Use of Risk Adjustment Models for Provider Reimbursements - Text Mining Techniques for Healthcare Provider Quality Determination - page 362

Table 1. Physician reimbursement by medication for treatment of osteoporosis

RevisedRXName N Obs Variable Mean Std Dev

Actonel® 6214 SUM OF REIMBURSEMENT

TOTAL CHARGE

139.4900000
261.0492787

634.5592923
892.4160990

Boniva® 1702 SUM OF REIMBURSEMENT

TOTAL CHARGE

98.6538977
211.6433980

182.7933705
1215.17

Evista® 2915 SUM OF REIMBURSEMENT

TOTAL CHARGE

138.7793274
346.3469526

414.3256163
1678.43

Fosamax® 11555 SUM OF REIMBURSEMENT

TOTAL CHARGE

154.8413638
286.5861503

307.8532491
729.2619202

reimbursement formulas

It is very difficult to find information concerning reimbursement formulas. However, information on
Medicare reimbursements is readily available, and we will discuss it in detail. In addition, the MEPS
data provides actual reimbursement information as well as actual payments from different providers. For
this reason, it is extremely valuable information. We can use this information to see if we can estimate
the relationship of payment to charges. We can also look at the cost-to-charge ratio to see which hos-
pitals greatly inflate their charges, and see if these are the same hospitals that shift patients into higher
categories of severity.
Another aspect of reimbursement is that approximately 5-6% of the patient population is designated
as outlier and the provider is eligible for an additional payment; the formulas for payment only allow
for 5-6% of the population to be so designated. As we noted in chapter 3, this percentage assumes that
the population is normally distributed. However, if the true distribution is more likely to be gamma, a
considerably higher proportion of the population should have this designation. (Anonymous-CMS, 2006)
Since it is fairly reasonable to assume a gamma distribution instead of a normal distribution, providers

Figure 3. Kernel density of reimbursements by medication