Risk factor identification and contribution to surgery-specific surgical site infection CURRENT STATUS: POSTED

Background: To identify surgical procedures, thoracic and and quantify what the marginal contribution of explaining the future occurrence of SSI is of risk factors specific to each group of surgeries. Methods: Retrospective data from serial point-prevalence studies, performed at the Erasmus MC University Medical Center, Rotterdam, The Netherlands were used, together with medical data, from 3.250 surgical procedures, during the period January 2013 to 29 June 2014. Common risk factors across three groups of surgical procedures were identified by scanning literature and univariate analysis. A multivariate forward-step logistic regression model was used to identify the marginal contribution of the risk factor specific to each group of surgical procedure using the increase in the Gini coefficient. Results: For digestive system surgical procedures, antibiotic use, temperature, smoking status, age, CRP, thrombocyte, during of surgery and surgical urgency are risk factors for SSI, where the last four were specific to the digestive system group of surgical procedures and increased the Gini coefficient by 9.5% (0.63 to 0.69). Preoperative length of stay, antibiotic use and Leukocyte are risk factors for orthopaedic surgical procedures. Temperature, age and the use of antibiotics were significant for thoracic surgical procedures. Conclusion: ASA class, body mass index (BMI), Preoperative length of stay, diabetes, antibiotic use, age, Leukocyte, temperature and smoking status as general risk factors common to digestive, orthopaedic and thoracic surgical procedures using existing literature and univariate logistic regression. Risk factors for SSI, specific to a surgical procedure group can increase the ability to explain the future occurrence of SSI by 9.5%. The models developed in this study can aid healthcare workers to improve preoperative counselling and help identify potentially modifiable risk factors.

measured in terms of increased length of stay in hospital, additional (surgical) procedures required, increased morbidity and mortality as well as in economic terms [3].
The use of surgical antimicrobial prophylaxis (SAP) has altered the outcome of surgery dramatically with early studies already showing odds of SSI being less than half after SAP was administered [4].
Risk factors relating to the patient, procedure and the environment also alter the odds of developing an SSI. Research has been done to identify risk factors for SSI with the hope that it will lead to preemptive action to reduce the incidence rate of SSI [5][6][7][8][9][10][11].
Patient related risk factors for SSI, such as obesity, diabetes, surgery duration and the American Society of Anaesthesiologists (ASA) score are risk factors for digestive system, thoracic and orthopaedic surgical procedures [15], [18][19][20][21][22][23][24][25][26][27][28]. Some risk factors are only relevant to certain types of surgical procedures, but the marginal contribution of these surgery specific risk factors to explain the occurrence of SSI has yet to be reported.
The ECDC reports the SSI surveillance data in 10 national health safety network (NHSN) categories, which combines ICD-10-CM surgical procedure codes together [1]. The literature, concerning risk factors for SSI, is segmented by the same categories and is done for good reason. Segmentation makes it possible to find useful and relevant risk factors unique to each segment. For this study, we focus on digestive system, thoracic and orthopaedic surgical procedures. Digestive system surgical procedures are more prone to SSI and the severity vary. The occurrence of SSI after thoracic and orthopaedic surgeries are both relatively low, but the severity of an SSI after a thoracic is much higher compared to orthopaedic surgeries. It has yet to be shown which risk factors are common across these three different group of surgical procedures and what the benefit is of including risk factors unique to them.
Risk factors common across three groups of surgeries are identified from existing literature and the marginal contribution of the risk factors unique to each of them when predicting SSI is quantified.

Methods
The Erasmus MC University Medical Center in Rotterdam, the Netherlands (Erasmus MC), is the largest university medical hospital in the Netherlands with more than 1.300 beds. [21] We used the verified output from the infection control practitioners (ICP) for the SSI output variable.
Weekly (alter two-weekly) prevalence surveys were performed during January 2013 until June 2014 using a semi-automated algorithm [21,22]. In short, an algorithm was used to calculate a nosocomial infection index (NII) which was then verified by infection control practitioners (ICP) in case of a positive outcome to determine whenever HAI was present or not. All patients with a NII > 7 are verified by an ICP, and a definite outcome SSI was concluded by the ICP using the electronic patient data system.
We included 3,250 surgical procedures, performed on 2,929 distinct patients. Data were extracted from a centralised database, containing cross-departmental data, and clinical synopsis reports, infectious disease consultation reports, laboratory results and imaging reports. Surgeries were included if they form part of the three types of surgeries under investigation in this study and had a point prevalence study within 30 days after the surgery took place. If the surgery took place within 30 days after an included surgery, then the latter surgery was excluded.
The data used for this study were anonymized in accordance to the Dutch Personal Data Protection Act (WBP). Approval of the Medical Ethical Committee was obtained (MEC-2018-1185). There was no need to obtain informed consent because this is a fully anonymous retrospective study and no further interaction with the patients were necessary.
Risk factors relating to digestive system surgical procedures, thoracic surgery and orthopaedic procedures were identified from existing literature using the corresponding medical subject headings (MeSH). For a risk factor to be included in this study, it had to be significant in the multivariate analysis of the literature considered. The risk factors which are common to all three groups of surgeries, according to our literature scan, are called "general risk factors".

