Use of Risk Adjustment Models for Provider Reimbursements - Text Mining Techniques for Healthcare Provider Quality Determination - page 359

measures (since it's much easier to control hypertension if the patient rarely had BPs above 130/80 to
begin with). Your average control of these diseases and costs per patient will soon run circles around
those of physicians who allow themselves to be hamstrung by more old-fashioned diagnostic criteria.
So, screen aggressively and diagnose generously!... Wouldn't it be better for all involved to just keep
taking her blood pressure until the blood pressure goal is met? If 2 or 3 checks don't do the trick, a
conscientious physician should be willing to sacrifice a Korotkoff sound or two for the sake of quality.
We have found that ultrarapid cuff deflation can be an equally effective nonpharmacologic remedy for
these resistant cases, and improve patient comfort at the same time. Be creative—you will find a way to
reach the recommended goal!

For example, patients were diagnosed with diabetes if fasting glucose values were above 140 mg/dl.
Then it was redefined at 126 mg/dl. Now 100 mg/dl is defined as pre-diabetes. Similarly, patients were
treated for high cholesterol at 220 mg/dl. Then it became 200 mg/dl and now it is moving to 170 mg/dl.
Each time the standard is lowered, more patients are diagnosed and treated. The number of diagnoses
for each patient increases.
Another concern with pay-for-performance is “adverse selection”, or restricting access to the most
severely ill, the uninsured, or the elderly with complex medical problems. (Bodrock & Mion, 2008)
Indeed, the above quote in Hayward & Kent goes on to suggest that a physician should refuse to treat
such patients, recommending another provider to the patient, who would be willing to treat the patient.
One study examining the unintended consequences of pay for performance indicated that providers with
a large minority patient population are penalized under the system either because minority patients are
harder to treat, or because they are more likely to be uninsured or to wait until their condition worsened.
(Karve, Ou, Lytle, & Peterson, 2008) This study also showed that hospitals with large minority popula-
tions were less likely to deliver evidence-based care, reducing their performance on input measures as
well as outcome measures. The question is, then, whether payers will take the patient's socioeconomic
status into consideration when defining reimbursements. There is no question that patient compliance is
an important issue, and more affluent patients are far more likely to comply with treatment requirements
compared to those who are less affluent.
Currently, reimbursement measures have focused on inputs into the system. For example, performance
measures proposed for surgeons include:(Lewis & Friesen, 2006)

• Prophylactic antibiotics delivered within one hour before surgery

• The prophylactic antibiotic as an appropriate cephalosporin

• Prophylactic antibiotics are discontinued within 24 hours of surgery ending

• Appropriate venous thromboembolism prophylaxis was ordered

While these are reasonable goals, there is no measure as to whether the surgeon is better at surgery
compared to other surgeons. Since the first three measures listed above involve antibiotic delivery, it
would be worthwhile to measure the infection rates of surgical patients to determine if there is a dif-
ference across surgeons, and whether there is a small misclassification rate in predicting infection just
using the input measures listed above.
Another contract for performance measures utilized by a private insurer included the following
measures, again focusing on inputs rather than outcomes: (Collier, 2007)