Why risk scoring models are essential for population health management

Population health management (PHM) is centered around delivering care and improving health outcomes in a specific group of people. It involves collecting, standardizing, and analyzing clinical or patient data to uncover ways to improve health of patients as well as financial outcomes for organizations.

For organizations wanting to implement effective population health management programs, they should have robust practices in place to assess population needs, integrate varied data, stratify populations by risk, and deploy targeted interventions. This is where integration of technology and data-driven approach can reduce costs, improve clinical outcomes, and help healthcare providers and organizations leverage evidence-based decision-making.

Let’s discover how.

A look at three different patient journeys

Jane, a 40-year-old female, has been experiencing persistent cough and shortness of breath for a few weeks. After visiting her physician and taking a series of lab tests, Jane finds that she is suffering from non-small cell lung cancer (NSCLC).

In another part of the world, 55-year-old Yang is also diagnosed with the same condition during one of his annual routine health checks at the hospital. He had no prior symptoms.

Fatima, aged 30, has been experiencing sudden weight loss, weak appetite, chronic fatigue, and difficulty swallowing. It was not long before Fatima too was diagnosed with NSCLC.

Jane, Yang, and Fatima are each put on a standard set of first-line treatments by their respective physicians. While their medications were the same, the journey towards their recovery differed widely.

Jane responded well to the first-line treatment, but things didn’t work as effectively for Yang and Fatima. They were both immediately moved to a second-line treatment where Yang responded well and showed signs of improvements. However, Fatima’s condition deteriorated. Though she was moved to a third-line treatment, there were no signs of recovery and she had to be hospitalized.

Figure 1: A typical patient care journey

Three individuals - all with the same condition, and placed on the same treatment plan - had very different outcomes. Understanding why each patient responded differently to the same treatment has a lot to do with their medical history, risk factors, demographics, and more.

In this example, Jane, Yang, and Fatima individually represent a large cohort of diagnosed and at-risk NSCLC patients with similar characteristics. Creating a comprehensive view of patient trends recorded for each cohort over time can help pharmaceutical organizations understand patterns of what is likely to happen to such patients and identify risk factors throughout their treatment journey. These insights can be used to alert and engage healthcare professionals (HCPs) when it matters the most, enabling them to curate the best, most effective treatment plans for patients and manage the overall health of the patient population.

Figure 2: Identifying risk factors along the patient care journey

How risk scoring models can help

Risk scores categorize patients into different risk groups based on their clinical and lifestyle characteristics. A score indicates the likelihood of an event during a patient's treatment journey, such as the possibility of switching to second- or third-line treatments, requiring hospitalization, future re-admissions, and so on.

Risk scores offer pharmaceutical organizations a clear-eyed view of individual patient journey along common disease trajectories. This enables organizations to:

• Accurately predict future patient health outcomes or risks

  Identify at-risk patients

  Identify HCPs who have recently engaged with patients

  Alert targeted HCPs about at-risk patients for timely interventions

  Ensure the continuous supply of treatments at the right time and place – driving brand visibility, trust, and engagement among HCPs along the way

Data shows that adopting risk-scoring models can help pharmaceutical organizations achieve:

Figure 3: Estimated outcomes with risk-scoring models

A step-by-step guide on how to build risk-scoring models

• Step 1: Define your target population

Define the patient cohort based on a set of inclusion and exclusion rules such as demographics, symptoms, comorbidities, drugs, procedures, visits, and other observations.

• Step 2: Build a predictive model

Based on a target cohort of patients with a certain medical history, create a model to identify the probability of a patient experiencing a specific event such as hospitalization, treatment switching, etc. within a given prediction window.

Figure 4: Predictive Modeling

• Step 3: Run new patients through the model and assign risk scores

Deploy the predictive model with an automated pipeline to create model-ready patient data. Run new patient details through the model and score them on the likelihood of disease advancement.

• Step 4: Map patients to their HCPs

Map each patient to their primary HCP based on the frequency of engagements within a specific period (for example: 12 months).

• Step 5: Target HCPs for engagement

Target HCPs based on precision-adjusted patient scores. A high adjusted score for an HCP implies that they have the highest number of at-risk patients, making them a key prospect for priority sales engagement. Finally, trigger alerts to sales representatives through email or an automated CRM system for timely communication and engagement.

Figure 5: Target HCPs through the adjusted patient score

Navigating challenges in implementing risk scoring models

A significant challenge may arise from unclear understanding of patient characteristics. This can result in an imbalanced ratio of healthy and high-need patients. Organizations, therefore, can collaborate with relevant entities and access population data including census, housing data, and others

Additionally, organizations may lack a robust digital platform which can restrict implementation of advanced analytics and interoperability across data systems. To address this, organizations should invest in modern and scalable data systems and work with skilled data scientists or IT professionals.

Conclusions

Creating risk scores will provide organizations the 360-degree insight necessary to bring HCPs and patients, who are in need of proactive and preventive care, together – driving positive outcomes across the healthcare journey.

Here are a few tips to keep in mind when you build an effective risk-scoring model:

• Ensure you have ready-to-use and objective data to achieve maximum prediction accuracy

  Keep data quality and standardization in check at all times

  Ensure you have a well-balanced team of advanced medical and technology experts

  Reassess individual risk levels regularly as they tend to change over time

  Adjust risk level's as a patients situation changes based on new information