The MidWest Clinicians’ Network (MWCN) is a nonprofit corporation consisting of approximately 100 community health centers, primary care associations, individual providers, and other partners within 10 Midwestern states [15
]. MWCN’s mission is to enhance professional and personal growth of clinicians so they can become effective leaders in their health centers and promoters of quality, community-based primary health care [15
]. MWCN organizational members may be federally qualified community health centers, non-Bureau of Primary Health Care supported National Health Service Corps sites, and primary care clinics [15
]. Our sample was composed almost entirely of federally qualified HCs; one center was not a federally qualified HC but had a patient population similar to that generally seen in federally qualified HCs.
The MWCN Research Committee is composed of clinicians, administrators, and clinician-researchers from health centers, as well as researchers from the University of Chicago. The Committee includes several general internists and nurse practitioners who have extensive experience conducting research in HCs and providing primary care to vulnerable populations, including migrant farm workers and inner-city, racial/ethnic minority populations. Other MWCN Research Committee members have experience in project management, health behavior, statistics, and community-based participatory research. Based on previous research and clinical experience caring for vulnerable HC populations, the MWCN Research Committee identified LHL as a major priority for research.
The provider survey was developed by the MWCN Research Committee members, and was based on their own experiences and a review of the literature. The survey and study design were approved by The University of Chicago’s Institutional Review Board. The survey contained 22 items, and domains included: 1) the perceived scope of LHL issues at the provider’s HC, 2) strategies the provider currently uses when assessing and assisting patients with LHL, 3) formal programs their HC has initiated at the organizational level, and whether or not these programs were successful, 4) barriers to implementing a formal health literacy program at their HC, 5) the provider’s perception of what would be a useful program to implement at their health center to address LHL, and 6) basic demographic information. All question response scale formats were either on a Likert-type scale, a “check all that apply” response format, or a simple “yes/no” response format, with the exception of one open-ended question at the end asking providers to add any additional comments or suggestions they had regarding LHL in HCs. The survey was pilot tested in a separate sample of 10 HC clinicians and revised for clarity and content and face validity based on their feedback.
The survey sample consisted of 803 providers from 49 different HCs. The mailing list used for this survey was generated from provider lists at each MWCN member organization that utilizes the MWCN Patient Satisfaction Survey. This Patient Satisfaction survey was optional for health centers and its purpose was to provide the organization with information and insight from the patient’s viewpoint about the services provided. At the time the mailing list was compiled, there were approximately 97 MWCN organizational members and affiliated clinics. Of those 97 members, 57 participated in the 2005 patient satisfaction surveys, and of those, 49 provided useable mailing information for their providers. Because it included both health center addresses and provider names, we used this sample of 49 MWCN member agencies to identify our pool of 803 providers for the initial mailing of the survey. The HCs surveyed were located in 10 Midwestern states: Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, Ohio, and Wisconsin. There was no difference in the distribution by state between MWCN HCs who participated or did not participate in the patient satisfaction survey. However, a higher percentage of rural HCs chose to participate in the survey, with 21 (72%) of 29 rural HCs using the patient satisfaction survey as compared to 28 (41%) of 68 urban HCs (p=0.005).
The survey sample included physicians, mid-level health care providers (e.g., nurse practitioners and physician assistants), dentists, dental hygienists, and registered nurses. We sent 803 surveys in the first of three mailings. Providers were informed in the cover letter that the survey was confidential and that their individual responses or names would not be shared with anyone including the HC where they were employed. We subsequently learned that 102 subjects were ineligible due to no longer being employed at the HC, being an employee of the HC but not a health care provider, or being a HC provider who treated only children. These 102 providers were removed, leaving 701 providers in the sample. If the survey was not returned after the initial mailing, we mailed up to two reminder letters with the survey and return envelope to non-responders, with a $2 bill enclosed as a small incentive for the third mailing.
This study was designed to generate primarily descriptive data. We ran descriptive statistics for each survey variable including mean and standard deviation for continuous variables, and frequency counts for categorical variables. Continuous variables were categorized according to the distribution of values for the variable in the dataset. We also examined associations between the binary outcomes of whether or not a specific technique, such as using a validated questionnaire to assess LHL or having patients repeat instructions back to assist patients with LHL, was used with the following independent variables: whether or not a provider had received LHL training, and provider type. For this purpose, we used generalized estimating equations (GEE) logistic regression [16
], which incorporates correlation due to nesting of respondents within HCs and provides standard errors that include this source of variation. Variables considered for model adjustment were the provider-level variables of age and gender, as well as the center-level variable of location (rural vs. urban). These covariates were included in models if they modified the odds ratios (OR) of interest by more than 10%. Data analysis was performed with SAS v.9 (SAS Institute Inc., Cary, NC).