Socioeconomic data: Meet a new kid on the predictive analytics block
Non-clinical life events and sociodemographic information, including street crime, domestic violence, illiteracy, income levels and lack of access to fresh, healthy food, can impact the stress severity and health outcomes of individuals across the United States. Such factors, also identified by the Institutes of Medicine as “social determinants of health,” exacerbate health conditions ranging from asthma, diabetes, and high blood pressure to depression, metabolic syndrome and chronic obstructive pulmonary disease.
Another critical factor in the equation is the damaging effects of stress on health. A quarter of Americans claim that stress strongly influences their physical health and 28 percent of adults say stress strongly impacts their mental health. Overall, 43 percent of adults suffer adverse health effects from stress.
Consumers with higher levels of stress have higher health risks, suggesting they might need behavioral, social or medical interventions. In addition, an individual’s socioeconomic status often drives conditions and behaviors, including depression, alcoholism, substance abuse, nutritional disorders, and hospital readmissions.
Exploring big data for this type of information is a new frontier in predictive modeling and can greatly benefit the healthcare industry. According to McKinsey & Company, big data can save $300 billion in healthcare largely through reductions in expenditures. Let us take a look at how socioeconomic data helps improve the accuracy of risk predictions, reveal hidden or unknown trends, and improve health outcomes.
What socioeconomic data knows that predictive models don’t
Imagine a single mother who has just gone through a divorce and had to move to a new neighborhood that has high crime rates. Up until then she was healthy – and from a clinical perspective, there is nothing to indicate otherwise – but living in a less safe place than she did before, having to find a job after she has been caring for her child at home, and dealing with emotional effects of divorce all make her susceptible to serious stress that can set off a series of adverse health challenges. It is this socio-determinant data that affords a unique view into her clinical risks and can allow either her plan or provider to be proactive in getting her the care she and her child need.
This type of insight is equally critical for diagnosing and treating low- or non-users of the healthcare system. In this case, a plan or provider has scarce clinical information and cannot adequately predict the individual’s health risks and would need to rely on a picture presented by the socioeconomic data.
Furthermore, socioeconomic data is vital to HCOs when it comes to caring for consumers with chronic conditions. Forty-five percent of Americans have at least one chronic condition, and half of them are not getting the recommended care. At the same time, 75 percent of every healthcare dollar goes toward treating those conditions. HCOs must manage the chronic conditions of an aging population, while honing in on factors that drive the majority of avoidable costs. These include specific conditions like diabetes and high cholesterol as well as more costly care settings.
Using the socioeconomic data, HCOs can strategically target consumers who require support for outcomes improvement. For example, an HCO might conclude that 45-year-old male diabetic who qualified for free glucose screening but failed to participate would benefit from a diabetes management program. The result is enhanced management of risk and reductions in avoidable costs.
HCOs can calculate their savings potential by adding avoidable costs related to emergency and outpatient care, inpatient stays, preventive measures and chronic care management. Based on the severity and progression of a disease, they can predict the timeframe for finding the greatest potential for savings.
Overall, socio-determinant data allows HCOs to more accurately predict risk and answer some of the industry’s most pressing questions, including:
- How much will a health plan pay – in hospital stays, procedures and prescriptions – for Bob, a 65-year old retiree with multiple chronic conditions, within the next 12 months? Why?
- Which patient costs are likely to increase, decrease or remain the same with the next three-to-five years? Why?
- How likely will adult Hispanics living on the west side of Chicago develop Type 2 diabetes within the next five years?
Motivation is key
Another key question that plays into the effectiveness of population health improvement efforts is how motivated individuals are to engage in the improvement of their own health. That may mean many things: from proactively maintaining a healthy diet to strict adherence to their care plans. Understanding which individuals in the population are motivated allows health plans to wisely allocate expensive resources like nurse case managers. A high-risk patient who is highly motivated may get as much benefit from a low-touch wellness program as he or she would from a high-touch program. An individual’s motivation – or willingness to engage in maintaining or improving their health is just as important as the data used to determine what puts that person at risk.
Predictive analytics drives this unique motivation score, which is validated by third parties including the Society of Actuaries (SOA) and Knowledge Discovery and Data Mining Organization (KDD). They take statistical methods, data mining and artificial intelligence and machine learning to provide actionable insight including:
- How likely is the patient to be highly engaged in their care? Who will respond best to interventions?
- Are they refilling supplies and taking their medications?
- Are they scheduling and doing follow-up visits or annual check-ups?
- Who are the patients frequently visiting the ER, but who might be able to self-manage, given some guidance?
- More specifically, how fully will Jill, a 53-year old receptionist who suffered a heart attack, engage with her care team and treatment plan in the next nine months?
To sum it up
By combining socioeconomic factors with medical and pharmacy claims, labs and health risk assessments in their predictive modeling, HCOs can obtain fresher, more expansive views of consumers at risk for avoidable healthcare costs. The industry overburdened by patients with chronic conditions, growing Medicare and Medicaid populations and skyrocketing costs cannot afford to rely on limited insights granted by traditional healthcare data. Personally identifiable information has long been in use by credit bureaus, law enforcement and pay-per-click advertisers, and when harnessed by predictive analytics for healthcare purposes, it holds a solid promise to help HCOs deliver better care to the right people at the right time, maintain higher levels of satisfaction and enhance value they bring to the communities they serve.
Trackback from your site.