Does This Woman Have an Acute Uncomplicated Urinary Tract Infection?

A 24-year-old healthy woman calls her primary care physician complaining of a burning pain when urinating and increased urinary frequency for several hours. She has had 2 prior urinary tract infections (UTIs), and this episode seems "just like the other ones." She is sexually active with one partner and uses a condom with spermicide. She denies fever, back pain, nausea, vomiting, vaginal discharge, and hematuria.

A 20-year-old woman presents to your office complaining of urinary frequency, burning on urination, and vaginal discharge. She states that she has had occasional fevers and chills but denies nausea, vomiting, and back pain. She is sexually active with one partner, takes oral contraceptive pills, and intermittently uses condoms. Physical examination shows her to be in mild discomfort and febrile but without tenderness in her costovertebral areas. Pelvic examination demonstrates no vaginal lesions or rashes, minimal white vaginal discharge, and no cervicitis. Her dipstick urinalysis result is negative for leukocyte esterase, nitrite, and blood.

Acute uncomplicated UTIs are common among women, accounting for more than 7 million office visits annually in the United States¹ and affecting half of all women at least once during their lifetime.² A recent study of sexually active young women found the incidence of cystitis to be 0.5% to 0.7% per year.³ In aggregate, the direct costs of these infections have been estimated to be $1.6 billion annually in the United States.⁴

One might anticipate that the management of acute uncomplicated UTI would be relatively uniform because the causative agents and in vitro susceptibilities are known, and therapeutic responses to antimicrobials have been studied carefully.^{2 ,5 - 7} Unfortunately, the evaluation and treatment of acute uncomplicated UTI in women vary substantially among physicians,⁸ likely reflecting the limitations of routine diagnostic assessments. If done carefully, however, the history taking and physical examination can likely be used in the initial evaluation of patients suspected of having an acute uncomplicated UTI and can guide the selection of additional diagnostic and therapeutic strategies.^{2 ,7}

Several types of UTI are possible, described by their location: urethritis, cystitis, pyelonephritis, and perinephric abscess. The usual reference standard for diagnosing UTI is the presence of "significant" bacteria in a clean-catch or catheterized urine specimen, most commonly defined as the isolation of at least 10⁵ colony-forming units (CFU) per milliliter of a single uropathogen.² In women who present with symptoms of cystitis or urethritis (lower UTI), it has been suggested that the best diagnostic criterion for clean-catch urine is the isolation of uropathogens in concentrations as low as at least 10² CFU/mL.⁹

Uncomplicated UTIs occur in individuals who have a normal urinary tract system. A UTI in an individual with a functional or anatomic abnormality of the urinary tract (including a history of polycystic renal disease, nephrolithiasis, neurogenic bladder, diabetes mellitus, immunosuppression, pregnancy, indwelling urinary catheter, or recent urinary tract instrumentation) is considered complicated and may have a higher risk of treatment failure.¹⁰ Differentiating between these types of UTIs is important because uncomplicated infections are usually cured with simple antimicrobial regimens.¹⁰

Previous studies have shown that the prevalence of asymptomatic bacteriuria (significant bacteriuria without symptoms of UTI) in women of reproductive age is approximately 5%.^{11 - 12} This value represents the pretest probability of disease (the probability of UTI before any diagnostic tests are applied). Several historical features, symptoms, and signs have been associated with acute UTI and may be useful as diagnostic tests, allowing the clinician to estimate the probability of UTI in a patient after taking a history and performing a physical examination (posttest probability). Historical features such as a previous history of UTI, recent sexual activity, or contraceptive use identify individuals at greater risk of developing a UTI. Symptoms of an acute infection include burning or pain on urination (dysuria), frequent voiding of small volumes of urine (frequency), the urge to void immediately (urgency), and the presence of blood in the urine (hematuria). Discomfort in the lower abdominal area is also consistent with a UTI. In contrast, patients who report vaginal discharge or irritation are less likely to have a UTI and more likely to have vaginitis or cervicitis. The presence of fever and suprapubic or costovertebral angle tenderness may indicate infection of the upper urinary tract.

