Healthgrades Methodology: Anatomy of a Rating

Our process used to develop performance ratings.

  1. First we determine what conditions a patient has or procedures performed (for example: total knee replacement) to identify patient like groups using ICD-9 and ICD-10 codes or in some cases DRGs. The model outcomes reflect clinical-based measures of patient disposition during and after care and include in-hospital complications, or in-hospital and 30-day post-admission mortality.

    A full list of ICD-9 and ICD-10 codes can be found in the Healthgrades ICD-10 Mapping Tool.

  2. Next we consider the current condition of the patient and how that could possibly affect the outcome of the condition/procedure.
    a. Develop list of potential risk factors (comorbid conditions, etc.)
    b. Use logistic regression to determine which of the potential risk factors statistically relate to the outcome as well as the strength of that relationship

  3. Then we determine the hospital's predicted outcome rate by summing the individual patient predicted values for each hospital.

  4. Lastly, using a statistical test, we compare the actual outcome for each hospital with that same hospital's predicted outcome.

  5. A star rating is then assigned based on whether the hospital performed statistically better than, the same as, or worse than the predicted outcome.

How We Inform Our Methodology