- Email: [email protected]
The role of preoperative blood parameters to predict the risk of surgical site infection
Accepted Manuscript The role of preoperative blood parameters to predict the risk of surgical site infection Edin Mujagic, Walter R. Marti, Michael Coslovsky, Jasmin Zeindler, Sebastian Staubli, Regula Marti, Robert Mechera, Savas D. Soysal, Lorenz Gürke, Walter P. Weber PII:
S0002-9610(17)31336-3
DOI:
10.1016/j.amjsurg.2017.08.021
Reference:
AJS 12489
To appear in:
The American Journal of Surgery
Received Date: 23 June 2017 Revised Date:
17 August 2017
Accepted Date: 21 August 2017
Please cite this article as: Mujagic E, Marti WR, Coslovsky M, Zeindler J, Staubli S, Marti R, Mechera R, Soysal SD, Gürke L, Weber WP, The role of preoperative blood parameters to predict the risk of surgical site infection, The American Journal of Surgery (2017), doi: 10.1016/j.amjsurg.2017.08.021. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Background Routine preoperative blood work is not recommended but selected biochemical markers may predict the risk of surgical site infection (SSI). This study examines the association between preoperative biochemical markers and the risk of SSI.
RI PT
Methods This observational cohort study, nested in a randomized controlled trial, was conducted at two tertiary referral centers in Switzerland.
SC
Results
122 (5.8%) of 2093 patients experienced SSI. Preoperative increasing levels of albumin (OR
M AN U
0.93), CRP (OR 1.34), hemoglobin (OR 0.87) and eGFR (OR 0.90) were significantly associated with the odds of SSI. The same accounts for categorized parameters. The highest area under the curve from ROC curves was 0.62 for albumin. Positive predictive values ranged from 6.4% to 9.5% and negative predictive values from 94.8% to 95.7%. The
TE D
association of CRP, mildly and moderately decreased eGFR and hemoglobin with the odds of SSI remained significant on multivariate analysis.
Conclusions
AC C
EP
Our results do not support generally delaying elective surgery based on preoperative blood results. However, it may be considered in situations with potentially severe sequelae of SSI.
ACCEPTED MANUSCRIPT 1
The role of preoperative blood parameters to predict the risk of surgical site infection
RI PT
Edin Mujagica*, Walter R Martib*, Michael Coslovskyc, Jasmin Zeindlera, Sebastian Staublia, Regula Martib, Robert Mecheraa, Savas D Soysala, Lorenz Gürkea, Walter P Webera *Contributed equally
AC C
EP
TE D
Corresponding author: Edin Mujagic, MD Department of Surgery University Hospital of Basel Spitalstrasse 21 4031 Basel Switzerland Email: [email protected] Phone: +41613286116 Fax: +41612657512
M AN U
SC
a Department of Surgery, University Hospital of Basel, Spitalstrasse 21, 4031 Basel, Switzerland [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected] b Department of Surgery, Kantonsspital Aarau, Tellstrasse, 5001 Aarau, Switzerland [email protected]; [email protected]; [email protected] c University of Basel, University Hospital, Department of Clinical Research, Clinical Trial Unit, Spitalstrasse 12, 4031 Basel, Switzerland [email protected]
ACCEPTED MANUSCRIPT 2
Abstract Background Routine preoperative blood work is not recommended but selected biochemical markers may
preoperative biochemical markers and the risk of SSI.
Methods
RI PT
predict the risk of surgical site infection (SSI). This study examines the association between
two tertiary referral centers in Switzerland.
M AN U
Results
SC
This observational cohort study, nested in a randomized controlled trial, was conducted at
122 (5.8%) of 2093 patients experienced SSI. Preoperative increasing levels of albumin (OR 0.93), CRP (OR 1.34), hemoglobin (OR 0.87) and eGFR (OR 0.90) were significantly associated with the odds of SSI. The same accounts for categorized parameters. The highest area under the curve from ROC curves was 0.62 for albumin. Positive predictive
TE D
values ranged from 6.4% to 9.5% and negative predictive values from 94.8% to 95.7%. The association of CRP, mildly and moderately decreased eGFR and hemoglobin with the odds
Conclusions
EP
of SSI remained significant on multivariate analysis.
AC C
Our results do not support generally delaying elective surgery based on preoperative blood results. However, it may be considered in situations with potentially severe sequelae of SSI.
Key words
Surgical site infection; Preoperative blood testing; Predictive value; biochemical markers
ACCEPTED MANUSCRIPT 3
Introduction Surgical site infections (SSI) account for up to 38% of all nosocomial infections in inpatients and are considered the most common type of nosocomial infections among surgical
RI PT
patients.(1, 2) In spite of numerous measures to prevent SSI, up to 5% of all patients
undergoing surgery are reported to experience SSI with a substantial increase in morbidity and mortality.(3-5) Furthermore, SSI are associated with prolonged hospital stay and cost.(6-
SC
8) The identification of patients at increased risk of SSI before surgery is of importance to endorse pre- and perioperative SSI prevention measures or delay surgery to minimize risk
M AN U
where indicated.
Routine preoperative laboratory blood testing was standard practice in the past, but is no longer recommended.(9-14) For example, it was shown that only hemoglobin levels, renal function and electrolytes were associated with postoperative complications, while normal blood tests did not reduce their likelihood. Furthermore, only 0 to 3 percent of abnormal test
TE D
results changed preoperative management.(10) However, there is evidence for the following selected preoperative laboratory tests to potentially predict the risk of postoperative complications including SSI. First, preoperatively
EP
elevated c-reactive protein (CRP) levels have been reported to be associated with the risk of postoperative surgical site and remote nosocomial infections (15-19) as well as with other
AC C
major postoperative complications in cardiac surgery, including renal dysfunction and death.(19, 20) Second, total white blood cell count (WBCC) has been reported to be associated with postoperative complications in surgical patients. One report showed a linear correlation of increasing preoperative WBCC values, even within the normal limits, with postoperative complications, but did not focus on SSI.(21) Third, malnutrition is a common problem in surgical patients, especially in such subsets as the elderly and chronically ill. It is known to adversely affect outcomes in those patients.(22, 23) Albumin has been found to be the most reliable indicator of the nutritional status.(24) Hypoalbuminemia as a marker for
ACCEPTED MANUSCRIPT 4 malnutrition has been established as an independent risk factor for SSI.(25-28) Fourth, end stage renal insufficiency is a known risk factor for surgical site infections.(29, 30). Although chronic kidney failure of earlier stages has been shown to be associated with an increased risk for SSI (31), that association has been shown to be less pronounced. Finally,
RI PT
preoperative anemia has been associated with risk of SSI. However, it was not identified as an independent predictor.(32)
The aim of this study was to examine whether a set of routine preoperative blood tests
SC
including WBCC, CRP, Creatinine (eGFR), albumin and hemoglobin predict the risk of SSI either alone or in combination. If so, these parameters could help identify patients at
M AN U
increased risk for SSI, thereby allowing to reinforce SSI prevention measures or to delay
AC C
EP
TE D
surgery in order to optimize the conditions of the patients and procedures.
ACCEPTED MANUSCRIPT 5
Methods Setting This observational cohort study was nested within a bicenter randomized controlled trial (RCT)(33) that was designed to evaluate the optimal timing of surgical antimicrobial
RI PT
prophylaxis. Data were prospectively collected between February 22nd, 2013 and July 31st, 2015 at the University Hospital of Basel and the hospital of Aarau, two tertiary referral
centers in Switzerland. The preoperative blood test parameters hemoglobin, creatinine,
SC
albumin, CRP and white blood cell count were collected in those patients considered to have an indication for preoperative blood testing as per clinical standards. The indication for
M AN U
preoperative blood testing at both sites was based on the general condition of the patient, the complexity of the procedure and individual judgment of the physicians on the ward. The analysis included all patients for whom the above-mentioned preoperative blood parameters were available from blood taken within 3 days before surgery. Hence, this is a
committees was obtained.
Patients
TE D
retrospective analysis of a prospectively collected dataset. Approval by the local ethics
The study protocol of the RCT has been published.(34) Briefly, all inpatients aged 18 years
EP
or older undergoing general, oncologic, vascular and orthopedic trauma procedures with an indication for surgical antimicrobial prophylaxis (cefuroxime, in colorectal surgery combined
AC C
with metronidazole) according to clinical standards were included. These standards were based on CDC guidelines for surgical wound classification.(2) Exclusion criteria included outpatient surgery, contraindication for cefuroxime and/or metronidazole, preexisting antibiotic therapy within 14 days prior to surgery, cognitive impairment, combined operations including other than the above specified surgical subspecialties and emergency procedures with planned incision within 2 hours after indicating the procedure. All patients in this study provided written consent to participate.
ACCEPTED MANUSCRIPT 6
Endpoints The primary endpoint was the occurrence of SSI within 30 days after surgery. SSI is defined and categorized according to CDC criteria.(2) In hospital SSI were diagnosed by the surgical team according to clinical standards. In addition, to increase sensitivity, inpatients were
RI PT
regularly seen by members of the RCT study team. Post discharge follow up was performed by trained members of the RCT study team by contacting the patients over the phone. When a post discharge SSI was suspected, hospital charts were reviewed and primary care
board certified infectious diseases specialist.
Statistical analysis
SC
physicians were contacted for additional information. All suspected SSI were validated by a
M AN U
The analysis included all patients with a complete set of pre-operative blood values for WBCC, CRP, Creatinine, albumin and hemoglobin and completed 30 day follow up. Descriptive statistics are provided for demographic and baseline characteristics and values. Categorical variables are summarized as frequency and percentage. Continuous variables
TE D
are summarized as mean and standard deviation (SD) and median and interquartile range (IQR) for blood values due to highly skewed distributions. In case of comparisons between groups, Fisher’s exact test was used for categorical variables, and t-tests or Kruskal-Wallis
EP
tests for continuous variables, as appropriate.
The association of each of the blood parameters with SSI was analyzed using logistic
AC C
regression models. In a first step, the association of the blood value as a continuous variable was investigated. Receiver operator characteristic (ROC) curves were plotted for each potential predictor, and the area under the curve (AUC) calculated – higher AUC values indicate better discriminative abilities for the potential predictor and an AUC value of 0.5 suggests a predictive ability not different from chance. Due to the wide range of hemoglobin and eGFR values, results are presented per 10-unit increases. For CRP, values are logtransformed and thus results refer to 10-fold increases in CRP levels. In a second step, the blood parameters were categorized into groups according to commonly used cut offs. Normal ranges were 35-52g/L for albumin, <10mg/L for CRP (SI conversion
ACCEPTED MANUSCRIPT 7 factor 9.524), 140-180g/L for hemoglobin in men, 120-160g/L for hemoglobin in women and 3.5x109/L- 10.0x109/L for leukocytes. For eGFR values, the CKD-EPI equation(35) was used and ≥90mL/min/1.73m2 was considered normal renal function, 60-89mL/min/1.73m2 mildly decreased, 30-59mL/min/1.73m2 moderately decreased, 15-29mL/min/1.73m2 severely
RI PT
decreased and <15mL/min/1.73m2 renal failure. For parameters with three categories, the “low” or the “high” category was combined with the “normal” category when they were not different in terms of outcome. This was the case for albumin, leukocytes and hemoglobin.
SC
Hence, these three parameters were grouped into two categories. Logistic regression
models were fitted with the categorized parameters as potential predictors. In addition, the
M AN U
diagnostic performance of the parameter is assessed by providing the sensitivity, specificity and positive and negative predictive values at the relevant cut point. To assess whether the predictive value of a parameter differed depending on the timing of surgical antimicrobial prophylaxis (SAP) (early SAP in the anesthesia room vs. late SAP in
TE D
the operating room) the timing and the interaction with the blood parameter categories were included in the model, followed by subgroup analyses, where the effect of SAP timing was examined in each blood parameter group. A similar approach was taken to examine whether a parameter’s predictive value differed depending on the surgical department
EP
(general/oncologic, orthopedic trauma or vascular).
AC C
Finally, to assess the independent effect of each parameter, adjusted for all other parameters, on the odds of SSI, a multiple logistic regression model was fit. Firth’s penalized likelihood was used for the multiple regression models due to low separability of rare subgroups(36). All the above-mentioned blood parameters were included, categorized by the above described cut points and with their interaction with timing of SAP. Non-significant interaction terms were removed from the model. For all logistic regression models, odds ratios and 95% confidence intervals are presented. The alpha level for all analyses was taken as 0.05. All analyses were performed using R (version 3.4.0; 2017)(37).
