Population Health Analytics

In the first two posts of this three-part Mythbusters series, we’ve examined several long-standing myths about population health analytics and the Johns Hopkins ACG^® System — from misconceptions about proactive care and frailty to assumptions about coding languages, real-world performance and data security. In this final installment, we’re addressing a few additional areas where confusion still arises and offering a clearer view of what the ACG System is designed to do.

Myth: All segmentation models are the same.

Effective population segmentation is already transforming how organizations design proactive, tailored health care interventions. Many segmentation models consider only surface-level data — such as demographics or additive condition counts — and some focus only on specific subsets of the population, such as individuals over age 65. In addition, many segmentation models assign individuals to multiple segments at once, making the results difficult to interpret and act on.

Not all segmentation models are created equal. Effective models, such as the Patient Need Groups (PNGs) in the ACG System, group individuals into mutually exclusive, clinically meaningful and actionable groups based on a comprehensive view of health needs. The ACG System incorporates data from both primary and secondary care settings and uses standardized, routinely collected datasets across pediatric, adult and elderly populations. The ACG System’s segmentation models identify individuals with similar clinical and demographic needs who can benefit from proactive interventions.

Myth: Risk stratification is the same as segmentation.

It’s a common misconception that risk stratification and segmentation are the same — when in reality, risk stratification and segmentation are complementary, but distinct approaches. Risk stratification prioritizes patients based on their relative health risk compared to a population average or their likelihood of experiencing a specific future health outcome. Segmentation, on the other hand, groups clinically similar patients into actionable categories, offering insights into why patients are at risk and how best to intervene.

When combined, segmentation and stratification can help prioritize individuals who need targeted care management programs and tailored, proactive engagement. For example, not all individuals who have a dominant major chronic condition have the same risk of being a high user of health care resources. Stratifying patients within a dominant major chronic condition segment can inform where to prioritize resources and interventions first.

Thoughtful segmentation — paired with stratification — can lead to informed decisions and better care. To learn more, read our Risk Stratification Handbook to discover how a robust population health management approach can transform care delivery.

Myth: A risk score is sufficient for population health management.

Risk scores are valuable for understanding relative resource use compared to a population average or estimating the likelihood of outcomes such as a hospital admission or emergency department visit. However, they represent only a single indicator of risk. To be most effective, risk scores should be integrated into a broader population health management strategy.

An effective strategy will consider additional information — such as an individual’s clinical profile, behavioral characteristics and social determinants of health —as well as identify those who may be moving toward higher risk before an acute event occurs.

Myth: Managing a high-risk population means segmentation won’t provide useful or actionable insights.

A common concern about the Johns Hopkins PNG segmentation framework is that the results may not be actionable when most patients are high risk and fall into PNG groups 10 and 11.

Even when a large proportion of the population is high-risk and has very high health needs, the segmentation model remains essential for creating distinct cohorts and identifying those patients who will drive significant cost and utilization so that relevant interventions and care coordination strategies can be implemented.

Segmentation can also reveal insights into other population groups that might otherwise go unnoticed. For example, PNG 1 — non-users — are often not healthy individuals, but those who are largely unknown to the health system due to social determinants of health. These individuals are likely to become high risk and have more complex social and clinical needs, making early identification a key opportunity for intervention and outreach.

Patient Need Groups provide a toolkit for segmenting an entire population, further risk stratifying each segment, and then identifying modifiable risk factors using Care Modifiers. These additional tools are essential for generating actionable insights that can drive meaningful impact at the patient level.

Beyond Care Modifiers and risk stratification, other ACG System models and markers  — such as frailty markers, Social Need Markers and predicted hospitalization risk — can provide additional insights within a single PNG segment. The key takeaway: a clinically grounded, mutually exclusive segmentation model like PNGs is essential for effective population health management. Combined with the ACG System’s broader set of risk models and markers, it can create a comprehensive picture of a patient’s health needs, enabling targeted care, informed resource allocation and optimized staffing strategies.

To learn more about how the ACG System can support your organization, visit hopkinsacg.org or contact us at acginfo@jh.edu. If you are a current ACG System customer, please reach out to your Account Manager.