Development and validation of the SLEOTB survey
The systemic lupus erythematosus observations of travel burden (SLEOTB) project developed a scale for measuring potential travel burden by conduct preliminary interviews to gain baseline knowledge of travel burden experienced by SLE patients in South Carolina, developing a survey instrument that to characterize observations, and pre-testing the survey through cognitive interviewing sessions. More detailed information on development of the survey tool-the background, literature search strategy applied, results of qualitative studies performed, and pre-testing for usability-is reported elsewhere . For a community perspective on issues of travel burden (e.g., costs of travel, health care impacts of travel burden, visitation frequency for various health care needs, etc.), individual interviews were conducted with ten (10) SLE patients enrolled in the SLE Database Project at the Medical University of South Carolina (MUSC). Patients were randomly selected from the over 1000 SLE patients currently being followed at MUSC. A coding tree was developed to assist with coding of this data. Interviews were coded using NVivo software (Qualitative Solutions and Research, Pty Ltd, Victoria, Australia), a qualitative analysis research tool. Knowledge gained through the preliminary interviews assisted in determining which categories of analysis from the preliminary interviews still held valid for the construction of the survey.
Major focus areas of the initial survey instrument were: (1) travel patterns for health care; (2) perceptions of travel burden for health care; (3) perception of discrimination and its impact on healthcare utilization (pooled from two validated sources-reactions to race module from the behavioral risk factor surveillance system [23, 24] and Victoroff’s Oppression Questionnaire ); and (4) perceptions of how travel burden impacts disease management and seeking health. To ensure the validity of the survey instrument, cognitive interviews were conducted with 15 randomly selected participants from the MUSC SLE Database who did not participated in the preliminary interviews. Cognitive interviewing has been shown to improve survey questionnaires [26–30]. In order to further refine data on thematic issues relevant and specific to the types of self-rated travel burden that SLE patients in South Carolina experience across various socio-economic strata, the cognitive interviews were analyzed using the cultural consensus model [31–34]. The cultural consensus model is a recent innovation in ethnographic methods that has been shown to be sensitive to intra-cultural diversity as well as effects of stressors by social and cultural context . Cultural consensus modeling was performed using Anthropac software (Columbia, Anakytic Technologies).
The resulting survey instrument assesses the impact that self-rated travel burden has on SLE patients regarding: (1) visitation frequency for primary care/rheumatologist/immunologist; (2) participation in clinical trials; (3) non-health related issues; and (4) how self-rated travel burden compares across urban/rural divides. The survey also includes questions concerning quality of life measures, costs for travel to health care, socio-demographic information, estimated time spent for travel to healthcare, and preferred mode of transport. Questions contained in the 55-item questionnaire are structured in a likert-scale manner, for measuring the strength of responses. Each component of the survey also provides space for open-ended response, allowing more in-depth responses and giving participants the opportunity to provide additional information they may deem relevant.
Patients and entry criteria
Patients invited for survey administration were SLE patients attending rheumatology clinics at MUSC. All SLE patients met at least four components of the 1997 American College of Rheumatology (ACR) revised criteria for SLE , were 18 years of age or older, and were residing in South Carolina at the time of the study. Patients invited to participate in the proposed study were lupus patients participating in a longitudinal observational web-based SLE Database at MUSC. There were 402 patients with lupus enrolled in the Database during enrollment in this study. Patients in the Database were characterized longitudinally for disease activity and quality of life. As part of the informed consent process, participants agreed to future re-contact regarding other research studies. MUSC’s SLE cohort is geographically diverse, representing more than 60 South Carolina, Georgia, and North Carolina counties. Of the 402 patients with lupus, 336 were African–American. This study was approved by the University of South Carolina (USC) and MUSC Institutional Review Boards and written consent was obtained prior to data collection.
For the current study, a link to a description of the study was placed on the MUSC lupus erythematosus (MUSCLE) group’s website and their listservs were used to email potential participants. Recruitment letters were also mailed and phone calls made to MUSC SLE database participants, and flyers were posted in corresponding lupus clinics. Survey administration was offered to participants in the most convenient format for them. This included telephone, online, or in-person administration. Patients were asked during recruitment which method they preferred. An online-version of the instrument was developed and made available on MUSC’s Research Electronic Data Capture (REDCap) system, a secure, web-based application designed exclusively to support data capture for research studies [37, 38]. Target enrollment for the survey was 148 participants. Although we were only able to secure roughly 28 % of participants from the original target, this relatively small sample should not have any impact on findings as SLE patients can be difficult to enroll in studies in general [39–41]. While the survey instrument obtained both qualitative and measures more readily handled with quantitative analyses, this paper focuses on quantitative analyses. For a review of qualitative work in this area, please see Ortiz, Flournoy-Floyd, & Williams, 2015 .
The main variables assessed to examine health-related travel burden among SLE patients included travel time (in minutes) to lupus-associated medical care, and distance (in miles) to rheumatologists of lupus patients. To characterize travel burden more broadly, participants were asked to rate several domains in which travel could produce burden: (1) difficulty keeping appointments; (2) difficulty with general medical care travel; (3) difficulty with primary care travel (travel to/from primary care physicians); and (4) difficulty with rheumatologist travel (travel to/from rheumatologist). In distinguishing between travel for various aspects related to lupus patients care, it was our intent to be able to isolate specific travel burden for rheumatologist considering this subspecialty is vitally important for lupus patients. Moreover, being able to assess travel in the domains of general medical travel (e.g., travel for medications) and travel for primary care allows us to more thoroughly characterize travel burden. Response options for each of these measure was a likert scale (1, very difficult; 2, difficult; 3, neither; 4, easy; 5, very easy). All measures rely on self-reported data from patients participating in this survey. We utilized four outcome measures to characterize travel burden: (1) whether travel affected appointments (yes/no); (2) whether travel caused medications to be discontinued (yes/no), (3) whether medical transportation increased stress (yes/no), and (4) number of appointments missed due to transportation problems in the past year (count measure). Additionally, we control for several sociodemographic characteristics which have been shown to be impactful along the casual pathway in contributing to stress among lupus: (1) race (white, black); (2) age (in years); (3) gender (male/female); (4) educational attainment (equal to or less than high diploma or equivalent/college degree or higher); (5) employment status (yes/no); self-rated health status (good/fair); annual household income (<$15,000; $15,000–$60,000; ≥$60,000); marital status (married, never married, other).
Input of data from telephone, online, and in-person surveys was completed throughout survey administration, and data was exported from the REDCap system in an excel format and manipulated using SAS statistical software. Survey data was analyzed utilizing statistical methods most appropriate for the sample size and descriptive statistics generated. First, we provide sociodemographic characteristics of participants. Then we describe travel burden by using measures to characterize various domains in which travel could produce burden by presenting descriptive statistics (Tables 2, 3, 4). For group comparisons of the interactions of travel burden and personal attitudes about travel burden, the Kruskal–Wallis test was applied. In order to investigate the association between travel burden and medical care, we focused on our four primary variables. For (1) travel affected appoints, (2) whether travel caused medications to be discontinued and (3) whether medical transportation increased stress we utilized logistic regression models. For the last outcome measure, number of appointments missed as a result of transportation problems in the past year, we utilized Poisson regression models. To further study the relationship between medical travel burden and selected outcomes (medication discontinuation and missed appointments) after adjusting for possible variables including education level, employment status, self-valuated health status, annual household income and marital status, multivariable logistic or Poisson regression models were constructed.