Surgical site infection incidence and surgical antibiotic prophylaxis in the ICU patient: a multicenter observational study Authors: Bruno Pastene, Nyl Chemli, Emmanuel Dudoignon, François Depret, Minh-Pierre Lê, Nicolas Mongardon, Anne-Claire Lukaszewicz, Romain Ronflé, Marie Le Baron, Laurent Zieleskiewicz, Ignacio Martin-Loeches, Gary Duclos, Marc Leone Published: 2026 ABSTRACT BackgroundSurgical site infection (SSI) is a major cause of postoperative morbidity. Surgical antibiotic prophylaxis (SAP), defined as the pre-incision administration of an antimicrobial agent targeting likely organisms in clean or clean-contaminated procedures without active infection at the operative site, is a cornerstone of prevention in surgery. However, critically ill patients represent a distinct population with profound susceptibility to infection. Data on SSI incidence and SAP practices in patients undergoing surgery during an intensive care unit (ICU) stay remain scarce.MethodsRetrospective multicenter cohort study in five French ICUs (March 2022–March 2023). Adult patients (≥18 years) with an ICU stay ≥ 48 h who underwent surgery during their ICU stay were included. Primary outcome was SSI incidence after surgery. Generalized estimating equations (GEE) with a prespecified adjustment set were used to estimate the adjusted association between antimicrobial spectrum and SSI, accounting for clustering within centers.Results338 patients were included (median age 56 years, IQR 35 – 67; 72% male). Overall SSI incidence was 21%. Procedures with suspected infection present at time of surgery (PATOS) (n = 52, 15%) had an SSI incidence of 37%, against 18% in PATOS-negative patients. In the primary adjusted GEE analysis, broad-spectrum vs. narrow-spectrum antimicrobials (OR 3.06; 95%CI 1.63–5.76; P < 0.001) and no antimicrobial vs. narrow-spectrum antimicrobials (OR 2.49; 95%CI 1.01–6.15; P = 0.048) were independently associated with SSI. Stratified analysis showed that antimicrobial spectrum also had significant effect on SSI risk on PATOS-positive and trauma patients. SSI was associated with prolonged ICU/hospital length of stay, increased organ support use, and worse outcomes at day 28 and day 90.ConclusionsSSI affects more than one in five critically ill patients. After adjustment for prespecified confounders, the association with broad-spectrum therapy likely reflected confounding by indication, whereas the absence of any antimicrobial coverage was independently associated with increased SSI risk. Conventional SAP guidelines derived from non-ICU populations may not apply to critically ill patients. ICU-specific prophylactic strategies, improved coordination between ICU and operating room teams, and prospective studies are needed to enhance SSI prevention in this high-risk population.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s40560-026-00916-0. Introduction Surgical site infection (SSI) is a major postoperative complication, with reported incidences ranging from 0.5% in knee prosthetic surgery to more than 10% in colorectal procedures [1]. SSIs are consistently associated with prolonged hospitalization, increased healthcare resource utilization, and excess mortality [1–3]. In the operating room, international guidelines frame the use of surgical antibiotic prophylaxis (SAP), a cornerstone intervention for SSI prevention [4]. SAP refers to the prophylactic administration of an antimicrobial agent targeting the most likely organisms for a given surgical site in clean or clean-contaminated procedures [1, 5]. In most cases, a single pre-incision dose is recommended. Contaminated and dirty-infected surgeries are generally excluded from SAP recommendations, except in specific settings such as trauma surgery with contaminated open fractures. The intensive care unit (ICU) represents a unique clinical environment where the most severely ill patients are managed, and a substantial proportion undergo surgical procedures during their ICU stay. These patients frequently present with multiple risk factors for infection, including prior or ongoing exposure to broad-spectrum antibiotics, colonization with multidrug-resistant (MDR) organisms, infection present at time of surgery (PATOS), organ dysfunction, and physiological frailty, all of which may profoundly modify SSI risk and SAP effectiveness [6, 7]. Despite these risks, data on SSI occurrence in ICU patients undergoing surgery during their ICU stay remain scarce. Available evidence is limited to selected populations, including a retrospective study reporting a high SSI incidence following surgery for intra-abdominal infection [8], and a narrative review highlighting the challenges of SAP stewardship in cardiac surgery ICU patients [9]. Importantly, while some guidelines recommend adapting SAP based on patient colonization status for specific procedures [10], there are currently no dedicated recommendations addressing SAP strategies in critically ill patients requiring surgery during their ICU stay. Preliminary single-center observational data suggest marked heterogeneity in SAP practices in this setting, underscoring significant uncertainty regarding optimal management [11]. We therefore hypothesized that ICU patients undergoing surgery during their ICU stay experience a substantial burden of SSIs and that SAP practices in this population are highly variable and frequently deviate from standard guideline recommendations. The primary objective of this multicenter retrospective observational study was to determine the incidence of SSIs in patients requiring surgical procedures during their ICU stay. The secondary objectives were to describe the current SAP strategies in this population, to assess their compliance with published guidelines [5] and the causal relationships between antimicrobial practices and SSI, thereby providing a foundation for future optimization of SAP stewardship in critically ill surgical patients. Methods Study design and setting We conducted a retrospective, multicenter cohort study across five academic and general intensive care units (ICUs): Marseille Hospital Nord, Aix-en-Provence Hospital, Paris Hospitals Saint Louis and Henri Mondor, and Lyon Hospital Edouard Herriot, from March 2022 to March 2023. In the Marseille center, all eligible patients were consecutively included, whereas in the other four centers, predefined cohorts of 20 to 31 consecutive patients were collected to ensure feasibility of data collection while maintaining multicenter participation. Adult patients (≥ 