This study was conducted from the responses of a national cross-sectional, online survey of patient's satisfaction (DrScore.com) that collected anonymous patient ratings of U.S. primary care physicians for patient advocacy research and to produce patient satisfaction report cards for physicians. The survey focused on the most recent outpatient visit and used a list of U.S. physicians that permitted patients to look up their doctors and access the survey. Participation in the survey was advertised to patients on a public radio show (The Peoples Pharmacy), through patient advocacy groups, and through on line search engines. The survey asked patients to both rate their physician on several dimensions of health care experiences, as well as provide specific comments about aspects of care that were most excellent or most in need of improvement. Questions were rated on a scale of 0 ('not al all satisfied') to 10 ('extremely satisfied'). Two patient satisfaction scores were considered as outcomes in this study: ratings of the provider (Physician Care, 9 items) on the thoroughness of care, physician communication and follow-up, listening, demeanor, discussion of test results, answering questions, treatment success, and including the patient in decision processes; a second rating was of the practice (Office Practice, 5 items) and included items on continuity of care, convenience of facility, referrals, hours, and ability to meet health care needs of the patient. For both scales, the summed scores were scaled from 0 to 100 by taking the item mean and multiplying it by 100, representing compete satisfaction on all characteristics measured.
Patient waiting time at the last office visit was measured by asking the patient to recall the amount of time he/she waited before being seen by the physician for a scheduled appointment. Response categories were: 1–5 minutes waiting time in office, 6–15 minutes, 16–30 minutes, 31–60 minutes, and more than 1 hour. The shorter time intervals at the start were chosen because pilot data showed that approximately 70% of the patients waited below 15 minutes. Perceived time spent with the physician was measured as < 5 minutes, 6–10 minutes, and > 10 minutes, also assessed by patient recall. No personal identifying information were collected in this study (e.g., name, address, of medical number) and expedited IRB approval was obtained to conduct analyses of de-identified data.
Statistical analysis
Multivariate regression and logistic regression models predicting the three satisfaction ratings were estimated using the Generalized Estimating Equations (GEE) method implemented in the SAS System v9 (GenMod procedure)[11] In order to adjust for clustering, an exchangeable working correlation matrix was specified where the observations were clustered according to clinic. The default robust standard errors in proc GenMod were used. All models were adjusted for patient reported age, gender, reason for visit, and first visit. Age was modeled as a continuous variable based on its observed close approximation to a linear response to an overall rating of patient satisfaction with physician seen. Assessments of covariates such as type of health care organization, severity of illness, and race were not collected in the study survey.