ACCEPTED MANUSCRIPT
RI PT
8
Results
A total of 5175 patients were included in the analysis for the RCT and hence assessed for
SC
the availability of preoperative levels of the blood parameters of interest. A complete set of preoperative WBCC, CRP, albumin, creatinine and hemoglobin measurements was present
M AN U
for 2480 of 5175 patients (48%). Demographics of all patients of the RCT, both with or without preoperative blood tests available, are presented in table 1. The 2480 patients with preoperative blood tests available were older, had higher ASA scores, had more secondary diagnoses, were more likely to be women, diabetics and on dialysis. Further characteristics are shown in table 1.
TE D
The current study population consisted of 2093 of the 2480 (84.4%) patients with available blood tests that had 30 day follow up. The 30 day SSI rate was higher in patients with preoperative blood work than in the group without (5.8% vs. 4.5%, p=0.044).
EP
Preoperative levels of the blood parameters by 30 day SSI status are shown in table 2. Analyzing the blood parameters as continuous variables using univariable logistic
AC C
regression, an increase in levels of all parameters but leukocytes was associated with significantly increased (for CRP) or decreased (for albumin, eGFR and hemoglobin) odds of surgical site infections (figure 1). Figure 1 also shows ROC curves for all 5 potential blood parameters. The highest area under the curve (AUC) value was 0.62 for albumin. Based on commonly used cut offs, low albumin, low hemoglobin, mildly and moderately decreased eGFR resulted in highly significant increases in the odds for SSI. High CRP increased the odds significantly whereas leukocytes and more severely decreased eGFR
ACCEPTED MANUSCRIPT 9 were not associated with significant changes in the odds for SSI (Figure 2). Measures of performance were calculated for all parameters categorized to two levels: Sensitivities for albumin, CRP, hemoglobin and leukocytes were 37.7, 34.4, 48.4 and 75.4%, respectively. Specificities were 76.2, 74.3, 71.4 and 32.0%, respectively. Positive predictive values were
RI PT
8.9, 7.7, 9.5 and 6.4%, respectively whereas negative predictive values were 95.2, 94.8, 95.7 and 95.5%, respectively.
A significant interaction between a categorized blood parameter and SAP timing was found
SC
for CRP (p<0.001) (table 3). The subgroup analysis showed that the odds ratio of SSI was 1.96 (95% CI 1.23; 3.17) for patients who had late SAP compared to early SAP if their CRP
M AN U
levels were normal. On the other hand, patients with high CRP levels had an OR of 0.49 (95% CI 0.25; 0.94) with late SAP compared to early SAP.
For the other blood parameters, interaction terms were not significant even though there were differences in the direction of odds ratios by timing of SAP for some of the subgroups. Specifically, in patients with normal or severely decreased eGFR, the odds of having SSI
TE D
were reduced when receiving SAP late in the operating room whereas in those with mildly and moderately decreased eGFR, the odds of having SSI were increased when receiving SAP late in the operating room.
EP
When performing subgroup analyses of the categorical blood parameters per surgical
AC C
department including interaction tests, there is a significant interaction between surgical department and eGFR only (p=0.013). In practice, in general, oncologic and orthopedic trauma surgery, decreased eGFR increased the odds of having SSI whereas in vascular surgery, it decreased the odds of having SSI (table 4). However, within the vascular surgery subgroup, the 95% confidence intervals for the odds ratios for SSI in each of the eGFR categories are wide and cross 1 significantly. Table 5 shows the results of the multiple logistic regression model. The odds ratios for SSI with mildly and moderately decreased eGFR, high CRP and low hemoglobin remained significant. Also, the interaction between CRP levels and timing of SAP remained significant
ACCEPTED MANUSCRIPT 10 (p-value of the interaction term 0.003). To assist interpretation of this interaction term, subgroup analysis based on CRP categories show that the odds of SSI with late SAP remain increased in those patients with normal CRP (OR 1.89, 95% CI 1.18; 3.08) and remain decreased in those with high CRP (OR 0.56, 95% CI 0.29; 1.07). Table 5 also shows that the
AC C
EP
TE D
M AN U
SC
RI PT
interaction of eGFR and surgical department remained significant (global p-value 0.0189).
11
ACCEPTED MANUSCRIPT
Discussion To our knowledge, this is the first study that specifically addresses the association between the preoperative levels of a set of five routine blood tests and the incidence of SSI. In 2093 patients that underwent general, oncologic, trauma and vascular surgery procedures, the
RI PT
odds of having SSI were significantly associated with preoperative albumin, CRP and
hemoglobin levels. The odds of SSI were significantly increased by mildly and moderately decreased kidney function. The lack of an association for severely decreased kidney function
SC
and kidney failure is most likely explained by the very limited number of patients with
severely decreased kidney function and kidney failure, as suggested by the wide 95%
M AN U
confidence interval. Finally, preoperative white blood cell count did not a have a significant impact on the odds of SSI. Importantly, despite the highly significant associations between several blood parameters and the risk of SSI, none of the tests showed very good abilities to predict SSI when using the blood test values as continuous variables as shown by the ROC curves. In this setting, albumin was the strongest predictor with an area under the curve of
TE D
0.62. Similarly, when looking at categorical values, the performance characteristics in terms of sensitivity and specificity were modest. However, the positive predictive values (PPV) ranging between 6% and 10% and the negative predictive values (NPV) of up to 96%
EP
showed better predictive abilities to identify patients at increased risk of experiencing SSI and even more so those at low risk, considering the overall SSI rate of 5.8% and its effect on
AC C
the PPV and NPV.
Another interesting finding of this study was the significant interaction between preoperative CRP level categories (normal vs. high) and the timing of surgical antimicrobial prophylaxis (SAP, early vs. late) in terms of SSI risk. The main results of the RCT showed no significant impact of timing of SAP on the risk of SSI.(33) Based on the present results, however, it could be argued that late administration of SAP before incision increases the odds of SSI in patients with normal preoperative CRP levels, while in patients with elevated preoperative CRP levels, the odds are decreased. Interpretation of this finding is difficult and speculative. Most importantly, this cohort study is exposed to all inherent bias of an observational study
12
ACCEPTED MANUSCRIPT
design even though it was nested in a RCT. It could be argued, however, that elevated CRP levels are indicative of an inflammatory state which could alter the effectiveness of cefuroxime and/or metronidazole. Furthermore, there was weak evidence of an interaction of renal function as calculated by
RI PT
eGFR and the timing of SAP in terms of SSI risk. Although not statistically significant, patients with normal renal function showed a trend towards lower rates of SSI with late
administration of SAP while in those with mildly and moderately decreased renal function, the odds of SSI were numerically higher with late administration. This association has not been
SC
previously described in the literature. A possible explanation for this weak interaction could be a change in the pharmacokinetics of the study drugs in patients with decreased renal
M AN U
function, inasmuch as SAP may require a longer time period to achieve the minimum inhibitory concentration in the soft tissue necessary to prevent SSI. Unfortunately, the data set did not allow to further test this hypothesis, for example by determining the percentage of chronic vs. acute kidney failure in patients with decreased eGFR levels.
TE D
The significant interaction seen between eGFR levels and surgical department is again almost impossible to interpret. In practice, our finding that in patients undergoing vascular surgery, decreased renal function as measured by eGFR levels seems to be protective of
EP
experiencing SSI, cannot be explained by any known facts. Although statistically significant, this finding could potentially be attributable to a type I error, eventually driven by the quite low
AC C
numbers of vascular surgery patients within this study (334 of a total of 2093). This hypothesis is backed by the wide 95% confidence intervals of the odds ratios for SSI within the vascular surgery subgroup. However, it remains a hypothesis. Our study has several strengths: First, the data of this analysis were collected in the setting of a RCT (33) in a prospective manner. The outcome of interest, surgical site infection, was validated by an independent infectious diseases specialist. Therefore, the study design allowed the generation of high quality data. Second, the large sample size of patients allowed to investigate the association between blood parameters and risk of SSI with high
13
ACCEPTED MANUSCRIPT
power. Finally, the participation of several different specialties increases the generalizability of our findings. However, we acknowledge the limitations of this study. First, although the dataset was prospectively collected within a RCT, the study was observational in nature. Second, the 5
RI PT
blood parameters were chosen based on existing evidence suggesting a possible association with the risk of SSI. We purposefully refrained from exploring other parameters for feasibility reasons. Third, the blood parameters were registered based on their availability in clinical routine, which parallels the selection of patients in need of routine blood work
SC
suggesting a more complex patient and procedure population. Indeed, compared to patients without blood work, the study population was older, had higher ASA scores, more secondary
M AN U
diagnoses, included a higher percentage of diabetics, patients on dialysis and was more likely to have undergone urgent surgery. Also, only procedures that include the administration of SAP as per clinical standards were analyzed here, suggesting a further selection of patients. It is possible that the administration of SAP masks the abilities of blood
TE D
tests to predict SSI risk. Accordingly, the role of blood workup to stratify patients undergoing procedures without SAP according to their risk of SSI could be different. However, the low incidence of SSI in most procedures without SAP suggests that any such role may be limited.
EP
Last, although this study includes patients undergoing a wide variety of general, oncologic, trauma and vascular procedures, there are other surgical specialties such as cardiac,
AC C
thoracic or neurosurgery that are not included in this study and hence, our findings potentially cannot be applied to such patient populations.
Conclusion
This study demonstrates strong associations between the levels of routine preoperative blood parameters and the risk of SSI. However, none of the blood parameters examined in this study showed striking predictive abilities in terms of surgical site infections and hence, the decision to postpone procedures to optimize patient factors that impact preoperative blood results cannot be generally recommended at a defined cut off. However,
14
ACCEPTED MANUSCRIPT postponement for optimization should be considered in the individual patient, particularly
Author Contributions
M AN U
SC
RI PT
when a potential SSI would have severe consequences.
Study conception and design: Mujagic, WR Marti, Weber
Acquisition of data: Mujagic, Brinkmann, Zeindler, Staubli, R Marti, Mechera, Soysal Analysis and interpretation of data: Mujagic, WR Marti, Coslovsky, Weber
TE D
Drafting of manuscript: Mujagic, WR Marti, Weber
Critical revision: Coslovsky, Brinkmann, Zeindler, Staubli, R Marti, Mechera, Soysal Final approval: Mujagic, WR Marti, Coslovsky, Brinkmann, Zeindler, Staubli, R Marti,
Funding
EP
Mechera, Soysal, Weber
AC C
This study - nested in a randomized controlled trial – was supported by the Swiss National Foundation, the “Kantonsspital Aarau”, the “Gottfried und Julia Bangerter-Rhyner-Stiftung”, the University of Basel, the “Freie Akademische Gesellschaft Basel”, and the “Nora van Meeuwen-Haefliger-Stiftung”. The authors report no conflicts of interest related to this study.
15
ACCEPTED MANUSCRIPT
References 1. Lewis SS, Moehring RW, Chen LF, et al. Assessing the relative burden of hospitalacquired infections in a network of community hospitals. Infect Control Hosp Epidemiol. 2013 Nov;34(11):1229-30. PubMed PMID: 24113613. Pubmed Central PMCID: PMC3977691.
RI PT
2. Mangram AJ, Horan TC, Pearson ML, et al. Guideline for prevention of surgical site infection, 1999. Hospital Infection Control Practices Advisory Committee. Infect Control Hosp Epidemiol. 1999 Apr;20(4):250-78; quiz 79-80. PubMed PMID: 10219875. 3. Coskun D, Aytac J, Aydinli A, Bayer A. Mortality rate, length of stay and extra cost of sternal surgical site infections following coronary artery bypass grafting in a private medical centre in Turkey. J Hosp Infect. 2005 Jun;60(2):176-9. PubMed PMID: 15866018.
SC
4. Coello R, Charlett A, Wilson J, et al. Adverse impact of surgical site infections in English hospitals. J Hosp Infect. 2005 Jun;60(2):93-103. PubMed PMID: 15866006.
