The dipstick test was accurate for diagnosing urinary tract infections in high-risk but not low-risk patients
ACP J Club. 1993 Jan-Feb;118:17. doi:10.7326/ACPJC-1993-118-1-017
Lachs MS, Nachamkin I, Edelstein PH, et al. Spectrum bias in the evaluation of diagnostic tests: lessons from the rapid dipstick test for urinary tract infection. Ann Intern Med. 1992 Jul 15;117:135-40.
To determine whether the leukocyte esterase and bacterial nitrite dipstick test for the rapid diagnosis of urinary tract infection (UTI) has different sensitivities and specificities for patients with different clinical manifestations.
Independent comparison of the diagnostic performance of the dipstick test with urine culture for consecutive patients with a high or low probability of infection.
University hospital adult emergency department and associated walk-in clinic.
366 patients (mean age 36 y, range 15 to 92 y, 83% women) who had a urinalysis done for clinical suspicion of UTI. 362 patients were divided into high (> 50%) (n = 103) or low (≤ 50%) (n = 259) prior probability of UTI based on the physician's estimate from the clinical examination, presenting signs and symptoms, and emergency department urinalysis (microscopic examination and dipstick test for protein, pH, blood, glucose, and ketones).
Description of test and diagnostic standard
Tests were done independently in the clinical microbiology laboratory. The dipstick test (Ames Multistix) results were read by an automated processor; urinary nitrite and leukocyte esterase were interpreted as positive or negative, and a positive reaction in either or both was deemed a positive dipstick result. The diagnostic standard for UTI was urine culture. The threshold for a positive culture was > 105 colony-forming units/mL.
Main outcome measures
Sensitivity and specificity for the dipstick test were compared with urine culture for patients with high or low probability of UTI.
The prevalence of costovertebral angle tenderness, dysuria, frequency or urgency, or both, was higher in the "high" probability group. Urine cultures were positive for 20% of patients overall, 53 (51%) of the high prior probability group, and 18 (7%) of the low group. Test properties for the overall, high probability, and low probability groups are summarized in the Table.
The dipstick test for urinary tract infections was highly sensitive in patients with a high prior probability of infection but insensitive in patients with a low prior probability of infection.
Source of funding: No external funding.
For article reprint: Dr. M.S. Lachs, Yale University School of Medicine, 333 Cedar Street TMP B15, New Haven, CT 06510-8025, USA. FAX 203-785-3876.
Table. Test properties of dipstick testing for diagnosis of urinary tract infections
|Patient probability||Sensitivity (95% CI)||Specificity (CI)||+LR* (CI)||-LR* (CI)|
|High||92% (82 to 98)||42% (28 to 57)||1.59 (1.28 to 2.11)||0.18 (0.07 to 0.45)|
|Low||56% (31 to 79)||78% (73 to 83)||2.53 (1.46 to 3.8)||0.57 (0.31 to 0.86)|
|All patients||83% (73 to 91)||71% (66 to 77)||2.92 (2.36 to 3.59)||0.23 (0.14 to 0.38)|
*+LR = likelihood ratio for the presence of disease if the test is positive; -LR = likelihood ratio if the test is negative. Both calculated from data in article
The study by Lachs and colleagues is extremely useful in 2 respects. First, it shows that the dipstick test for urinary tract infection lacks sensitivity precisely in those patients for whom an effective diagnostic test would be most useful: patients with vague symptoms for whom the diagnosis is not clear. By asking clinicians to rate prior probability of infection based on clinical presentation, the investigators showed that this test is not likely to add much useful information. When clinical findings point strongly to a diagnosis, a confirmatory test may not be needed. Alternatively, when the clinician is uncertain, the results of a good diagnostic test are most likely to help.
The main point of this paper has previously been made: Diagnostic tests are often studied in populations different from those to whom they are applied. If the study population is very sick, sensitivity may be higher than in typical office practices, particularly among patients for whom there is diagnostic uncertainty. This is "sensitivity bias." Similarly, specificity may be high in a healthy population (low probability). When replicated in patients who are sicker (and for whom there is more diagnostic uncertainty), more false-positive results are likely. This is "specificity bias" (1, 2).
Clinicians who are able to estimate probability of disease in their practices may find the appendix to this paper useful for calculating the sensitivity and specificity relevant to their patient population.
Martin F. Shapiro, MD, PhD
University of CaliforniaLos Angeles, California, USA