Predictive Factors of Surgical Site Infection in Prosthetic Joint Surgery. A Prospective Study on 760 Arthroplasties.

Purpose: Surgical site infection (SSI) is a debilitating complication of total joint arthroplasty (TJA) with signicant morbidity and increased costs. Aim of our study was to identify potential risk factors for SSIs in a population of patients undergoing TJA. Methods: TJA were prospectively recruited at Santa Maria Maddalena Hospital from February 2019 to April 2020. Age, sex, major comorbidities, American Society of Anesthesiologists (ASA) class, length of surgery, type of surgical suture, total hospital length of stay and clinical laboratory data were collected. The study population was then divided into two groups: Group A, normal post-operative course, and Group B, patients who developed SSI at follow-up (17-25 days). Results: 25/760 (3.3%) patients developed SSIs at follow-up. Clinical and demographic parameters were not different between the two groups. Total leucocyte and neutrophil values at discharge resulted to be signicatively higher in Group B compared to Group A (p=0.025 and p=0.016, respectively). Values of 7860/mL for total leucocyte, and 5185/mL for neutrophil count at discharge signicantly predicted the future development of SSI (AUC 0.623 and AUC 0.641, respectively; p<0.05) independently from confounding factors (total leukocytes: O.R.=3,69 [95% C.I. 1,63-8,32]; neutrophils: O.R.=3,98 [95% C.I. 1,76-8,97. Deep SSIs has been diagnosed signicantly before supercial SSIs (p=0,008), with a median advance of 9 days. Conclusion: Total leukocytes and neutrophils at discharge seem useful to identify a population at risk for the development of SSIs following TJA. Further studies on larger populations are needed to develop a predictive SSIs risk score that should include those variables.


Background
The number of total joint arthroplasty (TJA) procedures has steadily increased over the past decades, mainly due to demographic changes, with more aged people less willing to accept activity limitations (1).
This increased demand of operative procedures implies an increasing number of patients at postsurgical complications risk.
Surgical site infection (SSI) is the second most frequent complication after instability and prosthetic loosening (2,3) and represents one of the most common healthcare-associated infections (HAI) among orthopedic patients (4). According to some reports, the rate of SSI secondary to TJA is expected to increase by 2-6.5% for hip and 6.8% for knee arthroplasty respectively in the next decades (5), with a consequent increase in the total costs related to treatment (6).
Based on CDC classi cation (7), organ spaced SSIs, identi ed as periprosthetic joint infection (PJI), occurs in 1-2% of primary and in 4% of revision arthroplasties (8). It represents an extremely serious complication, due to the increased patient mortality and morbidity, and a burden that increases the hospital costs (9), estimated to be about twenty-four times higher than the cost of a primary implant (10).
Patients affected by PJI after joint replacement procedures require prolonged antibiotic therapy, revision, or removal of the prosthesis (11), and are at higher risk of impaired functional ability.
Around two thirds of PJI cases are caused through intra-operative inoculation of micro-organisms (12) and the di culty in their treatment depends on the ability of micro-organisms to grow and persist on the implant surface and on necrotic tissue in the form of a bio lm (13).
Bio lm micro-organisms are up to 1000 times more resistant to growth-dependent antimicrobial agents than their planktonic counterparts. In vitro susceptibility testing with the determination of minimal inhibitory concentrations (MIC) do not re ect the in vivo situation, and in the clinical setup the effective local concentration of many antibiotics cannot be reached by systemic application only (14).
Based on microbial virulence, PJI can be divided into "early" (within the rst four weeks after implantation) or "delayed" (typically between three months and three years). Early infections are sustained by high-virulent organisms (e.g. Staphylococcus aureus, streptococci, enterococci) and clinically occur local and systemic signs of in ammation. On the contrary, delayed infections are caused by low-virulent pathogens (e.g. coagulase-negative staphylococci or Cutibacterium species) and manifest with insidious clinical symptoms such as arthralgia or early implant failure (15).
A conservative treatment performed with DAIR (Debridment, Antibiotic, Implant retention) is only allowed if the infection is diagnosed and treated within one month of implantation of the prosthesis, or within three weeks of the onset of symptoms. To predict outcome of PJI treated with DAIR, the KLIC-score (KLICscore: Kidney, Liver, Index surgery, Cemented prosthesis and C-reactive protein value) has been formulated as a risk strati cation tool (16).
In the updated edition of the CDC guidelines on SSIs prevention (17), a whole chapter is dedicated to the prevention of PJI, remarking how this surgery play an important role especially regarding the burden of possible complications.
Nevertheless, none of the suggested interventions reached conclusive results, highlighting the lack of knowledge on SSIs prevention in arthroplasty.
The aim of our study was to identify potential risk factors for the development of SSIs in a population of patients undergoing TJA which could be used in future SSI prediction scores to guide the clinical followup of total joint replacements.