M AN U
5. Kirkland KB, Briggs JP, Trivette SL, et al. The impact of surgical-site infections in the 1990s: attributable mortality, excess length of hospitalization, and extra costs. Infect Control Hosp Epidemiol. 1999 Nov;20(11):725-30. PubMed PMID: 10580621. 6. Poulsen KB, Bremmelgaard A, Sorensen AI, et al. Estimated costs of postoperative wound infections. A case-control study of marginal hospital and social security costs. Epidemiol Infect. 1994 Oct;113(2):283-95. PubMed PMID: 7925666. Pubmed Central PMCID: PMC2271539.
TE D
7. Vegas AA, Jodra VM, Garcia ML. Nosocomial infection in surgery wards: a controlled study of increased duration of hospital stays and direct cost of hospitalization. Eur J Epidemiol. 1993 Sep;9(5):504-10. PubMed PMID: 8307135.
EP
8. Boyce JM, Potter-Bynoe G, Dziobek L. Hospital reimbursement patterns among patients with surgical wound infections following open heart surgery. Infect Control Hosp Epidemiol. 1990 Feb;11(2):89-93. PubMed PMID: 2107250.
AC C
9. Chung F, Yuan H, Yin L, et al. Elimination of preoperative testing in ambulatory surgery. Anesth Analg. 2009 Feb;108(2):467-75. PubMed PMID: 19151274. 10. Smetana GW, Macpherson DS. The case against routine preoperative laboratory testing. Med Clin North Am. 2003 Jan;87(1):7-40. PubMed PMID: 12575882. 11. Garcia-Miguel FJ, Serrano-Aguilar PG, Lopez-Bastida J. Preoperative assessment. Lancet. 2003 Nov 22;362(9397):1749-57. PubMed PMID: 14643127. 12. Macpherson DS. Preoperative laboratory testing: should any tests be "routine" before surgery? Med Clin North Am. 1993 Mar;77(2):289-308. PubMed PMID: 8441296. 13. Turnbull JM, Buck C. The value of preoperative screening investigations in otherwise healthy individuals. Arch Intern Med. 1987 Jun;147(6):1101-5. PubMed PMID: 3592875.
16
ACCEPTED MANUSCRIPT 14. Kaplan EB, Sheiner LB, Boeckmann AJ, et al. The usefulness of preoperative laboratory screening. JAMA. 1985 Jun 28;253(24):3576-81. PubMed PMID: 3999339.
15. Shimizu T, Ishizuka M, Kubota K. The preoperative serum C-reactive protein level is a useful predictor of surgical site infections in patients undergoing appendectomy. Surg Today. 2015 Nov;45(11):1404-10. PubMed PMID: 25480421. Epub 2014/12/07. eng.
RI PT
16. Ortega-Deballon P, Duvillard L, Scherrer ML, et al. Preoperative adipocytokines as a predictor of surgical infection after colorectal surgery: a prospective survey. Int J Colorectal Dis. 2014 Jan;29(1):23-9. PubMed PMID: 24132530.
SC
17. Mohri Y, Miki C, Kobayashi M, et al. Correlation between preoperative systemic inflammation and postoperative infection in patients with gastrointestinal cancer: a multicenter study. Surg Today. 2014 May;44(5):859-67. PubMed PMID: 23722282.
M AN U
18. Moyes LH, Leitch EF, McKee RF, et al. Preoperative systemic inflammation predicts postoperative infectious complications in patients undergoing curative resection for colorectal cancer. Br J Cancer. 2009 Apr 21;100(8):1236-9. PubMed PMID: 19319134. Pubmed Central PMCID: PMC2676538. 19. Cappabianca G, Paparella D, Visicchio G, et al. Preoperative C-reactive protein predicts mid-term outcome after cardiac surgery. Ann Thorac Surg. 2006 Dec;82(6):2170-8. PubMed PMID: 17126130. Epub 2006/11/28. eng.
TE D
20. Kim DH, Shim JK, Hong SW, et al. Predictive value of C-reactive protein for major postoperative complications following off-pump coronary artery bypass surgery: prospective and observational trial. Circ J. 2009 May;73(5):872-7. PubMed PMID: 19282606. Epub 2009/03/14. eng.
EP
21. Amaranto DJ, Wang EC, Eskandari MK, et al. Normal preoperative white blood cell count is predictive of outcomes for endovascular procedures. J Vasc Surg. 2011 Nov;54(5):1395-403.e2. PubMed PMID: 21802244. Epub 2011/08/02. eng.
AC C
22. Sungurtekin H, Sungurtekin U, Balci C, et al. The influence of nutritional status on complications after major intraabdominal surgery. J Am Coll Nutr. 2004 Jun;23(3):227-32. PubMed PMID: 15190047. 23. Inoue Y, Miki C, Kusunoki M. Nutritional status and cytokine-related protein breakdown in elderly patients with gastrointestinal malignancies. J Surg Oncol. 2004 May 1;86(2):91-8. PubMed PMID: 15112251. 24. Sung J, Bochicchio GV, Joshi M, et al. Admission serum albumin is predicitve of outcome in critically ill trauma patients. Am Surg. 2004 Dec;70(12):1099-102. PubMed PMID: 15663053. 25. Bohl DD, Shen MR, Kayupov E, Della Valle CJ. Hypoalbuminemia Independently Predicts Surgical Site Infection, Pneumonia, Length of Stay, and Readmission After Total Joint Arthroplasty. J Arthroplasty. 2016 Jan;31(1):15-21. PubMed PMID: 26427941.
17
ACCEPTED MANUSCRIPT
26. Walls JD, Abraham D, Nelson CL, et al. Hypoalbuminemia More Than Morbid Obesity is an Independent Predictor of Complications After Total Hip Arthroplasty. J Arthroplasty. 2015 Dec;30(12):2290-5. PubMed PMID: 26148837.
RI PT
27. Hennessey DB, Burke JP, Ni-Dhonochu T, et al. Preoperative hypoalbuminemia is an independent risk factor for the development of surgical site infection following gastrointestinal surgery: a multi-institutional study. Ann Surg. 2010 Aug;252(2):325-9. PubMed PMID: 20647925. 28. Neumayer L, Hosokawa P, Itani K, et al. Multivariable predictors of postoperative surgical site infection after general and vascular surgery: results from the patient safety in surgery study. J Am Coll Surg. 2007 Jun;204(6):1178-87. PubMed PMID: 17544076.
SC
29. Moran-Atkin E, Stem M, Lidor AO. Surgery for diverticulitis is associated with high risk of in-hospital mortality and morbidity in older patients with end-stage renal disease. Surgery. 2014 Aug;156(2):361-70. PubMed PMID: 24957670.
M AN U
30. Gajdos C, Hawn MT, Kile D, et al. The risk of major elective vascular surgical procedures in patients with end-stage renal disease. Ann Surg. 2013 Apr;257(4):766-73. PubMed PMID: 22982978. 31. Ackland GL, Moran N, Cone S, et al. Chronic kidney disease and postoperative morbidity after elective orthopedic surgery. Anesth Analg. 2011 Jun;112(6):1375-81. PubMed PMID: 20807976.
TE D
32. Weber WP, Zwahlen M, Reck S, et al. The association of preoperative anemia and perioperative allogeneic blood transfusion with the risk of surgical site infection. Transfusion (Paris). 2009 Sep;49(9):1964-70. PubMed PMID: 19453989.
EP
33. Weber WP, Mujagic E, Zwahlen M, et al. Timing of surgical antimicrobial prophylaxis: a phase 3 randomised controlled trial. Lancet Infect Dis. 2017 Apr 03. PubMed PMID: 28385346.
AC C
34. Mujagic E, Zwimpfer T, Marti WR, et al. Evaluating the optimal timing of surgical antimicrobial prophylaxis: study protocol for a randomized controlled trial. Trials. 2014;15:188. PubMed PMID: 24885132. Pubmed Central PMCID: Pmc4040488. Epub 2014/06/03. eng. 35. Levey AS, Stevens LA, Schmid CH, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009 May 05;150(9):604-12. PubMed PMID: 19414839. Pubmed Central PMCID: PMC2763564. 36. 38.
Firth D. Bias reduction of maximum likelihood esteimates. Biometrika. 1993;80(1):27-
37. Team RC. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2016.
18
ACCEPTED MANUSCRIPT Table 1. Baseline characteristics by availability of blood work Characteristic n
Blood tests available 2480
No blood tests available 2695
Sex – number (%) Male Female
1214 (49.0) 1266 (51.0)
1588 (58.9) 1107 (41.1)
Age – years (mean (SD))
60.1 (18.7)
54.4 (18.2)
ASA‡ score (n (%)) 1 2 3 4
304 (12.3) 1284 (51.8) 847 (34.2) 44 (1.8)
628 (23.3) 1448 (53.7) 601 (22.3) 17 (0.6)
3.5 (2.6)
2.7 (2.3)
<0.001
RI PT
<0.001
SC
<0.001
Number of secondary diagnoses (mean (SD))
2425 (97.9) 47 (1.9) 4 (0.2)
2678 (99.4) 16 (0.6) 0 (0.0)
Wound class (n (%)) I II III IV
1998 (80.8) 270 (10.9) 162 (6.6) 43 (1.7)
2070 (76.9) 526 (19.5) 74 (2.7) 21 (0.8)
260 (10.5) 2220 (89.5)
220 (8.2) 2475 (91.8)
26.5 (6.0)
27.6 (7.6)
TE D
Diabetes (n (%)) Yes No
M AN U
Dialysis (n (%)) No dialysis Hemodialysis CAPD†
EP
BMI** (kg/m2) (mean(SD))
<0.001
<0.001
<0.001
0.005
Urgent procedure (n (%))
<0.001 <0.001
549 (22.1)
371 (13.7)
1931 (77.9)
2324 (86.3)
Duration of surgery (mean (SD))
108.1 (71.7)
108.2 (78.0)
0.937
30 day SSI* (n (%)) General / oncologic surgery Orthopedic trauma surgery Vascular surgery
122 (5.8) 80 (8.4) 22 (2.7) 20 (6.0)
112 (4.5) 75 (5.7) 17 (1.8) 20 (7.8)
0.044
SSI status unknown
387 (15.6)
192 (7.1)
<0.001
No
AC C
Yes
p
Comparisons by t-test for continuous variables, and Fisher’s exact test for categorical variables. * For the 30 day SSI rate, total numbers correspond to the complete case numbers (where the 30 day SSI status is known): 2093 and 2503, respectively ASA‡ American Society of Anesthesiologists CAPD† Continuous Ambulatory Peritoneal Dialysis BMI** Body-mass index
19
ACCEPTED MANUSCRIPT Table 2. Levels of preoperative blood tests stratified by 30 day SSI status Level of blood predictors Median (IQR)
Mean (SD)
1971
3.6 (1.0 – 10.4)
14.3 (31.1)
SSI Total
122 2093
5.7 (1.5 – 15.8) 3.6 (1.1 – 10.8)
22.2 (49.2) 14.7 (32.5)
Leukocytes No SSI SSI Total
1971 122 2093
8.4 (6.6 – 11.0) 8.0 (6.5 – 9.9) 8.4 (6.5 – 10.9)
9.2 (3.6) 8.7 (3.3) 9.1 (3.6)
Albumin No SSI SSI Total
1971 122 2093
38.0 (35.0 – 40.0) 35.9 (33.0 – 38.5) 37.6 (35.0 – 40.0)
37.1 (4.8) 35.2 (5.4) 37.0 (4.8)
eGFR No SSI SSI Total
1971 122 2093
88.0 (70.6 – 101.8) 81.2 (63.4 – 92.5) 87.6 (70.3 – 101.3)
84.7 (25.7) 77.5 (24.4) 84.3 (25.7)
Hemoglobin No SSI SSI Total
1971 122 2093
137.0 (127.0 – 148.0) 130.0 (117.0 – 145.0) 137.0 (126.0 – 148.0)
135.9 (18.6) 130.7 (20.1) 135.5 (18.8)
p 0.03
RI PT
0.23
<0.0001
SC
CRP No SSI
n
M AN U
Variable
AC C
EP
TE D
Comparison between groups are done using Wilcoxon’s rank-sum test. IQR=interquartile range. SD=standard deviation CRP: C-reactive protein eGFR: estimated glomerular filtration rate
