An update review on risk factors and scales for prediction of deep sternal wound infections

Abstract

Surgical site infections are the most common nosocomial infections in surgical patients. The preventable and the unmodifiable risk factors for deep sternal wound infections (DSWI) have been amply assessed in the literature. The aim of this review was to describe the results of the numerous published studies to describe all the DSWI risk factors and the scales devised to predict SWI, with a view to providing an update on this issue. A comprehensive search of the Medline and Embase databases was performed (considering studies from January 1995 to April 2011); and a manual search was also conducted using references cited in original publications and relevant review articles. There are several risk factors associated with DSWI, which could be classified in four categories as demographic (e.g. sex and age), behavioural (e.g. smoking and obesity), baseline clinical conditions (e.g. diabetes, hypertension and COPD) and surgical operative risk factors (e.g. duration of operation and emergency operation). Six scales for predicting the risk of DSWI are described in the literature: they vary not only in accuracy but also in ease of application and they are applied at different times (some only preoperatively and others also postoperatively). This study provides a broad update on our knowledge of the risk factors for DSWI and the scales for prediction with a view to improving the management of infections at cardiosurgery units.

INTRODUCTION

Surgical site infections (SSIs) are the most common nosocomial infection in surgical patients and wound site infections are a major cause of postoperative illness, accounting for approximately one in four of all nosocomial infections (1). SSIs can be considered the most common preventable adverse outcome after major surgical procedures and the second most common adverse event occurring in hospitalised patients. An estimated 40–60% of these infections are thought to be preventable (2). The Center for Disease Control and Prevention defines deep sternal wound infections (DSWI) as infection involving incisional deep soft tissue within 30 days of the operation (3). The reported incidence of DSWI ranges from 0·7% 4, 5 to 2·3% (6). This is a potentially life‐threatening complication and is associated with a higher morbidity, including a higher incidence of postoperative myocardial infarction, stroke and low output syndrome (7) and a higher early mortality 8, 9, while its influence on long‐term mortality is debated in the literature 8, 10. These detrimental health outcomes are a primary concern and make it essential to consider all removable or containable risk factors for DSWI. The consequences of such infections also entail an increase in health care costs because of a greater need for parenteral antibiotic therapy, longer hospital stays, postoperative resternotomies. Both the preventable and the unmodifiable risk factors for DSWI have been amply discussed in the literature to shed light on possible prevention strategies as well as to identify patients carrying the greatest intrinsic risk of DSWI 4, 11, 12, 13. Several scales 14, 15, 16, 17, 18, 19, 20 for predicting the risk of SWI have been defined, in addition to the general National Nosocomial Infection Surveillance (NNIS) System risk index 21, 22, which is thought to perform well across a broad range of surgical procedures, but has been judged inadequate in stratifying patients undergoing coronary artery bypass surgery in terms of their risk of acquiring a SSI (23).

The aim of this review was to describe the results of the numerous published studies to pinpoint all DSWI risk factors and the scales for predicting them with a view to providing an update on this issue.

METHODS

Search method

A comprehensive search of the Medline and Embase database was performed (considering studies from January 1995 to April 2011) and a manual search was also conducted using references cited in original publications and relevant review articles. The search process involved combining the MeSH terms ‘deep wound infection’, ‘risk factors' and ‘prediction scale’. These keywords were expanded in the search so as to include all articles investigating the same issue using different terms (i.e. ‘sternal wound infection’, ‘prospective study’ and ‘epidemiology’).

Study selection and variables

Each publication identified by this process was reviewed and included in this study if the following criteria were met: (i) observational, not experimental studies; (ii) studies published after 1995; (iii) studies not designed to validate scales and (iv) studies published in English. When the results of a study were published more than once, only the latest and most complete article was included in the analysis.

The data were tabulated by name of the article's first author, year, place, study design, demographic characteristics of the sample (mean age, gender as the percentage of males), pathological and therapeutical conditions, behavioural factors, preoperative‐intra‐post‐surgical conditions. Finally, we tabulated any reported early mortality and mortality at 1 or more years.

