificity. Solutions that incorporate neighborhood-, household-, and patient-specific information are far more powerful.

The Unlimited Upside of the Right Predictive Analytics Solution

Automated predictive analytics solutions that use highly granular SDH data, modeling it with the most advanced machine learning methods, offer endless potential for tailoring clinical interventions to the unique risks and needs of each patient.

As these solutions come of age and care teams learn how to harness their full potential, health systems will see an array of benefits come to pass:

  • Vast improvements in risk-modeling and -management, with care teams consistently aware of non-clinical factors affecting cost, outcomes, and patient satisfaction
  • More efficient allocation of resources to the most impactable patients as the care management needs of all patients become more transparent and predictable
  • Richer accounts of clinical decision-making in medical records, particularly when it comes to complex cases, with SDH risk analysis built directly into care plans
  • Lower administrative costs and other efficiency improvements stemming from the greater automation of the care-planning process
  • Automated capture of data that health systems can use to research the broader influence of SDH at the population level and to develop new risk-mitigation strategies; given the continuous aggregation of data, predictive models will improve over time and be responsive to changes in social factors
Share.

About Author

Sathya “Sam” Rangaswamy, MS
Sathya “Sam” Rangaswamy, MS

Sathya “Sam” Rangaswamy, MS, is the founder and CEO of ZeOmega. Prior to founding ZeOmega, he worked at American Airlines and The Sabre Group where he led and delivered web-based solutions for Air New Zealand, Swiss Air, American Airlines, Carlson Travel and Travelocity. He has architected business process and content management solutions for Gartner Inc., servicing over 25 channel events such as RetailVision and Midsize Enterprise Vision.During the dot-com era he was vice president of technology for a medical management start-up where he architected and developed automated fax and web based preauthorization solutions using decision tree based algorithms for medical management. He has extensive experience in conceptualizing and reengineering business processes for medical management, event management and content management. Rangaswamy has several publications to his credit including “The design of fuzzy constraint based controller for a dynamic control system,” which was accepted at the International Joint Conference of the 4th IEEE conference on Fuzzy Systems held in Yokohama, Japan in 1995. Rangaswamy holds a Master’s Degree in Computer Science from North Carolina State University, Raleigh, North Carolina, and a Bachelor’s Degree in Industrial Engineering from The Manipal Institute of Technology.

Comments are closed.