18 years) were eligible if they had an ICU stay of at least 48 h and underwent any surgical procedure for any indication during their ICU stay. Patient data were collected and processed in strict compliance with the European General Data Protection Regulation (GDPR) and the French reference methodology MR-004. The study was conducted in accordance with the principles of the Declaration of Helsinki, registered in the GDPR/APHM registry under number PADS23-112, and received institutional review board (IRB) approval (Ethics Committee for Research in Anesthesia and Intensive Care IRB 00010254–2023-071). All patients were adequately informed of the use of their data [12]. The study was reported in accordance with the STROBE statement for strengthening the reporting of observational studies [13]. Data collection Data were extracted retrospectively from the electronic medical records of each participating center using a standardized data collection form. Baseline patient characteristics included age, sex, comorbidities, documented beta-lactam allergy, Simplified Acute Physiology Score (SAPS) II, and Sequential Organ Failure Assessment (SOFA) score at ICU admission [14, 15]. Admission diagnosis and baseline risk factors for MDR bacteria were also recorded [16]. Preoperative microbiological data included the presence, site, and microbiological profile of MDR bacterial colonization. Perioperative variables included the delay between ICU admission and surgery, PATOS status, Centers for Disease Control and Prevention (CDC) Surgical Wound Classification (SWC), American Society of Anesthesiologists (ASA) score on surgery day, type of surgery, duration of the surgical procedure, and SOFA score on the day of surgery [17]. Perioperative antimicrobial management variables included the type of antimicrobial strategy used (SAP, new antimicrobial therapy, or ongoing antimicrobial therapy), antimicrobial spectrum, and adequacy of treatment according to published guidelines [5, 18]. The occurrence and characteristics of SSIs were collected during the ICU stay, including delay between surgery and SSI diagnosis, CDC SSI classification, microbiological findings, and the presence of MDR bacteria or polymicrobial infection [4]. Postoperative outcomes included new MDR bacterial colonization (site and microbiology), requirement for renal replacement therapy (RRT), duration of invasive mechanical ventilation and vasopressor support during the ICU stay. ICU and hospital length of stay, ICU mortality, and administrative patient status at day 28 and day 90 following surgery were also collected. Outcomes The primary outcome was the incidence of SSIs. Secondary outcomes included identification of potential risk factors for SSI and the association between SSI occurrence and postoperative outcomes. Definitions SSI was defined according to CDC criteria as an infection occurring within 30 days after surgery and classified as superficial incisional SSI (involving skin or subcutaneous tissue), deep incisional SSI (involving deep soft tissues), or organ/space SSI (involving any anatomical structure opened or manipulated during the procedure) [4].Risk factors for MDR bacteria included broad-spectrum antimicrobial therapy within the preceding 3 months, history of extended-spectrum beta-lactamase (ESBL)-producing Enterobacterales carriage, hospitalization abroad within 12 months, residence in a nursing home with an indwelling device, or an ongoing MDR outbreak in the healthcare sector.PATOS was defined according to CDC criteria as an operative procedure with evidence of infection at the time of surgery and/or a surgery scheduled for infectious causes [19].Antimicrobial strategy was defined based on antimicrobial exposure at the time of surgery. Patients receiving antibiotics prior to surgery were classified as receiving “ongoing antimicrobial therapy”. Patients in whom an empirical curative antimicrobial treatment was initiated intraoperatively were classified as receiving “new antimicrobial therapy”. Patients not receiving antibiotics before surgery and who received antimicrobials in the operating room before surgical incision solely for prophylaxis were classified as receiving “SAP”.Assessment of treatment adequacy was independently performed by two authors (BP and NC) by comparing administered antimicrobial therapy with published good-practice guidelines for SAP and antimicrobial stewardship in the ICU [5, 18]. Non-compliance with guidelines was defined as any of the following: SAP administered when not indicated; SAP not administered when indicated; inappropriate choice of SAP (excessively broad or narrow spectrum, or use of beta-lactam allergy alternatives in non-allergic patients); SAP administered in addition to ongoing antimicrobial therapy; omission of ongoing antimicrobial therapy; or initiation of new antimicrobial therapy not adapted to the clinical situation. Compliance with guidelines was defined as administration of appropriate SAP, continuation of appropriate ongoing antimicrobial therapy, or initiation of new antimicrobial therapy adapted to the clinical context. Discrepancies between assessors were resolved by consensus after discussion.Antimicrobials active against both Gram-positive and Gram-negative bacteria (beta-lactam/beta-lactamase inhibitor combinations, third- and fourth-generation cephalosporins, antipseudomonal beta-lactams, carbapenems, fluoroquinolones) or combination regimens were classified as broad-spectrum therapies. Antimicrobials with a more limited spectrum of activity (first- and second-generation cephalosporins, clindamycin) were classified as narrow-spectrum. SSI was defined according to CDC criteria as an infection occurring within 30 days after surgery and classified as superficial incisional SSI (involving skin or subcutaneous tissue), deep incisional SSI (involving deep soft tissues), or organ/space SSI (involving any anatomical structure opened or manipulated during the procedure) [4]. SSI was defined according to CDC criteria as an infection occurring within 30 days after surgery and classified as superficial incisional SSI (involving skin or subcutaneous tissue), deep incisional SSI (involving deep soft tissues), or organ/space SSI (involving any anatomical structure opened or manipulated during the procedure) [4]. Risk factors for MDR bacteria included broad-spectrum antimicrobial therapy within the preceding 3 months, history of extended-spectrum beta-lactamase (ESBL)-producing Enterobacterales carriage, hospitalization abroad