Methods
We examined the results of an 88-item survey concerning each physician’s clinical approach to patients with a possible UTI. We mailed the survey in 1994 and 1995 to practicing primary care physicians in 4 specialties that commonly treat adults with uncomplicated UTIs: general internal medicine, family practice, obstetrics and gynecology, and emergency medicine. The emergency medicine physicians were excluded from this analysis, since most of these physicians are hospital based and do not have an office laboratory in the same sense as the other specialties. To obtain a geographically diverse sample, we surveyed physicians from 4 states: Alabama, Nebraska, Minnesota, and Pennsylvania. The names and addresses of potential respondents were obtained from the physician masterfile of the American Medical Association (AMA). Since the AMA masterfile has limited data on physician characteristics (eg, the physician’s specialty), all data in our analysis were based on self-reported survey results. We surveyed the entire population of eligible physicians in each state except Pennsylvania; a random sample was chosen from Pennsylvania because of the large number of practicing physicians in that state.
The survey asked detailed questions about each physician’s clinical approach to a 30-year-old woman with dysuria, a presentation suggestive but not diagnostic of a UTI.18,19 The main outcomes included whether specific tests (urine dipstick, microscopic urinalysis [UA], wet prep, and urine culture) were available in the office and whether these tests were used when diagnosing a UTI in the given patient. In both parts of the analysis we controlled for possible confounding variables. These included items reflecting the physician’s belief in the usefulness of clinical and laboratory information in diagnosing UTIs and physician and practice characteristics. The variables used in the analyses are shown in Table 1.
In the first part of the analysis we determined whether practicing in a regulated state (Pennsylvania) is associated with changes in the likelihood of having tests available in the office. The 4 outcome variables in this part of the analysis included the presence or absence of the dipstick, microscopic UA, wet prep, and urine culture. Thus, test availability was analyzed using binary dependent variables (yes=the physician reported the test in the office; no=otherwise). We did multivariate logistic regressions because of the binary nature of the dependent variables. Variables explaining the presence or absence of the test in the office included the key variable of interest: the physician’s state of residence (either in Pennsylvania or 1 of the previously unregulated states) and the group of variables reflecting the clinical beliefs concerning history, physical examination, and test usefulness in diagnosing UTI in our hypothetical patient, as well as physician, practice, and community characteristics. The explanatory variables were included in the regression models to control for any factors that might confound the effects of being from Pennsylvania. We hypothesized that tests would be found less frequently in Pennsylvania when controlling for other factors.
In the second part of the analysis, we examined whether the availability of tests in the office is related to their use in diagnosing UTIs in the hypothetical patient. The outcome variables in the second part were the self-reported frequency of ordering microscopic UA, urine culture, and the urine dipstick test (the leukocyte esterase and/or the nitrite test). Test use was analyzed as a binary variable (yes=the physician sometimes or usually performed the test; no=the physician rarely did the test). We did not have information on the physician’s use of the wet prep test, so we could not analyze the relationship between availablity and use for the wet prep. Because we analyzed the variable representing test use as a binary variable, we also used logistic regression for the analysis in the second part. In each regression, the variables explaining the frequency of using the test included 3 outcome measures from the first part of the analysis (whether the dipstick, micro UA, or culture was available in the office), and the same control variables reflecting physician beliefs and the personal and practice characteristics used in the first part of the analysis. We hypothesized that test availability would be related to its use in diagnosing UTIs after controlling for other relevant factors.
We used the Stata statistical software program20 for the analysis.
Results
A total of 8942 surveys were sent out. There were 2172 usable surveys returned. After excluding the responses from the 274 emergency medicine physicians, we analyzed the remaining 1898 responses. We were able to compare respondents and nonrespondents on 3 demographic characteristics: sex, board-certification status, and length of time since graduating from medical school. The survey responders were more likely to be board certified but otherwise were similar to nonrespondents Table 2. Comparison of response patterns in Pennsylvania and the other 3 states using logistic regression indicates that Pennsylvania respondents were less likely to be men than respondents from other states. The magnitude of these differences, however, appears small.