Septic Arthritis and Osteomyelitis in Children.ppt

Editor's Notes

• #1: Welcome to the Education Guides portion of The Rational Clinical Examination series. Teaching tips and notes for making sessions relevant and interactive will be included in the notes pages of selected slides. Throughout the Education Guides, the following ABC’S may be highlighted in the notes pages: A: Slide contains animation geared to increase interactivity. B: Slide contains basic principles related to teaching diagnosis. These slides are part of a uniform set that will be used throughout the Education Guides. C: Slide contains an opportunity to increase relevance and interactivity through use of cases or by asking the learners to commit to a specific answer that can be used for discussion and for anchoring their responses. One easy, inexpensive, and fun strategy is to hand out blank file cards at the start of the session and have learners use them to write down their numeric guesses of probabilities. These cards can be passed around the room in order to “blind” the process. Once the learners have written down their estimates, they can pass the cards around the room so that no one is holding their own card and can report on what is in front of them without identifying their own answers. This contributes to a safe learning environment in which learners are not afraid to make honest guesses. S: Slide contains a stumbling block.
• #2: C: You can enhance relevance and interactivity by always anchoring your session with 1 or more cases selected to illustrate key points (ie, select patient characteristics that are relevant to the question at hand). The Education Guides slides will always start with cases from The Rational Clinical Examination Articles and/or Updates. However, educators could choose to use cases from their own or their learners’ experiences.
• #3: ED indicates emergency department. Learners can be asked to identify what they think the relevant portions of this case are and whether these increase or decrease the likelihood of septic arthritis. Historical features they may consider include Monoarticular presentation Underlying rheumatoid arthritis Long-term steroid use Time course of presentation
• #4: ESR indicates erythrocyte sedimentation rate; WBC, white blood cell. What about physical examination findings and laboratory findings? Absence of fever Left knee effusion Leukocytosis Elevated ESR Before going on to the next slide, ask your learners whether they would do an arthrocentesis. Ask them to justify their choices and to specifically identify what principles and findings they are using. This discussion can serve as a starting point for teaching the concept of action threshold. When do we decide to take an action (eg, an arthrocentesis)? How do we balance the potential benefits with the potential risks?
• #5: WBC indicates white blood cell. In this scenario, the tap is performed, and we can consider the findings.
• #6: C: Ask learners to commit to a “guesstimate” of probability on a blank file card. Once learners have responded, you can consider the use of a stem and leaf plot to quickly record the distribution of probabilities guessed to come up with a “group” guesstimate. Hold on to this estimate as you can use it later. Try to get participants to identify both serious and more benign causes of acute arthritis, and use this to illustrate why making a diagnosis is important.
• #7: Ask participants to identify what they think the relevant portions of this case are, and whether these increase or decrease the likelihood of septic arthritis. Features to discuss here might include Monoarticular presentation Prior episodes diagnosed as gout
• #8: C indicates centigrade.
• #9: C: Ask learners to commit to a “guesstimate” of probability on a blank file card. Once learners have responded, you can consider the use of a stem and leaf plot to quickly record the distribution of probabilities guessed to come up with a “group” guesstimate. Hold on to this estimate as you can use it later. The relationship between gout and septic arthritis could serve as a good opportunity to discuss both prior probability and action thresholds. Although gout and septic arthritis can occur together, they most often do not. Thus, the presence of a common disorder such as gout actually should change your prior probability of a less common disorder such as septic arthritis. In addition, your learners will need to assess their action threshold either for arthrocentesis or for starting antibiotics in the context of a possible gouty flare, instead of septic arthritis.
• #12: A full listing of diseases that make up the differential diagnosis for acute monoarticular arthritis is presented in Box 65-1 of the Septic Arthritis chapter at www.jamaevidence.com; however, this slide represents the most common entities.
• #14: Many conditions including crystal arthritis and systemic diseases, such as rheumatoid arthritis, can present with fever, joint swelling, pain, and stiffness, mimicking the clinical presentation of septic arthritis.
