r i s k   a s s e s s m e n t   l a b


These days, with managed care, physicians must make important business decisions using concepts they never learned about in medical school.

For example, suppose you are a member of a group practice. The practice can handle more patients. A health plan approaches you with an offer to add hundreds of its enrollees to your practice. They will pay you $300 per patient per month, for which the practice must provide all physician services and all acute hospitalization needed by these patients. There are other details, but, to simplify matters, we will ignore them for the moment. Is this a reasonable offer? Should you accept it?

To answer this question, you might want to estimate the likely average monthly medical care costs for these new patients relative to your existing patient pool. How would you do this? To help you understand, we have developed a simple interactive example to illustrate the underlying concepts. Here is how it works.

Essentially, you use patient attributes as clues to estimate how costly any given patient is likely to be. A patientís age, gender, and medical history are examples of such clues. You have data on these clues for your current patients, and the health plan has similar data on their enrollees. You also have actual medical care cost data for your patients. Of course, no amount of such information is going to tell you how costly any one specific patient is going to be. However, you can get reliable estimates of the total (or average) costs for groups of patients that have similar cost profiles. You can use clues to assign patients to one of a reasonably small number of "expected cost groups" (or ECGs, for short). Then you can calculate the average cost for each ECG. Thus, a young woman with no history of disease will fall into a lower cost ECG than will an elderly woman with diabetes. Again, being in a low cost ECG does not necessarily mean that a patient will have low medical costs. It just means that she has a lower probability of high cost compared to someone in a higher cost ECG. She could still get hit by a bus. In our example, a patient's gender and age must be known in order to assign him/her to an ECG.

Letís say we have five ECGs: ECG1 through ECG5. You can think of each ECG as a multi-faceted die with a different cost range (eg. $0-$99, $100-$499, $500-$1999, etc.) on each face. A patientís actual experience works like a roll of the appropriate die. Each of the 5 ECG dice has the same cost ranges on the faces, but they are weighted differently. For example, the ECG1 die (on the "healthier" end of the scale) will have higher probabilities associated with the lower cost faces and lower probabilities associated with the higher costs faces than the ECG5 die (on the "sicker" end of the scale).


NOTE: You must have a JAVA1.1-enabled browser to view this lab.
  Microsoft Internet Explorer 5.01
(Go to Microsoft's download page).
  Netscape Navigator 4.7
(Go to Netscape's download page).

  Use your browser's BACK and FORWARD buttons to switch between this page and the lab.