Vaginal infections (eg, Gardnerella, Candida albicans, Trichomonas), sexually transmitted diseases that may lead to pelvic inflammatory disease (eg, Chlamydia trachomatis, Neisseria gonorrhoeae), and other sexually transmitted diseases (eg, herpes simplex virus) that may mimic symptoms of UTI are considered separate from UTIs. Reasons for this include the fact that they are caused by different microbes; limited to female genital structures with a unique set of complications if untreated; and require different forms of treatment.¹³ Differentiating between sexually transmitted diseases, vaginal infections, and UTIs can be difficult because symptoms and signs commonly overlap.¹³

We searched the English-language medical literature to determine the accuracy and precision of the clinical examination in women suspected of having an acute UTI. We searched MEDLINE for articles from 1966 through September 2001 with a search strategy similar to that used by other authors in this series.¹⁴ Search terms included urinary tract infection, diagnostic tests, physical examination, and sensitivity and specificity. (The complete search strategy is available from the authors on request.) This computerized search was supplemented with a manual review of the bibliographies of all identified articles, additional "core" articles (identified a priori as articles used to develop a recent guideline for treating acute uncomplicated UTI in women), 3 commonly used clinical skills textbooks,^{15 - 17} and contact with experts in the field. One of the authors (B.K.N.) initially screened the titles and abstracts of the search results. Two of the authors (S.B. and B.K.N.) then independently reviewed and abstracted data from articles identified as relevant.

We included studies in our review if they contained original data on the accuracy or precision of the history and/or physical examination in diagnosing acute uncomplicated UTI in healthy women. Articles were excluded if they evaluated infants, children or adolescents, pregnant women, nursing home patients, or patients with complicated UTI; or contained insufficient or incomplete data to allow calculation of likelihood ratios (LRs) for signs or symptoms of acute UTI.

We also chose to include articles on the dipstick test in this analysis because it is commonly used in the clinical setting and provides an immediate result that can be incorporated with other elements of the initial clinical assessment. During our search, we discovered that a previous systematic review evaluated the diagnostic accuracy of the dipstick test.¹⁸ Since this was a high-quality review (meeting all 6 criteria of a previously published guideline for evaluating systematic reviews),¹⁹ we chose to use the information regarding the accuracy of the dipstick test synthesized in that article.

The methodological quality of the included articles was assessed independently by 2 authors (S.B. and B.K.N.) using criteria adapted from other authors in this series.^{14 ,20} Disagreements were resolved by a third author (S.S.) Level I studies included those with an independent blind comparison of signs or symptoms with a gold standard among a large number (≥50) of consecutive patients suspected of having a UTI. Level II studies were similar to those in level I but involved a smaller number of patients (<50). Level III studies were retrospective chart reviews. Level IV studies included "grab" samples of patients (ie, nonconsecutive patients who obviously have the target condition plus, perhaps, healthy individuals without symptoms suggestive of the target condition) or made comparisons of signs or symptoms to diagnostic standards of uncertain validity among consecutive patients. Finally, level V studies included those that used a diagnostic standard of uncertain validity among "grab" samples of patients.

We used published raw data from the reported studies that met our criteria to calculate summary measures for the LRs for components of the clinical examination for UTI. Likelihood ratios are related to sensitivity and specificity [positive LR = sensitivity/(1 − specificity) and negative LR = (1 − sensitivity)/specificity], but are more clinically useful because they can be used to generate posttest probabilities.²¹ A random-effects model was used to generate conservative summary measures and confidence intervals (CIs) for the LRs and estimates of disease prevalence.^{22 - 23} Uncertainty in these measures is reflected in the broad CIs around the estimates. When a summary LR included studies of lower quality, we conducted sensitivity analyses to examine the impact of excluding lower-quality studies on the summary LR and the effectiveness score, a measure of the discriminatory power of a diagnostic test.²⁴

We found 9 studies of the 464 identified by the search that satisfied all inclusion criteria (Table 1). Six studies^{25 - 30} reported the accuracy of 1 or more symptoms in the diagnosis of UTI, 2 studies^{31 - 32} reported the accuracy of symptoms and physical examination signs, and 1 study reported the accuracy of self-diagnosis.³³