Statistical analysis
During the initial analysis of the data, a missing value analysis is performed. Using IBM SPSS Statistics for Windows, Version 25, we performed t-tests to determine if the differences in the averages of variables with missing values and those without were statistically significant. These tests, together with Little's MCAR test, convinced us that the missing values were not completely randomly missing and that we could not make use of more simple imputation methods. We chose to use conditional Markov chain Monte Carlo (MCMC) with multiple imputations for the imputation process. [23,24] Tests for significant differences in the occurrence of SSI across all three groups of surgeries were performed using the Wilcoxon-Kruskal-Wallis test for continuous variables and Pearson chi-square test for categorical variables. In this study we consider variables to be statistically significant if the corresponding p-value is less than 0.05.
In order to build a prognostic prediction model for SSI, Hosmer et al. suggests fitting a univariate logistic regression model to each variable separately and if the p-value is less than a certain p-value, 0.1 is this case, then consider the variable good enough to include in the multivariate logistic regression model. [26] The data were split according to the different groups of surgeries in order to evaluate them separately. A univariate analysis was performed for each of the three groups of surgeries using all the variables identified across the three different groups. Significant variables in the univariate analysis are then also added to the list of variables associated with each group of surgery, together with the variables identified from the literature. This results in an extended list of general risk factors as more risk factors are common across the three groups of surgeries.
A multivariate logistic regression model was applied using a forward stepwise approach.(28) The multivariate model was first built using the general risk factors and then, using the result as a starting point, the surgery group specific risk factors were added to the model in order of the Akaike information criterion (AIC) until convergence is reached. In this case, we chose the conversion of the model to imply that there are no additional variables which can be added which will be statistically significant with a p-value of less than 0.05 or an AIC of 3.8415.
Each variable contributes information about the SSI outcome variable, which in turn reduces the error or residual of the model. For a logistic regression model, the residual or error can be stated in terms of deviance. To determine the importance of the different variables in the final model, we calculate the difference in model deviance caused by the exclusion of each variable. Model performance are determined using the Gini coefficient after each step of the multivariate modelling process and the difference is reported as the marginal contribution of surgery group specific risk factors for this study. [27]