within 12 months, residence in a nursing home with an indwelling device, or an ongoing MDR outbreak in the healthcare sector. Risk factors for MDR bacteria included broad-spectrum antimicrobial therapy within the preceding 3 months, history of extended-spectrum beta-lactamase (ESBL)-producing Enterobacterales carriage, hospitalization abroad within 12 months, residence in a nursing home with an indwelling device, or an ongoing MDR outbreak in the healthcare sector. PATOS was defined according to CDC criteria as an operative procedure with evidence of infection at the time of surgery and/or a surgery scheduled for infectious causes [19]. PATOS was defined according to CDC criteria as an operative procedure with evidence of infection at the time of surgery and/or a surgery scheduled for infectious causes [19]. Antimicrobial strategy was defined based on antimicrobial exposure at the time of surgery. Patients receiving antibiotics prior to surgery were classified as receiving “ongoing antimicrobial therapy”. Patients in whom an empirical curative antimicrobial treatment was initiated intraoperatively were classified as receiving “new antimicrobial therapy”. Patients not receiving antibiotics before surgery and who received antimicrobials in the operating room before surgical incision solely for prophylaxis were classified as receiving “SAP”. Antimicrobial strategy was defined based on antimicrobial exposure at the time of surgery. Patients receiving antibiotics prior to surgery were classified as receiving “ongoing antimicrobial therapy”. Patients in whom an empirical curative antimicrobial treatment was initiated intraoperatively were classified as receiving “new antimicrobial therapy”. Patients not receiving antibiotics before surgery and who received antimicrobials in the operating room before surgical incision solely for prophylaxis were classified as receiving “SAP”. Assessment of treatment adequacy was independently performed by two authors (BP and NC) by comparing administered antimicrobial therapy with published good-practice guidelines for SAP and antimicrobial stewardship in the ICU [5, 18]. Non-compliance with guidelines was defined as any of the following: SAP administered when not indicated; SAP not administered when indicated; inappropriate choice of SAP (excessively broad or narrow spectrum, or use of beta-lactam allergy alternatives in non-allergic patients); SAP administered in addition to ongoing antimicrobial therapy; omission of ongoing antimicrobial therapy; or initiation of new antimicrobial therapy not adapted to the clinical situation. Compliance with guidelines was defined as administration of appropriate SAP, continuation of appropriate ongoing antimicrobial therapy, or initiation of new antimicrobial therapy adapted to the clinical context. Discrepancies between assessors were resolved by consensus after discussion. Assessment of treatment adequacy was independently performed by two authors (BP and NC) by comparing administered antimicrobial therapy with published good-practice guidelines for SAP and antimicrobial stewardship in the ICU [5, 18]. Non-compliance with guidelines was defined as any of the following: SAP administered when not indicated; SAP not administered when indicated; inappropriate choice of SAP (excessively broad or narrow spectrum, or use of beta-lactam allergy alternatives in non-allergic patients); SAP administered in addition to ongoing antimicrobial therapy; omission of ongoing antimicrobial therapy; or initiation of new antimicrobial therapy not adapted to the clinical situation. Compliance with guidelines was defined as administration of appropriate SAP, continuation of appropriate ongoing antimicrobial therapy, or initiation of new antimicrobial therapy adapted to the clinical context. Discrepancies between assessors were resolved by consensus after discussion. Antimicrobials active against both Gram-positive and Gram-negative bacteria (beta-lactam/beta-lactamase inhibitor combinations, third- and fourth-generation cephalosporins, antipseudomonal beta-lactams, carbapenems, fluoroquinolones) or combination regimens were classified as broad-spectrum therapies. Antimicrobials with a more limited spectrum of activity (first- and second-generation cephalosporins, clindamycin) were classified as narrow-spectrum. Antimicrobials active against both Gram-positive and Gram-negative bacteria (beta-lactam/beta-lactamase inhibitor combinations, third- and fourth-generation cephalosporins, antipseudomonal beta-lactams, carbapenems, fluoroquinolones) or combination regimens were classified as broad-spectrum therapies. Antimicrobials with a more limited spectrum of activity (first- and second-generation cephalosporins, clindamycin) were classified as narrow-spectrum. Statistical analysis Categorical variables were reported as counts and percentages, and continuous variables as mean ± standard deviation or median with interquartile range (IQR, 25th–75th percentile), depending on distribution assessed using the Shapiro–Wilk and Kolmogorov–Smirnov tests. Descriptive statistics were used to summarize baseline characteristics. Univariate analyses were performed to compare preoperative characteristics and postoperative outcomes between patients with and without SSI. Normally distributed continuous variables were compared using two-tailed Student’s t-tests, while non-normally distributed variables were analyzed using the Mann–Whitney U test. A two-sided p-value < 0.05 was considered statistically significant. To contextualize the SSI rate and assess clinically important confounders on the SSI risk, descriptive sensitivity analyses stratified the cohort by PATOS status and SAP spectrum within PATOS-negative patients. All sensitivity analyses were descriptive, and no formal subgroup testing was conducted. To estimate confounder-adjusted association of antimicrobial spectrum on SSI risk, a directed acyclic graph (DAG) (Supplemental Fig. 1) was constructed to formalize assumptions about the presumed causal structure among study variables. Pearl’s backdoor criterion was applied to identify the minimal sufficient adjustment set, which consisted of ASA score, congestive heart failure, hypertension, diabetes, age, SAPS II score at admission, NNIS (National Nosocomial Infections Surveillance) risk index, CDC SWC, PATOS status, preoperative MDR bacteria colonization.Fig. 1Distribution of antimicrobial prophylaxis strategies in the cohort (n = 338). Patients were categorized into three primary strategies: new