A retrospective study design was used, and 223 patients who underwent surgery for spinal tumors at our hospital between December 2024 and March 2026 were included. The sample size calculation was based on relevant literature [4]; allowing for a 20% attrition rate, the final sample size was determined to be 223 cases. Inclusion criteria were as follows: 1) pathologically confirmed spinal tumor; 2) first-time tumor lesion resection plus spinal internal fixation; 3) complete medical records, preoperative laboratory tests, operative records, and postoperative follow-up data; 4) postoperative follow-up ≥ 30 d, enabling definitive determination of incisional infection; and 5) informed consent provided by the patient and family members. This single-center retrospective case study extracted all data from archived electronic medical records of our hospital. No interventional procedures were performed on patients during the study. All personal information was anonymized and desensitized such that individual identities could not be traced back. This study was approved by the Institutional Ethics Committee of our hospital, and informed consent was waived. Exclusion criteria were as follows: 1) active infection of the skin at the surgical site preoperatively or systemic sepsis; and 2) concomitant end-stage multiple organ failure or congenital immunodeficiency. Patients were assigned to an infection group or a non-infection group according to whether postoperative incisional infection occurred.

With reference to the Clinical Evidence-Based Guideline for the Diagnosis and Treatment of Spinal Surgical Site Infection [5] and the Standard for the Prevention and Control of Surgical Site Infection: postoperative spinal surgical site infection (surgical site infection, SSI) refers to a surgery-related infection occurring within 30 days after procedures without implants and within 90 days after procedures with implants. Diagnostic Criteria: 1) Clinical manifestations: persistent postoperative fever, aggravated perincisional pain, erythema, wound dehiscence, and purulent drainage; 2) Microbiological evidence: positive bacterial culture of wound drainage or puncture fluid; 3) Imaging evidence: MRI or CT showing deep fluid collection with pus, intraspinal or paravertebral abscess. A confirmed diagnosis requires at least one of the following conditions: a) Clinical manifestations combined with positive bacterial culture; b) Clinical manifestations with supportive imaging findings, and pus confirmed via surgery or puncture. Superficial surgical site infection: erythema, increased local skin temperature and purulent discharge from the incision within 30 days after surgery, with positive bacterial culture of secretions. Deep surgical site infection: pus accumulation in the deep fascia and muscular layers, accompanied by fever and markedly elevated inflammatory markers; imaging reveals deep fluid and pus collection in the operative area.

Eighteen clinical variables were collected and categorized into three groups: 1) baseline characteristics: age, BMI, preoperative serum albumin, hemoglobin, peripheral white blood cell count, neutrophil level, sex, history of diabetes, history of hypertension, smoking history, and alcohol consumption history; 2) intraoperative variables: surgical site, surgical approach, intraoperative blood loss, operative duration, and internal fixation implantation; and 3) postoperative variables: postoperative drainage duration and occurrence of postoperative cerebrospinal fluid leakage. The outcome variable in this study was whether a postoperative incisional infection occurred. Examples of variable assignment and coding: Surgical site: thoracolumbar spine = 1, cervical spine = 0 (reference group); Surgical approach: posterior approach = 1, anterior approach = 0 (reference group).

Statistical analyses were performed using SPSS 26.0. Quantitative data conforming to a normal distribution are presented as mean ± standard deviation ( x ¯ ± s ) and were compared using the t test. Categorical data are presented as counts and percentages (%) and were compared using the χ² test. Between-group comparisons were performed using Fisher’s exact test. Given that there were only 11 infected cases, the expected frequencies of most cells were less than 5, rendering the chi-square test inappropriate. Considering the limited number of 11 infection events, univariate analysis was first conducted to avoid overfitting and unstable parameter estimates. Continuous variables were compared via the t-test or Mann-Whitney U test, while categorical variables were analyzed with Fisher’s exact test. Variables with P < 0.10 in univariate analysis or clinically important predictors (age, surgical site, surgical approach) were incorporated into the multivariable logistic regression model. To mitigate bias caused by sparse data, Firth’s penalized maximum likelihood estimation was adopted for parameter estimation. Model performance was assessed from three dimensions: 1) Discrimination was evaluated using the receiver operating characteristic (ROC) curve and the area under the curve (AUC). The bias-corrected AUC and corresponding 95% confidence intervals (CIs) were calculated via Bootstrap resampling with 1000 replicates; 2) Calibration was assessed by the Hosmer-Lemeshow test and calibration curves; 3) The optimal cutoff value was determined according to the Youden index to calculate sensitivity and specificity. All reported indices were derived from the original modeling cohort, and internal validation results were presented simultaneously. A two-sided P < 0.05 was defined as statistically significant, with the screening threshold loosened to P < 0.10 for univariate variable selection.