Matherial And Methods
Patients undergoing TJA surgery (hip, knee, shoulder) were prospectively recruited at Santa Maria Maddalena Hospital (Rovigo, Italy) from February 2019 to April 2020. Patient demographics (age and sex), major comorbidities and medications, American Society of Anesthesiologists (ASA) class (18), length of surgery, type of surgical incision, type of surgical suture, wound dressing and total length of hospital stay were collected from clinical and surgical records.
Laboratory tests such as white blood cells count, neutrophils count, hematocrit, hemoglobin, CRP, ESR, total protein and albumin values had been recorded one week before the surgical procedure and at hospital discharge.
SSIs were diagnosed according to the CDC diagnostic criteria (7) and classi ed in super cial, deep or PJI. Each enrolled patient was evaluated by an infectious disease specialist between the 17th and 25th day after surgery for SSIs surveillance.
The study population was then divided into two groups according to the diagnosis of SSI: Group A, patients with a normal post-operative course, and Group B, patients who developed SSI at follow-up.
For each case of SSI, the clinical symptoms, etiology and antibiotic resistance pattern (when available), type of revision surgery (if performed), and type of antibiotic therapy administered were collected. For patients with PJI treated with DAIR the risk of failure was calculated by the KLIC-score. Data regarding outcome and one year follow-up were also reported.
The research was carried out in accordance with the ethical principles of the Declaration of Helsinki. The Local Ethics Committee "CESC VR-RO" approved the design of this study and all participants gave their written informed consent.

Statistical Analysis
Normality of distribution of continuous variables was assessed by Kolmogorov-Smirnov test. Normally distributed variables were expressed as mean ± standard deviation and compared by t-test. Not normally distributed variables were expressed as median (interquartile range) and compared with Mann-Whitney U test. Frequencies were reported as percentage and compared by Chi-square test or Fisher's exact test in the presence of a 2x2 contingency table. A mixed ANOVA was used to observe whether the values measured at pre-and post-operative occasions were different between the two groups of patients. The diagnostic accuracy for total leukocytes and neutrophil's count in detecting SSI was determined by a ROC (Receiver Operator Characteristic) analysis. The Youden's J Index (speci city+sensitivity-1), which maximizes the speci c and sensitivity, was used to nd the cut-off values. The cut-off values were used to calculate the positive predictive value (PPV), the negative predictive value (NPV) and accuracy. A multivariate logistic regression was used to determine the association between high total leukocytes (>7860 cells/µL) or high total neutrophil's count (>5185 cells/µL) at discharge and the presence of SSI; sex, age and BMI were included in the analysis as possible confounding factors. The analyses were performed by SPSS (IBM) for Windows, version 25. A p-value < 0.05 was considered signi cant.
Clinical and demographic characteristics of Group A and B are summarized in Table 1. A cefazoline-based antibiotic prophylaxis was used in most cases. Patients who were allergic to cefazoline or with known risk factors for MRSA colonization received vancomycin-based antibiotic prophylaxis.
Laboratory test parameters differ in those patients contracting SSI Results from laboratory tests performed in both pre-operative period and at discharge are reported in  We tested the diagnostic accuracy of total leukocyte and neutrophil's count at discharge to detect those subjects at risk to develop SSI by ROC curves analysis and area under the curve (AUC) calculation ( Figure  1). AUC values were 0.623 (p=0.025) and 0.641 (p=0.016), respectively. By calculating the Youden's J Index, which maximizes the sensitivity and speci city, we found a cut-off value for total leukocyte of 7860 cells/µL (sensitivity/speci city: 56%/75%) and for neutrophil's count of 5185 cells/µL (sensitivity/speci city: 48%/80%). The two variables showed a high negative predictive value (NPV; leukocytes: 98.03%; neutrophils: 97.85%); however, they showed a poor positive predictive value (PPV; leukocytes: 6.97%; neutrophils: 7.69%).
Nonetheless, by a multivariate binary logistic regression analysis we demonstrated that patients with values of leukocytes or neutrophils higher than the cut-off had a higher odd of developing SSI, independently from confounding factors like sex, age and BMI. In fact, patients with a value of total leucocyte at discharge higher than the cut-off had a 3. Finally, to further investigate the relation between these two parameters and the increased risk of developing SSI, we performed a mixed ANOVA analysis. By this mean, we were able to detect a SSIrelated growth of values from baseline (pre-operative) to discharge. As expected, the values of total leukocytes, ESR and PCR in the whole population are increased at discharge compared to the preoperative value (p=0.002, p=0.001 and p=0.001, respectively), whereas neutrophils almost reached the statistical signi cance (p=0.056). However, we found a signi cant interaction between the pre-and postoperative changes of total leukocytes and future development of infection. As displayed in Figure 2A, patients that will develop infection (Figure 2A, squares) had a steeper increase in total leukocytes compared with not infected subjects (p=0.006). Such interaction still exists for neutrophil's count, although it is less evident ( Figure 2B, p=0.019).