The studies were published between 1965 and 2001 and generally involved patients with 1 or more symptoms of a UTI who presented to outpatient clinics. The summary prevalence of UTI in the 5 studies that included only symptomatic patients and used an appropriate gold standard was 48% (95% CI, 41%-55%),^{25 - 28 ,30} indicating a high probability of disease for women who met the studies' inclusion criteria. In all of the included studies, UTI was defined by the presence of at least 10 000 or 100 000 CFU/mL of a single uropathogen, except for the most recent study, which used a cutoff of at least 100 CFU/mL.³³

Five^{25 - 28 ,30} of the 8 studies describing the accuracy of symptoms were of high quality (level I). Both studies^{31 - 32} describing the accuracy of the physical examination were of lower quality (level III and IV) as was the study examining self-diagnosis (level V).³³ Reasons for quality scores lower than level I are shown in Table 1. Two of the lower-quality studies^{29 ,31} received lower scores because they included patients with vaginal discharge but without symptoms of UTI and therefore did not specifically address the diagnostic accuracy of signs and symptoms exclusively in women suspected of having a UTI.

The precision of a symptom or sign refers to the degree to which different examiners report the same finding (eg, dysuria present or absent) when interviewing or examining the same patient. None of the identified studies described the precision of the history or physical examination in the diagnosis of UTI, possibly because the questions and examination procedures were considered to be unambiguous. For example, most of the historical items consist of asking yes or no questions such as "Are you having burning or pain with urination?" Variations in interview style and the phrasing of questions may affect results, but there is no information from the identified studies to suggest particular wording of questions or specific ways to examine patients for the 2 relevant physical examination signs (costovertebral angle tenderness and vaginal discharge).

Symptoms. Eight studies^{25 - 32} examined the accuracy of 9 different symptoms for predicting the presence of UTI. These symptoms and the corresponding positive and negative LRs from each study are shown in Table 2a. Three of the symptoms (flank pain, abdominal pain, fever) had both positive and negative summary LRs with CIs overlapping 1.0 and are therefore not useful as diagnostic tests.

Four symptoms significantly increased the probability of UTI: dysuria, frequency, hematuria, and back pain. Four symptoms significantly decreased the probability of UTI: absence of dysuria, absence of back pain, a history of vaginal discharge, and a history of vaginal irritation. The symptoms with the greatest diagnostic power were a history of vaginal discharge (LR, 0.3) and a history of vaginal irritation (LR, 0.2); both of these symptoms substantially reduced the probability of UTI.

Self-diagnosis. One study examined the accuracy of self-diagnosis and included 172 women in a university-based practice with recurrent UTI (more than 2 UTIs in the past year).³³ During the study period, 88 of the women reported 172 episodes of self-diagnosed UTI; 144 of these episodes (84%, 95% CI, 77%-90%) were found to have positive urine cultures. Additionally, 64 women reported mild symptoms that they did not self-diagnose as UTI and another 20 women never had symptoms. In this population of patients, the positive predictive value of self-diagnosis was very high (84%). Likelihood ratios for self-diagnosis can be calculated assuming that the women with mild symptoms or no symptoms correctly self-diagnosed with no infection (these women did not have a urine culture, but all symptoms resolved spontaneously). If this assumption is true, the LR for a positive self-diagnosis is 4.0, while the LR for a negative self-diagnosis is 0.0 (Table 2).

Combinations of Symptoms. One study²⁹ provided information to calculate the LRs for combinations of symptoms in the diagnosis of UTI (Table 3). In this study, the presence of dysuria and frequency without vaginal discharge or irritation was associated with a very high LR (24.6). Conversely, the LR for the combination of vaginal discharge or irritation without dysuria was low (0.3). While the LRs from this study must be interpreted with caution due to the study's low quality score (level IV), the observed LRs were similar to those calculated by combining the individual summary LRs from the other studies (Table 3).