antimicrobial therapy (n = 11, 3%), surgical antimicrobial prophylaxis/SAP (n = 248, 73%), and ongoing antimicrobial therapy (n = 79, 24%). Within each strategy, patients were classified by adherence to institutional guidelines. Green boxes indicate guideline-compliant administration; red boxes indicate non-compliant prescribing. Percentages represent either the proportion of the total cohort or the proportion within each strategy category. SAP = surgical antimicrobial prophylaxis, ICU = intensive care unit, % (total) = percentage of entire cohort, % (group) = percentage within strategy group Distribution of antimicrobial prophylaxis strategies in the cohort (n = 338). Patients were categorized into three primary strategies: new antimicrobial therapy (n = 11, 3%), surgical antimicrobial prophylaxis/SAP (n = 248, 73%), and ongoing antimicrobial therapy (n = 79, 24%). Within each strategy, patients were classified by adherence to institutional guidelines. Green boxes indicate guideline-compliant administration; red boxes indicate non-compliant prescribing. Percentages represent either the proportion of the total cohort or the proportion within each strategy category. SAP = surgical antimicrobial prophylaxis, ICU = intensive care unit, % (total) = percentage of entire cohort, % (group) = percentage within strategy group The primary confounder-adjusted analysis was performed using Generalized Estimating Equations (GEE) using the prespecified adjustment set, with an exchangeable correlation structure accounting for patient clustering within centers. To assess whether the trauma over-representation in our cohort may impact our findings, stratified GEE analyses by admission diagnosis were performed. For stratified subgroup analyses with fewer contributing clusters, we reported 95% confidence intervals (CI) derived from Firth-penalized logistic regression to avoid underestimating uncertainty in small-cluster settings. DAG and GEE analyses were performed in R version 4.5.2 (RStudio version 2026.05.0 + 218) using the geepack and dagitty packages. Descriptive analyses were performed using Jamovi version 2.3.19.0 Results Study population and baseline characteristics During the study period, 338 patients were included. Median (IQR) age was 56 years (35–67), and 245 (72%) were male. Trauma accounted for 52% of ICU admission diagnoses. Median (IQR) SAPS-II and SOFA scores at ICU admission were 35 (22–48) and 4 (2–7), respectively. A documented beta-lactam allergy was reported in 14 (4%) patients. Risk factors for MDR bacteria colonization were identified in 60 (18%) patients. Preoperative MDR colonization was documented in 29 (9%) patients, predominantly due to ESBL-producing Enterobacterales (n = 17). Among colonized patients, colonization status was assessed using rectal swabs in 16 (55%) (Table 1).Table 1Cohort description and univariate analysis on the primary outcome for demographics, surgery characteristics and antimicrobial treatment features during surgeryVariablesTotaln = 338SSIn = 70No SSIn = 268PDemographicsAge, y56 (35–67)57 (38–65)55 (35–67)0.993Gender, n (%) Male245 (72)50 (71)195 (73)0.824 Female93 (28)20 (29)73 (27) BMI, kg.m−224 (22–27)24 (22–30)24 (22–27)0.184ICU admission diagnosis, n (%) Elective surgery32 (9)8 (12)24 (9)0.055 Emergency surgery87 (26)19 (27)68 (25) Medical44 (13)15 (21)29 (11) Severe trauma175 (52)28 (40)147 (55)Center, n (%)0.007 Marseille Nord237 (70)45 (64)192 (72) Aix-en-Provence20 (6)6 (9)14 (5) Paris Mondor20 (6)8 (12)12 (4)* Paris St Louis30 (9)1(1)29 (11)* Lyon Edouard Herriot31 (9)10 (14)21 (8) SAPS II at ICU admission35 (22–48)36 (28–51)34 (21–48)0.040SOFA score at ICU admission4 (2–7)5 (3–7)4 (1–7)0.035Comorbidities, n (%) Any238 (70)50 (71)188 (70)0.835 Diabetes41 (12)12 (17)29 (11)0.149 Hypertension91 (27)25 (36)66 (25)0.063 Congestive cardiac failure19 (6)7 (10)12 (5)0.074 Stroke16 (5)4 (6)12 (5)0.664 Beta-lactam allergy, n (%)14 (4)4 (6)10 (4)0.458Risk factors for MDR bacteria, n (%) 0278 (82)52 (74)226 (84)0.143 149 (15)15 (21)34 (13) 211 (3)3 (4)8 (3) Preoperative MDR bacteria colonization, n (%)29 (9)5 (7)24 (9)0.630Site of MDR bacteria colonization, n (%) Rectal16 (55)2 (40)14 (58)0.618 Nasal5 (17)1 (20)4 (17) Both2 (7)02 (8) Other6 (21)2 (40)4 (17)Type of MDR bacteria colonization, n (%)0.676 ESBL-E17 (59)3 (60)14 (58) MRSA3 (10)03 (13) Other9 (31)2 (40)7 (29)Surgery characteristics Time from ICU admission to surgery, d1 (1–3)1 (0–3)1 (0–3)0.919 PATOS-positive surgery, n (%)52 (15)19 (27)33 (12)0.002 Surgery duration, min120 (75–190)120 (85–193)115 (70–190)0.483CDC SWC, n (%) < 0.001 1–2196 (58)21 (30)175 (65)* 3–4142 (42)49 (70)93 (35)*SOFA score on surgery day4 (1–7)5 (4–7)4 (1–7)0.031ASA score on surgery day, n (%) 113 (4)3 (4)10 (4)0.345 246 (13)6 (9)40 (15) 3134 (40)24 (34)110 (41) 499 (29)26 (37)73 (27) 546 (14)11 (16)35 (13)Type of surgery, n (%)0.008 Neurosurgery (brain)63 (19)7 (10)56 (21) Neurosurgery (spine)27 (8)027 (10)* Orthopedic77 (23)18 (26)59 (22) Uro-visceral73 (22)26 (37)47 (18)* Vascular4 (1)2 (3)2 (1) Endovascular13 (4)1 (1.5)12 (5) Thoracic30 (9)5 (7)25 (9) Endoscopy7 (2)1 (1.5)6 (2) Interventional radiology17 (5)3 (4)14 (5) ENT11 (3)3 (4)8 (3) Mixed12 (3)3 (4)9 (3) Other4 (1)1 (1.5)3 (2)Antimicrobial treatment features during surgeryTreatment group, n (%)0.003 SAP248 (73)41 (59)207 (77) Ongoing antimicrobial therapy79 (24)27 (38)52 (20) New antimicrobial therapy11 (3)2 (3)9 (3)* Treatment group and compliance, n(%) < 0.001 Compliant SAP184 (55)26 (37)158 (59)* Non-compliant SAP64 (19)15 (21)49 (19) Compliant ongoing antimicrobial therapy41 (12)19 (27)22 (8)* Non-compliant ongoing antimicrobial therapy38 (11)8 (12)30 (11) New antimicrobial therapy11 (3)2 (3)9 (3) Antimicrobial spectrum of the treatment, n (%) < 0.001 Narrow147 (43)14 (20)133 (50)* Broad97 (29)36 (51)61 (23)* No treatment administered94 (28)20 (29)74 (27)Primary outcome Surgical site infection, n (%)70 (21)–––*Post hoc standardized residuals > 2All continuous variables are expressed in median and interquartile rangeSSI surgical site infection, IQR interquartile range, BMI body mass index, ASA American Society of Anesthesiologists, ICU intensive care unit, SAPS Simplified Acute Physiology Score, SOFA Sequential Organ Failure Assessment, MDR multi-drug resistant, ESBL-E extended-spectrum beta-lactamase Enterobacterales, MRSA methicillin-resistant Staphylococcus aureus, PATOS present at time of surgery, CDC SWC Centers for Disease Control and Prevention Surgical Wound Classification, ENT ear–nose–throat, SAP