0.0012
0.0016
20
ACCEPTED MANUSCRIPT
Table 3. Interaction terms between levels of categorical blood parameters and late administration of SAP before incision Subgroup Albumin Normal Low
n
OR
95% CI
1577 516
1.26 1.23
[0.79; 2.02] [0.67; 2.28]
CRP Normal High
1544 549
1.96 0.49
[1.23; 3.17] [0.25; 0.94]
Hemoglobin Normal Low
1470 623
1.45 1.00
[0.87; 2.44] [0.59; 1.72]
p 0.953
RI PT
<0.001
1433 660
eGFR Normal Mildly decreased Moderately decreased Severely decreased/kidney failure
918 831 275 69
1.36 0.94
0.80 1.56 1.66 0.59
0.40
[0.89; 2.10] [0.45; 1.98]
M AN U
Leukocytes Normal High
SC
0.33
0.352
[0.40; 1.57] [0.92; 2.69] [0.70; 4.11] [0.07; 5.13]
AC C
EP
TE D
OR=odds ratio. CI=confidence interval. Odds ratios refer to the odds of SSI for each predictor subgroup with late SAP. P values refer to the interaction terms between categorical predictor levels and timing of SAP (early vs. late). CRP: C-reactive protein. eGFR: estimated glomerular filtration rate
21
ACCEPTED MANUSCRIPT
Table 4. Interaction terms and subgroup analyses between surgical department and categorical blood parameters. Predictor
Department
n
OR
95% CI
High CRP
0.123 General / oncologic
646
1.17
Orthopedic trauma
621
1.23
[0.43,3.03]
Vascular
277
3.61
[1.35,9.18]
[0.72,1.88]
0.315
RI PT
High leukocytes General / oncologic
636
0.57
[0.32,0.96]
Orthopedic trauma
528
0.86
[0.33,2.07]
Vascular
269
1.41
[0.45,3.80]
General / oncologic
752
2.71
[1.66,4.37]
0.302
586 239
General / oncologic
Moderately decreased Severely decreased / failure Mildly decreased
Orthopedic trauma
Moderately decreased Severely decreased / failure Mildly decreased
Vascular
TE D
Moderately decreased
1.50
[0.59,3.55]
1.38
[0.51,3.49]
M AN U
eGFR Mildly decreased
Orthopedic trauma Vascular
SC
Low albumin
Severely decreased / failure
356
2.27
0.013
[1.35, 3.88]
107
3.29
[1.67, 6.35]
13
2.19
[0.23, 9.69]
324
3.23
[1.16, 10.81]
102
2.84
[0.63, 11.86]
19
17.13
[3.58, 76.51]
151
0.65
[0.24, 1.83]
66
0.71
[0.19, 2.33]
37
0.13
[0.00, 1.12] 0.121
General / oncologic Orthopedic trauma
703 570
2.31 5.31
[1.44,3.69] [2.21,14.08]
Vascular
197
1.47
[0.59,3.69]
EP
Low hemoglobin
p
AC C
OR=odds ratio. CI=confidence interval. Odds ratios refer to the odds of SSI for each predictor per surgical department. Normal blood parameter levels serve as reference for all analyses. P values refer to the interaction terms between categorical predictor levels and surgical department. Since the number of patients in some of the eGFR levels was very small, and thus the number of events also small (or zero), Firth’s penalized likelihood was used to reduce the small sample bias in the estimated odds ratio and confidence intervals. CRP: C-reactive protein. eGFR: estimated glomerular filtration rate
22
ACCEPTED MANUSCRIPT
Table 5. Results of multiple logistic regression examining the combined effects of all potential predictors using Firth’s penalized likelihood. 95% CI [1.24, 4.00] [0.50, 1.24] [1.28, 2.94] [0.98, 2.34] [1.24, 3.62] [1.18, 4.65] [0.13, 6.05] [1.18, 3.08] [0.07, 0.56] [0.68, 3.97] [0.13, 0.66] [0.42, 5.05] [0.17, 4.21] [0.74, 93.76] [0.09, 0.91] [0.06, 1.05] [0.00, 1.28]
p 0.008 0.326 0.002 0.059 0.005 0.016 0.785 0.007 <0.001 0.232 0.003 0.615 0.858 0.091 0.035 0.058 0.070
RI PT
OR 2.23 0.80 1.94 1.53 2.10 2.37 1.29 1.89 0.22 1.74 0.30 1.36 0.87 6.75 0.29 0.26 0.06
SC
M AN U
Variable High CRP High leukocytes Low hemoglobin Low albumin Mildly decreased eGFR Moderately decreased eGFR Severely decreased eGFR / renal failure Late SAP Orthopedic trauma surgery department Vascular surgery department High CRP : Late SAP Mildly decreased eGFR : Orthopedic trauma surgery Moderately decreased eGFR : Orthopedic trauma surgery Severely decreased eGFR / failure : Orthopedic trauma surgery Mildly decreased eGFR : Vascular surgery Moderately decreased eGFR : Vascular surgery Severely decreased eGFR / failure : Vascular surgery
AC C
EP
TE D
OR=odds ratio. CI=confidence interval. CRP: C-reactive protein. GFR: estimated glomerular filtration rate For the vascular departments, the general and oncologic surgery department serves as reference. For the timing of SAP, early SAP serves as reference. For all blood parameters, the normal range serves as reference. “:” is the symbol for interaction terms.
23
ACCEPTED MANUSCRIPT Figure 1 ROC Curves and odds ratios for blood parameters to predict the risk of surgical site infection Odds ratios (OR) refer to an increase of 1 unit for leukocytes and albumin, 10 units for hemoglobin and eGFR and a ten-fold increase for CRP due to log transformation. CRP: C-reactive protein eGFR: estimated glomerular filtration rate
RI PT
Figure 2 Univariate logistic regressions for each blood parameter– categorized
AC C
EP
TE D
M AN U
SC
CRP: C-reactive protein eGFR: estimated glomerular filtration rate
1.0
Specificity
0.
S
0.2
0.0
0.0 1.0
1.0
0.8 0.6 0.4
Sensitivity
0.6
0.2
0.2
0.2
0.4
Sensitivity
0.8 0.6 0.4
0.6 0.4
Sensitivity
0.2
AUC 0.587 [0.538, 0.637]
0.0
0.0
0.0
0.2
0.0 1.0
0.8
0.6
0.4
0.2
0.0
Specificity
EGFR Hemoglobin
0.8
RI PT
1.0
1.0 0.8
1.0
0.4 0.0
Specificity
Hemoglobin AlbuminFigure 1: ROC EGFR for SSI. Albumin curve for potential predictors
0.8
0.2
Specificity
p=0.0029
EGFR
0.8
1.0
1.0 0.8
1.0 0.6 0.4
Sensitivity
0.2 0.0
0.0
SpecificitySpecificity
Specificity
1.0
1.0 0.8
Specificity Specificity
0.8
1.0 0.8
0.8
0.8 0.6 0.4
Sensitivity
0.2
0.8 1.00.2 0.6 0.80.0 0.4 0.6 0.2 0.4 0.0 0.2 0.4 1.0 0.0 0.81.00.2 0.60.80.0 0.40.6 1.0 0.8 0.6 0.4 1.00.20.4 0.80.00.2 0.6
Specificity
0.8
AUC 0.585 [0.529, 0.641][0.569, 0.670]AUC 0.619 [0.569, 0.670] AUC 0.619 0.4 0.8
Specificity
0.2 0.6
0.0 1.00.2 0.80.0 0.6 0.4
Specificity
0.4
1.0 0.8
Figure 1: ROC curve for potential predictors for SSI. Hemoglobin 1.0
0.6
0.6 0.4
Sensitivity
0.2 0.0
0.0
0.21.0 0.00.8
Specificity
0.8
0.6
AUC 0.587 [0.538, 0.637]AUC 0.587 [0.538, 0.637] AUC 0.585 [0.529, 0.641]
0.0
0.0
0.6 1.0
0.4
Sensitivity
0.6
0.2
0.2
0.2
0.4
Sensitivity
10 / 17
0.4 1.00.2 0.80.01.0 0.6
0.4 0.8
0.2 0.6
0.0 0.4
0.2
0.0
SC
1.0
0.6
Sensitivity
0.6 0.0
0.0
0.2
0.4
0.4
Sensitivity
0.6
ch10WeberAN2_bloodPredictors (rev: 70554) – Clinical Trial Unit – Spitalstrasse 12 – CH–4031 Basel – www.clinicaltrialunit.ch
0.2
Sensitivity
0.0
1.0 0.6
0.8
eGFR 0.6 0.4
OR 0.90; 95% CI 0.85-0.97
OR 0.96; 95% CI 0.91-1.01 p=0.16
Specificity Specificity
10
1.0
1.0 0.6 0.2 0.0
0.0
0.4
Sensitivity
0.8 0.6 0.4
Sensitivity
0.2 0.0
0.8
AUC0.613] 0.587AUC [0.538, 0.637] AUC 0.559 [0.505, AUC0.641] 0.533AUC [0.482, 0.583] 0.559 [0.505, AUC0.670] 0.533 [0.482, 0.583] AUC0.613] 0.585 [0.529, 0.619 [0.569,
0.0
0.2
AUC 0.533 [0.482, 0.583] 1.0
EGFR log (CRP) MANUSCRIPT HemoglobinLeukocytes Albumin Leukocytes ACCEPTED
0.8
1.0
1.0 0.8 0.6 0.4 0.2
0.4
OR 0.87; 95% CI 0.80-0.96 p=0.0031
Specificity
log10(CRP)
AUC 0.619 [0.569, 0.670] 0.6
0.2 0.2
p=0.022
Albumin
0.8
AUC 0.559 [0.505, 0.613] AUC 0.587 [0.538, 0.637]
Hemoglobin Leukocytes 1.01.0 0.80.8 0.60.6 0.40.4 0.20.2 0.00.0
OR 1.34; 95% CI 1.04-1.73 Specificity
p<0.001
1.0
AUC 0.533 [0.482, 0.583] AUC 0.619 [0.569, 0.670]
CRP
1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0.0 0.0 1.0
Sensitivity
0.0
0.0 0.0
0.0 0.0
Albumin 0.6 0.4 0.2
0.4
0.8
OR 0.93; 95% CI 0.90-0.96 Specificity
Sensitivity
S S 0. 0.
S S 0. 0 0.2 0.2
0
S
0.2 0.0
AUC 0.559 [0.505, 0.613] 1.0
Specificity
Specificity Specificity
Hemoglobin Figure 1: ROC curve for potential predictors for SSI.
M AN U 0.6
10 / 17
0.2
0.4
Sensitivity
0.4 0.2
Sensitivity
10 / 17
0.6
ch10WeberAN2_bloodPredictors (rev: 70554) – Clinical Trial Unit – Spitalstrasse 12 – CH–4031 Basel – www.clinicaltrialunit.ch ch10WeberAN2_bloodPredictors (rev: 70554) – Clinical Trial Unit – Spitalstrasse 12 – CH–4031 Basel – www.clinicaltrialunit.ch
0.0
0.0
AUC 0.585 [0.529, 0.641]AUC 0.585 [0.529, 0.641] 1.0
0.8
0.6
0.4 1.00.2 0.80.0 0.6
Specificity
0.4
0.2
0.0
Specificity
Figure 1: ROC curveFigure for potential for SSI. predictors for SSI. 1: ROCpredictors curve for potential
TE D
ch10WeberAN2_bloodPredictors (rev: 70554) – Clinical Trial(rev: Unit 70554) – Spitalstrasse 12Trial – CH–4031 Basel – www.clinicaltrialunit.ch ch10WeberAN2_bloodPredictors – Clinical Unit – Spitalstrasse 12 – CH–4031 Basel – www.clinicaltrialunit.ch
AC C
EP
10 / 17
10 / 17
ACCEPTED MANUSCRIPT Level
Albumin
Normal Low Normal High Normal Low Normal High Normal Mildly decreased Moderately decreased Severely decreased/failure
CRP Hemoglobin Leukocytes
1.0 OR
2.0
4.0
OR
CI
p
1.93
[1.31; 2.82]
<0.001
1.52
[1.02; 2.22]
0.035
2.34
[1.61; 3.38]
<0.001
0.69
[0.45; 1.05]
0.090
1.96 2.30 1.55
[1.29; 3.04] [1.32; 3.94] [0.45; 4.04]
0.002 0.003 0.418
EP
TE D
M AN U
SC
0.25 0.50
AC C
eGFR
Events 76 / 1577 46 / 516 80 / 1544 42 / 549 63 / 1470 59 / 623 92 / 1433 30 / 660 35 / 918 60 / 831 23 / 275 4 / 69
RI PT
Variable
S0002-9610(17)31336-3
DOI:
10.1016/j.amjsurg.2017.08.021
Reference:
AJS 12489
To appear in:
The American Journal of Surgery
Received Date: 23 June 2017 Revised Date:
17 August 2017
Accepted Date: 21 August 2017
Please cite this article as: Mujagic E, Marti WR, Coslovsky M, Zeindler J, Staubli S, Marti R, Mechera R, Soysal SD, Gürke L, Weber WP, The role of preoperative blood parameters to predict the risk of surgical site infection, The American Journal of Surgery (2017), doi: 10.1016/j.amjsurg.2017.08.021. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Background Routine preoperative blood work is not recommended but selected biochemical markers may predict the risk of surgical site infection (SSI). This study examines the association between preoperative biochemical markers and the risk of SSI.