Physical Examination. Two studies^{31 - 32} reported the accuracy of 2 physical examination signs for the presence of UTI. Both studies were of relatively low quality, and therefore the summary data do not represent strong evidence of the true accuracy of these signs (Table 2). The presence of costovertebral angle tenderness increases the likelihood of infection, but the LR is only weakly predictive and similar in magnitude to the related symptom (back pain). The presence of vaginal discharge on examination decreases the likelihood of UTI (LR, 0.7) although it is less powerful than the LR for the symptom of vaginal discharge reported by the patient (0.3).

Dipstick Urinalysis. Since a high-quality systematic review examining the accuracy of the dipstick urinalysis for the prediction of UTI exists, we used the data synthesized in the report by Hurlbut and Littenberg.¹⁸ Those authors identified and summarized 51 studies and generated summary receiver operating characteristic (ROC) curves for combinations of the nitrite and leukocyte esterase dipstick tests. They found that the nitrite positive or leukocyte esterase positive combination had the greatest area under the ROC curve, indicating the most accurate test. The point on the summary ROC curve with the best accuracy represents a sensitivity of 75% and a specificity of 82%. Using these values, the positive LR for a urinalysis is 4.2 and the negative LR is 0.3 (Table 2). A range of similar points on the ROC curve that was supported by the largest number of studies was also examined, and the resulting LRs were similar in magnitude. Although other combinations of the nitrite and leukocyte esterase test will increase either sensitivity or specificity (eg, requiring both to be positive will decrease sensitivity and increase specificity), the nitrite or leukocyte esterase positive combination was the most accurate test.¹⁸

Because the largest study to examine the accuracy of symptoms was also of lower quality,²⁹ we performed a sensitivity analysis to determine the effect of this study on the summary LRs. Inclusion of this study always made the symptoms (dysuria, frequency, vaginal irritation, and vaginal discharge) appear to be more powerful diagnostic tests. However, in no case did inclusion of this study improve a test with marginal discriminatory power into the highly effective range (effectiveness score ≥3.0).²⁴ The positive and negative LRs for dysuria and frequency excluded 1.0, whether or not the study was included, with 1 exception. The positive LR for increased urinary frequency was 1.8 (95% CI, 1.1-3.0) when all studies were included vs 1.4 (95% CI, 1.0-1.9) when the study was excluded. That study²⁹ has a larger impact on the diagnostic value of vaginal symptoms because fewer studies were involved. The absence of vaginal discharge, a feature reported in only 3 studies, makes a UTI more likely whether or not this study²⁹ is included (LR, 3.1 [95% CI, 1.0-9.3] for all studies vs LR, 1.7 [95% CI, 1.3-2.2] when excluded). The presence of vaginal discharge still lowers the likelihood of a UTI whether or not the study by Komaroff et al²⁹ is included (LR, 0.3 [95% CI, 0.1-0.9] for all studies vs LR, 0.6 [95% CI, 0.4-0.9] when the study is excluded). The impact on the efficiency of using the symptom of vaginal irritation is seen from Table 2 as only 2 studies evaluated this feature.

Symptoms suggestive of UTI are common complaints of young women seeking urgent medical care. Although textbooks of clinical medicine^{15 - 17} routinely mention many of the symptoms and signs of UTI, the overall accuracy of these symptoms and signs has not previously been critically and systematically evaluated. A clear understanding of the value of each of these diagnostic tests may enable physicians to make more informed decisions about the choice of specific tests and management options.

The initial step is to be certain that the patient does not have a complicated UTI as defined by the factors listed earlier (see "Definitions"). The probability of UTI in patients with risk factors for a complicated infection is not known because these patients were not included in the studies identified by our search. Such patients may be at greater risk of treatment failure,¹⁰ and clinicians may want to consider early urine culture and empirical treatment as shown at the top of the proposed algorithm (Figure 1).

Using a standard evidence-based technique,²¹ a clinical encounter begins with an estimation of the pretest probability of disease followed by the application of 1 or more diagnostic tests to determine the posttest probability of disease. We consider the pretest probability of UTI to be equal to the prevalence observed in studies of asymptomatic bacteriuria, or approximately 5%.^{11 - 12} In this review, 5 studies reported the prevalence of UTI in patients presenting with 1 or more symptoms of acute UTI, and the summary prevalence was 48% (95% CI, 41%-55%).