surgical antibiotic prophylaxis Table 1Cohort description and univariate analysis on the primary outcome for demographics, surgery characteristics and antimicrobial treatment features during surgeryVariablesTotaln = 338SSIn = 70No SSIn = 268PDemographicsAge, y56 (35–67)57 (38–65)55 (35–67)0.993Gender, n (%) Male245 (72)50 (71)195 (73)0.824 Female93 (28)20 (29)73 (27) BMI, kg.m−224 (22–27)24 (22–30)24 (22–27)0.184ICU admission diagnosis, n (%) Elective surgery32 (9)8 (12)24 (9)0.055 Emergency surgery87 (26)19 (27)68 (25) Medical44 (13)15 (21)29 (11) Severe trauma175 (52)28 (40)147 (55)Center, n (%)0.007 Marseille Nord237 (70)45 (64)192 (72) Aix-en-Provence20 (6)6 (9)14 (5) Paris Mondor20 (6)8 (12)12 (4)* Paris St Louis30 (9)1(1)29 (11)* Lyon Edouard Herriot31 (9)10 (14)21 (8) SAPS II at ICU admission35 (22–48)36 (28–51)34 (21–48)0.040SOFA score at ICU admission4 (2–7)5 (3–7)4 (1–7)0.035Comorbidities, n (%) Any238 (70)50 (71)188 (70)0.835 Diabetes41 (12)12 (17)29 (11)0.149 Hypertension91 (27)25 (36)66 (25)0.063 Congestive cardiac failure19 (6)7 (10)12 (5)0.074 Stroke16 (5)4 (6)12 (5)0.664 Beta-lactam allergy, n (%)14 (4)4 (6)10 (4)0.458Risk factors for MDR bacteria, n (%) 0278 (82)52 (74)226 (84)0.143 149 (15)15 (21)34 (13) 211 (3)3 (4)8 (3) Preoperative MDR bacteria colonization, n (%)29 (9)5 (7)24 (9)0.630Site of MDR bacteria colonization, n (%) Rectal16 (55)2 (40)14 (58)0.618 Nasal5 (17)1 (20)4 (17) Both2 (7)02 (8) Other6 (21)2 (40)4 (17)Type of MDR bacteria colonization, n (%)0.676 ESBL-E17 (59)3 (60)14 (58) MRSA3 (10)03 (13) Other9 (31)2 (40)7 (29)Surgery characteristics Time from ICU admission to surgery, d1 (1–3)1 (0–3)1 (0–3)0.919 PATOS-positive surgery, n (%)52 (15)19 (27)33 (12)0.002 Surgery duration, min120 (75–190)120 (85–193)115 (70–190)0.483CDC SWC, n (%) < 0.001 1–2196 (58)21 (30)175 (65)* 3–4142 (42)49 (70)93 (35)*SOFA score on surgery day4 (1–7)5 (4–7)4 (1–7)0.031ASA score on surgery day, n (%) 113 (4)3 (4)10 (4)0.345 246 (13)6 (9)40 (15) 3134 (40)24 (34)110 (41) 499 (29)26 (37)73 (27) 546 (14)11 (16)35 (13)Type of surgery, n (%)0.008 Neurosurgery (brain)63 (19)7 (10)56 (21) Neurosurgery (spine)27 (8)027 (10)* Orthopedic77 (23)18 (26)59 (22) Uro-visceral73 (22)26 (37)47 (18)* Vascular4 (1)2 (3)2 (1) Endovascular13 (4)1 (1.5)12 (5) Thoracic30 (9)5 (7)25 (9) Endoscopy7 (2)1 (1.5)6 (2) Interventional radiology17 (5)3 (4)14 (5) ENT11 (3)3 (4)8 (3) Mixed12 (3)3 (4)9 (3) Other4 (1)1 (1.5)3 (2)Antimicrobial treatment features during surgeryTreatment group, n (%)0.003 SAP248 (73)41 (59)207 (77) Ongoing antimicrobial therapy79 (24)27 (38)52 (20) New antimicrobial therapy11 (3)2 (3)9 (3)* Treatment group and compliance, n(%) < 0.001 Compliant SAP184 (55)26 (37)158 (59)* Non-compliant SAP64 (19)15 (21)49 (19) Compliant ongoing antimicrobial therapy41 (12)19 (27)22 (8)* Non-compliant ongoing antimicrobial therapy38 (11)8 (12)30 (11) New antimicrobial therapy11 (3)2 (3)9 (3) Antimicrobial spectrum of the treatment, n (%) < 0.001 Narrow147 (43)14 (20)133 (50)* Broad97 (29)36 (51)61 (23)* No treatment administered94 (28)20 (29)74 (27)Primary outcome Surgical site infection, n (%)70 (21)–––*Post hoc standardized residuals > 2All continuous variables are expressed in median and interquartile rangeSSI surgical site infection, IQR interquartile range, BMI body mass index, ASA American Society of Anesthesiologists, ICU intensive care unit, SAPS Simplified Acute Physiology Score, SOFA Sequential Organ Failure Assessment, MDR multi-drug resistant, ESBL-E extended-spectrum beta-lactamase Enterobacterales, MRSA methicillin-resistant Staphylococcus aureus, PATOS present at time of surgery, CDC SWC Centers for Disease Control and Prevention Surgical Wound Classification, ENT ear–nose–throat, SAP surgical antibiotic prophylaxis Cohort description and univariate analysis on the primary outcome for demographics, surgery characteristics and antimicrobial treatment features during surgery *Post hoc standardized residuals > 2 All continuous variables are expressed in median and interquartile range SSI surgical site infection, IQR interquartile range, BMI body mass index, ASA American Society of Anesthesiologists, ICU intensive care unit, SAPS Simplified Acute Physiology Score, SOFA Sequential Organ Failure Assessment, MDR multi-drug resistant, ESBL-E extended-spectrum beta-lactamase Enterobacterales, MRSA methicillin-resistant Staphylococcus aureus, PATOS present at time of surgery, CDC SWC Centers for Disease Control and Prevention Surgical Wound Classification, ENT ear–nose–throat, SAP surgical antibiotic prophylaxis Surgical characteristics Surgery was performed at a median (IQR) of 1 day (1–3) following ICU admission, with a median (IQR) operative duration of 120 min (75–190). PATOS-positive procedure accounted for 52 (15%) cases. On the day of surgery, median (IQR) SOFA score was 4 (1–7). According to the CDC SWC, 196 (58%) procedures were classified as clean or clean-contaminated (SWC 1–2) (Table 1). Perioperative antimicrobial management Overall, 248 (73%) patients received SAP, 79 (24%) were receiving ongoing antimicrobial therapy at the time of surgery, and 11 (3%) received a new antimicrobial therapy initiated intraoperatively. During surgery, 147 (43%), 97 (29%), and 94 (28%) patients received narrow-spectrum antimicrobials, broad-spectrum antimicrobials, no antimicrobial, respectively. SAP administration was considered compliant with guidelines in 184 (55%) patients and non-compliant in 64 (19%). Management of ongoing antimicrobial therapy was compliant in 41 (12%) patients and non-compliant in 38 (11%). All patients receiving a new antimicrobial therapy intraoperatively were managed in compliance with guidelines (n = 11 (3%)) (Table 1). Detailed antimicrobial strategies are presented in Fig. 1. Primary outcome SSI occurred in 70 patients, corresponding to an incidence of 21%. SSI features are presented in Table 2. Median (IQR) time for SSI diagnosis was 7 days (1–16). SSI were mainly represented by organ/space SSI (n = 48, 68%), followed by deep incisional (n = 20, 29%) and superficial (2, 3%). Staphylococcus aureus accounted for 16% of SSI pathogens, while other Gram-positive cocci (47%) and Enterobacterales (59%) were also frequently identified. Forty (57%) SSIs were polymicrobial and 15 (21%) involved MDR bacteria.Table 2Features of surgical site infectionsVariablesSSIn = 70Time from surgery to SSI diagnosis, day (IQR)7 (1–16)CDC SSI classification, n (%) Superficial2 (3) Deep incisional20 (29) Organ/Space48 (68)Microbiology of