How Well the Kocher Criteria Predict a Septic Hip
Pus trapped in the hip joint capsule can destroy a child's femoral head and growth plate when drainage is delayed.
Key terminology
- Hip joint capsule
- This fibrous sleeve encloses the femoral head and most of the femoral neck. Infection inside it fills the joint with pus under pressure.
- Transient synovitis
- This self-limited inflammation of the hip lining is the most common cause of acute limp in young children. It also makes an effusion, so it mimics a septic hip.
- Kocher criteria
- This score counts 4 findings in a child with an irritable hip. More positive findings mean a higher probability of septic arthritis.
- Prediction rule
- This tool combines a few findings into a probability of disease. It is built on one group of patients and should be retested on new ones.
- Area under the curve (AUC)
- This summary of a test ranges from 0.5 (a coin flip) to 1.0 (perfect). A drop from 0.96 to 0.86 means the rule sorted new patients less well.
- External validation
- This testing step applies a rule to patients from a different hospital than the one that built it. It shows whether the rule works beyond its home center.
- Kingella kingae
- This gram negative bacterium is a common cause of bone and joint infection in children <4 y/o. It often produces mild symptoms and near normal labs.
Takeaways
- The Kocher criteria are 4 predictors of a septic hip, built from a retrospective review at Boston Children's from 1979 to 1996. They are a history of fever, non-weight-bearing, ESR ≥40 mm/hr and serum WBC >12,000 cells/mm³. Later studies define fever as an oral temperature >38.5°C. (Kocher et al, J Bone Joint Surg Am 1999;81:1662-1670, PMID 10608376).
- Predicted probability of septic arthritis climbs steeply with each predictor. It was <0.2% with 0 predictors, 3% with 1, 40% with 2, 93.1% with 3 and 99.6% with all 4. These numbers came from the hospital that built the rule. (Kocher et al, J Bone Joint Surg Am 1999;81:1662-1670, PMID 10608376).
- The rule lost accuracy in the same hospital's next patients, with the area under the curve (AUC) falling from 0.96 to 0.86. This prospective cohort of 51 septic hips and 103 transient synovitis cases found the same 4 predictors. The authors called the result diminished but very good, since a prediction rule typically performs worse on new data. (Kocher et al, J Bone Joint Surg Am 2004;86:1629-1635, PMID 15292409).
- At St Louis Children's, all 4 predictors gave a 59% probability of septic arthritis, against 99.6% in Boston. This external validation found the best local model used fever, WBC >12,000 and a previous health-care visit, and it reached only 71%. The authors concluded neither model appeared valid at other hospitals. (Luhmann et al, J Bone Joint Surg Am 2004;86:956-962, PMID 15118038).
- Caird added CRP >2.0 mg/dL (>20 mg/L) as a 5th predictor in a prospective cohort of 48 children who had hip aspiration in Philadelphia. Fever (oral temperature >38.5°C) was the best single predictor, and CRP was the strongest independent risk factor. The probability of septic arthritis was 98% with 5 predictors, 93% with 4 and 83% with 3. (Caird et al, J Bone Joint Surg Am 2006;88:1251-1257, PMID 16757758).
- In a UK general hospital, all 5 predictors gave only a 59.9% probability of septic arthritis. Fever was again the best predictor. Septic arthritis is less common outside tertiary referral centers, and positive predictive value falls with prevalence, so the same findings produce more false positives. (Sultan et al, J Bone Joint Surg Br 2010;92:1289-1293, PMID 20798450).
- Kingella kingae septic hips in children <4 y/o often score low on the rule. In 140 confirmed cases, temperature and ESR passed the cutoffs in <50%, and CRP was the most sensitive marker at 76.2%. The authors concluded the Kocher and Caird criteria cannot confidently exclude a Kingella septic hip in preschool children. (De Marco et al, Microorganisms 2025;13:2323, PMID 41156783).
- A 2026 prospective cohort found the probability plateaued at 66.7% with 3 or 4 criteria. Septic arthritis and osteomyelitis overlapped heavily, and each made up 39.4% of children with 2 criteria. The authors suggested ≥2 criteria should prompt MRI rather than a trip straight to the OR. In Lyme-endemic areas, 56% of Lyme hips also had a synovial WBC ≥50,000. (Rogie et al, J Pediatr Orthop 2026, PMID 42802642; Dart et al, Pediatrics 2018;141:e20173810, PMID 29669751).