Dec 2004

A Fever of Unknown Source, Commentary 3

Timothy E. Quill, MD
Virtual Mentor. 2004;6(12):533-535. doi: 10.1001/virtualmentor.2004.6.12.ccas1-0412.


Mr. and Mrs. Kendall were quite excited about the newest addition to their family, 2-month-old Janna. But they were anxious on this trip to the doctor. Janna had developed a fever in the middle of the night, and the Kendalls decided to take her to the emergency room. Dr. Stinton, the ER physician on service, heard of the Kendalls' story and decided she'd better see their daughter straight away.

Upon questioning, Dr. Stinton learned that Mrs. Kendall's pregnancy with Janna was perfectly normal, as was her delivery at just over 38 weeks. In fact, until last night, the Kendalls had little reason to be concerned about Janna's health. As they reported, she'd passed all her doctor's visits with flying colors and had begun immunizations according to schedule. In other words, she was overall a healthy infant.

Last night, however, they awoke to cries unlike those they had heard before and ran in to check on Janna. She had no interest in being fed and seemed warm to her mother. She had been irritable throughout the day and less interested in nursing. Mrs. Kendall took Janna's temperature twice rectally, and both times it read 38.3° C. That's when they bundled her up and went to the ER.

On physical exam, Janna was alert and active in her mother's arms, but appeared ill. Her temperature registered 38.4° C. The physical was normal, revealing no obvious foci of infection. A complete blood count in the ER uncovered a slightly elevated white blood cell count (15.5 x 109 cells / Liter) but no other abnormalities. On the basis of these findings, Janna was admitted to the hospital for fever of unknown source; her workup included urinalysis, urine culture, chest radiography, blood cultures, and lumbar puncture. In lieu of lab results, antibiotics were started.

Distressed by the invasiveness of the tests (which lasted well into the early morning hours) and the hospital admission, the Kendalls asked for more information about the likelihood of a serious illness. Dr. Stinton responded that it was hard to say. "The information is quite complex, and we don't want to confuse you. Let's not take any chances." Willing to do whatever it took for Janna to be well, they agreed.

After Janna reached the hospital ward, Mr. Alstadt, a medical student on his clinical clerkship began learning Janna's story through the family's account and ER notes. He diligently prepared for presenting his new patient on rounds later that day.

When the attending asked Mr. Alstadt the reason for Janna's admission, Mr. Alstadt recalled the story and was poised with several sets of clinical guidelines and algorithms from various sources that he had researched overnight [1,2]. He quoted beautifully from them, and most of them seemed to support admitting Janna for a full sepsis workup.

The attending physician was not convinced. "Really? Who recommends a full sepsis workup for a moderately ill, 2-month-old infant likely to have reliable follow-up with her parents? In my clinical judgment, this patient should have been sent home last night, maybe after taking a blood culture, but definitely without a lumbar puncture. I think medical school should spend less time teaching you those guidelines and give you more time to really see patients. That's where you learn to practice medicine. Well, she's here now. Let's go see her."

The team walked to Janna's room, where they found an infant whose temperature was now 37.3° C, sleeping pleasantly. Her parents appeared tired, but relieved.

The attending physician glanced at Mr. Alstadt and smiled.

Commentary 3

Not being a pediatrician, I may be demonstrating a certain hubris discussing clinical reasoning around a febrile infant where the possibility of meningitis has been raised. I remember certain "rules of thumb" from my training: if you simply thought of doing a lumbar puncture (LP) as part of a fever work-up, you must proceed. The implicit message was that meningitis was so dangerous that any possibility, no matter how small, warranted the procedure. Our only choice when we felt an LP was clearly not indicated was to not think of the possibility, or at least not talk about it. Clearly there must be a better way.

The development of algorithms or guidelines has been part of an effort to use best available evidence to counteract some of the irrationalities and irregularities that result when one relies totally on the clinical judgment and inherent biases of individual physicians. These more sophisticated "rules of thumb" usually rely on characteristics of population-based studies as well as on expert opinion to guide physicians to make rational choices more on the basis of available clinical evidence than on individual clinical experience. In many instances, such guidelines have improved the quality of care in areas where the evidence is clear (for example, the use of ACE inhibitors in heart failure [1] and not using antibiotics for uncomplicated sinusitis [2]) and where practice variation has been wide and irrational.

Algorithms and guidelines are well suited to common dilemmas such as the febrile infant [3], where there may be a low probability of a dreaded disease like bacterial meningitis, where there may be clinical markers that increase or decrease the probabilities, and where the consequences of delay in diagnosis and treatment can be lethal. Thus, when one thinks of meningitis in the differential diagnosis of a febrile infant, one must think in terms of thresholds (probability of meningitis). Since the consequences of missing the diagnosis of bacterial meningitis are so profound, the threshold for treatment must be relatively low. But the threshold cannot be so low that we return to the level of "if you think of meningitis, you must do a lumbar puncture," or so high that we put some infants unnecessarily at risk for the consequences of untreated infection because we're reluctant to intervene. If we say the probability of meningitis in this infant was about 5 percent, then we would do 19 unnecessary LPs for every case of meningitis. Importantly, although the case may appeal to our sentiments with its happy ending, this misses the point. The fact that this baby looks fine the next day does not tell us anything about the appropriateness of the LP and antibiotics in the first place.