• #16: S: Slide contains a frequent stumbling block. Prior probability: Many learners will stumble when trying to come up with a prior probability. This is the first opportunity to go back to the learners' original file card “guesstimates” to discuss how they came up with those numbers. For history and physical examination (H&P) in The Rational Clinical Examination series, the pre-“test” probability is the pre-“H&P” probability or the prevalence. This will be true for all of The Rational Clinical Examination Education Guides. Frequently, learners will try to include elements of the H&P in their assessment, rather than using prevalence. If they do this, they will ultimately overestimate the effect of the clinical examination. The range of prior probabilities for septic arthritis comes from 2 prospective studies of septic arthritis. 1. Shmerling RH, Delbanco TL, Tosteson AN, Trentham DE. Synovial fluid tests: what should be ordered? JAMA. 1990;264(80):1009-1014. [PMID: 2198352] 2. Jeng GW, Wang CR, Liu ST, et al. Measurement of synovial tumor necrosis factor-alpha in diagnosing emergency patients with bacterial arthritis. Am J Emerg Med. 1997;15(7):626-629. [PMID: 9375540]
• #17: Synovial fluid culture is the generally accepted reference standard, but it is not perfect (sensitivity 75%-95%).1 Thus, a positive gram stain, positive blood cultures in a patient with acutely swollen and painful joint, aspiration of pus from the joint space, or response to antibiotics are sometimes used in clinical practice as the reference standard for septic arthritis 1. Shmerling RH. Synovial fluid analysis. A critical reappraisal. Rheum Dis Clin North Am. 1994;20(2):503-512. [PMID: 8016423]
• #18: B: Slide contains basic principles related to teaching diagnosis. These slides are part of a uniform set that should be used throughout the Education Guides to review and clarify these core principles.   The commonly used mnemonics SnNout for sensitivity and SpPin for specificity are helpful in guiding learners to remember the direction of sensitivity and specificity with respect to a 2 × 2 table and in defining how a test with high sensitivity or specificity operates. It is important to remember that the test characteristic itself does not “rule in” or “rule out” anything. Rather, a test result modifies a pretest probability (also known as prior probability) to change the likelihood of disease in a particular individual. Sensitivity SnN(–)OUT: High-sensitivity negative tends to “rule out” disease. Specificity SpP(+)IN: High-specificity positive tends to “rule in” disease.
• #19: B: Slide contains basic principles related to teaching diagnosis: Likelihood Ratio. S: When teaching the principles of LRs, a frequent stumbling block relates to the fact that an LR is calculated for each test result. For categorical tests in which the results fall into a number of non-numeric categories (eg, yes/no or present/absent), an LR will be calculated for each test result. For continuous test results (eg, measurement of jugular venous pulse [JVP] in cm), cutoffs are defined and a different LR is calculated for each cutpoint defined (eg, JVP > 3 cm above the sternal notch).
• #20: B: Slide contains basic principles related to teaching diagnosis: Likelihood Ratio. Learners will frequently want to have an overall sense of what defines a “good” LR vs a “bad” LR. However, values of LRs are more accurately part of a spectrum in which an LR of 1 has no impact on pretest probability and both high numbers for LR and small fractions for LR have a greater impact on the pretest probability.
• #21: B: Slide contains basic principles related to teaching diagnosis: Likelihood Ratio. A: Slide uses animation to graphically illustrate the concept that LR has greater impact on pretest probability as it moves away from the central line of LR = 1, where the test result will not differentiate disease from no disease. Note that animation in this slide is fully automated (ie, you do not need to do anything) and is used to provide a different way of describing how to interpret LRs that may appeal to visual learners.
• #22: B: Slide contains basic principles related to teaching diagnosis: Likelihood Ratio. This follow-up slide allows you to clarify that the LRs on each side of 1 have equal differentiating power (eg, LR = 10 is just as powerful as LR = 0.1). LR = 10 increases the likelihood of disease for a given test result, whereas LR = 0.1 decreases likelihood of disease with the same power for a given test result. S: When teaching LR, a frequent stumbling block is comparing LRs and creating rank order. From this slide and the prior slide it can be seen that, in order to directly compare LRs in terms of their diagnostic power, a clinician would have to compare LRs that are >1 to the inverse of LRs that are <1. However, LRs >1 can continue to infinity, while LRs <1 have a much narrow range of 0-0.99. Thus, these scales are not directly comparable. Fortunately, it is easy to compare because all you need to do is take the value 1/LR for values <1 and compare it to the LR for values >1. The value that is greater (1/LR vs LR) has greater diagnostic power in that it will move the posttest probability further from the pretest probability. For example, LR = 0.01 is far more powerful than LR = 5, because 1/0.01 = 100, which is much greater than the LR of 5.
