Assessing Disparities in Adult Vaccination Using Multimodal Approaches in Primary Care Offices: Methodology

Abstract

Racial disparities in invasive pneumococcal disease and pneumococcal polysaccharide vaccination (PPV) persist despite significant progress. One reason may be that minority patients receive primary care at practices with fewer resources, less efficient office systems, and different priorities. The purposes of this paper are: (1) to describe the recruitment of a diverse array of primary care practices in Pittsburgh, Pennsylvania serving white and minority patient populations, and the multimodal data collection process that included surveys of key office personnel, observations of practice operations and medical record reviews for determining PPV vaccination rates; and (2) to report the results of the sampling strategy. During 2005, 18 practices participated in the study, six with a predominantly minority patient population, nine with a predominantly white patient population, and three with a racial distribution similar to that of this locality. Eight were solo practices and 10 were multiprovider practices; they included federally qualified health centers, privately owned practices and faculty and University of Pittsburgh Medical Center community practices. Providers represented several racial and ethnic groups, as did office staffs. PPV rates determined from 2,314 patients’ medical records averaged 60.3 ± 22.6% and ranged from 11% to 97%. Recruitment of practices with attention to location, patient demographics, and provider types results in a diverse sample of practices and patients. Multimodal data collection from these practices should provide a rich data source for examining the complex interplay of factors affecting immunization disparities among older adults.

APPENDIX

Sampling Scheme

A proportional random cluster sample will be conducted to achieve about 165 patients per practice, nested by clinician.

• Step 1. Site or UPMC will develop list of names of persons aged 65 or older seen in the last year as outpatients. We will assume N patients appear on the list. Inpatient visits are not to be included.

• Step 2. Honest broker will take list and assign each patient to a clinician based on the clinician seen most frequently (over last 3 years).

• Step 3. The list will be stratified by clinician.

• Step 4. If a clinician has <15 patients, then delete clinician and clinician’s patients (revise N to be the number of patients remaining).

• Step 5. If, and only if, N <175, then take all patients for the practice (inclusive sample randomization not necessary).

• Step 6. If N ≥ 175, each clinician specific list of patients will be randomized by honest broker as follows below.

• Step 7. Calculate the proportion p of patients to be sampled; p = 175/N.

• Step 8. The number of patients on a particular clinician’s list is n. For each clinician in a practice, calculate n × p. If n × p > 15 for each clinician, then randomly sample n × p patients and sampling is complete. If n × p ≤ 15, then go to Step 9.

• Step 9. If n × p ≤ 15 for a particular clinician, then randomly sample 15 patients for that clinician and sample n × p patients for the other clinicians.