SSI, n (%) Non-fermenting Gram-negative bacteria5 (7) Enterobacterales41 (59) Other Gram-negative bacteria3 (4) S. aureus11 (16) Other Gram-positive cocci33 (47) Other13 (19) Polymicrobial SSI, n (%)40 (57) MDR bacteria SSI, n (%)15 (21)IQR interquartile range, SSI surgical site infection, CDC Centers for Disease Control and Prevention, MDR multi-drug resistant Table 2Features of surgical site infectionsVariablesSSIn = 70Time from surgery to SSI diagnosis, day (IQR)7 (1–16)CDC SSI classification, n (%) Superficial2 (3) Deep incisional20 (29) Organ/Space48 (68)Microbiology of SSI, n (%) Non-fermenting Gram-negative bacteria5 (7) Enterobacterales41 (59) Other Gram-negative bacteria3 (4) S. aureus11 (16) Other Gram-positive cocci33 (47) Other13 (19) Polymicrobial SSI, n (%)40 (57) MDR bacteria SSI, n (%)15 (21)IQR interquartile range, SSI surgical site infection, CDC Centers for Disease Control and Prevention, MDR multi-drug resistant Features of surgical site infections IQR interquartile range, SSI surgical site infection, CDC Centers for Disease Control and Prevention, MDR multi-drug resistant Secondary outcomes and univariate analysis In univariate analysis, severWty scores at ICU admission and on the day of surgery, type of surgery, surgery for infectious indications, CDC SWC, antimicrobial treatment group, compliance with antimicrobial guidelines, and antimicrobial spectrum were associated with SSI occurrence (Table 2). SSI occurrence was associated with prolonged ICU length of stay, including time from surgery to ICU discharge, longer hospital length of stay, increased duration of vasopressor support and invasive mechanical ventilation, and greater use of RRT. ICU mortality did not differ between patients with and without SSI; however, outcomes at day 28 and day 90 were significantly worse in patients who developed SSI (Table 3).Table 3Cohort description and univariate analysis on the primary outcome for outcomesVariablestotal(N = 338)SSI(n = 70)no SSI(n = 268)POutcomesPostoperative MDR bacteria colonization, n (%)41 (12)13 (19)28 (10)0.064Site of postoperative MDR bacteria colonization, n (%) Rectal24 (58)7 (54)17 (61)0.902 Nasal2 (5)1 (8)1 (3) Both4 (10)1 (8)3 (11) Other11 (27)4 (30)7 (25)Type of postoperative MDR bacteria colonization, n (%)0.247 ESBL-E26 (63)11 (85)15 (54) MRSA2 (5)02 (7) Both2 (5)02 (7) Other11 (27)2 (15)9 (32) Time from surgery to ICU discharge, d (IQR)6 (3–13)10 (4–28)5 (3–11) < 0.001 Postoperative RRT, n (%)38 (11)18 (26)20 (8) < 0.001 Postoperative MV duration, d (IQR)1 (0 – 6)2 (0–12)1 (0–5)0.009 Postoperative vasopressor duration, d (IQR)1 (0–2)1 (0–4)1 (0–2)0.018 ICU LOS, d (IQR)7.5 (4–17)14 (7–32)7 (4–15) < 0.001 Hospital LOS, d (IQR)19 (11–39)40 (24–61)17 (10–27) < 0.001 ICU death, n (%)54 (16)8 (11)46 (17)0.244Patient situation at day 28, n (%) < 0.001 Home76 (22)2 (3)74 (28)* Rehabilitation center67 (20)9 (13)58 (22) Hospitalized, ICU52 (15)24 (34)28 (10)* Hospitalized, IMCU2 (1)02 (1) Hospitalized, ward80 (24)28 (40)52 (19)* Dead55 (16)6 (9)49 (18) Unknown6 (2)1 (1)5 (2)Patient situation at day 90, n (%)0.004 Home159 (47)24 (34)135 (50) Rehabilitation center55 (16)18 (26)37 (14) Hospitalized, ICU6 (2)1 (1)4 (2) Hospitalized, IMCU001 (1) Hospitalized, ward22 (7)11 (16)11 (4)* Dead68 (20)12 (17)56 (21) Unknown28 (8)4 (6)23 (8)SSI surgical site infection, MDR multi-drug resistant, ESBL-E extended-spectrum beta-lactamase Enterobacterales, MRSA methicillin-resistant Staphylococcus aureus, RRT renal replacement therapy, MV mechanical ventilation, ICU intensive care unit, LOS length of stay, IMCU intermediate care unit, IQR interquartile range Table 3Cohort description and univariate analysis on the primary outcome for outcomesVariablestotal(N = 338)SSI(n = 70)no SSI(n = 268)POutcomesPostoperative MDR bacteria colonization, n (%)41 (12)13 (19)28 (10)0.064Site of postoperative MDR bacteria colonization, n (%) Rectal24 (58)7 (54)17 (61)0.902 Nasal2 (5)1 (8)1 (3) Both4 (10)1 (8)3 (11) Other11 (27)4 (30)7 (25)Type of postoperative MDR bacteria colonization, n (%)0.247 ESBL-E26 (63)11 (85)15 (54) MRSA2 (5)02 (7) Both2 (5)02 (7) Other11 (27)2 (15)9 (32) Time from surgery to ICU discharge, d (IQR)6 (3–13)10 (4–28)5 (3–11) < 0.001 Postoperative RRT, n (%)38 (11)18 (26)20 (8) < 0.001 Postoperative MV duration, d (IQR)1 (0 – 6)2 (0–12)1 (0–5)0.009 Postoperative vasopressor duration, d (IQR)1 (0–2)1 (0–4)1 (0–2)0.018 ICU LOS, d (IQR)7.5 (4–17)14 (7–32)7 (4–15) < 0.001 Hospital LOS, d (IQR)19 (11–39)40 (24–61)17 (10–27) < 0.001 ICU death, n (%)54 (16)8 (11)46 (17)0.244Patient situation at day 28, n (%) < 0.001 Home76 (22)2 (3)74 (28)* Rehabilitation center67 (20)9 (13)58 (22) Hospitalized, ICU52 (15)24 (34)28 (10)* Hospitalized, IMCU2 (1)02 (1) Hospitalized, ward80 (24)28 (40)52 (19)* Dead55 (16)6 (9)49 (18) Unknown6 (2)1 (1)5 (2)Patient situation at day 90, n (%)0.004 Home159 (47)24 (34)135 (50) Rehabilitation center55 (16)18 (26)37 (14) Hospitalized, ICU6 (2)1 (1)4 (2) Hospitalized, IMCU001 (1) Hospitalized, ward22 (7)11 (16)11 (4)* Dead68 (20)12 (17)56 (21) Unknown28 (8)4 (6)23 (8)SSI surgical site infection, MDR multi-drug resistant, ESBL-E extended-spectrum beta-lactamase Enterobacterales, MRSA methicillin-resistant Staphylococcus aureus, RRT renal replacement therapy, MV mechanical ventilation, ICU intensive care unit, LOS length of stay, IMCU intermediate care unit, IQR interquartile range Cohort description and univariate analysis on the primary outcome for outcomes SSI surgical site infection, MDR multi-drug resistant, ESBL-E extended-spectrum beta-lactamase Enterobacterales, MRSA methicillin-resistant Staphylococcus aureus, RRT renal replacement therapy, MV mechanical ventilation, ICU intensive care unit, LOS length of stay, IMCU intermediate care unit, IQR interquartile range Sensitivity analysis PATOS status stratification showed different SSI incidences across subgroups. The SSI incidence was 37% in the PATOS-positive subgroup and 18% in the PATOS-negative subgroup. The PATOS-negative patients treated with narrow-spectrum SAP had an 8% SSI incidence, while those who did not receive SAP had an SSI incidence at 22%. Both SSI incidences in these patients were higher than those reported in a benchmark non-ICU patient population [1] (Table 4).Table 4PATOS stratification of SSI incidencePatient subgroupnSSI rateOverall cohort33821%PATOS-positive5237%PATOS-negative28618%PATOS-negative + narrow-spectrum prophylaxis1408%PATOS-negative + no treatment administered6922%General surgery benchmark (non-ICU)–0.5–10% [1] Table 