Interestingly, the probability of UTI changes substantially when a patient presents to a clinician, increasing from 5% (in historical controls without symptoms) to approximately 50% (in patients in the included studies who presented with 1 or more symptoms). This change in probability corresponds to an LR of 19, representing a very powerful "diagnostic test." Clinically, it is useful to know that patients who present with 1 or more symptoms of UTI have a very high probability of infection. Since all of the studies included in this review evaluated the diagnostic value of symptoms and signs after patients presented to a clinician, the relevant pretest probability for these tests is 50%.

Although the pretest probability of UTI in the average patient who presents with 1 or more symptoms is approximately 50%, this varies considerably according to the individual's risk profile. There are 3 well-established risk factors for acute UTI in young women: recent sexual intercourse,^{3 ,34 - 38} use of spermicide (on condoms or with diaphragms) during sexual intercourse,^{3 ,34 - 36 ,39 - 40} and history of UTI.^{3 ,36} Other risk factors, including a maternal history of UTI,³⁴ a history of childhood onset of UTI,³⁴ and the presence of bacterial vaginosis,⁴¹ also have been found to be associated with UTI. The presence of any of these risk factors increases the pretest probability of UTI and should be considered when evaluating patients. Unfortunately, the diagnostic power of these risk factors (sensitivity, specificity, or LRs) is not known, as the majority of studies assessing these risk factors used a case-control design or did not present sufficient data to calculate LRs.^{3 - 4 ,35 - 39 ,42} Further research is needed to determine the diagnostic power of these risk factors so that the information can be used during the clinical encounter to estimate the pretest probability of disease.

In the included studies, all diagnostic tests were evaluated by their ability to change the already very high (50%) probability of UTI in the study population. Because these patients initially presented with at least 1 symptom, some of the power of each symptom was already "used up" by the time the patient presented to a clinician (and the probability of UTI increased from 5% to 50%). In a sense, the diagnostic power of the symptom is being "used" twice. Initially, the presenting symptom (most commonly dysuria or frequency) caused the patient to present to a clinician and was at least partially responsible for raising the probability of UTI from 5% to 50%. Subsequently, the value of the presenting symptom and all other potentially relevant symptoms was assessed after presentation to a clinician.

It is therefore not surprising that most of the individual symptoms and signs have LRs relatively close to 1.0 and therefore do not have great additional diagnostic power after presentation. The main exception to this finding is the history of vaginal discharge or vaginal irritation, which substantially reduces the probability of UTI.

One study found that back pain and costovertebral angle tenderness were useful for predicting the presence of UTI.³² This study was a retrospective chart review of patients who had a urine culture in an emergency department, and it is possible that back pain and costovertebral angle tenderness were predictive of upper UTI (pyelonephritis). However, since none of the included studies performed a gold standard test for upper UTI, we were unable to determine if individual symptoms and signs were more predictive of upper vs lower UTI. Most patients with symptoms suggestive of UTI and features classically associated with upper UTI (back pain, fever) are evaluated and treated for presumed pyelonephritis (Figure 1), even though the diagnostic accuracy of these signs and symptoms for predicting upper UTI is not known. Since most patients in the included studies did not have back pain and fever, we believe that the other symptoms evaluated in our review are most useful for predicting lower UTI (cystitis).

In contrast to the value of individual tests, certain combinations of symptoms result in large changes in the probability of UTI and represent powerful diagnostic tests. The combination of dysuria and frequency without vaginal discharge or irritation corresponds to an LR of 24.6. Although the combined LRs were generated from only one study of lower quality,²⁹ these LRs were similar to those found when multiplying the summary LRs for the individual symptoms, suggesting that they are reasonable estimates of the true diagnostic power of these combinations. In addition, another study⁴³ that was excluded from our analysis (because it included an unknown number of asymptomatic patients) used the same combinations of symptoms and found similar positive predictive values and LRs.

Although evaluated in only one study,³³ self-diagnosis appears to be a useful diagnostic test (LR, 4.0) in women with recurrent UTI. Because this study did not perform urine cultures for women with mild or no symptoms, there is some uncertainty in the LR estimates. Similarly, the study population consisted of mostly highly educated single white women, and it is not clear if the results apply to other groups of women. Nonetheless, these findings suggest that women learn to recognize the symptoms of UTI and are able to accurately diagnose a new infection, a finding that deserves further study and may have important implications for management of this large group of patients.

