SUNNYVALE, Calif. (Apr. 01, 2019) – To improve population health management, the importance of stratifying patients by expected risk is widely understood among most leading healthcare organizations. It’s certainly an intuitive concept: By identifying which patients are likely to be at highest risk of utilization in the future, healthcare organizations (HCOs) can focus their resources on the patients who will most benefit from early interventions that would reduce future utilization and improve outcomes. The tricky part, of course, is predicting which patients are most likely to need those interventions, and the cost savings the interventions are likely to generate.
Traditionally, HCOs have relied upon little more than retrospective claims data to inform this analysis. Conventional wisdom has held that the same 5 percent of the population that generated the most utilization and cost one year would be the same 5 percent that would do so the following year.
Recent advances in the sophistication of healthcare data analytics have demonstrated the fallacy of this “wisdom.” Our analysis reveals that, of that 5 percent, just 1 percent of those patients over the following year are most responsible for driving utilization and costs. The other 4 percent comes from portions of the population that today have been identified only as low- or moderate-risk, and have yet to transition to the highrisk cohort.
For HCOs, the problem is that they often lack visibility into that 4 percent; this cohort have not been high utilizers, so the substantial amount of claims data that HCOs traditionally use to perform risk stratification is simply not available for this group of patients. So the key question becomes: How do HCOs take the guesswork out of population health management?
The big threat: Unknown risk
Broadly speaking, most HCOs essentially manage two patient population cohorts: those with known risk, and those with unknown risk. For known risk, the situation is largely positive because HCOs are doing a lot to effectively manage these patients. Generally, we know who these patients are, and we’re effectively managing their conditions, in addition to their associated costs. For known-risk patients, we are effectively identifying them prior to the expression of disease, by evaluating chronic conditions, acute conditions and trauma risks for which there is an intervenable opportunity to both avoid costs and drive superior clinical outcomes.
With unknown risk, however, the picture is not so rosy. Previous approaches to risk stratification, which traditionally rely on claims data, have failed to segment the population with enough accuracy and specificity largely because claims data is retrospective, and, on its own, cannot provide insight on the prospective migration of patients among risk groups.
Far beyond claims data
Today’s most advanced analytics platforms can help HCOs overcome the challenge of an over-reliance on claims data for risk stratification. These platforms specialize in identifying unknown risk for patients within a population that have insignificant claims history, and a correspondingly low risk stratification, but are nevertheless highly likely to utilize significant healthcare resources.
Leading analytics platforms are capable of finding these patients via a data-driven, artificial intelligence evidence-based predictive health engine that, when given all known data about an individual patient, computes a precise prediction of absolute disease-risk and disease evolution. While these platforms take advantage of claims data when available, they are not reliant on it, instead drawing from medical data, laboratory data, biometric data, pharmacy data, social and family history data, behavioral data, and other sources.
Further, leading predictive analytics platforms can help provider groups target patients with specific prescriptive clinical and health management interventions that will be most impactful and deliver the greatest return on investment. These prescriptive insights can help providers mitigate high-cost utilization not only in the long-term, but over the near-term, as well.
Low-risk today, high-risk tomorrow
In the context of risk stratification, HCOs’ objective of collecting and analyzing such a wide variety of data is to obtain predictive insights that identify which low- and moderate-risk patients today will become high-risk patients tomorrow. On the clinical side, characteristics such as high cholesterol, blood pressure and resting heart rates and associated trends are most indicative of such transitions.
Additionally, lifestyle characteristics such as high body mass index, smoking, alcohol use and even number of clinic visits often signify the transition. Age, gender, other demographics and social determinants of health also play a role. Modern analytics platforms take all of these factors into consideration to accurately predict the patients most likely to transition from low- and moderate-risk to high-risk.
Here’s a more concrete example: Consider the scenario of a 75-year-old patient who had labs that in 2017 indicated high blood pressure, borderline high cholesterol, high triglycerides, and close to 100 resting heart rate. In that year, this patient’s total-cost-of-care was around $500, primarily attributed to clinical visits. However, in the following year total-cost-of-care rose sharply to around $20,000 because the patient developed coronary artery disease (CAD).
