Urinary tract infections (UTIs) are commonly over-diagnosed in emergency departments and ambulatory settings. As many as half of all women presumptively treated with a UTI do not, in fact, have one (here and here). In some settings, the rate of over-diagnosis may be dramatically higher. Some over-diagnosis is necessary; but in ambulatory or acute-care settings, a pattern of over-diagnosis may lead to several problems.
- First, the patient’s actual problem may go untreated (such as an STI or other causes of the presenting symptoms).
- Second, unnecessary antibiotic exposure may later harm the patient, even leading to C. diff infections in the hospital setting or other severe complications.
- Third, the provider, who incorrectly diagnosed the UTI, may learn to misattribute particular symptoms to UTIs that should not be, leading to more over-diagnosis in the future with other patients.
Why are UTIs over-diagnosed? One main reason is the inappropriate utilization of urinalyses (UA). Rather than order UAs based on symptoms, which increases the pretest probability of a UTI and, therefore, the utility of the urinalysis in the first place, many patient-care settings order routine urinalyses, almost as if it were a screening test. One study found that more than half of UAs ordered in the hospital were inappropriate. The rates may be higher in ambulatory settings.
Better understanding how to appropriately diagnosis a UTI, and the pitfalls involved in over-diagnosis, will help us understand the cognitive tools necessary to correctly diagnosis any pathology. Let’s look at what knowledge and skills are necessary.
- We need a knowledge of how the disease presents. The more we understand what the actual symptoms of a UTI are, who is more at risk for a UTI, and other relevant risk factors, then the better we will be able to formulate an idea of the pretest probability of a UTI.
- We need to understand how well the tests for the disease work, and how they compare to the gold-standard test. In the case of a UTI, we need to know which parts of a UA are informative (nitrites and leukocyte esterase) and what the positive predictive values of each part are compared to the gold standard of a urine culture with more than a 100k colony-forming-units.
- We at least need to understand how well the test works in patients with a low, intermediate, or high pretest probability. This allows us to integrate the pretest probability with the test and arrive at a clinically useful post-test probability in the real world setting.
- We need to understand what level of certainty is necessary before initiating treatment. This varies with the disease and the risks of the treatment; it varies with how important it is to be right or how dangerous it is to be wrong. This last step balances the risk-benefit equation that dictates all therapeutics.
How do UTIs present?
UTIs present as dysuria with or without other symptoms like urgency, frequency, hematuria, or suprapubic pain. These symptoms would encompass the most common UTIs, including urethritis and cystitis. In the case of pyelonephritis, additional symptoms of fever, chills, flank pain, nausea, vomiting, and costovertebral-angle (CVA) tenderness might be present. Also, the timing of these symptoms, in most cases, should be acute and progressive.
What symptoms are commonly misattributed to UTIs? In ambulatory settings, a variety of pain symptoms, including general pelvic pain and lower back pain are often falsely attributed to UTIs. Also, fevers of unknown origin may be falsely attributed to UTIs. Note well that flank pain and CVA tenderness should only be attributed to a UTI in the case of a patient with fevers and chills. Conversely, fevers should only be attributed to UTIs when other symptoms (apart from fever) are not present.
Risk factors for UTIs include sexual intercourse, new sexual partners, use of spermicides, history of a prior UTI, or history of UTI in a first-degree family relative. The following are not risk factors for recurrent UTI: precoital or postcoital voiding patterns, daily beverage consumption, frequency of urination, delayed voiding habits, wiping patterns, tampon use, douching, use of hot tubs, type of underwear, or body-mass index (see here).
The differential diagnosis for UTI symptoms include: vaginitis, urethritis, interstitial cystitis, PID, herpes, chlamydia, bladder cancer, post-menopausal atrophic vaginitis, etc. This differential diagnosis assumes that symptoms like low back pain or generalized abdominal and pelvic pain are not falsely attributed to UTI.
More than 50% of women who acutely have one or more symptoms of a UTI will have a UTI on urine culture, and more than 90% of women who have one or more symptoms of a UTI without symptoms of vaginal irritation will have a UTI on urine culture (see here).
How good are the tests for UTIs?
A urine culture revealing > 100k colony-forming-units is the gold standard for the diagnosis of a UTI. The disadvantage of this test is time and cost. Point-of-care urinalysis (UA) is the most commonly employed diagnostic tool. Only three parts the UA have any potential importance: Leukocyte esterase, nitrites, and blood. Of these, leukocyte esterase and nitrites are the most important. How do they perform?
- Presence of nitrites alone yields a positive likelihood ratio or LR+ = 6.5 and a negative likelihood ratio or LR- = 0.58.
