Research Article

Maximizing Interpretability and Cost-Effectiveness of Surgical Site Infection (SSI) Predictive Models Using Feature-Specific Regularized Logistic Regression on Preoperative Temporal Data

Table 1

Summary table of articles focused on pre- and/or postoperative data in the prediction of SSI.

ArticleTaskDataPreoperative/postoperativeMethod

Medina-Fernández et al. [1]Use of CRP identifies postoperative infectious complications in patients undergoing colorectal surgeryBlood tests: CRP, neutrophil-to-lymphocyte ratio (NLR), and white blood cell (WBC)Postoperative (2 POD)Multiple linear regression

Dutta et al. [16]Examine WCC, albumin, and CRP following esophagogastric cancer resection as a predictor of postoperative surgical site infectious complicationsBlood tests: White cell count (WCC), albumin, and CRPPostoperative (7 POD)Friedman test and Wilcoxon signed-rank test on medians and ranges

Platt et al. [17]Analyze postoperative WCC, albumin, and CRP and their diagnostic accuracy in case of infectious complications.Blood tests: WCC, albumin, and CRPPostoperative (7 POD)Friedman test and Wilcoxon signed-rank test on medians and ranges

Silvestre et al. [18]Assess the value of serum CRP and PCT time course in the postoperative setting of elective colorectal surgery with primary anastomosis and its potential in detecting infectious postoperative complications.Blood tests: CRP and procalcitonin (PCT)Pre- and postoperative (9 POD)Student’s t test, Mann–Whitney U test, logistic regression

Soguero-ruiz et al. [20]Prediction of SSI with individual blood tests and in a joint model considering linear and nonlinear classifiers, both before and after surgeryBlood tests: hemoglobin, leucocytes, CRP, potassium, sodium, creatinine, ALAT, thrombocytes, albumin, and ALPPreoperative (20 days) and postoperative (20 POD)Gaussian process, linear and nonlinear SVM

Gans et al. [21]Systematic review

Ke et al. [22]The use of dynamic wound data for SSI risk prediction is explored, by exploiting the low-rank property of the spatial-temporal data via the bilinear formulationProcedure related data as well as clinical data such as smoking, diabetes mellitus, or alcohol use, among othersPostoperative (21 POD)Bilinear prediction model, projected gradient descent, bounded matrix completion

Shimizu et al. [27]The authors investigated the risk factors for SSI in patients who had undergone appendectomyBlood tests such as CRP, albumin, NLR; the length of the operation, the number of intra-abdominal drains, the term of antibiotic use, the hospital stay, among others; also, clinical background features were consideredPreoperativeChi-squared test and the Mann–Whitney U test; odds ratio (OR)

Ortega-deballon et al. [28]The aim of this study was to look for a relationship between the fatty tissue metabolism measured by adipocytokine levels and the risk of postoperative infectionBlood tests: preoperative plasma levels of eight adipocytokines, cholesterol, triglycerides, insulin, and CRP; furthermore, patient-specific and intraoperative risk factor for infection such as age and sex, among othersPre- and postoperativeChi-squared tests or Fisher’s exact tests, Wilcoxon test, Spearman’s correlation coefficients, and the odds ratios

Mohri et al. [29]The aim of this study was to examine the association between postoperative infection and preoperative systematic inflammation in patients undergoing resection of gastrointestinal cancerBlood tests: white cell count, hemoglobin, albumin, CRP; furthermore, age, sex, tumor site, operative approach, and the American Society of Anesthesiologists (ASA) gradePre- and postoperativeChi-squared tests, Wilcoxon rank test; a multiple logistic regression analysis was also considered

Moyes et al. [30]The aim was to examine the relationship between the preoperative mGPS (the glasgow prognostic score) and the development of postoperative complications in patients undergoing potentially curative resection for colorectal cancerBlood tests: white cell count, albumin, and C-reactive protein and clinicopathological characteristics such as age, gender, tumor site, and nodal involvement, among others.Pre- and postoperativeMantel–Haenszel (χ2) test for trend, logistic regression analysis

Cappabianca et al. [31]The study objective was to evaluate the effect of CRP on short-term and midterm outcome after cardiac surgeryPreoperative patient profile, including features such as diabetes, body mass index, and smoking history, among others; this type of surgery was also consideredPre- and postoperativeχ2 test, shapiro–Wilk test, kaplan–meier curves, and the log-rank test, logistic regression and cox model

Mujagic et al. [32]This study examines the association between preoperative biochemical markers and the risk of SSIBlood tests: hemoglobin, creatinine, albumin, CRP, and white blood cell count; and other baseline features such as ASA and diabetes, among others.Pre- and postoperativeFisher’s exact test, t-test, kruskal-Wallis test, logistic regression