POEMs Research Summaries
History alone can correctly classify slightly more than half the women with suspected urinary tract infections (UTIs). (LOE = 2b)
Study Design: Cross-sectional Funding: Unknown/not stated
Setting: Outpatient (primary care) Allocation: Unknown
The 196 women in this study complained of dysuria for less than 1 week, were 12 years or older, and were recruited from primary care practices in the Netherlands. The researchers excluded: pregnant or lactating women; those with symptoms of pyelonephritis; recent antibiotic use; recent urologic procedures; known structural or functional anomalies; and an immunocompromised state. Each patient underwent a structured clinical assessment and submitted a urine sample for dipstick testing, a urinalysis, and culture. The gold standard for UTI diagnosis was more than 10^3 colony-forming units of a single uropathogen per milliliter. Additionally, the authors polled practicing clinicians and learned that clinicians believe that probabilities of less than 30% and more than 70% are clinically meaningful for guiding UTI treatment decisions. The authors then did a bunch of statistical stuff to identify a range of factors that might discriminate women with UTIs from those without UTIs. This generated several models using only the history or various combinations of the history and various urine tests. As in many other studies, the prevalence of UTI in these women was 61%. Using only the history correctly classified more than half of the women.
Three factors arose as important:
- a positive response to “Do you think you have a UTI?”;
- a positive response to having significantly severe dysuria; and
- a negative response to having vaginal irritation.
These items correctly classified 56% of women with a less than 30% or more than 70% likelihood of having a UTI. Adding a urine dipstick test increased the likelihood to 73%. However, performing a dipstick test only for women whose responses are mixed (placing their likelihood of UTI between 30% and 70%) increased the diagnostic accuracy to 83%. Finally, these authors found that neither examining the urine sediment nor the dipslide test were very useful.
Henry C. Barry, MD, MS
Michigan State University
East Lansing, MI