- Presences of leukocyte esterase alone, LR+ = 1.4 and a LR- = 0.44.
- The presence of either leukocyte esterase or nitrites on UA is 75% sensitive and 82% specific for a UTI.
So how well does this test work in a patient with low, intermediate, and high pretest probability for a UTI?
Often, what we really want to know when we order a diagnostic test is, What is the post-test probability if the test comes back positive? Recall that this is the post-test probability or positive predictive value, which is,
- Positive post-test probability = (post-test odds)/(post-test odds +1)
- Post-test odds = pretest odds X positive likelihood ratio
- Pretest odds = pretest probability / (1 – pretest probability)
So let’s consider three ranges of pretest probabilities:
- Low pretest probability, such as an asymptomatic pregnant woman or a patient who receives a “screening UA.” The prevalence of asymptomatic bacteriuria in pregnancy is about 6%, so we can use this as our pretest probability.
- Intermediate pretest probability, such as woman presenting with acute onset of one or more symptoms of UTI. As we saw above, this gives a 50% pretest probability.
- High pretest probability, such as the woman presenting with one more symptoms of UTI and no symptoms of vaginal irritation, giving a pretest probability of 90%.
Using a simple calculator, like this one, here are the positive post-test probabilities or positive predictive values (PPV):
- Low pretest probability (6%):
- Nitrites, PPV = 29.3%
- Leukocyte esterase, PPV = 8.2%
- Intermediate pretest probability (50%):
- Nitrites, PPV = 86.67%
- Leukocyte esterase, PPV = 58.34%
- High pretest probability (90%):
- Nitrites, PPV = 98.32%
- Leukocyte esterase, PPV = 92.65%
So we see that in a patient with a high pretest probability, performing a UA doesn’t do much to change our pretest probability, which is so high to begin with that we might harm the patient in the event of a false negative. The presence of nitrites, in the case of the low and intermediate pretest probabilities groups, does significantly change our confidence about the diagnosis, but the presence of leukocyte esterase does not.
Tests are also sometimes important to help exclude a disease. But if you play around with the numbers for negative likelihood ratios and the value of a negative result or the negative predictive value, you will find that a UA is of little value to exclude UTI; a urine culture is necessary for this.
What level of certainty is necessary to begin treatment?
This question is the most complex question we have to consider, and there are many variables. Overall, the interventions we use (Bactrim or Macrobid usually) carry extraordinarily low risks of unintended harm; still, they can cause allergic reaction and even death, as well as antibiotic resistance, alteration of GI flora, or simply the risk of missing the correct diagnosis due to misdiagnosis.
What are the risks of not treating, or delaying treatment while awaiting the results of a urine culture? Well, more pain and suffering on the part of the patient, as well as the potential for worse infections (like pyelonephritis or sepsis). In the case of pregnant or immune-compromised women, the risks of not-treating can be devastating.
Thus, overall, it makes sense that we do not need an extraordinarily high level of confidence to begin treatment, and how much confidence we need varies with the patient. For an immune-competent, non-pregnant young woman who is symptomatic, treatment based on symptoms alone (giving a predictive value of at least 50-90%) is sufficient. We would likely do more harm than good in the event that the UA came back normal (false negative) and we withheld treatment.
In the case of an asymptomatic pregnant woman, given the potential negative sequelae of untreated asymptomatic bacteriuria (like preterm labor or pyelonephritis), it seems reasonable to treat any patient who has nitrites present on UA (PPV of 29.3%) but probably not leukocyte esterase alone (PPV of 8.2%). Furthermore, ordering a urine culture for the asymptomatic pregnant woman with nitrites makes sense because there is a 70% chance that she is being treated mistakenly, and knowing this may influence her prenatal care later on (such as the decision to give continuous antibiotic prophylaxis for recurrent UTIs).
The most common source of over-diagnosis or misdiagnosis of UTI is in the ambulatory care or acute care setting, where a UA is performed without sufficient pretest probability (often as part of “standard” labs), and leukocyte esterase is discovered. The patient’s pretest probability may be even lower than the 6% we have considered here in our calculations. The patient is then treated for a UTI, even though there is a 90-95% chance that she does not have one, with her presenting symptoms falsely attributed to a UTI. Her real condition often goes undiagnosed.
The most important skill necessary to prevent this from happening is a correct appreciation of the symptoms of a UTI in a non-febrile patient: new-onset dysuria, urgency, frequency, hematuria, and/or suprapubic pain. Don’t order a UA if one or more of these symptoms are not present. If they are present, consider not ordering a UA anyway and treating empirically unless something complicates the picture.