Characteristics of SSI classi cation
The time-to-diagnosis was compared between the three SSI classi cation categories (super cial, deep and PJI) through an ANOVA analysis. As displayed in Figure 3, deep SSI was diagnosed signi cantly before super cial ones (p=0.008), with a median advance of 9 days. This difference loose statistical signi cance when comparing deep SSI and PJIs (p=0.145).
According to the clinical presentation of SSI following arthroplasty, dehiscence of the surgical wound occurred in 10 patients (40%), 8 patients (32%) developed fever and 8 cases (32%) required a surgical revision. Three out of four cases (75%) of PJI were treated with DAIR; a KLIC score < 4 was registered in all cases with an expected failure rate between 4.5% and 19.4%.
A two-stage revision was carried only in one case (0.25%).
Twelve cases of SSI (48%) needed a new hospital admission and a parenteral antibiotic therapy; a microbiological diagnosis was disclosed only in 9 cases out of 25 (36%), from synovial uid and tissue samples (n=8) and bloodstream (n=1), respectively.
In 5 cases (56%) the etiology of the SSI has been attributed to gram-positive organisms, in 3 (33%) cases to gram-negative ones. A polymicrobial infection was detected only in one case (11%). The most frequently isolated bacterium was S. epidermidis (40%), methicillin and uoroquinolones resistant. Isolated pathogens and resistance patterns are shown in Table 3. All patients completed a full course of antibiotic therapy (2 weeks for super cial/deep SSIs and 10-12 weeks for PJI). None of the patients with a super cial or deep SSIs relapsed during a one-year follow-up. Two out of 4 PJI (50%) relapsed after a conservative DAIR approach (patients relapsed at 40 days and 3 months from the end of the therapy, respectively); relapsed PJI infections were caused by MSSA and P. mirabilis.