The criteria for defining the accuracy of prediction scales were drawn from arbitrary guidelines (based on a suggestion by Swets 1988) (24), distinguishing between non informative (AUC^0·5), scarcely accurate (0·5 < AUC < 0·7), moderately accurate (0·7 < AUC < 0·9), highly accurate (0·9 < AUC < 1) and perfect tests (AUC^1). For the Goodman–Kruskal non parametric correlation coefficient, Haley suggests that a risk index with a G value of less than 0·3 has a poor predictive power, 0·3–0·5 has a moderate predictive power and more than 0·6 has a high predictive power (25).

RESULTS

The characteristics of the studies considered in the analysis are given in Table 1.

Table 1.

Characteristics of the studies considered in the analysis

Author (year of publication)	Type of study	Period	Location	Number of patients	Control group	Mean age	Male (%)	Number of deep sternal wound infections (%)	Early mortality (30‐day mortality)	Mortality at 1 year
Sofer (1999) (26)	Retrospective observational	1996/04 to 1997/08	Israel	545		65	79	9 (1·7)	11 (2%)
Nakano (2008) (27)	Prospective cohort study	2000 to 2005	Japan	1500		67·8	72·7	12 (0·8)
Olsen (2002) (11)	Retrospective	1996/01 to 1999/06	USA	1980		61	64·9	37 (1·9)		8 (21·6%)
Trick (2000) (13)	Case control	1995/01 to 1998/03	USA	1796	90		n.r.	30 (1·7)
Toumpoulis (2005) (8)	Retrospective study	1992/01 to 2002/03	USA	3760		64·1	69·1	40 (1·1)	6 (15%)
Zacharias (1996) (28)	Retrospective study	1991/01 to 1994/12	Canada	2317		63	n.r.	21 (0·9)
Borger (1998) (4)	Retrospective study	1990/02 to 1995/12	Canada	12 267			72·1	92 (0·7)	1 (1%)
Hollenbeak (2000) (29)	Case control/cost analysis	1996/04 to 1997/08	USA	1519	160	64·8	77·7	41 (2·7)
Lu (2003) (5)	Retrospective study	1997/04 to 2001/03	UK	4228		63·5	82·4	28 (0·7)	7 (25%)
Ridderstolpe (2001) (30)	Retrospective cohort study	1996/01 to 1999/12	Sweden	3008		65·4	72·5	47 (1·6)	1 (1%)	7 (7·2%)
Kohli (2003) (31)	Prospective cohort study	1990/04 to 1995/12	Canada	11 508		60	73	133 (1·2)	(1.8%)	(5.6%)
Filsoufi (2009) (9)	Retrospective cohort study	1998/01 to 2005/12	France	5798		64	62	106 (1·8)	15 (14·2%)	27·6%
Fakih (2007) (32)	Retropective + case control	2000/01 to 2004/09	USA	3578		67·2	65·1	70 (1·9)
Steingrimsson (2004) (33)	Case control	1997/01 to 2004/12	Iceland	1650	163	66·2	72·4	41 (2·5)	4 (9·8%)	7 (17%)
Sakamoto (2003) (12)	Retrospective study	1987/01 to 1999/12	Japan	863		61·2	65·5	17 (2·0)
Savage (2007) (34)	Retrospective cohort study	2002 to 2004	USA	120 793		64·6	67·7	2018 (1·7)
Sachithanandan et al. (2008) (35)	Prospective cohort study	2001/01 to 2005/12	UK	4586		63·4	74	76 (1·7)	7 (9·2%)
The Parisian Mediastinitis Study Group (1996) (6)	Prospective cohort study	1993/03 to 1993/04	France	1830			68·6	42 (2·3)

The risk factors associated with DSWI are shown in 2, 3.

Table 2.

Demographic, baseline clinical condition and behavioural factors

Table 3.

Preoperative, intraoperative, and postoperative factors

The six SWI risk prediction scales found in literature are represented in Table 4.

Table 4.

Scales for predicting the risk of SWI

DISCUSSION

Many studies described four categories of risk factors that raise the risk of DSWI, that is, patients' demographic features, clinical conditions, behavioural factors and preoperative– intra‐post‐surgical conditions.

Demographic features

Age

A correlation between the risk of DSWI and age emerged in many studies, older age being associated with a higher probability of preoperative comorbidities and postoperative complications (36). Many age‐associated comorbidities increase a patient's sensitivity to infections, although malnutrition seems to be the main cause of a worse immune function in the elderly (37).

Sex

Males have more hair follicle, which can facilitate major microbial colonisation of the skin, while preoperative shaving can cause local skin irritation.

