The multi-method QI intervention added practice site visits (for academic detailing and QI facilitation) and network meetings (for sharing of best practices) to the approach of guideline dissemination and audit and feedback, employed in a less intensive intervention. Ten sites received the intensive multi-method QI intervention, and ten sites received the less intensive intervention. The study was conducted in a practice-based research network (PPRNet) among users of a common electronic medical record (Practice Partner Patient Records, Seattle WA), which historically provided audit and feedback to its practice members.
As a supplement to the original study, we were also interested in whether minority patients were more, less, or just as likely to benefit from the intervention as non-minorities. The study presented here focused on outcome and process measures for minorities within 3 primary care practices, all of which received the intensive intervention. These 3 practices (labeled A, B, and C) were selected because they each had a significant (i.e. > 10%) proportion of minority patients and had recorded patient ethnicity in their electronic medical record. Practice A is an urban internal medicine practice in the Midwestern U.S. with 5 healthcare providers. Practice B is a rural family medicine practice in the Northeastern U.S. with 8 healthcare providers. Practice C is an urban family medicine practice in the Southeastern U.S.
A total of 21 study indicators (see Table 1) were obtained from each practice at baseline (fourth quarter 2000) and at the end of the study (fourth quarter 2002). These indicators were derived from published sources [4, 5, 13–16] and were deemed to be the most appropriate indicators for measuring quality of prevention and management of cardiovascular disease and stroke. Fourteen of the study indicators are process measures, reflecting whether recommended tests were done, appropriate diagnoses made or medication prescribed. Seven indicators are outcome measures, reflecting whether patients achieved recommended treatment goals. Some of the measures represent primary prevention, e.g., screening for hypertension or hyperlipidemia. Others represent secondary prevention, e.g., reaching treatment goals for glycosylated hemoglobin, low-density lipoprotein (LDL) cholesterol, and blood pressure in patients with diabetes. The institutional review board at the Medical University of South Carolina approved the study.
To determine practice performance on the study indicators, participating practices ran a computer program to extract patient activity during the previous quarter from their electronic medical record. To protect patient confidentiality, the extract program assigned an anonymous numerical identifier unique to each patient. The extract program obtained demographic information such as age, ethnicity, and gender, and diagnoses, medications, laboratory data, and vital signs. Text of consultation reports, progress notes, and discharge summaries were not extracted. The data were copied to diskettes and mailed to PPRNet or sent electronically via a secure server. In the PPRNet offices, data were bridged to standard data dictionaries and converted to SAS® (Statistical Analysis System, Cary NC) data sets on standard microcomputers for analyses.
In each patient's electronic medical record, ethnicity was recorded as white, black/African American, American Indian/Alaskan native, Asian, native Hawaiian/other Pacific islander, and "some other ethnicity", while ethnicity was recorded as Hispanic/Latino and non-Hispanic/Latino, all in concordance with the 2000 U.S. Census ethnicity categories. Currently, these physician practices allow the patient to designate their ethnicity categorization. However, because this process for collecting ethnicity data began in the middle of our study, some ethnicity categorizations were made by the office staff within each of the practices. Ethnicity data was only available on approximately 42% of patients, due to the fact that the electronic medical record software program did not require physicians to enter patients' ethnicity data until its most recent version was released, which occurred during the study time frame. Improvements in process and outcome measures were compared between minority and non-minority patients. Minority was defined as any ethnic designation other than white non-Hispanic.
Changes in the process and outcome measures were of primary interest in this study. Within each practice, these changes were compared between the minority patient population and the white patient population. Measures were deemed eligible for comparison if at least 10 minority patients were included in the rate calculations. For example, if practice A only had 8 minority patients with a diagnosis of having had myocardial infarction (MI), then the measure of the percentage of MI patients who had been prescribed a beta blocker could not be compared between the minority and white patient population. The proportion of eligible measures in which minority patients exhibited greater improvement was calculated for each practice and for all 3 practices combined. A Wilcoxon signed rank test (the non-parametric equivalent of the paired t-test) was used to test the hypotheses that minority patients exhibited changes similar to those of the non-minority patients. This study had approximately 80% power (2-sided hypothesis testing, α = 0.05) to detect a 6.6 percentage point difference between average improvement in the study indicators among all minority and non-minority patients.