Use of Risk Adjustment Models for Provider Reimbursements - Text Mining Techniques for Healthcare Provider Quality Determination - page 383

Table 18. continued

Cluster # Code Combinations Translations Nosocomial?

31 V02.59, 311, 715.90,

300.00, 733.00, 682.2

Other specified bacterial diseases, Depressive disorder, not elsewhere classified,
Osteoarthrosis, unspecified whether generalized or localized, Anxiety states,
unspecified, Osteoporosis, unspecified, Other cellulitis and abscess, trunk

No

32 412, 428.0, 411.1,

v45.82, 272.4, v45.01

Old myocardial infarction, Congestive heart failure, unspecified, Intermediate
coronary syndrome, Percutaneous transluminal coronary angioplasty status, Other
and unspecified hyperlipidemia, Cardiac pacemaker

Yes

is not specifically identified.
Once we have identified those clusters that represent nosocomial infections versus community ac-
quired infections, as we have done in previous chapters, we can determine which hospitals have a high
proportion of patients in the nosocomial categories compared to those hospitals with more patients in the
community acquired clusters. There are hospitals, too, with patients in both types of clusters, suggesting
that poor infection control procedures can result in nosocomial infection acquired from those patients
admitted with community acquired infection.

future trends

The trend to using provider “report cards” to define reimbursements will only accelerate. Therefore, it
is important to validate the risk adjustment models that are used to define these report cards. Unfortu-
nately, validation has tended to lag behind the use of these models. As we have seen, there can be very
different results generated using different measures of patient severity. Because of these problems, it
will be necessary in the future to define multi-dimensional measures of quality, and to drill down into
the reasons that providers have different results on these report cards.

dIscussIon

As shown by the different examples of public reporting, there are already several states that have moved
to the web display of hospital quality, some by individual hospitals and some by general region. Private
companies such as healthgrades.com also have quality reports available to the public. Therefore, there
has to be considerable concern that the methodology used to develop the quality rankings is adequate to
the task. The more dollars that are at stake in the reimbursement formulas, the greater the incentive to
“game” the system by upcoding; by taking advantage of problems with the current methodology. The
only way to prevent such gaming is to use methods that do not require that we assume conditions that
we know to be false.