Guidelines have been very helpful in standardizing best practices, but they cannot be applied rigidly and unthinkingly. They define standards of care in highly restricted circumstances and force clinicians to justify why they recommend variation, if they choose to do so. Of course, one can also "unbundle" a guideline (ie, by following 1 part of the guideline but choosing not to follow another) in this case by performing an LP without empirically starting antibiotics if the pretest probability is low and the initial cell counts appear negative. Conversely, if the child looked sicker or had an exposure to a known case of meningitis, such that the pretest probability approached 50 percent, empirically starting antibiotics while doing the LP might have been the best course. Thus, the existence of a guideline or algorithm does not preclude the need to exercise clinical judgment. Part of that judgment requires that probabilities of alternate outcomes be estimated based on existing studies in light of clinical experience, and then the reasons for particular actions are clearly explained.

Communicate Evidence to Parents

The area where the physicians in this vignette were most inadequate was not in the decision to do an LP and start empirical antibiotics but in their inability or unwillingness to make every effort to make their thinking transparent to the parents first, and to the medical student second. With all the work that has been done promoting evidence-based decision making, there has been a minimal amount of empirical work discovering how best to communicate about this evidence; this is necessary for making the best possible decisions with patients and families [4]. In this low-probability situation, the parents might have been told about the odds of bacterial meningitis using either lay terms ("small chance") or numbers ("5 percent chance"), and then told that the physicians' recommendation is that empirical treatment be started even though the odds are much higher (95 percent) that there is no such infection.

Most parents would accept such recommendations, but some might want to have further input in this situation based on their values and personal experiences. For example, if one of the parents had had a serious allergic reaction to antibiotics, or held a belief system that precluded the use of Western medicine, he or she might ask some hard questions about risks and benefits of empirical treatment. Physician and family would agree that they all had the infant's best interests in mind, and they would then seek common ground around the best possible treatment rather than entering into a power struggle [5]. Physicians should not shy away from making recommendations when the evidence about the best approach is clear from a medical point of view [6], but they must also learn to explain their recommendations in easily understandable terms. They must listen carefully to, and learn from, patients and families to be sure both parties understand one another. When a child is involved, additional ethical obligations of representing a patient with no actual voice in the discussion are added to the mix. The communication skills needed to conduct this part of an interview are not well taught in medical school, and most clinicians have not been observed or evaluated on their skill level. Therefore, practice variation is probably quite large, further adding to the gap between best evidence and best practice.

In a similar vein, the medical student should not get the message that guidelines in general are useless, just that they need to be interpreted in light of clinical judgment. The attending might ask the students about their view of the odds that the patient had meningitis the evening before when they were engaged in decision making and then discuss the probabilistic thresholds for doing an LP and for starting antibiotics. More sophisticated trainees might then be asked how they would present their recommen-dations to the parents and how they would engage them as partners. Blindly following published guidelines should be discouraged, but using guidelines as a starting point for clinical decision making is clearly on the right track. "Really seeing patients," as suggested by the attending physician, should be encouraged, but the clinical thinking needs to be quantified, and the potential consequences of action and inaction explored in light of available evidence. Then, making one's thinking transparent with patients, families, colleagues and trainees, and engaging them fully in the process, finally closes the loop of delivering the best possible treatment.


  1. Jong P, Demers C, McKelvie RS, et al. Angiotensin receptor blockers in heart failure: a meta-analysis of randomized controlled trials. J Am Coll Cardiol. 2002;39(3):463-470.
  2. Williams JW Jr, Aguilar C, Cornell J, et al. Antibiotics for acute maxillary sinusitis. Cochrane Database Syst Rev. 2003(2);CD000243.

  3. Harper MB. Update on the management of the febrile infant. Clinical Pediatric Emergency Medicine. 2004;5(1):5-12.
  4. Epstein RM, Alper BS, Quill TE. Communicating evidence for participatory decision making. JAMA. 2004;291(19):2359-2366.
  5. Fisher R, Ury W, Patton B. Getting to Yes: Negotiating Agreement Without Giving In. Boston: Houghton-Mifflin; 1981.

  6. Quill TE, Brody H. Physician recommendations and patient autonomy: finding a balance between physician power and patient choice. Ann Intern Med. 1996;125(9):763-769.


Virtual Mentor. 2004;6(12):533-535.



The people and events in this case are fictional. Resemblance to real events or to names of people, living or dead, is entirely coincidental. The viewpoints expressed on this site are those of the authors and do not necessarily reflect the views and policies of the AMA.