• #23: These risk factors were identified from a prospective cohort study (Kaandorp CJ, Van Schaardenburg D, Krijnen P, Habbema JD, van de Laar MA. Risk factors for septic arthritis in patients with joint disease: a prospective study. Arthritis Rheum. 1995;38:1.819-1825. [PMID: 8849354]). The LRs for the absence of these findings can be seen in Table 65-4 in the chapter on Septic Arthritis at www.jamaevidence.com. However, the LRs are not very useful (LRs are close to 1), thus, we have not listed them here. You can explain to your learners that they should assess both the magnitude of the summary LR and also the range of the confidence interval. Findings that have wide confidence intervals that include 1 should be viewed with caution.
• #24: Ask your learners how to interpret this slide. Why would a study present only sensitivities? Consideration of the 2x2 table would remind your learners that sensitivity applies only to patients who have disease, in this case, patients who have been diagnosed with septic arthritis. Thus, 85% of patients with septic arthritis have joint pain; 57% have fever. Frequently we have studies in which the authors are studying only people who have been diagnosed with disease (eg, septic arthritis) without a group of patients who do not have disease. In such a case you can only calculate sensitivity. Thus, these numbers do not tell us anything about patients without septic arthritis.
• #25: CRP indicates C-reactive protein; WBC, white blood cell. One single study assessed the LR of fever (Kortekangas P, Aro HT, Tuominen J, Toivanen A. Synovial fluid leukocytosis in bacterial arthritis vs. reactive arthritis and rheumatoid arthritis in the adult knee. Scand J Rheumatol. 1992;21(6):283-288. [PMID: 1475638]). In this study, perhaps counter to intuition, the presence of fever actually decreased the likelihood of septic arthritis while the absence of fever increased the likelihood and both of these confidence intervals included 1. No studies evaluated other findings such as tenderness to palpation, edema or range of motion of the affected joint. Similarly the serum laboratory tests were not very powerful diagnostic tests, with likelihood ratios that are close to 1. At this point you can go back and reflect on the learners’ initial responses to the cases. Learners may have thought that the presence of fever, elevated peripheral WBC, or elevated ESR were associated with an increased likelihood of septic arthritis. You can reinforce the importance of knowing which clinical findings are not particularly helpful, in addition to knowing which findings are. Although the LRs for peripheral WBC count <10 000/μL (0.28, 0.07-1.10) and ESR < 30 ( 0.17, 0.20-1.30) are small (far from 1), they both have confidence intervals that are fairly wide that include 1.
• #26: PMNs indicates polymorphonuclear cells. Two trends should be noted: (1) The higher the synovial WBC count (above 25 000/μL), the more useful this test result is in increasing the probability of septic arthritis, and (2) a synovial fluid WBC count below 25 000 or a PMN count below 90% are modestly helpful in decreasing the probability of septic arthritis. No LR confirms or excludes the diagnosis as the final probability will depend on where you started, that is, the pretest probability.
• #27: Handouts are both necessary and helpful for hands-on learning. For this exercise the following handouts may be considered: The nomogram (next slide) Slide(s) with prior probability information Slide(s) with LRs
• #28: B: Slide contains basic principles related to teaching diagnosis. The nomogram allows a clinician to move directly from pretest probability (on the left side of the nomogram), through the LR, to the posttest probability on the right, without having to convert from probability to odds and back again. Many educators use the nomogram to visually illustrate the impact of an LR on a pretest probability. For naïve learners who may never have seen the nomogram before, be sure to orient them to the 3 vertical axes of the nomogram. Note that most clinicians do not carry around the nomogram. Some use LR calculators that can be downloaded onto a handheld personal digital computer. Many simply have a sense from experience with the nomogram as to how much a certain LR will affect pretest probability. See the online LR nomogram calculator available in JAMAevidence at http://www.jamaevidence.com/calculators/9000026.