A leading analytics platform, however, would have predicted the onset of CAD for this patient based on patients with a similar profile and clinical trajectory. Additionally, the platform would have recommended early interventions that would help prevent, delay or reduce the severity of CAD, in turn reducing the patient’s total-cost-of-care.
Centralizing analytics to overcome fragmentation
A barrier HCOs often encounter that hinders improvement in their population health management programs is a fragmented, decentralized approach to analytics delivery models that leads to inefficiency and excessive costs. As a result of this approach, data analysis happens in different business units that do not share assumptions, analytics methods or insights broadly.
In contrast, under a centralized delivery model, an experienced team of data analysts report to one function at the enterprise level, even if they are assigned to serve different business units, based on strategic priorities set at the corporate level. This business-oriented team of analysts meets the need of organizational stakeholders while maintaining and developing intelligence in-house.
A centralized analytics delivery model is important, in large part, because it offers an improvement to the fragmented, incomplete data governance models that too many provider groups still operate. For example, it’s not uncommon for large health systems to contract with multiple vendors to analyze population health risk for groups of patients with different conditions, such as diabetes, osteoarthritis and others.
This lack of a single source of truth in analytics can lead to different answers to the same question, such as conflicting guidance on levels of risk, and in turn, on the highest-priority patients to target for interventions. As a result of this fragmented and potentially conflicting information, when prioritizing care plans and interventions, the health system cannot build a consistent, holistic clinical profile with a 360-degree view of each patient that accounts for the same factors.
After establishing a solid foundation for data, the HCO will be ready to adopt a single analytics platform that delivers actionable information to decision makers. Today’s leading analytics platforms often employ machine-learning systems to automatically extract important insights that may not be otherwise apparent to human analysts.
Ultimately, the goal is to create one internal, centralized professional services group within the HCO that delivers analytics-asa-service to other organizational stakeholders. By structuring its analytics functions in this manner, the HCO can eliminate the fragmentation and cacophony of multiple systems that offer conflicting advice and prevent leadership from understanding the organization’s full analytics spend.
The integral role of care management teams
Traditionally, care management teams have consisted of a patient’s primary care physician and that physician’s nurses and staff. Today, however, providing care for high-risk patients, such as those who have multiple comorbid conditions and socioeconomic issues, often requires HCOs to assemble teams of greater size and sophistication.
Consequently, many care management teams have grown to include specialty physicians, care management nurses, therapists, social workers, nutritionists and educators.
To achieve success, all care team members must accept that they are a smaller part of a larger, diverse team dedicated to patient service, and that patients and their families are valued team members. Frequent, effective communication among team members is critical so that all develop an understanding of each team member’s responsibilities and each one’s role in preventing care gaps and preventing duplication.
Providing such a high level of coordinated care by a diverse group of professionals brings about its own challenges. However, HCOs can mitigate many of those challenges through the adoption of an advanced centralized analytics system that streamlines workflows, prioritizes daily tasks for team members in a predictive way, and directs action to the interventions that will positively affect the most patients in the most efficient way.
A prerequisite of such a centralized system is that it must have access to all relevant patient data – electronic health records, claims, laboratory, biometric, prescriptions and more. Additionally, the system must analyze large, disparate sets of data to ascertain trends and patterns that lead to better patient outcomes and impact clinicians’ daily workflows. Advanced predictive capabilities, such as artificial intelligence and machine learning, may play an important role in this analysis.
More insight, more opportunity
If HCOs are to successfully take on additional risk, they cannot accomplish it by greater revenue alone; they must seek out insights that identify opportunities to avoid costs. To discover those opportunities, HCOs must first understand the populations that are transitioning from low- and moderate- into high-risk categories, then determine where there is the greatest potential for clinical interventions that mitigate costs and drive savings.
And that requires a lot more than retrospective claims data
For the full article, please go to the Population Health News.
The Power of Prediction: What You Don’t Know About Your Patients Can Drive You to Bankruptcy
Insights and Strategies