RI PT
Methods This observational cohort study, nested in a randomized controlled trial, was conducted at two tertiary referral centers in Switzerland.
SC
Results
122 (5.8%) of 2093 patients experienced SSI. Preoperative increasing levels of albumin (OR
M AN U
0.93), CRP (OR 1.34), hemoglobin (OR 0.87) and eGFR (OR 0.90) were significantly associated with the odds of SSI. The same accounts for categorized parameters. The highest area under the curve from ROC curves was 0.62 for albumin. Positive predictive values ranged from 6.4% to 9.5% and negative predictive values from 94.8% to 95.7%. The
TE D
association of CRP, mildly and moderately decreased eGFR and hemoglobin with the odds of SSI remained significant on multivariate analysis.
Conclusions
AC C
EP
Our results do not support generally delaying elective surgery based on preoperative blood results. However, it may be considered in situations with potentially severe sequelae of SSI.
ACCEPTED MANUSCRIPT 1
The role of preoperative blood parameters to predict the risk of surgical site infection
RI PT
Edin Mujagica*, Walter R Martib*, Michael Coslovskyc, Jasmin Zeindlera, Sebastian Staublia, Regula Martib, Robert Mecheraa, Savas D Soysala, Lorenz Gürkea, Walter P Webera *Contributed equally
AC C
EP
TE D
Corresponding author: Edin Mujagic, MD Department of Surgery University Hospital of Basel Spitalstrasse 21 4031 Basel Switzerland Email: [email protected] Phone: +41613286116 Fax: +41612657512
M AN U
SC
a Department of Surgery, University Hospital of Basel, Spitalstrasse 21, 4031 Basel, Switzerland [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected] b Department of Surgery, Kantonsspital Aarau, Tellstrasse, 5001 Aarau, Switzerland [email protected]; [email protected]; [email protected] c University of Basel, University Hospital, Department of Clinical Research, Clinical Trial Unit, Spitalstrasse 12, 4031 Basel, Switzerland [email protected]
ACCEPTED MANUSCRIPT 2
Abstract Background Routine preoperative blood work is not recommended but selected biochemical markers may
preoperative biochemical markers and the risk of SSI.
Methods
RI PT
predict the risk of surgical site infection (SSI). This study examines the association between
two tertiary referral centers in Switzerland.
M AN U
Results
SC
This observational cohort study, nested in a randomized controlled trial, was conducted at
122 (5.8%) of 2093 patients experienced SSI. Preoperative increasing levels of albumin (OR 0.93), CRP (OR 1.34), hemoglobin (OR 0.87) and eGFR (OR 0.90) were significantly associated with the odds of SSI. The same accounts for categorized parameters. The highest area under the curve from ROC curves was 0.62 for albumin. Positive predictive
TE D
values ranged from 6.4% to 9.5% and negative predictive values from 94.8% to 95.7%. The association of CRP, mildly and moderately decreased eGFR and hemoglobin with the odds
Conclusions
EP
of SSI remained significant on multivariate analysis.
AC C
Our results do not support generally delaying elective surgery based on preoperative blood results. However, it may be considered in situations with potentially severe sequelae of SSI.
Key words
Surgical site infection; Preoperative blood testing; Predictive value; biochemical markers
ACCEPTED MANUSCRIPT 3
Introduction Surgical site infections (SSI) account for up to 38% of all nosocomial infections in inpatients and are considered the most common type of nosocomial infections among surgical
RI PT
patients.(1, 2) In spite of numerous measures to prevent SSI, up to 5% of all patients
undergoing surgery are reported to experience SSI with a substantial increase in morbidity and mortality.(3-5) Furthermore, SSI are associated with prolonged hospital stay and cost.(6-
SC
8) The identification of patients at increased risk of SSI before surgery is of importance to endorse pre- and perioperative SSI prevention measures or delay surgery to minimize risk
M AN U
where indicated.
Routine preoperative laboratory blood testing was standard practice in the past, but is no longer recommended.(9-14) For example, it was shown that only hemoglobin levels, renal function and electrolytes were associated with postoperative complications, while normal blood tests did not reduce their likelihood. Furthermore, only 0 to 3 percent of abnormal test
TE D
results changed preoperative management.(10) However, there is evidence for the following selected preoperative laboratory tests to potentially predict the risk of postoperative complications including SSI. First, preoperatively
EP
elevated c-reactive protein (CRP) levels have been reported to be associated with the risk of postoperative surgical site and remote nosocomial infections (15-19) as well as with other
AC C
major postoperative complications in cardiac surgery, including renal dysfunction and death.(19, 20) Second, total white blood cell count (WBCC) has been reported to be associated with postoperative complications in surgical patients. One report showed a linear correlation of increasing preoperative WBCC values, even within the normal limits, with postoperative complications, but did not focus on SSI.(21) Third, malnutrition is a common problem in surgical patients, especially in such subsets as the elderly and chronically ill. It is known to adversely affect outcomes in those patients.(22, 23) Albumin has been found to be the most reliable indicator of the nutritional status.(24) Hypoalbuminemia as a marker for
ACCEPTED MANUSCRIPT 4 malnutrition has been established as an independent risk factor for SSI.(25-28) Fourth, end stage renal insufficiency is a known risk factor for surgical site infections.(29, 30). Although chronic kidney failure of earlier stages has been shown to be associated with an increased risk for SSI (31), that association has been shown to be less pronounced. Finally,
RI PT
preoperative anemia has been associated with risk of SSI. However, it was not identified as an independent predictor.(32)
The aim of this study was to examine whether a set of routine preoperative blood tests
SC
including WBCC, CRP, Creatinine (eGFR), albumin and hemoglobin predict the risk of SSI either alone or in combination. If so, these parameters could help identify patients at
M AN U
increased risk for SSI, thereby allowing to reinforce SSI prevention measures or to delay
AC C
EP
TE D
surgery in order to optimize the conditions of the patients and procedures.
ACCEPTED MANUSCRIPT 5
Methods Setting This observational cohort study was nested within a bicenter randomized controlled trial (RCT)(33) that was designed to evaluate the optimal timing of surgical antimicrobial
RI PT
prophylaxis. Data were prospectively collected between February 22nd, 2013 and July 31st, 2015 at the University Hospital of Basel and the hospital of Aarau, two tertiary referral
centers in Switzerland. The preoperative blood test parameters hemoglobin, creatinine,
SC
albumin, CRP and white blood cell count were collected in those patients considered to have an indication for preoperative blood testing as per clinical standards. The indication for
M AN U
preoperative blood testing at both sites was based on the general condition of the patient, the complexity of the procedure and individual judgment of the physicians on the ward. The analysis included all patients for whom the above-mentioned preoperative blood parameters were available from blood taken within 3 days before surgery. Hence, this is a
committees was obtained.
Patients
TE D
retrospective analysis of a prospectively collected dataset. Approval by the local ethics
The study protocol of the RCT has been published.(34) Briefly, all inpatients aged 18 years
EP
or older undergoing general, oncologic, vascular and orthopedic trauma procedures with an indication for surgical antimicrobial prophylaxis (cefuroxime, in colorectal surgery combined
AC C
with metronidazole) according to clinical standards were included. These standards were based on CDC guidelines for surgical wound classification.(2) Exclusion criteria included outpatient surgery, contraindication for cefuroxime and/or metronidazole, preexisting antibiotic therapy within 14 days prior to surgery, cognitive impairment, combined operations including other than the above specified surgical subspecialties and emergency procedures with planned incision within 2 hours after indicating the procedure. All patients in this study provided written consent to participate.
ACCEPTED MANUSCRIPT 6
Endpoints The primary endpoint was the occurrence of SSI within 30 days after surgery. SSI is defined and categorized according to CDC criteria.(2) In hospital SSI were diagnosed by the surgical team according to clinical standards. In addition, to increase sensitivity, inpatients were
RI PT
regularly seen by members of the RCT study team. Post discharge follow up was performed by trained members of the RCT study team by contacting the patients over the phone. When a post discharge SSI was suspected, hospital charts were reviewed and primary care
board certified infectious diseases specialist.
Statistical analysis
SC
physicians were contacted for additional information. All suspected SSI were validated by a
M AN U
The analysis included all patients with a complete set of pre-operative blood values for WBCC, CRP, Creatinine, albumin and hemoglobin and completed 30 day follow up. Descriptive statistics are provided for demographic and baseline characteristics and values. Categorical variables are summarized as frequency and percentage. Continuous variables
TE D
are summarized as mean and standard deviation (SD) and median and interquartile range (IQR) for blood values due to highly skewed distributions. In case of comparisons between groups, Fisher’s exact test was used for categorical variables, and t-tests or Kruskal-Wallis
EP
tests for continuous variables, as appropriate.
The association of each of the blood parameters with SSI was analyzed using logistic
AC C
regression models. In a first step, the association of the blood value as a continuous variable was investigated. Receiver operator characteristic (ROC) curves were plotted for each potential predictor, and the area under the curve (AUC) calculated – higher AUC values indicate better discriminative abilities for the potential predictor and an AUC value of 0.5 suggests a predictive ability not different from chance. Due to the wide range of hemoglobin and eGFR values, results are presented per 10-unit increases. For CRP, values are logtransformed and thus results refer to 10-fold increases in CRP levels. In a second step, the blood parameters were categorized into groups according to commonly used cut offs. Normal ranges were 35-52g/L for albumin, <10mg/L for CRP (SI conversion
ACCEPTED MANUSCRIPT 7 factor 9.524), 140-180g/L for hemoglobin in men, 120-160g/L for hemoglobin in women and 3.5x109/L- 10.0x109/L for leukocytes. For eGFR values, the CKD-EPI equation(35) was used and ≥90mL/min/1.73m2 was considered normal renal function, 60-89mL/min/1.73m2 mildly decreased, 30-59mL/min/1.73m2 moderately decreased, 15-29mL/min/1.73m2 severely
RI PT
decreased and <15mL/min/1.73m2 renal failure. For parameters with three categories, the “low” or the “high” category was combined with the “normal” category when they were not different in terms of outcome. This was the case for albumin, leukocytes and hemoglobin.
SC
Hence, these three parameters were grouped into two categories. Logistic regression
models were fitted with the categorized parameters as potential predictors. In addition, the
M AN U
diagnostic performance of the parameter is assessed by providing the sensitivity, specificity and positive and negative predictive values at the relevant cut point. To assess whether the predictive value of a parameter differed depending on the timing of surgical antimicrobial prophylaxis (SAP) (early SAP in the anesthesia room vs. late SAP in
TE D
the operating room) the timing and the interaction with the blood parameter categories were included in the model, followed by subgroup analyses, where the effect of SAP timing was examined in each blood parameter group. A similar approach was taken to examine whether a parameter’s predictive value differed depending on the surgical department
EP
(general/oncologic, orthopedic trauma or vascular).
AC C
Finally, to assess the independent effect of each parameter, adjusted for all other parameters, on the odds of SSI, a multiple logistic regression model was fit. Firth’s penalized likelihood was used for the multiple regression models due to low separability of rare subgroups(36). All the above-mentioned blood parameters were included, categorized by the above described cut points and with their interaction with timing of SAP. Non-significant interaction terms were removed from the model. For all logistic regression models, odds ratios and 95% confidence intervals are presented. The alpha level for all analyses was taken as 0.05. All analyses were performed using R (version 3.4.0; 2017)(37).