4PATOS stratification of SSI incidencePatient subgroupnSSI rateOverall cohort33821%PATOS-positive5237%PATOS-negative28618%PATOS-negative + narrow-spectrum prophylaxis1408%PATOS-negative + no treatment administered6922%General surgery benchmark (non-ICU)–0.5–10% [1] PATOS stratification of SSI incidence Confounder-adjusted analysis: effect of antimicrobial spectrum on SSI risk In the primary confounder-adjusted GEE analysis, fitted incorporating the adjustment set identified by the DAG (ASA score, congestive heart failure, hypertension, diabetes, age, SAPS-II score at admission, NNIS risk index, CDC SWC, PATOS status, preoperative MDR bacteria colonization) and using center as the exchangeable correlation structure, antimicrobial spectrum was independently associated with SSI risk (Table 5):Broad-spectrum SAP vs. narrow-spectrum SAP: OR 3.06 (95%CI 1.63–5.76, P < 0.001)No SAP vs. narrow-spectrum: OR 2.49 (95%CI 1.01–6.15, P = 0.048).Table 5Primary confounder-adjusted analysis: generalized estimating equations (GEE) model results for antimicrobial spectrum and SSI riskVariableBeta (log-OR)Std. errorz-statisticPAdjusted OR (95% CI)Intercept−2.5701Exposure Broad vs. narrow1.11940.32223.474 < 0.0013.06 (1.63–5.76) No treatment vs. narrow0.91210.46131.9770.0482.49 (1.01–6.15)Model: GEE binomial logistic regression with exchangeable correlation structure, clustering by hospital site. Adjusted for: age, SAPS-II score at surgery, ASA classification, NNIS risk index, diabetes, hypertension, heart failure, MDR colonization, PATOS status, and CDC surgical wound classificationOR odds ratio, CI confidence interval, GEE generalized estimating equations, SAPS-II Simplified Acute Physiology Score, ASA American Society of Anesthesiologists, NNIS National Nosocomial Infections Surveillance, MDR multi-drug resistant, PATOS: present at time of surgery, CDC Centers for Disease Control and Prevention Broad-spectrum SAP vs. narrow-spectrum SAP: OR 3.06 (95%CI 1.63–5.76, P < 0.001) Broad-spectrum SAP vs. narrow-spectrum SAP: OR 3.06 (95%CI 1.63–5.76, P < 0.001) No SAP vs. narrow-spectrum: OR 2.49 (95%CI 1.01–6.15, P = 0.048). No SAP vs. narrow-spectrum: OR 2.49 (95%CI 1.01–6.15, P = 0.048). Table 5Primary confounder-adjusted analysis: generalized estimating equations (GEE) model results for antimicrobial spectrum and SSI riskVariableBeta (log-OR)Std. errorz-statisticPAdjusted OR (95% CI)Intercept−2.5701Exposure Broad vs. narrow1.11940.32223.474 < 0.0013.06 (1.63–5.76) No treatment vs. narrow0.91210.46131.9770.0482.49 (1.01–6.15)Model: GEE binomial logistic regression with exchangeable correlation structure, clustering by hospital site. Adjusted for: age, SAPS-II score at surgery, ASA classification, NNIS risk index, diabetes, hypertension, heart failure, MDR colonization, PATOS status, and CDC surgical wound classificationOR odds ratio, CI confidence interval, GEE generalized estimating equations, SAPS-II Simplified Acute Physiology Score, ASA American Society of Anesthesiologists, NNIS National Nosocomial Infections Surveillance, MDR multi-drug resistant, PATOS: present at time of surgery, CDC Centers for Disease Control and Prevention Primary confounder-adjusted analysis: generalized estimating equations (GEE) model results for antimicrobial spectrum and SSI risk Model: GEE binomial logistic regression with exchangeable correlation structure, clustering by hospital site. Adjusted for: age, SAPS-II score at surgery, ASA classification, NNIS risk index, diabetes, hypertension, heart failure, MDR colonization, PATOS status, and CDC surgical wound classification OR odds ratio, CI confidence interval, GEE generalized estimating equations, SAPS-II Simplified Acute Physiology Score, ASA American Society of Anesthesiologists, NNIS National Nosocomial Infections Surveillance, MDR multi-drug resistant, PATOS: present at time of surgery, CDC Centers for Disease Control and Prevention In accordance with confounder-adjusted inference reporting standards [20], only exposure ORs are shown. Complete model output is presented in Supplementary Table S1. Stratified GEE analysis of admission diagnosis (trauma vs. non-trauma) showed that the adjusted OR for broad vs. narrow-spectrum SAP remained substantially elevated and consistent across both subgroups (non-trauma: OR 5.90, 95%CI 2.84 – 12.25; trauma: OR 5.05, 95%CI 1.83 – 15.51). However, when comparing with a narrow-spectrum SAP, the non-administration of antimicrobial therapy was associated with higher SSI risk in trauma patients (OR 5.47, 95%CI 1.69 – 18.71)) and remained non-significant in non-trauma patients (OR 1.51, 95%CI 0.49 – 4.66) (Table 6).Table 6Stratified GEE analysis on admission diagnosisAnalysisComparisonAdjusted OR (95% CI)PNon-traumaBroad vs. narrow5.90 (2.84–12 .25) < 0.0001Non-traumaNo treatment vs. narrow1.51 (0.49–4.66)0.477TraumaBroad vs. narrow5.05 (1.83–15.51)* < 0.0001TraumaNo treatment vs. narrow5.47 (1.69–18.71)* < 0.0001OR odds ratio, CI confidence interval*For these subgroups (2 contributing hospitals), CI derived from Firth-penalized logistic regression due to limited clusters Table 6Stratified GEE analysis on admission diagnosisAnalysisComparisonAdjusted OR (95% CI)PNon-traumaBroad vs. narrow5.90 (2.84–12 .25) < 0.0001Non-traumaNo treatment vs. narrow1.51 (0.49–4.66)0.477TraumaBroad vs. narrow5.05 (1.83–15.51)* < 0.0001TraumaNo treatment vs. narrow5.47 (1.69–18.71)* < 0.0001OR odds ratio, CI confidence interval*For these subgroups (2 contributing hospitals), CI derived from Firth-penalized logistic regression due to limited clusters Stratified GEE analysis on admission diagnosis OR odds ratio, CI confidence interval *For these subgroups (2 contributing hospitals), CI derived from Firth-penalized logistic regression due to limited clusters Discussion In this multicenter cohort of 338 critically ill patients undergoing surgery during their ICU stay, we observed a high incidence of SSI, affecting more than one in five patients. This burden substantially exceeds that reported in most non-ICU surgical populations and underscores the profound susceptibility of critically ill patients exposed to surgical interventions in an already high-risk environment. Despite widespread use of perioperative antibiotics, SSI occurrence was strongly driven by surgical contamination severity and, notably, by the complete absence of antimicrobial administration during surgery, rather than by deviations from standard SAP guidelines. These findings challenge the direct transposability of conventional SAP