Commentary & Perspective

Commentary & Perspective on
"Differentiation Between Septic Arthritis and Transient Synovitis of the Hip in Children with Clinical Prediction Algorithms"
by Scott J. Luhmann, et al.

Commentary & Perspective by
Mininder S. Kocher, MD, MPH*,
Department of Orthopaedic Surgery, Children's Hospital, Boston, Massachusetts

Clinical prediction rules attempt to make the art of diagnosis more objective by estimating the probability of a diagnostic outcome and allowing the clinician to classify patients according to the risk of disease¹. The task of differentiating septic arthritis and transient synovitis of the hip in children is important but often vexing. We previously published a clinical prediction rule for the differentiation of septic arthritis and transient synovitis of the hip based on four independent, multivariate clinical predictors of septic arthritis²: history of fever, non-weight-bearing, an erythrocyte sedimentation rate greater than or equal to 40 mm/hr, and a serum white blood-cell count >12,000/mm³ (12.0 × 10⁹/L). This prediction rule was derived from data obtained between 1979 and 1996 at Children's Hospital in Boston, from a study of eighty-two patients with septic arthritis and eighty-six patients with transient synovitis. The prediction rule demonstrated excellent diagnostic performance² in the original patient population with an area under the receiver-operating characteristic curve of 0.96.

Clinical prediction rules typically demonstrate diminished performance in a new patient population because they are optimally modeled to the original data set¹. Therefore, the validation of a clinical prediction rule is essential. A clinical prediction rule can be validated within the original dataset by selective sampling, known as bootstrapping. Ideally, however, a clinical prediction rule should be validated by examining its performance prospectively in a new patient population.

In their study, Luhmann et al. evaluated the performance of our clinical prediction rule retrospectively in 165 hips (163 children) at St. Louis Children's Hospital. They found diminished diagnostic ability of the prediction rule in their patient population. The best prediction model for their patient population included two of our previous variables, history of fever and serum white blood-cell count >12,000/mm³ (12.0 × 10⁹/L), and a third, new variable, previous health-care visit.

Others have also examined the utility of clinical and laboratory variables in the differentiation of septic arthritis and transient synovitis of the hip in children. Jung et al. performed a study that was similar to our derivation study and to Luhmann et al.'s validation study in ninety-seven children with transient synovitis of the hip and twenty-seven children with septic arthritis of the hip³. They identified differences between the two groups with use of univariate analysis and identified five independent multivariate predictors of septic arthritis: temperature >37° C, erythrocyte sedimentation rate >20 mm/hr, C-reactive protein level >1.0 µg/dL (10 µg/L), serum white blood-cell count >11,000 per high-power field, and a radiographic difference in joint space of >2 mm. They developed a different algorithm based on the thirty-two combinations of the five predictors. The area under the receiver-operating characteristic curve for their prediction rule was 0.986. Other investigators have emphasized the utility of C-reactive protein⁴, erythrocyte sedimentation rate⁵, and oral temperature⁶.

Differential performance of the prediction rule in these two studies is most likely due to expected diminished performance in a new patient population. Because clinical prediction rules are optimized to their original data set, they typically do demonstrate diminished diagnostic performance in a new patient population. The best summary measure of performance of a clinical prediction rule is the area under the receiver-operating characteristic curve, in which 1.00 is perfect prediction and 0.50 is random guessing. In applying our prediction rule prospectively in a new patient population at Children's Hospital in Boston, we also found diminished but nevertheless good overall diagnostic performance with an area under the receiver-operating characteristic curve of 0.86 (compared with the area under the receiver-operating characteristic curve of 0.96 from the original study)⁷. Even in the study by Luhmann et al., our prediction rule demonstrated good diagnostic performance, with an area under the receiver-operating characteristic curve of 0.80.

In summary, it is important to remember that a clinical prediction rule is neither a rigid guideline nor a replacement for clinical judgment. The performance of clinical prediction rules may vary in new clinical settings, and different variables may be more predictive in different clinical settings. The goal of our clinical prediction rule was to aid in the often vexing differentiation of septic arthritis and transient synovitis of the hip in children by stratifying patients according to the risk of septic arthritis. Clinical judgment is still essential in the management of these patients. Aspiration of the hip is still mandatory to establish the diagnosis of septic arthritis of the hip.

*The author did not receive grants or outside funding in support of his research or preparation of this manuscript. He did not receive payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity. No commercial entity paid or directed, or agreed to pay or direct, any benefits to any research fund, foundation, educational institution, or other charitable or nonprofit organization with which the author is affiliated or associated.

References

1. Wasson JH, Sox HC, Neff RK, Goldman L. Clinical prediction rules: applications and methodological standards. N Engl J Med. 1985;313:793-9.
2. Kocher MS, Zurakowski D, Kasser JR. Differentiating between septic arthritis and transient synovitis of the hip in children: An evidence-based clinical prediction algorithm. J Bone Joint Surg Am. 1999;81:1662-70.
3. Jung ST, Rowe SM, Moon ES, Song EK, Yoon TR, Seo HY. Significance of laboratory and radiologic findings for differentiating between septic arthritis and transient synovitis of the hip. J Pediatr Orthop. 2003;23:368-72.
4. Levine MJ, McGuire KJ, McGowan KL, Flynn JM. Assessment of the test characteristics of C-reactive protein for septic arthritis in children. J Pediatr Orthop. 2003;23:373-7.
5. Del Beccaro MA, Champoux AN, Bockers T, Mendelman PM. Septic arthritis versus transient synovitis of the hip: The value of screening laboratory tests. Ann Emerg Med. 1992;21:1418-22.
6. Kunnamo I, Kallio P, Pelkonen P, Hovi T. Clinical signs and laboratory tests in the differential diagnosis of arthritis in children. Am J Dis Child. 1987;141:34-40.
7. Kocher MS, Mandiga R, Zurakowski D, Barnewolt C, Kasser JR. Validation of a clinical prediction rule for the differentiation of septic arthritis and transient synovitis of the hip in children. J Bone Joint Surg Am. In press.

Copyright © 2004 by the The Journal of Bone and Joint Surgery, Inc.