© Copyright 2007 Health Grades, Inc. All rights reserved.
Hospital Report CardsTM Methodology 2
May not be reprinted or reproduced without permission from Health Grades, Inc.
However, since the appendectomy cohort includes very few cases over 65 years of age, all-payer state data
were used to rate hospitals in those states where state data are available.
For Multivariate Logistic Regression-Based Ratings (see below), HealthGrades conducted a series of data
quality checks to preserve the integrity of the ratings. Based on the results of these checks, we excluded a
limited number of cases because they were inappropriate for inclusion in the database or miscoded.
Examples of excluded patient records were:
·
Patients under the age of 65 (except appendectomy)
·
Patients who left the hospital against medical advice or who were transferred to another acute care
hospital
·
Patients discharged alive with a length of stay equal to or less than one day (except for coronary
interventional procedures, heart attack, heart failure, resection / replacement of abdominal aorta,
carotid endarterectomy, back and neck surgery (spinal fusion), back and neck surgery (except
spinal fusion), chronic obstructive pulmonary disease, community acquired pneumonia, peripheral
vascular bypass, and atrial fibrillation)
·
Patients who were still in the hospital when the Medicare claim was filed
·
Patients with an invalid gender
Methodology for Ratings
Our methodology takes into account patient characteristics such as age, gender, and underlying medical
conditions that could increase the patient's risk of mortality or complication. Specifics about the statistical
methods used are provided here and include:
·
Multivariate Logistic Regression-Based Ratings
·
APR-DRG-Based Ratings
Multivariate Logistic Regression-Based Ratings
The inhospital data for 28 diagnoses and procedures on the HealthGrades Web site represent three years
of patient discharges from 2003 to 2005 for MedPAR and three years of patient discharges from 2002 to
2004 for state data.
In the initial analysis of the data, a separate data set was created for each group of patients having a
specific procedure or diagnosis based on ICD-9-CM coding (e.g., coronary bypass surgery, total hip
replacement). Each group of patients was defined by using the information on diagnoses and procedures
coded in the patient records. See Appendix A for a list of the diagnosis and procedure codes that define
each patient cohort. The quality measure for some cohorts was mortality, whereas for other cohorts the
quality measure was major complications.
For each patient cohort, we developed a list of specific procedures (e.g., quadruple bypass surgery), a list of
risk factors (Appendix C), and a list of post-surgical complications. These latter two lists were developed in
two steps:
1
We identified all diagnoses occurring in more than .5 percent of the patients for the current analysis and
the previous analysis.
2
We used a team of clinical and coding experts to identify the complications in the list created in Step
One.
In some cases an ICD9 code can be either a risk or a complication. In these cases, a code is differentiated
by the presence of a 900 post-operative complication code. For example in the case where a patient record
