In Pursuit of the Patient Stratification Gold Standard: Getting There with Healthcare Analytics


As health systems take on risk and shift to value-based payment models, providing effective care for a population of patients is imperative. Care Management is a pillar of population health management strategies—succeeding in this arena is increasingly important to thriving in this risk-based industry.

However, care management is a costly endeavor requiring significant investments in infrastructure and expenditure of resources to achieve targeted clinical and financial outcomes. Even the healthiest among us benefits from some degree of care management, but limited resources lead to an unfortunate reality: patients must be stratified in ways that facilitate prioritized enrollment into care management programs.

This article explains why identifying patients who are truly impactable is key to maximizing the cost-effectiveness and sustainability of health systems’ care management programs—and why leveraging healthcare analytics in conjunction with clinical judgement will help systems achieve this patient stratification gold standard.

Problems with the Common Care Management Approach

Today, many health systems rely on qualitative assessments to identify and enroll patients in care management programs. Some systems identify patients by calling every individual within two days of discharge. At other systems, primary care providers are responsible for referring patients to care management. Although there is value in using qualitative means (clinical judgement) to determine which patients need care management, relying solely on such mechanisms isn’t likely to create the optimal patient panels for two key reasons:

  1. Patients who are part of an at-risk population, but aren’t receiving all necessary care within the health system, are likely to be missed. This may be due to them not qualifying for any single provider’s patient panel, or because their fragmented records make it difficult for a provider to appropriately ascertain the full continuum and complexity of their needs.
  2. A myriad of biases and heuristics are at play when systems rely entirely on qualitative inputs. For example, patients asking for more attention may get enrolled instead of those who are reticent to engage independently, but who would benefit more from care management.

How to Improve Patient Stratification with Healthcare Analytics: Three Levels of Maturity

Health systems can use healthcare analytics for incorporating quantitative information to more effectively stratify patients into the appropriate care management program:

Level 1: Identify high-cost, high-risk patients in a population

The first level of maturity in using healthcare analytics to improve patient stratification is identifying high-cost, high-risk patients in a population. Given that the top five percent of patients are responsible for nearly half of the dollars spent on healthcare, and that those with multiple chronic conditions cost up to seven times as much as patients with only one, focusing care management resources on these individuals is often a logical way to quickly leverage an analytics system.

Level 2: Identify patients with rising-risk profiles

The next level is identifying patients with rising-risk profiles. If systems can recognize these individuals earlier, then it may be possible to intervene before their health status worsens and they become part of the high-cost, high-risk group.

Level 3: Identify patients who are truly impactable (the patient stratification gold standard)

The highest level of maturity in using analytics to improve patient stratification, is identifying patients who are truly impactable. This is the primary goal of patient stratification—the gold standard. Achieving this gold standard requires more than simply identifying patients with a high-cost or rising-risk status. Many of the highest cost patients will remain extraordinarily expensive regardless of intervention, such as those with end-stage renal disease. Others with high costs in a given time frame will regress to the mean without any additional care management (e.g., someone with a routine, successful joint replacement). The gold-standard patient stratification process identifies patients for whom care management support results in improved clinical and financial outcomes that would have otherwise not occurred.

Why Healthcare Analytics Is Key to Achieving the Patient Stratification Gold Standard

Achieving the gold standard of patient stratification capabilities, in which impactable patients are effectively identified, relies on a hybrid quantitative/qualitative approach. Analytics can provide visibility into the full continuum of care by drawing from disparate healthcare data sources. Combining claims data (which many health systems only recently started getting access to) with clinical records greatly enhances the precision with which patient profiles can be created. Stratification algorithms can also bring in sources, such as socioeconomic data, when available.

While data is imperative, a key method to quantitatively derive impactability is employing analytics to illuminate trends of patients care management programs have successfully impacted in the past. By creating this learning system, common traits among patients who respond to intervention begin to emerge:

  • Clinical similarities (e.g., conditions or treatments).
  • Programmatic strengths (e.g., specific care plans or care managers with community ties who connect with patients).
  • Social factors (e.g., IMPACT score or demographic characteristics).

As analytics help health systems isolate these characteristics, they can be turned into variables and fed back into the stratification algorithm to enhance its precision, making the patient stratification gold standard a reality.

Healthcare Analytics Should Enhance (not Replace) Clinician Judgement

There is no question that clinician knowledge is a critical component of any successful patient stratification program. Many health systems are not going after the entire high-cost or rising-risk population—instead, they’re using their own prioritization based on experience and understanding of their market to further segment that population. For example, a care management program may be targeted at the highest cost diabetics with several co-morbid conditions, excluding cancer or end-stage renal disease. Stratification algorithms may soon do more to drive these targeted decisions, but in the short-term, qualitative determination of where to begin remains important.

Qualitative determination (clinical judgement) is key because any stratification algorithm is ultimately no more than a highly refined assumption about who the most impactable patients are. There are many nonclinical, nuanced factors that influence clinician-led enrollment today that are unlikely to become standard data elements, either because of privacy concerns, limited relevance to standard clinical care, or simply being items that aren’t easily quantifiable (e.g., willingness to engage or social support needed to successfully manage health).

While some cutting-edge health systems are using creative proxies for these factors—identifying patients with addresses that change often, for example—many of these determinations will happen on a case-by-case basis. Any stratification list needs to go through a workflow in which individuals, such as care managers, can remove patients they know are not appropriate for enrollment, and add patients the algorithm has missed.

The Analytics-Driven Pursuit of Patient Stratification Will Improve Outcomes

Identifying patients who will benefit most from care management—achieving the patient stratification gold standard—is an iterative process. Although there is incredible value in health systems’ current strategies, creating the most effective patient stratification process requires employing healthcare analytics in an increasingly sophisticated fashion.

Health systems can start by using analytics to identify high-cost, high-risk, and rising-risk patients fairly quickly. Ultimately, however, health systems must work toward employing a learning system—in tandem with qualitative information— to best deduce where the opportunities lie as they pursue patient stratification that truly improves health outcomes.

Powerpoint Slides

Would you like to use or share these concepts?  Download this presentation highlighting the key main points,

Click Here to Download Slides

Loading next article...