• #32: C: We can now return to these estimates to discuss in light of the data that has been presented.
• #33: A: Slide contains animation geared to increase interactivity.
• #34: The pretest probability of 18% was selected as the midpoint of the range of prevalence estimates (8%-27%). Ask your learners if this is reasonable. What would happen to the posttest probability if we used 8%? What if we used 27%? The answer is illustrated in the next slide.
• #35: This slide visually illustrates the range of prior probabilities. The posttest probability would range from 18% to 48%. Are they comfortable with that range?
• #36: This slide brings up the topic of sequential testing. Ask learners why the starting probability is 36%, rather than 18% as in the last slide. Here we are making the assumption that the history of RA has increased the probability of septic arthritis from 18% to 36%. Because the synovial fluid WBC count is a test that is independent from the history of RA, we can perform sequential testing and assume that the LR of WBC > 25 000/μL can be applied to the 36% probability of septic arthritis in a patient with RA. Sequential testing is done to improve our probability estimates as long as the tests we are using in sequence are independent. If they are not independent (eg, 2 physical findings) it would not be valid to perform sequential testing. Learners who wish to delve deeper into this topic can be referred to Education Guide for the chapter, A Primer on the Precision and Accuracy of the Clinical Examination. Some learners might not agree that these tests are independent. It could be argued that patients with RA might have an elevated synovial fluid WBC on the basis of their RA. If this comes up, it is a great opportunity to discuss the role of uncertainty in clinical decision making.
• #38: You can ask your learners about their action threshold, the probability at which they would start antibiotics. For many clinicians a posttest probability of 62% is already high enough to warrant action. If that is the case, then the additional information provided by the synovial fluid differential WBC may not alter your actions. Also, this is an opportunity to return to the question of sequential testing. If you did get a differential WBC, what would you use for the pretest probability, 18%, 36%, or 62%? In this case you could use the 36% but could not justifiably use 62% because the synovial WBC count and the synovial WBC differential are not independent of each other. Using a pretest probability of 36% and an LR+ 3.4, you would get a posttest probability of 66%. Arguably, a low synovial fluid PMN count (<90%) might raise your awareness of different diagnoses (eg, cancer or tuberculosis instead of bacterial infection).
• #41: We can now return to the original estimates provided by the learners and apply the LRs and probabilities that have been presented.
• #42: A: Slide contains animation geared to increase interactivity.
• #43: Once again, the pretest probability of 18% is the midpoint between 8% and 27%.
• #44: Ask your learners whether they think it is valid to use these 2 likelihood ratios sequentially. In this case, it is not valid because the risk factors of advancing age and the presence of diabetes mellitus are not likely to be independent, so you would be overestimating impact if you used them sequentially. Thus, for both this slide and the prior slide, the pretest probability used is 18%.
• #45: The posttest probability of septic arthritis is estimated to be approximately 40% (43% posttest probability using age as a risk factor and 37% using diabetes mellitus as a risk factor).
• #46: The practice of medicine includes elements of decision making in the context of clinical experience. In this case, most experienced providers would factor in the history of gout as an alternative diagnosis and would lower their estimate that the presentation was due to septic arthritis. Even though we do not have a LR to numerically define how much we should decrease our estimate, we use our judgment. For septic arthritis we were presented with a range of prior probabilities from 8% to 27%. Thus, it seems reasonable to assign this patient to a pretest probability at the low end of this range, 8%. If we use 8%, the posttest probability is ~20%, which most clinicians would consider too high to risk the severe consequences of untreated infection. Arthrocentesis should be performed to help confirm the diagnosis of gout and exclude infection.
• #47: For visual learners, this is the nomogram that illustrates the probability estimates described in the prior slide.
• #49: PMNs indicates polymorphonuclear cells; WBC, white blood cell.
• #50: This Education Guide slide set has been created as a part of The Rational Clinical Examination series. The content for each Education Guide slide set comes directly and exclusively from relevant chapters in The Rational Clinical Examination book. You can link to this chapter if you are on a networked computer with access to the JAMAevidence Web site through the hyperlinks contained in the references to the chapter on the next slide.