recommendations to the ICU setting and suggest that the pathophysiology of SSI in critically ill patients may differ fundamentally from that of elective surgical populations, warranting ICU-specific risk stratification and prophylactic strategies. Our study highlights several important points regarding SSI and SAP in ICU patients. The incidence of SSI was 21%, which was higher than the incidences reported for non-ICU patients (0.5–10%) [1]. However, previous studies reported a 23% incidence of SSI in severe trauma patients [21] and in tertiary peritonitis patients admitted to ICU [8]. Our results are consistent with current literature and confirmed that ICU patients are at high risk for SSI. Nevertheless, the PATOS-stratified sensitivity analysis suggests that SSI risk is most likely attributable to case-mix complexity, notably the high prevalence of highly contaminated (CDC-SWC ≥ 3) and PATOS-positive procedures. Among PATOS-negative patients receiving appropriate narrow-spectrum prophylaxis, SSI rates approached 8%, directly comparable to non-ICU general surgery benchmarks [1, 21]. Our findings also suggest that SSI is associated with impaired outcomes, notably prolonged hospital stay, underlining the susceptibility of ICU patients and the need to define preventive strategies. In the confounder-adjusted analysis, the elevated SSI risk in patients receiving broad-spectrum SAP (OR 3.06) presumably reflects residual confounding by indication rather than causality [22, 23]. However, patients receiving no SAP had significantly higher SSI risk (OR 2.49, P = 0.048), compared to those receiving narrow-spectrum SAP. In this critically ill population, the absence of any perioperative prophylaxis, even guideline-concordant, is associated with impaired outcomes, suggesting that the prophylactic action of SAP might be of greater importance here. In stratified analysis on admission diagnosis, the adverse effect of the absence of any perioperative prophylaxis seems more important in trauma patients when compared with non-trauma patients. These new findings highlight a critical discrepancy between existing prophylactic frameworks and the biological susceptibility of critically ill patients, suggesting that strict adherence to standard SAP recommendations may be insufficient in selected ICU populations. Our findings are in contrast with findings in general surgery where strict adherence in SAP guidelines typically improves outcomes [24]. The increased susceptibility to infection of ICU patients, likely related to immunosuppression, presence of invasive devices, organ dysfunction, and frequent exposure to broad-spectrum antimicrobials, may partly explain our findings and suggests that the risk–benefit balance of SAP in this population may differ from that observed in general surgery patients [25]. Dedicated randomized controlled studies may be required to further evaluate this issue. Of note, 10% of the entire cohort did not receive their ongoing antimicrobial therapy during surgery, highlighting the need to improve communication between ICU and operating room teams. This study has several notable strengths. Its multicenter design, including both academic and general ICUs, provides a comprehensive overview of SSI risk factors and SAP practices in critically ill patients. By capturing a wide spectrum of patient profiles, surgical indications, and antimicrobial strategies, the study reflects real-world practice and enhances the external validity and generalizability of the findings. In addition, data collection was standardized across participating centers, limiting inter-center variability and improving the internal consistency of the analyses. Several limitations should also be acknowledged. Owing to its retrospective observational design, the study is susceptible to missing data, unmeasured confounding, and information bias. As a result, some causal relationships, particularly regarding the effect of SAP strategies on SSI occurrence, cannot be definitively established. Although efforts were made to harmonize data collection across centers, residual heterogeneity in clinical practice and documentation may persist. Next, more than half of the cohort consisted of trauma patients, representing a specific case mix that may influence SSI risk. Also, postoperative antimicrobial therapy may have treated early SSIs, therefore leading to an underestimation of their true incidence. Finally, although crude SSI rates varied across participating centers, we used GEE with center-level clustering to appropriately account for within-center correlation. However, this approach does not eliminate the observed between-center heterogeneity in baseline SSI risk, which may reflect differences in practices or case-mix factors not captured in our adjustment set. Center-level variations therefore remain a potential source of residual confounding. Conclusion In this multicenter cohort of critically ill patients undergoing surgery during their ICU stay, SSIs were frequent, affecting more than one in five patients and confirming SSI as a major and underappreciated complication in this population. PATOS and antimicrobial spectrum stratification showed that overall SSI incidence was largely explained by case-mix complexity and surgical contamination severity. After adjustment for center and key confounders, the absence of any perioperative antimicrobial administration was independently associated with SSI occurrence when compared with administration of a narrow-spectrum SAP. These findings suggest that, in general, the main principles of standard SAP remain valid, and failure to administer prophylaxis (i.e., not using antimicrobials) is a key preventable cause of SSI. Further research is needed to establish whether SAP should be administered to ICU patients undergoing interventions for which the SAP guidelines recommend against antibiotic prophylaxis. Together, our results highlight a critical gap between existing guidelines and real-world ICU practice and underscore the need for ICU-specific prophylactic strategies, improved coordination between ICU and operating room teams, and prospective studies to define optimal antimicrobial approaches tailored to this high-risk population. Institutional review board statement This study was conducted in accordance with the Declaration of Helsinki and approved by an institutional review board (Ethics Committee for Research in Anesthesia and Intensive Care IRB 00010254–2023-071). Supplementary Information Additional file 1. Additional file 1.