Background: Total hip arthroplasty is a commonly performed procedure in the United States and Canada that is associated with a definite risk of postoperative infection. Moreover, diagnosing an infection after total hip arthroplasty can present a challenge as there are no preoperative tests that are consistently sensitive and specific for infection in patients who need a revision arthroplasty. The present prospective study was performed to evaluate a variety of investigations for the diagnosis of infection at the site of a previous arthroplasty in order to determine if any combination of diagnostic studies could be used to determine which patients are at risk for a postoperative wound infection.Methods: We prospectively analyzed the preoperative and intraoperative investigations used for the diagnosis of infection in 178 patients who had a total of 202 revision hip replacements. Clinical data were collected preoperatively. Investigations to determine the presence or absence of infection included a white blood-cell count, measurement of the erythrocyte sedimentation rate, measurement of the level of C-reactive protein, preoperative aspiration of the joint, intraoperative gram-staining and culture of periprosthetic tissue, a white blood-cell count in synovial fluid, and examination of intraoperative frozen sections. Frozen sections were analyzed in a blinded fashion without knowledge of clinical or laboratory data. Patients receiving antibiotics at the time of aspiration or collection of specimens for intraoperative culture were excluded from the analysis of those investigations, regardless of the results of the cultures. A positive result (suggestive of infection) was clearly defined for each of the investigations.Results: Thirty-five hips (17 percent) were determined to be infected on the basis of clinical findings and positive results, according to the defined criteria, of investigations. With inflammatory conditions excluded, the sensitivity, specificity, positive predictive value, and negative predictive value were 0.82, 0.85, 0.58, and 0.95, respectively, for the erythrocyte sedimentation rate and 0.96, 0.92, 0.74, and 0.99, respectively, for the level of C-reactive protein. All patients who had a periprosthetic infection had an elevated erythrocyte sedimentation rate or level of C-reactive protein, but not always both. When patients who were receiving antibiotics were excluded, the results of aspiration of the joint were 0.86 for sensitivity, 0.94 for specificity, 0.67 for the positive predictive value, and 0.98 for the negative predictive value. Intraoperative studies revealed sensitivities, specificities, positive predictive values, and negative predictive values of 0.19, 0.98, 0.63, and 0.89, respectively, for gram-staining of specimens of the most inflamed-appearing tissue; 0.36, 0.99, 0.91, and 0.90, respectively, for the white blood-cell count in synovial fluid; and 0.89, 0.85, 0.52, and 0.98, respectively, for a neutrophil count in synovial fluid of more than 80 percent. The sensitivity, specificity, positive predictive value, and negative predictive value were 0.80, 0.94, 0.74, and 0.96, respectively, for the frozen sections and 0.94, 0.97, 0.77, and 0.99, respectively, for the intraoperative cultures.Conclusions: The combination of a normal erythrocyte sedimentation rate and C-reactive protein level is reliable for predicting the absence of infection. Aspiration should be used when the erythrocyte sedimentation rate or the C-reactive protein level is elevated or when a clinical suspicion of infection remains. We found the gram stain to be unreliable. Examination of intraoperative frozen sections is useful in equivocal cases or when hematological markers may be falsely elevated because of an inflammatory or other condition.