Dipstick urinalysis alone is a moderately powerful diagnostic test (Table 2). If the dipstick is used alone, the posttest probabilities for women with symptoms of a UTI are 81% (positive result) and 23% (negative result).

Figure 1 shows a proposed algorithm for evaluating patients with symptoms of UTI. Although the algorithm itself has not been prospectively studied, the recommendations are based on the posttest probabilities of UTI generated from the summary LRs in the current analysis (Table 2). In women with risk factors for a complicated UTI or with back pain, fever, or malaise (suggesting possible pyelonephritis), a urine culture with initial empirical treatment is recommended. If a woman reports a history of vaginal discharge, the posttest probability of UTI from this single historical item is reduced to 23%, and a pelvic examination to rule out a vaginal infection should be considered in addition to a dipstick urinalysis and urine culture.

The algorithm highlights the finding that the history and physical examination alone can substantially increase the posttest probability of UTI, effectively "ruling in" the diagnosis. Since the only physical examination finding that increases the probability of UTI is costovertebral angle tenderness, the physical examination may be omitted without a substantial loss of diagnostic power in patients without a history of vaginal discharge or irritation. Using individual summary LRs, a patient with dysuria, frequency, and hematuria (but no back pain at this point in the algorithm) has a posttest probability of UTI of 81%; using the combined LR estimate of dysuria and frequency without vaginal discharge (LR, 24.6), the posttest probability of UTI is 96%. Given these very high probabilities of UTI, clinicians should consider empirical treatment without urine culture or dipstick urinalysis.

Conversely, even mostly negative history, physical examination findings, and dipstick urinalysis cannot reliably rule out the diagnosis of UTI in women without a history of vaginal discharge or irritation. For example, to generate the lowest possible posttest probability of disease, a woman must still present with at least 1 symptom. If she presents with frequency (LR, 1.8) with no dysuria (LR, 0.5) and no back pain (LR, 0.8) (the only 2 negative symptoms other than vaginal symptoms), a negative dipstick result (LR, 0.3), and no other positive symptoms, her posttest probability of disease is still 18%, which is considerably higher than the prevalence of asymptomatic bacteriuria in the population (5%). Although we do not address the optimum management of such patients, we believe that the relatively high probability of UTI (~20%) warrants a urine culture (Figure 1), an approach that has been supported by others.¹⁰ Clinicians may also want to consider performing a pelvic examination, especially in patients at high risk for sexually transmitted disease or if the urine culture is negative and symptoms persist. As noted, it is theoretically possible to rule out UTI in women who present with vaginal discharge, where the lowest possible posttest probability of disease is 6% (if they also have no dysuria, no back pain, a negative dipstick result, and no other positive symptoms). We recommend that clinicians consider obtaining a urine culture in patients with at least 1 urinary symptom and vaginal discharge, since the posttest probability of disease will only rarely reach this lowest possible 6%.

If the history and physical examination are neither strongly positive nor negative, a positive dipstick result still results in a high posttest probability of disease (approximately 80%), and empirical therapy should again be considered without urine culture. In all of the scenarios in the algorithm urine culture may be indicated, without regard to the posttest probabilities, if the patient has experienced recurrent infection and antibiotic resistance is suspected.

Older guidelines for the evaluation of patients with suspected UTI recommend urine culture in all patients, even in those found to have a high probability of UTI after the history and physical examination.^{29 ,44} More recent reviews and management strategies suggest that a diagnosis of UTI can be established in women who present with typical symptoms and are found to have a positive dipstick or urinalysis result (without obtaining a urine culture).^{10 ,45 - 48}