Univariate statistical analyses were performed for all 18 clinical variables. Categorical variables were analyzed using Fisher’s exact test, while continuous variables were compared with the t-test or Mann-Whitney U test. The results showed that age (t = 2.391, P = 0.0175), surgical site (χ² = 10.165, P = 0.0015), and surgical approach (χ² = 23.111, P < 0.001) differed significantly between groups (P < 0.05). In contrast, BMI, preoperative albumin, hemoglobin, white blood cell count, neutrophil count, intraoperative blood loss, operative time, postoperative drainage duration, sex, diabetes, hypertension, smoking and alcohol history, internal fixation, and cerebrospinal fluid leakage yielded P > 0.05, and no association with postoperative incisional infection after malignant spinal tumor surgery was identified. Therefore, only these three variables were included in the multivariable logistic regression analysis. See Table 1.

After incorporating variables identified in the univariable screening into the multivariable adjusted analysis, age (Coef. = −0.0483, S.E. = 0.023, Wald = 4.528, P = 0.033, OR = 0.953, 95% CI: 0.911 - 0.996) was an independent protective factor and surgical approach (Coef. = 1.8932, S.E. = 0.758, Wald = 6.230, P = 0.013, OR = 6.647, 95% CI: 1.501 - 29.430) was an independent factors associated with postoperative spinal infection. See Table 2.

Table 1. Results of the univariate analysis (including t/χ² values).

Table 2. Multivariable binary logistic regression analysis (including S.E., Wald, OR, and 95% CI).

The predictive model was constructed based on the above regression coefficients, with the formula as follows: Logit(P) = −3.9001 − 0.0483 × age + 0.6116 × surgical site (thoracolumbar spine = 1) + 1.8932 × surgical approach (posterior approach = 1). The constructed prediction model achieved an AUC of 0.9419, with 100% sensitivity and 88.37% specificity, demonstrating good discriminatory performance and clinical predictive value. See Table 3 and Figure 1. After internal Bootstrap validation with 1000 resamples, the bias-corrected AUC was 0.921 (95% CI: 0.867 - 0.956), indicating no obvious overfitting of the model. The calibration curve demonstrated consistency between predicted probabilities and actual observed incidence. The Hosmer-Lemeshow test yielded χ² = 5.23, P = 0.437, which suggested favorable calibration performance.

Table 3. Diagnostic performance of the predictive model.

Figure 1.Receiver operating characteristic curve of the nomogram model for postoperative surgical site infection in spinal surgery, AUC = 0.942.

The results of this study showed that among 223 patients, 11 had positive surgical wound infections, yielding an overall postoperative incisional infection rate of 4.93%. This rate was higher than that reported by Liu Bin [7]et al. (0.93%) and by Ma Jiaojiao [8]et al. (0.43%).

4.2.1. Age

Age was identified as a protective factor for postoperative infection (OR = 0.953), with the risk of infection decreasing slightly for each additional year of age, which differs from the conclusions of most studies on routine spinal surgery [7][9]. Possible explanations are as follows: older patients with spinal tumors typically undergo comprehensive preoperative systemic evaluation, and those with concomitant diabetes or malnutrition receive early nutritional support and chronic disease management before surgery, with elective surgery scheduled only after relevant indices meet target thresholds. In contrast, tumors in young and middle-aged patients often progress rapidly, and surgery is frequently performed on an urgent, time-limited basis, leaving insufficient time for preoperative optimization. Therefore, infection prevention should not focus exclusively on older patients; postoperative incision care in young and middle-aged patients with malignant spinal tumors also requires close monitoring.

4.2.2. Surgical Approach

This study confirmed that the surgical approach is an independent risk factor for postoperative incisional infection after malignant spinal tumor surgery (OR = 6.647), indicating a 6.647-fold increase in infection risk among patients undergoing this type of approach. Related studies have also shown that [10][11]: the infection rate is relatively high after posterior internal fixation, and clinically high-risk procedures are predominantly posterior spinal tumor resections. Posterior approaches for spinal tumors require extensive dissection of the paraspinal muscles, and tumor lesions invade surrounding soft tissues; consequently, intraoperative wound oozing is substantial and tissue damage is severe, and hematoma readily becomes a medium for colonization and growth of pathogenic bacteria. In contrast, anterior approaches can avoid extensive muscle dissection. For patients undergoing posterior surgery, preoperative optimization of skin preparation and correction of abnormal indicators such as hypoproteinemia and anemia are recommended; intraoperatively, meticulous hemostasis and avoidance of unnecessary soft-tissue dissection should be emphasized; postoperatively, ensure unobstructed drainage and select prophylactic antibiotics on an individualized basis according to guidelines to reduce the risk of infection.