Baseline clinical variable and therapeutical conditions

Diabetes

Diabetes predisposes patients to DSWI by two main routes, first by delaying the healing of the surgical wound, and second by reducing the patient's immune response. Diabetes has a negative impact on the healing process through a number of mechanisms. During the inflammatory phase, diabetes‐related vascular changes such as capillary basement membrane thickening, limited vasodilation and platelet dysfunction result in poor leukocyte infiltration, inadequate oxygenation and clotting dysfunction. In the proliferative phase, a decreased presence of fibroblasts, collagen and growth factors within the wound further delays the essential matrix formation needed to fill the wound (38). Delayed wound closure during the maturation phase is evident, with prolonged wound granulation and reduced myoblast contractility. Diabetes makes the tissues a more favourable environment for micro‐organisms by damaging lymph, nerve and blood vessels.

Some research has also shown that diabetic patients have an impaired immune function: their neutrophil function is depressed, affecting adherence to the endothelium, chemotaxis and phagocytosis (39). The antioxidant system involved in bactericidal activity is weakened and cell‐mediated immunity is depressed. These impairments are exacerbated by hyperglycaemia and acidaemia, but are substantially (if not entirely) reversed by normalising pH and blood glucose levels (40).

Diabetes control demands a contribution from more than one specialist. The anaesthetist needs to monitor patients' serum glucose before the surgical procedure and plan suitably adjusted intraoperative insulin therapy (41). Surgeons should ensure that proper glucose control is maintained after the procedure (for least 48 hours). Nurses should educate patients to maintain glycaemic control, especially patients whose hyperglycaemic is only discovered at the preoperative workup.

Immunosuppressive treatments or immunosuppressed conditions

It is common knowledge that the immune system plays a crucial part in controlling infection and its suppression for therapeutic purposes (e.g. in patients on chronic corticosteroid or cyclosporine treatment, or chemotherapy) seems to raise the risk of a SWI (33).

Heart failure and COPD

During wound healing, the damaged tissues' continuity and function are re‐established, but this can only happen if the microcirculation is restored and the tissues are nourished (42). A continuous supply of oxygen to the tissues via the microcirculation is vital to the healing process and for resisting infection. A poor cardiac pump function (NYHA ≥3) and chronic obstructive pulmonary disease do not help proper tissue perfusion and oxygenation, they delay the process of wound healing. Oxygen delivery to wounds depends on blood oxygenation and perfusion and the diffusion distance from the blood to the tissue, which relates to the partial oxygen pressure 43, 44. Oxygenation is fundamental to cell motility and activity, and involved in healing inflammation, proliferation, collagen synthesis and angiogenesis. In inflammatory processes, hypoxia leads to anaerobic metabolism and causes acidosis and an ATP production insufficient for maintaining normal cell function, especially because the site of a wound is metabolically so active (45). Neutrophils and macrophages produce enzymes that require oxygen to function properly, and hypoxia consequently blunts their oxidative killing activity (46). In proliferation phases, hypoxia stimulates the expression of type IV collagenase and reduces the expression of laminin‐5, which inhibits keratin cell motility. It was also showed that low levels of ROS, produced when a cell is in hypoxia, as in hydrogen peroxide solution, inhibited the migration and proliferation of keratin cells (47). Molecular oxygen is also essential during collagen synthesis. Hydroxylation of proline and lysine in procollagen is a crucial step in collagen maturation. Procollagen molecules cannot form stable triple helices without hydroxyproline. Hydroxylation requires high amounts of oxygen (48).

Behavioural factors

Smoking

Smokers tend to have a poor wound healing function caused by a reduction in both local flow and tissue oxygen tension, which can delay the healing process, as described earlier (49). Wound oxygenation is also reduced by microvascular obstructions caused by platelet aggregation. Some studies also found that smokers are often more prone to bacterial infections and inflammatory diseases than non smokers, because of the hundreds of toxic components contained in cigarettes (50) and their impairment of the immune system.