ACCEPTED MANUSCRIPT
RI PT
8
Results
A total of 5175 patients were included in the analysis for the RCT and hence assessed for
SC
the availability of preoperative levels of the blood parameters of interest. A complete set of preoperative WBCC, CRP, albumin, creatinine and hemoglobin measurements was present
M AN U
for 2480 of 5175 patients (48%). Demographics of all patients of the RCT, both with or without preoperative blood tests available, are presented in table 1. The 2480 patients with preoperative blood tests available were older, had higher ASA scores, had more secondary diagnoses, were more likely to be women, diabetics and on dialysis. Further characteristics are shown in table 1.
TE D
The current study population consisted of 2093 of the 2480 (84.4%) patients with available blood tests that had 30 day follow up. The 30 day SSI rate was higher in patients with preoperative blood work than in the group without (5.8% vs. 4.5%, p=0.044).
EP
Preoperative levels of the blood parameters by 30 day SSI status are shown in table 2. Analyzing the blood parameters as continuous variables using univariable logistic
AC C
regression, an increase in levels of all parameters but leukocytes was associated with significantly increased (for CRP) or decreased (for albumin, eGFR and hemoglobin) odds of surgical site infections (figure 1). Figure 1 also shows ROC curves for all 5 potential blood parameters. The highest area under the curve (AUC) value was 0.62 for albumin. Based on commonly used cut offs, low albumin, low hemoglobin, mildly and moderately decreased eGFR resulted in highly significant increases in the odds for SSI. High CRP increased the odds significantly whereas leukocytes and more severely decreased eGFR
ACCEPTED MANUSCRIPT 9 were not associated with significant changes in the odds for SSI (Figure 2). Measures of performance were calculated for all parameters categorized to two levels: Sensitivities for albumin, CRP, hemoglobin and leukocytes were 37.7, 34.4, 48.4 and 75.4%, respectively. Specificities were 76.2, 74.3, 71.4 and 32.0%, respectively. Positive predictive values were
RI PT
8.9, 7.7, 9.5 and 6.4%, respectively whereas negative predictive values were 95.2, 94.8, 95.7 and 95.5%, respectively.
A significant interaction between a categorized blood parameter and SAP timing was found
SC
for CRP (p<0.001) (table 3). The subgroup analysis showed that the odds ratio of SSI was 1.96 (95% CI 1.23; 3.17) for patients who had late SAP compared to early SAP if their CRP
M AN U
levels were normal. On the other hand, patients with high CRP levels had an OR of 0.49 (95% CI 0.25; 0.94) with late SAP compared to early SAP.
For the other blood parameters, interaction terms were not significant even though there were differences in the direction of odds ratios by timing of SAP for some of the subgroups. Specifically, in patients with normal or severely decreased eGFR, the odds of having SSI
TE D
were reduced when receiving SAP late in the operating room whereas in those with mildly and moderately decreased eGFR, the odds of having SSI were increased when receiving SAP late in the operating room.
EP
When performing subgroup analyses of the categorical blood parameters per surgical
AC C
department including interaction tests, there is a significant interaction between surgical department and eGFR only (p=0.013). In practice, in general, oncologic and orthopedic trauma surgery, decreased eGFR increased the odds of having SSI whereas in vascular surgery, it decreased the odds of having SSI (table 4). However, within the vascular surgery subgroup, the 95% confidence intervals for the odds ratios for SSI in each of the eGFR categories are wide and cross 1 significantly. Table 5 shows the results of the multiple logistic regression model. The odds ratios for SSI with mildly and moderately decreased eGFR, high CRP and low hemoglobin remained significant. Also, the interaction between CRP levels and timing of SAP remained significant
ACCEPTED MANUSCRIPT 10 (p-value of the interaction term 0.003). To assist interpretation of this interaction term, subgroup analysis based on CRP categories show that the odds of SSI with late SAP remain increased in those patients with normal CRP (OR 1.89, 95% CI 1.18; 3.08) and remain decreased in those with high CRP (OR 0.56, 95% CI 0.29; 1.07). Table 5 also shows that the
AC C
EP
TE D
M AN U
SC
RI PT
interaction of eGFR and surgical department remained significant (global p-value 0.0189).
11
ACCEPTED MANUSCRIPT
Discussion To our knowledge, this is the first study that specifically addresses the association between the preoperative levels of a set of five routine blood tests and the incidence of SSI. In 2093 patients that underwent general, oncologic, trauma and vascular surgery procedures, the
RI PT
odds of having SSI were significantly associated with preoperative albumin, CRP and
hemoglobin levels. The odds of SSI were significantly increased by mildly and moderately decreased kidney function. The lack of an association for severely decreased kidney function
SC
and kidney failure is most likely explained by the very limited number of patients with
severely decreased kidney function and kidney failure, as suggested by the wide 95%
M AN U
confidence interval. Finally, preoperative white blood cell count did not a have a significant impact on the odds of SSI. Importantly, despite the highly significant associations between several blood parameters and the risk of SSI, none of the tests showed very good abilities to predict SSI when using the blood test values as continuous variables as shown by the ROC curves. In this setting, albumin was the strongest predictor with an area under the curve of
TE D
0.62. Similarly, when looking at categorical values, the performance characteristics in terms of sensitivity and specificity were modest. However, the positive predictive values (PPV) ranging between 6% and 10% and the negative predictive values (NPV) of up to 96%
EP
showed better predictive abilities to identify patients at increased risk of experiencing SSI and even more so those at low risk, considering the overall SSI rate of 5.8% and its effect on
AC C
the PPV and NPV.
Another interesting finding of this study was the significant interaction between preoperative CRP level categories (normal vs. high) and the timing of surgical antimicrobial prophylaxis (SAP, early vs. late) in terms of SSI risk. The main results of the RCT showed no significant impact of timing of SAP on the risk of SSI.(33) Based on the present results, however, it could be argued that late administration of SAP before incision increases the odds of SSI in patients with normal preoperative CRP levels, while in patients with elevated preoperative CRP levels, the odds are decreased. Interpretation of this finding is difficult and speculative. Most importantly, this cohort study is exposed to all inherent bias of an observational study
12
ACCEPTED MANUSCRIPT
design even though it was nested in a RCT. It could be argued, however, that elevated CRP levels are indicative of an inflammatory state which could alter the effectiveness of cefuroxime and/or metronidazole. Furthermore, there was weak evidence of an interaction of renal function as calculated by
RI PT
eGFR and the timing of SAP in terms of SSI risk. Although not statistically significant, patients with normal renal function showed a trend towards lower rates of SSI with late
administration of SAP while in those with mildly and moderately decreased renal function, the odds of SSI were numerically higher with late administration. This association has not been
SC
previously described in the literature. A possible explanation for this weak interaction could be a change in the pharmacokinetics of the study drugs in patients with decreased renal
M AN U
function, inasmuch as SAP may require a longer time period to achieve the minimum inhibitory concentration in the soft tissue necessary to prevent SSI. Unfortunately, the data set did not allow to further test this hypothesis, for example by determining the percentage of chronic vs. acute kidney failure in patients with decreased eGFR levels.
TE D
The significant interaction seen between eGFR levels and surgical department is again almost impossible to interpret. In practice, our finding that in patients undergoing vascular surgery, decreased renal function as measured by eGFR levels seems to be protective of
EP
experiencing SSI, cannot be explained by any known facts. Although statistically significant, this finding could potentially be attributable to a type I error, eventually driven by the quite low
AC C
numbers of vascular surgery patients within this study (334 of a total of 2093). This hypothesis is backed by the wide 95% confidence intervals of the odds ratios for SSI within the vascular surgery subgroup. However, it remains a hypothesis. Our study has several strengths: First, the data of this analysis were collected in the setting of a RCT (33) in a prospective manner. The outcome of interest, surgical site infection, was validated by an independent infectious diseases specialist. Therefore, the study design allowed the generation of high quality data. Second, the large sample size of patients allowed to investigate the association between blood parameters and risk of SSI with high
13
ACCEPTED MANUSCRIPT
power. Finally, the participation of several different specialties increases the generalizability of our findings. However, we acknowledge the limitations of this study. First, although the dataset was prospectively collected within a RCT, the study was observational in nature. Second, the 5
RI PT
blood parameters were chosen based on existing evidence suggesting a possible association with the risk of SSI. We purposefully refrained from exploring other parameters for feasibility reasons. Third, the blood parameters were registered based on their availability in clinical routine, which parallels the selection of patients in need of routine blood work
SC
suggesting a more complex patient and procedure population. Indeed, compared to patients without blood work, the study population was older, had higher ASA scores, more secondary
M AN U
diagnoses, included a higher percentage of diabetics, patients on dialysis and was more likely to have undergone urgent surgery. Also, only procedures that include the administration of SAP as per clinical standards were analyzed here, suggesting a further selection of patients. It is possible that the administration of SAP masks the abilities of blood
TE D
tests to predict SSI risk. Accordingly, the role of blood workup to stratify patients undergoing procedures without SAP according to their risk of SSI could be different. However, the low incidence of SSI in most procedures without SAP suggests that any such role may be limited.
EP
Last, although this study includes patients undergoing a wide variety of general, oncologic, trauma and vascular procedures, there are other surgical specialties such as cardiac,
AC C
thoracic or neurosurgery that are not included in this study and hence, our findings potentially cannot be applied to such patient populations.
Conclusion
This study demonstrates strong associations between the levels of routine preoperative blood parameters and the risk of SSI. However, none of the blood parameters examined in this study showed striking predictive abilities in terms of surgical site infections and hence, the decision to postpone procedures to optimize patient factors that impact preoperative blood results cannot be generally recommended at a defined cut off. However,
14
ACCEPTED MANUSCRIPT postponement for optimization should be considered in the individual patient, particularly
Author Contributions
M AN U
SC
RI PT
when a potential SSI would have severe consequences.
Study conception and design: Mujagic, WR Marti, Weber
Acquisition of data: Mujagic, Brinkmann, Zeindler, Staubli, R Marti, Mechera, Soysal Analysis and interpretation of data: Mujagic, WR Marti, Coslovsky, Weber
TE D
Drafting of manuscript: Mujagic, WR Marti, Weber
Critical revision: Coslovsky, Brinkmann, Zeindler, Staubli, R Marti, Mechera, Soysal Final approval: Mujagic, WR Marti, Coslovsky, Brinkmann, Zeindler, Staubli, R Marti,
Funding
EP
Mechera, Soysal, Weber
AC C
This study - nested in a randomized controlled trial – was supported by the Swiss National Foundation, the “Kantonsspital Aarau”, the “Gottfried und Julia Bangerter-Rhyner-Stiftung”, the University of Basel, the “Freie Akademische Gesellschaft Basel”, and the “Nora van Meeuwen-Haefliger-Stiftung”. The authors report no conflicts of interest related to this study.
15
ACCEPTED MANUSCRIPT
References 1. Lewis SS, Moehring RW, Chen LF, et al. Assessing the relative burden of hospitalacquired infections in a network of community hospitals. Infect Control Hosp Epidemiol. 2013 Nov;34(11):1229-30. PubMed PMID: 24113613. Pubmed Central PMCID: PMC3977691.
RI PT
2. Mangram AJ, Horan TC, Pearson ML, et al. Guideline for prevention of surgical site infection, 1999. Hospital Infection Control Practices Advisory Committee. Infect Control Hosp Epidemiol. 1999 Apr;20(4):250-78; quiz 79-80. PubMed PMID: 10219875. 3. Coskun D, Aytac J, Aydinli A, Bayer A. Mortality rate, length of stay and extra cost of sternal surgical site infections following coronary artery bypass grafting in a private medical centre in Turkey. J Hosp Infect. 2005 Jun;60(2):176-9. PubMed PMID: 15866018.
SC
4. Coello R, Charlett A, Wilson J, et al. Adverse impact of surgical site infections in English hospitals. J Hosp Infect. 2005 Jun;60(2):93-103. PubMed PMID: 15866006.
M AN U
5. Kirkland KB, Briggs JP, Trivette SL, et al. The impact of surgical-site infections in the 1990s: attributable mortality, excess length of hospitalization, and extra costs. Infect Control Hosp Epidemiol. 1999 Nov;20(11):725-30. PubMed PMID: 10580621. 6. Poulsen KB, Bremmelgaard A, Sorensen AI, et al. Estimated costs of postoperative wound infections. A case-control study of marginal hospital and social security costs. Epidemiol Infect. 1994 Oct;113(2):283-95. PubMed PMID: 7925666. Pubmed Central PMCID: PMC2271539.
