Luhmann et al developed a simple 3-item clinical prediction rule for identifying children likely to have septic arthritis of the hip. This can help separate these patients from children with transient synovitis. The authors are from St. Louis Children's Hospital and Washington University Medical Center.
Clinical features shared by the 2 syndromes (page 956):
(1) spontaneous onset of progressive hip, groin or thigh pain
(2) limp or inability to bear weight on the ipsilateral leg
(3) fever
(4) irritability
Parameters for algorithm:
(1) history of fever >=38.5°C
(2) peripheral white blood cell (WBC) count
(3) recent history of a previous health-care visit
|
Parameter
|
Finding
|
Points
|
|
history of fever
|
no
|
0
|
|
|
yes
|
1
|
|
peripheral WBC count
|
<= 12,000 per µL
|
0
|
|
|
> 12,000 per µL
|
1
|
|
recent history of previous health care visit
|
no
|
0
|
|
|
yes
|
1
|
where:
• "Recent history of health-care visit" is somewhat nebulous to me as a risk factor. It might make sense as an indicator of a more chronic condition (viz, it lasted more than a few days and so is not "transient").
number of risk factors =
= SUM(points for all 3 parameters)
Interpretation:
• minimum number of risk factors: 0
• maximum number of risk factors: 3
Performance:
• If all 3 parameters are present, then the probability of septic arthritis is 71%.
• The probability with fewer parameters was not listed in the print of the article.
Limitations:
• The authors claim that the prediction rule is better than that of Kocher et al, with 71% probability vs 59%. However, (1) Kocher's area under the ROC is slightly higher (80% vs 77%) and (2) Kocher et al give 2 categories for septic arthritis (true/definite, presumed), yet it appears as if only the true is referred to.
• The likelihood ratios for the terms are low (1.0, 1.5 and 1.2).
• A prediction probability of 71% would be of marginal use to me.
