## What are regression trees?

Regression trees are a way to partition your explanatory variables to (potentially) better predict an outcome of interest.  Regression trees start with a an outcome (let’s call it y) and a vector of explanatory variables (X).   Simple Example For instance, let y be health care spending, X=(X1,X2) where X1 is the patient’s age and X2 is the patient’s…

## Regional Variation in Medicare Spending and Risk Adjustment

What is the key driver of regional variation in Medicare spending? According to a paper by Reschovsky Hadley, and Romano (2013), the answer is regional variation in patient health. Two key casemix adjustment methods—controlling for patient conditions obtained from diagnoses on claims and expenditures of those at the end of life—were evaluated. We failed to…

Medicare beneficiaries have the option to enroll in private plans to have them operate their benefits rather than use the tradiational Medicare Fee-for-services (FFS) program.  Medicare pays these private plans, known as Medicare Advantage (MA) plans, premiums based on the health status of their enrollees.   Medicare uses a risk score to measure beneficiary health status.…

## Risk Adjustment: K-fold Cross Validation

Oftentimes, researchers will use an existing risk adjustment model and apply it to new data.  The question is, which risk adjustment model should be used?  The most popular method is to calibrate the model (i.e., estimate model parameters or coefficients) using one sample and validate the model using a separate data file.  This standard approach,…

## Risk Adjustment: Overfitting the Model

When creating risk adjustment models to predict health care spending, many researchers aim to maximize the goodness-of-fit of the model.  Maximize the goodness of fit, however, can produce the problem of overfitting. Overfitting occurs when a model describes random error or noise instead of the underlying relationship. Overfitting generally occurs when a model is excessively…

## Risk Adjustment: Measuring Model Fit

You are the CEO of a health plan with millions of beneficiaries.  If you have a demographic and health status information on these beneficiaries, can you predict how much each beneficiary will cost you?  If so, how accurate will your prediction be? Most likely, you will use a risk adjustment model to predict how much…

## Measuring Patient Case Mix in Medicare

How does Medicare measure patient case mix?  For the most part, Medicare uses the Hierarchical Condition Category (HCC) model.  A recent CMS presentation describes the HCC model in more detail.  Today I review where CMS applies the HCC model, provide an overview of the HCC methodology, briefly describe its performance, and give some background on…

## Risk Scores and provider size

At a recent AcademyHealth presentation, Cheryl Damberg discussed her research to design a P4P program for implementation for Integrated Healthcare Association (IHA).  One thing I noticed about the presentation was that smaller provider groups had patients with lower risk scores (i.e., healthier patients).  Is it really the case that small providers treat much healthier patients?…

To evaluate providers based on the health outcomes or the cost of care, one must attempt to evaluate dimensions of care which are strictly within the providers control. For instance, if a physicians treats two patients with breast cancer, but one patient has a more advanced form of breast cancer, one should take this difference…

## Risk Adjustment: Predicting Future Expenditures

Many states rely on managed care organizations (MCOs) to provide medical services for their Medicaid beneficiaries.  Contracting out medical services to private providers relies on the government’s capacity to accurately predict expected cost of care for each beneficiary.  This is typically done through risk-adjusted capitation rates. Which risk adjustment strategy works best?  The answer of…