1. Inclusion and exclusion criteria
This study was approved by the Ethics Committee of our Institute (NO.2021-K-241-01) in accordance with the guiding principles of the Declaration of Helsinki. All electronic medical records and image data were anonymised and personal identifiers were completely removed.
Patients who underwent surgical treatment for Pilon fractures in our hospital from January 2012 to June 2021 were included in this retrospective study. The inclusion criteria were: 1) age ≥ 18 years; 2) the Arbeitsgemeinschaft für Osteosynthesefragen/Orthopaedic Trauma Association (AO/OTA) 43 pilon fracture; 3) closed fracture; 4) underwent open reduction and internal fixation (ORIF); 5) complete clinical data. Exclusion criteria were as follows: 1) open fracture; 2) pathological fracture; 3) tibia shaft fracture; 4) trimalleolar ankle fracture; 5) conservative treatment; 6) kirschner wire or external fixation. Finally, a total of 516 pilon fracture patients were enrolled in our study.
2. Risk factors and outcome measures
Demographic information including, age, gender, hemoglobin, serum albumin, c-reactive protein (CRP), blood platelet, leukocyte, preoperative blood sugar, waiting time for surgery, current smoking status and drinking status were extracted from the medical records. Among the causes of injury were falling from height, traffic accident, hit by heavy object and other. Polytrauma was defined as trauma to more than one of the following systems: musculoskeletal, abdominal, cardiothoracic, urogenital, vascular, and central nervous systems. Fractures were classified as extra-articular (43A), partially articular (43B), and intra-articular (43C) according to the AO/OTA system [7]. The degree of soft tissue injury was assessed using the Tscherne classification: Grade 0 represents minimal tissue damage associated with simple fracture pattern; Grade 1 involves superficial abrasion or contusion; Grade 2 involves deep abrasion of skin or muscle contusion; Grade 3 presents with extensive skin and muscle damage or crush injury, subcutaneous avulsion, and/or compartment syndrome [8]. Where there was conflflict in classifification, group discussion was used to reach consensus. Factors related to surgery were also assessed, including operative time, intraoperative blood loss, surgical approach, bone graft, drainage and number of people in the operating room.
A staged approach was used for pilon fractures with severe soft tissue damage, first with external fixation of the tibia and/or restoration of fibula length, and then with delayed tibial open reduction and internal fixation after soft tissue improvement. We defined surgical site infection as any infection that occured at the surgical incision site or deep tissue within 30 days of surgery (within one year of implant used) according to the U.S. Centers for Disease Control and Prevention (CDC) [19]. SSI including superficial and deep infection, with or without positive cultures. The surgeon decided to use antibiotics, wound treatment and surgical treatment based on patient clinical symptom and wound condition.
3. Statistical analysis
Patients were randomly divided into a training group and a validation group (3:1). Measurement data are expressed as mean ± standard deviation, and count data are expressed as n (%). In the training group, univariate analysis using Mann-Whitney U and Chisquared tests as appropriate was performed to assess the association between different variables and surgical site infection. Multivariate analysis of variables with P < 0.1 was then performed to determine the independent risk factors for infection [20]. Based on the regression coefficients of independent risk factors, we established a nomogram model to predict the relationship between surgical site infection and Pilon fracture.
Discrimination of dichotomous results was most often evaluated by calculating the area under the curve (AUC) of the receiver operating characteristic (ROC) curve. Generally, an AUC between 0.5 and 0.7 indicates low accuracy, 0.70-0.9 is considered acceptable, and AUC>0.9 means that the model shows excellent discriminative power [20]. ROC curves were undertaken in both the training and validation group. The calibration curve was the image comparison of predicted probabilities and actual probabilities, which was assessed using the Hosmer-Lemeshow test. Statistical analyses were carried out using Empower[1]Stats (http://www.empowerstats.com, X&Y Solutions, Inc., Boston, MA) and R version 4.0.2 for Windows (R Foundation for Statistical Computing, Vienna, Austria). Two-tailed analysis with P value less than 0.05 indicated that the difference was statistically significant.