An Evidence-Based Clinical Prediction Algorithm for the Musculoskeletal Infection Society Minor Criteria.

BACKGROUND

The diagnosis of a periprosthetic joint infection (PJI) remains a clinical challenge, as there is no uniformly accepted gold standard. In 2011, the Musculoskeletal Infection Society (MSIS) convened a work group to create a standardized definition for a PJI that could be universally adopted. Based on the MSIS criteria, the diagnosis of a PJI can be made with 1 of the 2 major criteria, or 3 of the 5 minor criteria. The purpose of this study was to determine the likelihood of having a PJI based on the number of positive minor criteria and thereby develop a prediction algorithm for differentiating between a chronic PJI and a non-PJI based on the number of positive MSIS minor criteria.

METHODS

We retrospectively reviewed 297 patients who presented to a tertiary care center between 2004 and 2014 with a failed total joint arthroplasty and subsequently underwent a PJI workup to exclude chronic PJI. Patients were divided into 2 groups: (1) PJI group and (2) non-PJI group. Patients who had a positive PJI workup and subsequently underwent a 2-stage revision for infection were included in the PJI group. Patients who had a negative clinical and diagnostic workup were included in the non-PJI group. One hundred eighty-two patients met the criteria for inclusion in the study, 91 in each group. Univariate and multiple logistic regression analyses were used to evaluate 21 independent variables in each of the 2 groups. A prediction algorithm for differentiating between a chronic PJI and a non-PJI based on independent multivariate variables was created.

RESULTS

Patients who had a PJI differed significantly (P < .05) from those who did not have a PJI with regard to 10 independent variables, which included all the MSIS minor criteria we evaluated. Five independent multivariate variables were identified to differentiate between the 2 groups: positive cultures, elevated synovial white blood cell count, elevated synovial polymorphonuclear neutrophil percentage, elevated erythrocyte sedimentation rate, and elevated C-reactive protein. The predictive probability of a PJI for all 32 combinations of these 5 variables was: 3.6% for 1 positive variable, 19.3% for 2, 58.7% for 3, 83.8% for 4, and 97.8% for 5. The chi-squared test for trend and the area under the receiver-operating characteristic curve (0.977) suggest that the model is highly predictive, with an excellent diagnostic performance in identifying a PJI.

CONCLUSIONS

Diagnosing a PJI remains a clinical challenge as there is no gold standard for diagnosis. The development of the MSIS criteria, which is based on a consensus of over 400 of the world's experts in musculoskeletal infection, was a major step forward in defining the diagnosis of a PJI. However, to our knowledge, the likelihood of having a PJI based on the number of positive minor criteria has yet to be validated or quantified. Of the 20 independent variables that were evaluated, 10 were found to be significantly associated with a PJI, including all the MSIS minor criteria evaluated. In addition, a diagnostic prediction algorithm was constructed to determine the likelihood of a PJI based on 5 binary independent multivariate variables. The relationship was also examined with a receiver-operating characteristic curve analysis. The area under the curve was 0.98, indicating excellent diagnostic performance for the MSIS minor criteria in identifying a PJI.

LEVEL OF EVIDENCE

III.