Discussion
Surgical site infections represent the most frequent and expensive infectious complication in Europe and USA, as they often require a prolonged hospitalization and sometimes a new surgery, signi cantly impacting on mortality and morbidity (9).
In the present study, we enrolled a relatively large cohort of patients (n=760) undergoing TJA (hip, knee and shoulder) at Santa Maria Maddalena Hospital, a third level regional reference center for orthopedic surgery.
The comparison between our study population and the Italian epidemiological data available (19) shows a lower percentage of deep SSIs and PJI compared to the national infection rates. In fact, in the Italian case series, 54% of SSI diagnosed in orthopedic surgery are classi ed as deep or PJI, while in our population only 40% of SSI match this classi cation.
It is worth noting that none of the revision surgeries carried out at our institution developed a SSI, although literature data points out that the incidence of this complication is signi cantly higher following a revision compared to a primary implant (20). The reason for this difference, however, could be found in the small number of revisions performed (4%), compared to primary arthroplasties (96%).
Both sex and mean age of the study population are demographically homogeneous with the results reported in the "National surveillance system of SSIs" (19); the mean BMI of the enrolled population is 29.10 kg/m 2 , explaining how obesity plays a key role in joint arthrosis.
In the comparison of the two study groups, none of the clinical variables considered reached statistical signi cance. According to literature data, some of the considered variables (eg. BMI, diabetes, steroid and anticoagulant therapy, number of surgeons, duration of surgery, staples) have been associated with an increased risk of developing SSI, an association not found in our study.
On the contrary, the analysis of the main laboratory tests brought some signi cant results: the total leukocytes and neutrophils count at the time of hospital discharge were signi cantly higher in group B (p=0,025; p=0,016, respectively). To our knowledge, the role of total leukocytes and neutrophils at discharge has never been highlighted in literature data as an independent risk factor or the development of SSI following joint replacement.
Cut-off values for total leukocytes>7860/µL and neutrophils> 5185/µL may be able to discriminate the two groups. Although the diagnostic value of these tests appears weak to recommend them as diagnostic tests for SSI, their high negative predictive value (total leukocytes: 98.03%, OR: 3.69; neutrophils: 97.85%, OR: 3.98) suggest their potential role during the post-operative screening, aimed at highlighting patients with a higher risk of developing SSI, alone or as a part of a cluster of variables or a predictive score.
The limit of these two tests lies in the very low positive predictive value (total leukocytes: 6.97%; neutrophils: 7.69%), which suggests the need to integrate further diagnostic exams in the selected population to discriminate infected patients from false positive cases. However, given the crucial onemonth deadline to perform a successful DAIR, we believe that rapidly excluding patients at lower risk and concentrating the efforts on the others could be a useful (and potentially cost-effective) solution in the clinical pathway of TJA follow-ups.
By comparing the timing of SSI diagnosis to the CDC classi cation (7), we found an earlier SSI diagnosis for deep infections compared to the super cial ones (p=0.008) and PJI (p = 0.145), (Figure 3). Even though in the latter case statistically signi cance is not reached, a relationship between the time of SSI diagnosis and the tissue depth of the infection itself hasn't been described yet in literature reports yet, except for PJI that rise later in time (7).
The outpatient examination performed by an infectious disease specialist for SSI surveillance was carried out between the 17th and 25th day post-surgery. Data from the enrolled population show that all diagnosed super cial SSI and PJI fall within this observation period. Deep SSI, on the other hand, anticipated the screening period, since many cases were diagnosed before the hospital discharge.
According to the clinical presentation, hyperpyrexia was recorded mainly in patients with deep or periprosthetic SSI.
Among SSI with microbiological diagnosis (n=9), 4/9 cases were supported by a Gram negative bacterium. This element suggests that contamination of the surgical site may have occurred at a time other than the surgical act, presumably due to wrong management of the dressing after discharge. Risk factors for the development of SSIs from Gram negative bacteria are diabetes mellitus and the presence of urinary catheter (21); none of the patients who developed SSI caused by Gram negative bacteria were diabetic, whereas data regarding the placement of a urinary catheter were not collected in this study. In literature, DAIR treatment for PJI shows success rates varying between 32 and 100% (22). DAIR represents an attractive surgical modality for treatment of PJI although protocols differ in several retrospective series and randomized controlled or prospective trials on this topic are lacking. It is well known that the increase in the chances of success of this procedure depends on a careful selection of patients associated to radicality of the surgical toilette.
In the present study, the KLIC score used to stratify the risk of relapse in patients undergoing DAIR has shown poor accuracy for outcome prediction even in patients de ned at "low risk" (KLIC score <4); according to our opinion, the main limit of the score is represented by the lack of a microbiologic parameter. Relapsed periprosthetic infections, in fact, were sustained by virulent pathogens (MSSA and P.mirabilis, respectively).
Main limitation of our study is represented by the poor PPV of total leukocytes and neutrophils value at discharge that may discourage its application as possible clinical tests to discriminate patients at risk to develop future SSI. This may be partially due to the low sample size of the group of patients presenting with SSI, which unbalanced the study. However, we consider it as a good starting point for studies on larger populations to replicate our ndings.

Conclusions
Two clinical variables, total leukocytes and neutrophils at discharge, seem useful to identify a population at risk for the development of SSIs. Further studies on larger populations are needed to develop a predictive SSIs risk score that should include those variables. In addition, our case series, though limited, seems to discourage the use of KLIC score in its present form as as risk strati cation tool in DAIR. Further studies are needed to validate a successful prognostic score for DAIR procedure that consider the addition of bacterial-related variables, such us type of isolated strains and antimicrobial resistance pro les, when available.

Declarations Fundings
This research did not receive any speci c grant from funding agencies in the public, commercial, or notfor-pro t sectors.
Ethics approval and consent to participate The study was approved by the local ethic committee "CESC VR-RO". Written informed consents were obtained from all participants before their enrolment.

Consent for publication
All the enrolled subjects had signed a written informed consent for publication.

Availability of data and materials
All the data included in the manuscript are available from the authors upon reasonable request.

Competing interests
The authors declare that they have no competing interests. ROC curve for the two tests: total leucocytes at discharge and neutrophils at discharge.

Figure 2
Variation of total leucocytes (A) and neutrophils (B) in the two groups between the pre-operative and at discharge measurements. Time to diagnosis of SSI (days from the operation) in relation to SSI classi cation.