TE D
7. Vegas AA, Jodra VM, Garcia ML. Nosocomial infection in surgery wards: a controlled study of increased duration of hospital stays and direct cost of hospitalization. Eur J Epidemiol. 1993 Sep;9(5):504-10. PubMed PMID: 8307135.
EP
8. Boyce JM, Potter-Bynoe G, Dziobek L. Hospital reimbursement patterns among patients with surgical wound infections following open heart surgery. Infect Control Hosp Epidemiol. 1990 Feb;11(2):89-93. PubMed PMID: 2107250.
AC C
9. Chung F, Yuan H, Yin L, et al. Elimination of preoperative testing in ambulatory surgery. Anesth Analg. 2009 Feb;108(2):467-75. PubMed PMID: 19151274. 10. Smetana GW, Macpherson DS. The case against routine preoperative laboratory testing. Med Clin North Am. 2003 Jan;87(1):7-40. PubMed PMID: 12575882. 11. Garcia-Miguel FJ, Serrano-Aguilar PG, Lopez-Bastida J. Preoperative assessment. Lancet. 2003 Nov 22;362(9397):1749-57. PubMed PMID: 14643127. 12. Macpherson DS. Preoperative laboratory testing: should any tests be "routine" before surgery? Med Clin North Am. 1993 Mar;77(2):289-308. PubMed PMID: 8441296. 13. Turnbull JM, Buck C. The value of preoperative screening investigations in otherwise healthy individuals. Arch Intern Med. 1987 Jun;147(6):1101-5. PubMed PMID: 3592875.
16
ACCEPTED MANUSCRIPT 14. Kaplan EB, Sheiner LB, Boeckmann AJ, et al. The usefulness of preoperative laboratory screening. JAMA. 1985 Jun 28;253(24):3576-81. PubMed PMID: 3999339.
15. Shimizu T, Ishizuka M, Kubota K. The preoperative serum C-reactive protein level is a useful predictor of surgical site infections in patients undergoing appendectomy. Surg Today. 2015 Nov;45(11):1404-10. PubMed PMID: 25480421. Epub 2014/12/07. eng.
RI PT
16. Ortega-Deballon P, Duvillard L, Scherrer ML, et al. Preoperative adipocytokines as a predictor of surgical infection after colorectal surgery: a prospective survey. Int J Colorectal Dis. 2014 Jan;29(1):23-9. PubMed PMID: 24132530.
SC
17. Mohri Y, Miki C, Kobayashi M, et al. Correlation between preoperative systemic inflammation and postoperative infection in patients with gastrointestinal cancer: a multicenter study. Surg Today. 2014 May;44(5):859-67. PubMed PMID: 23722282.
M AN U
18. Moyes LH, Leitch EF, McKee RF, et al. Preoperative systemic inflammation predicts postoperative infectious complications in patients undergoing curative resection for colorectal cancer. Br J Cancer. 2009 Apr 21;100(8):1236-9. PubMed PMID: 19319134. Pubmed Central PMCID: PMC2676538. 19. Cappabianca G, Paparella D, Visicchio G, et al. Preoperative C-reactive protein predicts mid-term outcome after cardiac surgery. Ann Thorac Surg. 2006 Dec;82(6):2170-8. PubMed PMID: 17126130. Epub 2006/11/28. eng.
TE D
20. Kim DH, Shim JK, Hong SW, et al. Predictive value of C-reactive protein for major postoperative complications following off-pump coronary artery bypass surgery: prospective and observational trial. Circ J. 2009 May;73(5):872-7. PubMed PMID: 19282606. Epub 2009/03/14. eng.
EP
21. Amaranto DJ, Wang EC, Eskandari MK, et al. Normal preoperative white blood cell count is predictive of outcomes for endovascular procedures. J Vasc Surg. 2011 Nov;54(5):1395-403.e2. PubMed PMID: 21802244. Epub 2011/08/02. eng.
AC C
22. Sungurtekin H, Sungurtekin U, Balci C, et al. The influence of nutritional status on complications after major intraabdominal surgery. J Am Coll Nutr. 2004 Jun;23(3):227-32. PubMed PMID: 15190047. 23. Inoue Y, Miki C, Kusunoki M. Nutritional status and cytokine-related protein breakdown in elderly patients with gastrointestinal malignancies. J Surg Oncol. 2004 May 1;86(2):91-8. PubMed PMID: 15112251. 24. Sung J, Bochicchio GV, Joshi M, et al. Admission serum albumin is predicitve of outcome in critically ill trauma patients. Am Surg. 2004 Dec;70(12):1099-102. PubMed PMID: 15663053. 25. Bohl DD, Shen MR, Kayupov E, Della Valle CJ. Hypoalbuminemia Independently Predicts Surgical Site Infection, Pneumonia, Length of Stay, and Readmission After Total Joint Arthroplasty. J Arthroplasty. 2016 Jan;31(1):15-21. PubMed PMID: 26427941.
17
ACCEPTED MANUSCRIPT
26. Walls JD, Abraham D, Nelson CL, et al. Hypoalbuminemia More Than Morbid Obesity is an Independent Predictor of Complications After Total Hip Arthroplasty. J Arthroplasty. 2015 Dec;30(12):2290-5. PubMed PMID: 26148837.
RI PT
27. Hennessey DB, Burke JP, Ni-Dhonochu T, et al. Preoperative hypoalbuminemia is an independent risk factor for the development of surgical site infection following gastrointestinal surgery: a multi-institutional study. Ann Surg. 2010 Aug;252(2):325-9. PubMed PMID: 20647925. 28. Neumayer L, Hosokawa P, Itani K, et al. Multivariable predictors of postoperative surgical site infection after general and vascular surgery: results from the patient safety in surgery study. J Am Coll Surg. 2007 Jun;204(6):1178-87. PubMed PMID: 17544076.
SC
29. Moran-Atkin E, Stem M, Lidor AO. Surgery for diverticulitis is associated with high risk of in-hospital mortality and morbidity in older patients with end-stage renal disease. Surgery. 2014 Aug;156(2):361-70. PubMed PMID: 24957670.
M AN U
30. Gajdos C, Hawn MT, Kile D, et al. The risk of major elective vascular surgical procedures in patients with end-stage renal disease. Ann Surg. 2013 Apr;257(4):766-73. PubMed PMID: 22982978. 31. Ackland GL, Moran N, Cone S, et al. Chronic kidney disease and postoperative morbidity after elective orthopedic surgery. Anesth Analg. 2011 Jun;112(6):1375-81. PubMed PMID: 20807976.
TE D
32. Weber WP, Zwahlen M, Reck S, et al. The association of preoperative anemia and perioperative allogeneic blood transfusion with the risk of surgical site infection. Transfusion (Paris). 2009 Sep;49(9):1964-70. PubMed PMID: 19453989.
EP
33. Weber WP, Mujagic E, Zwahlen M, et al. Timing of surgical antimicrobial prophylaxis: a phase 3 randomised controlled trial. Lancet Infect Dis. 2017 Apr 03. PubMed PMID: 28385346.
AC C
34. Mujagic E, Zwimpfer T, Marti WR, et al. Evaluating the optimal timing of surgical antimicrobial prophylaxis: study protocol for a randomized controlled trial. Trials. 2014;15:188. PubMed PMID: 24885132. Pubmed Central PMCID: Pmc4040488. Epub 2014/06/03. eng. 35. Levey AS, Stevens LA, Schmid CH, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009 May 05;150(9):604-12. PubMed PMID: 19414839. Pubmed Central PMCID: PMC2763564. 36. 38.
Firth D. Bias reduction of maximum likelihood esteimates. Biometrika. 1993;80(1):27-
37. Team RC. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2016.
18
ACCEPTED MANUSCRIPT Table 1. Baseline characteristics by availability of blood work Characteristic n
Blood tests available 2480
No blood tests available 2695
Sex – number (%) Male Female
1214 (49.0) 1266 (51.0)
1588 (58.9) 1107 (41.1)
Age – years (mean (SD))
60.1 (18.7)
54.4 (18.2)
ASA‡ score (n (%)) 1 2 3 4
304 (12.3) 1284 (51.8) 847 (34.2) 44 (1.8)
628 (23.3) 1448 (53.7) 601 (22.3) 17 (0.6)
3.5 (2.6)
2.7 (2.3)
<0.001
RI PT
<0.001
SC
<0.001
Number of secondary diagnoses (mean (SD))
2425 (97.9) 47 (1.9) 4 (0.2)
2678 (99.4) 16 (0.6) 0 (0.0)
Wound class (n (%)) I II III IV
1998 (80.8) 270 (10.9) 162 (6.6) 43 (1.7)
2070 (76.9) 526 (19.5) 74 (2.7) 21 (0.8)
260 (10.5) 2220 (89.5)
220 (8.2) 2475 (91.8)
26.5 (6.0)
27.6 (7.6)
TE D
Diabetes (n (%)) Yes No
M AN U
Dialysis (n (%)) No dialysis Hemodialysis CAPD†
EP
BMI** (kg/m2) (mean(SD))
<0.001
<0.001
<0.001
0.005
Urgent procedure (n (%))
<0.001 <0.001
549 (22.1)
371 (13.7)
1931 (77.9)
2324 (86.3)
Duration of surgery (mean (SD))
108.1 (71.7)
108.2 (78.0)
0.937
30 day SSI* (n (%)) General / oncologic surgery Orthopedic trauma surgery Vascular surgery
122 (5.8) 80 (8.4) 22 (2.7) 20 (6.0)
112 (4.5) 75 (5.7) 17 (1.8) 20 (7.8)
0.044
SSI status unknown
387 (15.6)
192 (7.1)
<0.001
No
AC C
Yes
p
Comparisons by t-test for continuous variables, and Fisher’s exact test for categorical variables. * For the 30 day SSI rate, total numbers correspond to the complete case numbers (where the 30 day SSI status is known): 2093 and 2503, respectively ASA‡ American Society of Anesthesiologists CAPD† Continuous Ambulatory Peritoneal Dialysis BMI** Body-mass index
19
ACCEPTED MANUSCRIPT Table 2. Levels of preoperative blood tests stratified by 30 day SSI status Level of blood predictors Median (IQR)
Mean (SD)
1971
3.6 (1.0 – 10.4)
14.3 (31.1)
SSI Total
122 2093
5.7 (1.5 – 15.8) 3.6 (1.1 – 10.8)
22.2 (49.2) 14.7 (32.5)
Leukocytes No SSI SSI Total
1971 122 2093
8.4 (6.6 – 11.0) 8.0 (6.5 – 9.9) 8.4 (6.5 – 10.9)
9.2 (3.6) 8.7 (3.3) 9.1 (3.6)
Albumin No SSI SSI Total
1971 122 2093
38.0 (35.0 – 40.0) 35.9 (33.0 – 38.5) 37.6 (35.0 – 40.0)
37.1 (4.8) 35.2 (5.4) 37.0 (4.8)
eGFR No SSI SSI Total
1971 122 2093
88.0 (70.6 – 101.8) 81.2 (63.4 – 92.5) 87.6 (70.3 – 101.3)
84.7 (25.7) 77.5 (24.4) 84.3 (25.7)
Hemoglobin No SSI SSI Total
1971 122 2093
137.0 (127.0 – 148.0) 130.0 (117.0 – 145.0) 137.0 (126.0 – 148.0)
135.9 (18.6) 130.7 (20.1) 135.5 (18.8)
p 0.03
RI PT
0.23
<0.0001
SC
CRP No SSI
n
M AN U
Variable
AC C