Unlike these treatment recommendations, our proposed algorithm (Figure 1) suggests that, in selected patients with mostly positive symptoms, the probability of UTI is so high (~90%) that empirical treatment may be considered without dipstick testing or urinalysis. A similar strategy was recently evaluated in a randomized trial comparing management via telephone with office evaluation in 72 women with suspected UTI.⁴⁹ The investigators found no difference in symptom scores or patient satisfaction with the 2 strategies. Prior studies examining the effect of symptom-based management of patients with suspected UTI (after a telephone call or office visit to a health care provider) have shown that empirical therapy decreases costs without increasing adverse outcomes.^{50 - 51} However, the main purposes of the current algorithm are to define the posttest probabilities of disease from specific clinical scenarios and to allow clinicians to make informed testing and treatment decisions based on their clinical judgment. Further research is needed to determine clinical outcomes, costs, and patient satisfaction associated with different testing and treatment strategies for managing patients who present with specific constellations of symptoms of UTI.

In the first case, the woman has 2 symptoms of UTI (dysuria and frequency), no vaginal discharge, and believes that her current symptoms are similar to prior episodes. These features all increase her probability of UTI, which is greater than 90%. Her sexual history does not suggest that she is at high risk for a sexually transmitted disease. Using the algorithmic approach, the patient should be asked about risk factors for complicated infection as well as symptoms classically associated with pyelonephritis (fever, back pain, nausea, vomiting). As has been shown, telephone evaluation and treatment of similar patients may be an appropriate strategy.^{49 - 50} In this patient, a positive dipstick urinalysis result would further increase the probability of UTI, while a negative result would not rule out infection.

In the second case, the woman has 2 symptoms of UTI (dysuria and frequency) as well as vaginal discharge (which decreases the probability of UTI and increases the probability of vaginal infection). A pelvic examination does not suggest a specific diagnosis and the dipstick urinalysis result is negative. The posttest probability of UTI is approximately 20%, illustrating that even a negative physical examination and dipstick test result are insufficient to rule out UTI in a patient with 1 or more symptoms. A urine culture will help determine the need for treatment, and cervical cultures are indicated to rule out chlamydia and gonorrhea and help determine the cause of her symptoms.

In a woman who presents with 1 or more symptoms of UTI, the probability of infection is high (approximately 50%). Four symptoms (dysuria, frequency, hematuria, and back pain) and 1 sign (costovertebral angle tenderness) increase the probability of UTI when present. Combinations of symptoms can substantially increase the likelihood of UTI, effectively ruling in the disease based on the history alone. Patients with recurrent infection may be able to accurately self-diagnose UTI.

In contrast, the history and physical examination cannot reliably rule out UTI in women who present with urinary symptoms. Although 4 symptoms (absence of dysuria, absence of back pain, and a history of vaginal discharge or vaginal irritation) and 1 sign (vaginal discharge) decrease the probability of UTI, even combinations of symptoms, signs, and a negative dipstick result rarely lower the probability of UTI below 20%. A urine culture and pelvic examination should be considered in patients who present with some symptoms of UTI but with mostly negative history and physical examination findings.

Dipstick urinalysis, which is a simple and inexpensive test, is moderately powerful and should be considered in women with appropriate urinary tract symptoms. If the dipstick result is positive, the probability of UTI is high, especially when combined with other positive findings from the history and physical examination. If the dipstick result is negative, the probability of disease is still relatively high (23%) and a urine culture should be considered to rule out infection.

Care should be taken to identify women with vaginal discharge or vaginal symptoms. If present, a pelvic examination and cervical culture are indicated to rule out infection due to chlamydia⁵² or gonorrhea, as well as other vaginal infections that require definitive therapy. Similarly, in women with back pain, fever, or significant malaise, an office examination, combined with dipstick urinalysis and urine culture, may aid in the diagnosis of pyelonephritis, although the accuracy of individual tests for establishing upper UTI is not known.

Knowledge of the LRs for specific symptoms, signs, and diagnostic tests used to evaluate patients with suspected UTI may improve the ability of clinicians to more accurately predict the probability of infection in individual patients. It seems reasonable to offer empirical treatment when the probability of infection is high and to pursue additional diagnostic testing (eg, urine culture, pelvic examination, and cervical cultures) when the probability of UTI is low or intermediate. However, the actual cost-effectiveness of specific testing and treatment strategies is not clearly established, and prospective studies examining clinical benefits, adverse effects, costs, and patient satisfaction with specific approaches are needed.