Results
In total, 55 risk factors were identified across the three types of surgical procedures from literature [15], [18][19][20][21][22][23][24][25][26][27][28]. Of the risk factors identified, ASA class, body mass index (BMI), Preoperative length of stay and diabetes were identified as general risk factors. The complete list of risk factors identified from literature can be found in Table 6 as part of the Appendix. There are 24 risk factors unique to digestive system surgical procedures, 14 to orthopaedic procedures and 15 to thoracic surgeries. Of the 55 risk factors identified from the literature, 23 could be extracted from our own data to use in this study. These 23 risk factors included in this study, together with the number of surgery groups they occurred in (N), are shown in Table 1. The test for differences for risk factors in Table 1 are shown in Table 2. The variables indicating the type of surgery was statistically significant but there was no significant difference between the ASA classes of patients for which SSI occurred. Diabetes and BMI were also not significant. Preoperative length of stay was the only general risk factor which was statistically significant. Antibiotics use, temperature and CRP were significant during initial analysis across the groups of surgeries with pvalue of less than 0.001. The significant univariate results of digestive system, orthopaedic and thoracic surgical procedures are shown in Table 3 and the full results can be found in Table 5, Table 6 and Table 7 in the Appendix.
Antibiotic use, age, Leukocyte, temperature and smoking status were added to the list of general risk factors after being found statistically significant in the univariate analysis -increasing the number of general risk factors to 9. BMI and diabetes were two risk factors identified as general Risk factors from the literature but were not found significant in any of the three univariate analyses in our own study. The general risk factors which were significant for digestive system surgical procedures were antibiotic use, temperature, smoking status and age according to the multivariate analysis results shown in Table 4. The additional significant risk factors, indicated in bold, which were unique to the  The resultant equations of the multivariate logistic regression models are provided in Equation 1 to Equation 3. The abbreviations of the variable names are provided in Table 1.
We summarise the findings concerning the statistical significance of the risk factors by surgical procedure in Table 5. The stage of significance indicates the last stage of the analysis where this risk factor was statistically significant for corresponding group of surgeries. The source indicates where the reason why each risk factor was considered for each group of surgeries. In the last section we see the risk factors we found in literature but were not significant for any of the groups of surgeries for both the univariate and multivariate analysis. Risk factors which were significant in the multivariate analysis, but not the univariate analysis are shown in the first part of Table 5. This study produced three prognostic prediction models (Equation 1 -3) for the future occurrence of SSI within a 30 days period after undergoing one of the three group of surgical procedures. Using existing literature and the results of the univariate analysis, we identified general risk factors across three groups of surgical procedures and showed that the marginal contribution for models predicting SSI could be up to 9.5% in terms of discriminatory power by including surgery group specific risk factors. We also showed that this benefit can be measured in terms of reduced model error and ability to discriminate using the Gini as performance measure.
The initial results of the missing data analysis revealed that the variables relating to the CRP and Leukocyte levels have a sizeable amount of missing values. The difference in the preoperative length of stay led us to believe that most of these missing values were caused by patients who did not stay sufficiently long in the hospital to have the laboratory tests before the surgical procedure took place.
A possible explanation could be that most of the missing values were due to short-stay elective or emergence surgery patients or the absence of signs of infections which would lead to not requesting this chemical test. These laboratory tests are usually only done in case the patient is suspected of already having an infection. Our conclusion was that final model might be prone to overestimate the probability of SSI since the missing values were imputed using data from patients with greater risk of infection.
We used a 30-day outcome period in which we observe if an SSI was present or not, but according the CDC definition, this outcome period should be 1 year for surgical implantation procedures. Since our data only spans over 18 months, it was not possible to use a 12-month outcome windows for all surgical implantation procedures, which is a limitation of this study. Another limitation is that for all three groups of surgical procedures, the use of antibiotics is predictive of occurrence of SSI. This could mean that the patient is already fighting off some kind infection and is receiving treatment for a present condition which has nothing to do with the surgery under investigation. This study was performed in as a single centre study, which limits the generalisability of this study. Future research can repeat this research on multiple patient populations in order produce more generalisable results.
The predictive models were designed with the aim to use it to investigate more variables as potential risk factors for the future occurrence of SSI, 30 days after a surgical procedure. The administration of prophylaxis and the optimal timing thereof is an important risk factor however, these data were not available, but in future we recommend that these data should also be included in similar studies.
Although some research has been done in this regard, to our knowledge, it has not been done for this patient population and confounders identified in this study.
The length of stay before a surgical procedure had mixed outcomes as a risk factor in the literature.
Abuzaid et al. found it to be not significantly associated with SSI after isolated coronary artery bypass grafting, while Sang et al. found it to be highly significant [5,7]. Preoperative length of stay can also be a proxy for the severity of a patient's condition, if we assume that patients in worse conditions are more likely to have been in hospital for a longer period before undergoing a surgical procedure. This variable was significant in both the univariate and multivariate case for the orthopaedic surgeries group. The cause of the increased lengths of stay could be investigated as a potential modifiable risk factor.
Another opportunity for future research is to investigate which risk factors are predictive for the occurrence of SSI over different time periods. Doing this will enable healthcare workers to identify which risk factors explain the occurrence of SSI soon after surgery and which are better explain the occurrence of SSI nearing the end of the 30-day period or even later for implantation surgeries. This can help set guidelines to determine the vigilance necessary to mitigate the risk of SSI on a patient level.

Conclusion
We identified ASA class, body mass index (BMI), Preoperative length of stay, diabetes, antibiotic use, age, Leukocyte, temperature and smoking status as general risk factors common to digestive, orthopaedic and thoracic surgical procedures using existing literature and univariate logistic regression. We used a systematic approach to quantify the marginal contribution of the surgical procedure group specific risk factors using the results of a serial prevalence study and retrospective medical data. For digestive system surgical procedures, it is possible to increase the predictive power to explain the future occurrence of SSI by 9.5% by including CRP, thrombocyte, duration of the surgery and if the surgery will be elective or not. Pre-operative length of stay, antibiotic use and Leukocyte were significant general risk factor for the orthopaedic group of surgeries. Temperature, age and antibiotic use were significant general risk factors for the thoracic group of surgical All authors read and approved the manuscript.

Availability of data and material
The data that support the findings of this study are available from Erasmus MC but restrictions apply to the availability of these data, which were used under license for the current study, and so are not