Obesity

Obesity is strongly associated with diabetes mellitus and can give rise to technical difficulties and prolonged operating times that can contribute to the onset of DSWI (51). Even in an analysis adjusted for diabetes and duration of surgery, it was found that obesity remained an independent risk factor for DSWI. In the obese, the abundant adipose tissues produce and release a variety of proinflammatory and anti‐inflammatory factors (52). Differences in serum (and tissue) concentrations have been reported for various antibiotics in obesity. The underlying causes of these alterations are due to the physiologic changes that can alter both Vd and drug Cl. The changes observed in GFR are particularly difficult to predict with current methods. Obese patients may be incorrectly dosed with the use of fixed (underdosed) or TBW‐based dosing (overdosed) when the contribution of pharmacokinetic alterations in obesity is unrecognised (53). There is also a greater risk of dehiscence because of straining of subcutaneous sutures. It has been reported that a personalised education and dietary measures before surgery could have positive long‐term effects and reduce the obesity‐related risk of infections (54).

Surgical operative risk factors

Internal thoracic artery. Patients needing a CABG using the internal thoracic artery (ITA) carry a higher risk of DSWI. The perforating branches of the ITA perfuse the sternal region so using the ITA reduces subsequent sternal perfusion, as showed in animal as well as human studies.

Prolonged operating times. A poor tissue oxygenation during a lengthy surgical procedure and ventilation treatment could be a cause of a higher postoperative risk of DSWI, as suggested in some studies (55). A prolonged duration of the procedure could in itself increase the risk of DSWI in cells damaged by drying on exposure to air and surgical retractors. It also predisposes patients to infections directly via the higher likelihood of contamination of the incision or tissue desiccation (56). Longer procedures are more likely to be associated with blood loss, contributing to tissue hypoxaemia and shock, and reducing the patient's general resistance.

Blood transfusions. A large meta‐analysis showed that patients who receive a blood transfusion have a higher risk of postoperative infectious complications than patients who do not (57). Blood transfusions temporarily induce an acquired immune suppression and therefore also a potentially increased risk of postoperative infectious complications. In particular, four possible mechanisms have been proposed (58), including an immunomodulatory effect mediated by immunologically active white blood cells (WBCs) that downregulate the recipient's immune function; soluble biological response modifiers released from WBCs during the storage of blood; and soluble human leukocyte antigen peptides or other soluble mediators that circulate in allogeneic plasma. The fourth possible mechanism is not an immunomodulatory effect, but involves a mechanism by which blood transfusion leads to postoperative organ dysfunction, which in turn predisposes to infection.

Re‐exploration. Re‐exploration for bleeding after surgery and other reasons for an unplanned return to the operating room can increase the risk of infection because they expose patients to further wound attack from airborne microbial agents. Haematoma is also a good pabulum for microbial growth.

Postoperative hypoxaemia or ventilatory support. Low cardiac output syndrome or respiratory impairment after surgery could cause inadequate tissue oxygenation, interfering with the healing process as explained earlier.

Postoperative CPR. Cardiopulmonary resuscitation places the sternum at risk of fracture or dehiscence and the body only diverts the blood to the vital organs (59).

Prediction scales

Our literature review identified six scales for predicting the risk of SWI and another general tool for predicting nosocomial infections (Table 4), which can also be applied to SWIs, that is, the NNIS risk index. The value of these different tools was evaluated by means of the receiver operating characteristic (ROC) curve, which is a fundamental tool for assessing prognostic scales. In a ROC curve, the true positive rate (sensitivity) is plotted as a function of the false positive rate (100 specificity) for different cut‐off points on a scale (60). The area under the ROC curve is a measure of how well a cut‐off can distinguish between two diagnostic groups (SWI+/SWI−). Two of the risk indexes and the NNIS showed a limited accuracy, while the other four specific indexes proved moderately accurate. The latters are more complicated, however, and not very easy to apply. Two of the moderately accurate scales, one of the less accurate and the NNIS all need variables inherent in the surgical and/or postoperative phases, however, and so they can only assess patients after they have undergone surgery. Risk indexes that have to be used postoperatively can only enable a risk adjustment for the purposes of comparing the performance postoperatively, whereas scales for use preoperatively facilitate the preoperative design of prevention strategies (e.g. optimising diabetes control prior to surgery, or adopting a different prophylactic antibiotic regimen). In any case, the only risk scale that enabled a straightforward preoperative assessment was found moderately accurate, but a different statistical approach (not area under ROC curve) was used, preventing a direct comparison of its accuracy with other scales.

In conclusion, this study provides a broad update on what we know about the risk factors for SWI and the risk scales that can be used to manage infection control at cardiosurgery units.