EP
TE D
Comparison between groups are done using Wilcoxon’s rank-sum test. IQR=interquartile range. SD=standard deviation CRP: C-reactive protein eGFR: estimated glomerular filtration rate
0.0012
0.0016
20
ACCEPTED MANUSCRIPT
Table 3. Interaction terms between levels of categorical blood parameters and late administration of SAP before incision Subgroup Albumin Normal Low
n
OR
95% CI
1577 516
1.26 1.23
[0.79; 2.02] [0.67; 2.28]
CRP Normal High
1544 549
1.96 0.49
[1.23; 3.17] [0.25; 0.94]
Hemoglobin Normal Low
1470 623
1.45 1.00
[0.87; 2.44] [0.59; 1.72]
p 0.953
RI PT
<0.001
1433 660
eGFR Normal Mildly decreased Moderately decreased Severely decreased/kidney failure
918 831 275 69
1.36 0.94
0.80 1.56 1.66 0.59
0.40
[0.89; 2.10] [0.45; 1.98]
M AN U
Leukocytes Normal High
SC
0.33
0.352
[0.40; 1.57] [0.92; 2.69] [0.70; 4.11] [0.07; 5.13]
AC C
EP
TE D
OR=odds ratio. CI=confidence interval. Odds ratios refer to the odds of SSI for each predictor subgroup with late SAP. P values refer to the interaction terms between categorical predictor levels and timing of SAP (early vs. late). CRP: C-reactive protein. eGFR: estimated glomerular filtration rate
21
ACCEPTED MANUSCRIPT
Table 4. Interaction terms and subgroup analyses between surgical department and categorical blood parameters. Predictor
Department
n
OR
95% CI
High CRP
0.123 General / oncologic
646
1.17
Orthopedic trauma
621
1.23
[0.43,3.03]
Vascular
277
3.61
[1.35,9.18]
[0.72,1.88]
0.315
RI PT
High leukocytes General / oncologic
636
0.57
[0.32,0.96]
Orthopedic trauma
528
0.86
[0.33,2.07]
Vascular
269
1.41
[0.45,3.80]
General / oncologic
752
2.71
[1.66,4.37]
0.302
586 239
General / oncologic
Moderately decreased Severely decreased / failure Mildly decreased
Orthopedic trauma
Moderately decreased Severely decreased / failure Mildly decreased
Vascular
TE D
Moderately decreased
1.50
[0.59,3.55]
1.38
[0.51,3.49]
M AN U
eGFR Mildly decreased
Orthopedic trauma Vascular
SC
Low albumin
Severely decreased / failure
356
2.27
0.013
[1.35, 3.88]
107
3.29
[1.67, 6.35]
13
2.19
[0.23, 9.69]
324
3.23
[1.16, 10.81]
102
2.84
[0.63, 11.86]
19
17.13
[3.58, 76.51]
151
0.65
[0.24, 1.83]
66
0.71
[0.19, 2.33]
37
0.13
[0.00, 1.12] 0.121
General / oncologic Orthopedic trauma
703 570
2.31 5.31
[1.44,3.69] [2.21,14.08]
Vascular
197
1.47
[0.59,3.69]
EP
Low hemoglobin
p
AC C
OR=odds ratio. CI=confidence interval. Odds ratios refer to the odds of SSI for each predictor per surgical department. Normal blood parameter levels serve as reference for all analyses. P values refer to the interaction terms between categorical predictor levels and surgical department. Since the number of patients in some of the eGFR levels was very small, and thus the number of events also small (or zero), Firth’s penalized likelihood was used to reduce the small sample bias in the estimated odds ratio and confidence intervals. CRP: C-reactive protein. eGFR: estimated glomerular filtration rate
22
ACCEPTED MANUSCRIPT
Table 5. Results of multiple logistic regression examining the combined effects of all potential predictors using Firth’s penalized likelihood. 95% CI [1.24, 4.00] [0.50, 1.24] [1.28, 2.94] [0.98, 2.34] [1.24, 3.62] [1.18, 4.65] [0.13, 6.05] [1.18, 3.08] [0.07, 0.56] [0.68, 3.97] [0.13, 0.66] [0.42, 5.05] [0.17, 4.21] [0.74, 93.76] [0.09, 0.91] [0.06, 1.05] [0.00, 1.28]
p 0.008 0.326 0.002 0.059 0.005 0.016 0.785 0.007 <0.001 0.232 0.003 0.615 0.858 0.091 0.035 0.058 0.070
RI PT
OR 2.23 0.80 1.94 1.53 2.10 2.37 1.29 1.89 0.22 1.74 0.30 1.36 0.87 6.75 0.29 0.26 0.06
SC
M AN U
Variable High CRP High leukocytes Low hemoglobin Low albumin Mildly decreased eGFR Moderately decreased eGFR Severely decreased eGFR / renal failure Late SAP Orthopedic trauma surgery department Vascular surgery department High CRP : Late SAP Mildly decreased eGFR : Orthopedic trauma surgery Moderately decreased eGFR : Orthopedic trauma surgery Severely decreased eGFR / failure : Orthopedic trauma surgery Mildly decreased eGFR : Vascular surgery Moderately decreased eGFR : Vascular surgery Severely decreased eGFR / failure : Vascular surgery
AC C
EP
TE D
OR=odds ratio. CI=confidence interval. CRP: C-reactive protein. GFR: estimated glomerular filtration rate For the vascular departments, the general and oncologic surgery department serves as reference. For the timing of SAP, early SAP serves as reference. For all blood parameters, the normal range serves as reference. “:” is the symbol for interaction terms.
23
ACCEPTED MANUSCRIPT Figure 1 ROC Curves and odds ratios for blood parameters to predict the risk of surgical site infection Odds ratios (OR) refer to an increase of 1 unit for leukocytes and albumin, 10 units for hemoglobin and eGFR and a ten-fold increase for CRP due to log transformation. CRP: C-reactive protein eGFR: estimated glomerular filtration rate
RI PT
Figure 2 Univariate logistic regressions for each blood parameter– categorized
AC C
EP
TE D
M AN U
SC
CRP: C-reactive protein eGFR: estimated glomerular filtration rate
1.0
Specificity
0.
S
0.2
0.0
0.0 1.0
1.0
0.8 0.6 0.4
Sensitivity
0.6
0.2
0.2
0.2
0.4
Sensitivity
0.8 0.6 0.4
0.6 0.4
Sensitivity
0.2
AUC 0.587 [0.538, 0.637]
0.0
0.0
0.0
0.2
0.0 1.0
0.8
0.6
0.4
0.2
0.0
Specificity
EGFR Hemoglobin
0.8
RI PT
1.0
1.0 0.8
1.0
0.4 0.0
Specificity
Hemoglobin AlbuminFigure 1: ROC EGFR for SSI. Albumin curve for potential predictors
0.8
0.2
Specificity
p=0.0029
EGFR
0.8
1.0
1.0 0.8
1.0 0.6 0.4
Sensitivity
0.2 0.0
0.0
SpecificitySpecificity
Specificity
1.0
1.0 0.8
Specificity Specificity
0.8
1.0 0.8
0.8
0.8 0.6 0.4
Sensitivity
0.2
0.8 1.00.2 0.6 0.80.0 0.4 0.6 0.2 0.4 0.0 0.2 0.4 1.0 0.0 0.81.00.2 0.60.80.0 0.40.6 1.0 0.8 0.6 0.4 1.00.20.4 0.80.00.2 0.6
Specificity
0.8
AUC 0.585 [0.529, 0.641][0.569, 0.670]AUC 0.619 [0.569, 0.670] AUC 0.619 0.4 0.8
Specificity
0.2 0.6
0.0 1.00.2 0.80.0 0.6 0.4
Specificity
0.4
1.0 0.8
Figure 1: ROC curve for potential predictors for SSI. Hemoglobin 1.0
0.6
0.6 0.4
Sensitivity
0.2 0.0
0.0
0.21.0 0.00.8
Specificity
0.8
0.6
AUC 0.587 [0.538, 0.637]AUC 0.587 [0.538, 0.637] AUC 0.585 [0.529, 0.641]
0.0
0.0
0.6 1.0
0.4
Sensitivity
0.6
0.2
0.2
0.2
0.4
Sensitivity
10 / 17
0.4 1.00.2 0.80.01.0 0.6
0.4 0.8
0.2 0.6
0.0 0.4
0.2
0.0
SC
1.0
0.6
Sensitivity
0.6 0.0
0.0
0.2
0.4
0.4
Sensitivity
0.6
ch10WeberAN2_bloodPredictors (rev: 70554) – Clinical Trial Unit – Spitalstrasse 12 – CH–4031 Basel – www.clinicaltrialunit.ch
0.2
Sensitivity
0.0
1.0 0.6
0.8
eGFR 0.6 0.4
OR 0.90; 95% CI 0.85-0.97
OR 0.96; 95% CI 0.91-1.01 p=0.16
Specificity Specificity
10
1.0
1.0 0.6 0.2 0.0
0.0
0.4
Sensitivity
0.8 0.6 0.4
Sensitivity
0.2 0.0
0.8
AUC0.613] 0.587AUC [0.538, 0.637] AUC 0.559 [0.505, AUC0.641] 0.533AUC [0.482, 0.583] 0.559 [0.505, AUC0.670] 0.533 [0.482, 0.583] AUC0.613] 0.585 [0.529, 0.619 [0.569,
0.0
0.2
AUC 0.533 [0.482, 0.583] 1.0
EGFR log (CRP) MANUSCRIPT HemoglobinLeukocytes Albumin Leukocytes ACCEPTED
0.8
1.0
1.0 0.8 0.6 0.4 0.2
0.4
OR 0.87; 95% CI 0.80-0.96 p=0.0031
Specificity
log10(CRP)
AUC 0.619 [0.569, 0.670] 0.6
0.2 0.2
p=0.022
Albumin
0.8
AUC 0.559 [0.505, 0.613] AUC 0.587 [0.538, 0.637]
Hemoglobin Leukocytes 1.01.0 0.80.8 0.60.6 0.40.4 0.20.2 0.00.0
OR 1.34; 95% CI 1.04-1.73 Specificity
p<0.001
1.0
AUC 0.533 [0.482, 0.583] AUC 0.619 [0.569, 0.670]
CRP
1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0.0 0.0 1.0
Sensitivity
0.0
0.0 0.0
0.0 0.0
Albumin 0.6 0.4 0.2
0.4
0.8
OR 0.93; 95% CI 0.90-0.96 Specificity
Sensitivity
S S 0. 0.
S S 0. 0 0.2 0.2
0
S
0.2 0.0
AUC 0.559 [0.505, 0.613] 1.0
Specificity
Specificity Specificity
Hemoglobin Figure 1: ROC curve for potential predictors for SSI.
M AN U 0.6
10 / 17
0.2
0.4
Sensitivity
0.4 0.2
Sensitivity
10 / 17
0.6
ch10WeberAN2_bloodPredictors (rev: 70554) – Clinical Trial Unit – Spitalstrasse 12 – CH–4031 Basel – www.clinicaltrialunit.ch ch10WeberAN2_bloodPredictors (rev: 70554) – Clinical Trial Unit – Spitalstrasse 12 – CH–4031 Basel – www.clinicaltrialunit.ch
0.0
0.0
AUC 0.585 [0.529, 0.641]AUC 0.585 [0.529, 0.641] 1.0
0.8
0.6
0.4 1.00.2 0.80.0 0.6
Specificity
0.4
0.2
0.0
Specificity
Figure 1: ROC curveFigure for potential for SSI. predictors for SSI. 1: ROCpredictors curve for potential
TE D
ch10WeberAN2_bloodPredictors (rev: 70554) – Clinical Trial(rev: Unit 70554) – Spitalstrasse 12Trial – CH–4031 Basel – www.clinicaltrialunit.ch ch10WeberAN2_bloodPredictors – Clinical Unit – Spitalstrasse 12 – CH–4031 Basel – www.clinicaltrialunit.ch
AC C
EP
10 / 17
10 / 17
ACCEPTED MANUSCRIPT Level
Albumin
Normal Low Normal High Normal Low Normal High Normal Mildly decreased Moderately decreased Severely decreased/failure
CRP Hemoglobin Leukocytes
1.0 OR
2.0
4.0
OR
CI
p
1.93
[1.31; 2.82]
<0.001
1.52
[1.02; 2.22]
0.035
2.34
[1.61; 3.38]
<0.001
0.69
[0.45; 1.05]
0.090
1.96 2.30 1.55
[1.29; 3.04] [1.32; 3.94] [0.45; 4.04]
0.002 0.003 0.418
EP
TE D
M AN U
SC
0.25 0.50
AC C
eGFR
Events 76 / 1577 46 / 516 80 / 1544 42 / 549 63 / 1470 59 / 623 92 / 1433 30 / 660 35 / 918 60 / 831 23 / 275 4 / 69
RI PT
Variable