What a Readmission Risk Score Doesn’t Tell You

My husband and I just bought a house, which required that we go through the tedious exercise of applying for a mortgage. As part of the process our lender requested a credit report for both of us. This report included a credit score.

A credit score is a number ranging from 300 to 850. It is calculated based on five categories: payment history, amounts owed, length of credit history, new credit and types of credit used. However the number itself bears no relevance to these categories. It is a single measurement, like weight, that can be either relatively high or low along a scale.

In other words, my credit score does not offer any insight into why I got this score or how it could be changed. Furthermore, it does not take into consideration one significant factor–my husband, whose credit score is different from mine.

Fortunately our lender was satisfied with our credit scores. We closed on our home last month. The experience provides a real example of the limited usefulness of risk scores and other regression-based predictors.

For example, common models for predicting hospital readmissions are based on a logistic regression analysis. They take a set of variables (such as length of stay and diagnosis) and calculate a score that corresponds to high, medium, or low risk that a patient will be readmitted within 30 days. The risk score is a useful measure of one thing: risk of readmission.

What a regression-based risk score does not do is tell a hospital why patients with a particular condition are at greater risk for readmission compared with another diagnosis. Or help a physician understand how changes in severity of illness affect readmission rates among his patients. Other assessment tools, such as rules-based clinical logic, are necessary to provide this information.

Rather than a logistic regression analysis, 3M uses a clinical categorical model to analyze hospital readmissions. It assigns each patient to a single clinical category that reflects diagnoses, measures the interaction of multiple health conditions, and identifies avoidable, unplanned readmissions. The results allow health organizations to better understand the risk of readmission within a broad clinical context. You can find a complete breakdown of the clinical categorical model in the 3M Clinical and Economic Researcher’s blog on the fundamentals of classification systems.

A probability or risk score is valuable to predict risk and track trends. But a risk score alone will not help clinicians understand why patients are sick or how to help them avoid readmission. A clinical categorical model gives them information that can help answer these questions.

If only the banking industry classified consumers according to a similar categorical model. Then maybe I would better understand my credit score.

Kristine Daynes is a product marketing manager for payer and regulatory markets at 3M Health Information.

2 responses to “What a Readmission Risk Score Doesn’t Tell You

  1. Do you see this being adopted by those using the regression model currently? If so, do you see this as a potential benefit for the hospital?

  2. Yes, 3M has worked with many hospitals using a clinical categorical model to identify the types of patients at risk for readmission, particularly types of potentially preventable readmissions. The approach allows facilities to identify ways to improve the delivery of care, not just flag a patient at risk.

    Allina Health in Minnesota went as far as to include the 3M PPRs in their patient census dashboard. In looking to reduce over $40 million in costs associated with avoidable readmissions, Allina has been able to maintain their ratio of actual to expected readmissions below their target. They are participating in the Minnesota RARE Campaign, a project of the Minnesota Hospital Association which also targets preventable readmissions throughout the state using the 3M PPRs.

    Here is a link to an article about Allina Healthcare in Healthcare Informatics that discusses their patient census dashboard:
    http://www.healthcare-informatics.com/article/allina-s-pioneering-move-forward-population-health-risk-stratification-and-management-0

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s