Risk Adjustment Based Upon Resource Utilization - Text Mining Techniques for Healthcare Provider Quality Determination - page 240

Therefore, the exact methodology used to compute the indices is proprietary and unavailable for direct
examination.
Background
Studies that describe the relationship between healthcare utilization and patient condition and outcome
are scarce. The Kessner Adequacy of Prenatal Care Index was developed to examine the relationship
between prenatal care and birth outcomes.(Kotelchuck, 1994) It was regarded as flawed because of a
heavy reliance on the timing of prenatal care. A second index was developed to overcome these prob-
lems. However, the timing of prenatal care was also used to develop this second index, although its
importance was reduced.
Another study examined the relationship between compliance with medications and the need for
additional resources such as emergency room visits.(Tu et al., 2005) This study suggested that patient
compliance is an important factor in resource utilization, indicating that there may be problems with a
reliance on such utilization to determine the difference in quality between providers. Patients with dia-
betes who routinely take their medication and test their blood sugar levels are compliant; patients who
do not take their medication regularly enough are not. These patients who are not compliant may have
more episodes of uncontrolled blood sugar that require emergency treatment.
A second study examined the relationship between patient body mass index (BMI) and resource
utilization, with the result that patients with a BMI of 30 or greater had higher wound infection rates.
(Thomas et al., 1997) This study used the following diseases as indicator variables in the regression
model: degenerative joint disease, hypertension, cancer, coronary artery disease, peripheral vascular
disease, diabetes, emphysema/asthma, congestive heart failure, stroke, liver disease, and renal failure.
Note that while this list has some commonalities with the Charlson Index, there are differences. In
particular, all of the co-morbidities are given equal weight, and there is no discussion of the validity
of this list of co-morbid diseases, or justification as to why the BMI would be the ultimate marker for
resource utilization.
Another study, for example, showed that two providers that treated patients with COPD had patients
with very different characteristics.(Mapel et al., 2000) One provider has an average age that was ap-
proximately 7 years younger compared to the second provider and the patients had fewer co-morbidities.
However, the resource utilization was very similar between the two groups, suggesting that the first
provider was over-utilizing resources given its patient population.
resources In tHe mePs data
The MEPS dataset contains very detailed information about reimbursements from payers, including
payments by patients, insurers, and government agencies. Therefore, we can look at the relationship be-
tween actual reimbursements to patient condition. Moreover, the patient conditions are very detailed and
include any ICD9 code that was used to diagnose a patient in the course of a one-year period. Therefore,
we can examine actual resource utilization in relationship to patient diagnoses, to the other indices, and
to primary and secondary procedures. We can also examine the resource utilization by state and federal
governments, by private insurers, and by the patient.