Risk Factors for PJI (Limited Evidence)
Limited strength evidence supports that patients in which one or more of the following criteria are present are at an increased risk of periprosthetic joint infection (PJI) after hip and knee arthroplasty:

• Cardiac disease (arrhythmia, CAD, congestive heart failure, other)
• Immunocompromised status (other than HIV), including transplant, cancer
• Peripheral vascular disease
• Inflammatory arthritis
• Prior joint infection
• Renal disease
• Liver disease (hepatitis, cirrhosis, other)
• Mental health disorders (including depression)
• Alcohol use
• Anemia
• Tobacco use
• Malnutrition
• Diabetes
• Uncontrolled diabetes

Rationale

Out of 143 studies which met inclusion criteria for this recommendation, there were no high strength studies, and only ten were considered moderate strength. The remaining studies were considered low strength. Many of these studies evaluated multiple variables as it applied to either PJI directly, or to other perioperative complications, necessitating the use of various statistical methods to attempt to control for these variables. These methods were frequently constrained in doing so. Additionally, based on the available literature, whether a risk factor is “modifiable” or if optimization of a listed condition affects the risk of infection remained unclear. No specific threshold value could be endorsed for most listed conditions, but instead the simple binary presence or absence of the condition as defined by the individual study criteria affected the risk of PJI. Furthermore, many of the studies were based on national payor databases or registries, whose data is only as accurate as the data being input. As such, there is often no way to verify that an individual diagnosis or the definition of the diagnosis is accurate, though the sheer numbers in such databases may help correct for errors in diagnosis. Finally, definitions and clarity of location of infection introduced ambiguity between studies, some clearly indicating PJI, and others with variations of “deep infection,” “deep surgical site infection,” or “involves deep soft tissue,” among other delineations.

 

Limited Strength:

Alcohol
Only low-quality studies met inclusion criteria. Four of these six studies suggested an increased risk of hip and knee PJI in patients with history of alcohol abuse. Two others demonstrated no effect. Wu and associates performed a multivariate conditional logistic regression analysis on a cohort of 297 patients, identifying a nearly 3-fold increase (odds ratio 2.95, p=0.039) in risk of PJI in patients with alcohol abuse. However, the study did not provide definition of alcohol abuse, including whether this was current abuse, or a history of abuse. Rotevatn et al queried the Danish Anesthesia Database, identifying 30,799 patients, and stratified self-reported current alcohol consumption. Their findings suggest that patients who consume 168 to 252 grams of alcohol per week (14 grams of alcohol is typical of 1 beer or 1 glass of wine for reference) had a 1.55 hazard ratio for PJI versus those who did not report consumption of alcohol. But there was a significant portion missing data for tobacco use, not fully controlled for by the study.

Preoperative anemia
One moderate quality study by Greenky and associates (Greenky M, 2012) demonstrated increased risk of PJI in patient with anemia, defined as hemoglobin <12 g/dL in women, and <13 g/dL in men. This single-center study of 15,221 hip and knee patients identified an odds ratio of 1.95 for PJI in patients who were anemic, using a propensity score adjusted model, though the study does not disclose the specific variables that were controlled for. Of seven additional low-quality studies, Lu et al (Lu, M, 2017) identified a higher risk of deep infection in both hip and knee patients with anemia, two (Bozic 2012 and Lee 2015) showed increased risk in knee patients alone, and one (Bozic 2012) demonstrated increased PJI risk in hip patients alone. But one (Bozic 2014) demonstrated no effect in hip patients, and two others identified no effect on either hip or knee patients. These conflicting conclusions between low-quality studies, and paucity of any better-quality studies qualifies as limited evidence.

Cardiac disease (arrhythmia, CAD, congestive heart failure, other)
Fourteen low quality studies evaluating cardiac disease as it applies to PJI met inclusion criteria. Of these, four identified no increased risk of PJI in hip and knee patients, two in hip patients alone, and two in knee patients alone. However, this was in conflict with five other studies that suggested higher risk of PJI. Three studies demonstrated an increased risk in knee patients, including Lee and associates who suggested a 5 times higher risk of PJI in patients with “cardiac disease”; Long et al, who identified 2.1 higher risk in patients with atrial fibrillation; and Bozic and associates, who identified a 1.28 hazard ratio in patients with congestive heart failure. In another study, Bozic et al (Bozic 2012) identified an adjusted hazard ratio 1.30 in favor of PJI in hip patients with cardiac arrhythmia. Aggarwal and associates (Aggarwal 2013) showed a 9 times higher risk of hospital admission for PJI in both hip and knee patients with history of atrial fibrillation versus those without atrial fibrillation. Finally, one retrospective study (Tabatabae 2015) actually demonstrated a reduced risk of in-hospital wound healing complications (hematoma or seroma) in patients having undergone coronary revascularization (CABG, stenting).

Diabetes
One moderate strength study and 36 low strength studies evaluating the effect of diabetes and/or uncontrolled diabetes met inclusion criteria. In the 2010 moderate strength study, Pedersen and colleagues evaluated the effect of diabetes and diabetes with associated comorbidities on the rate of revision for deep infection in 57,575 patients in the Danish Hip Arthroplasty Registry, who underwent THA from 1996 to through 2005. Type 1 diabetics had a clinically insignificant higher risk of revision for deep infection, versus those without (rate ratio 1.01), while type 2 diabetics had a 1.49 times higher risk. Diabetic patients with history of complications in general related to their disease state had a 2.11 times higher risk than patients without diabetes, and those with cardiovascular comorbidities and diabetes had a 2.35 higher risk. It was unclear as to whether all important confounding variables were accounted for sufficiently.

Four low strength studies revealed increased risk of PJI in both diabetic hip and knee patients. Jiang and associates identified a 1.32 times higher risk of PJI based on diagnosis of diabetes in a study based on national and state-level databases of over 800,000 THA and TKA patients. In contrast, Wu and colleagues performed a hospital-based case-control study, including 297 patients, predicting a 5.47 odds ratio in favor of PJI for patients with diabetes, versus case controls in patients undergoing THA or TKA in China. In insulin-dependent diabetics, the odds ratio was lower than diabetics in general, at 3.69. In a study evaluating the risk of PJI in hip and knee arthroplasty patients with preoperative asymptomatic eucocyturia, Gou and associates incidentally found that 6 out of the 7 patients with early PJI – a out of a total of 739 patients – had diabetes, representing a significant logistic regression odds ratio of 69.65. Finally, Jamsen and associates reviewed the one year incidence of PJI in a single-center series of 7181 primary hip and knee arthroplasties performed for osteoarthritis, evaluating the effect of obesity, diabetes, and preoperative hyperglycemia. They found that patients with preoperative diagnosis of diabetes had an associated odds ratio of 2.31 for PJI. Even patients without diagnosis of diabetes, but with preoperative blood glucose of 124 mg/dL or higher had an adjusted odds ratio for PJI of 3.3 versus those with a blood of less than 124. Aside from this one numeric value, none of these four studies provided parameters on the diagnosis and severity of disease (e.g. blood glucose, hemoglobin A1C, etc.).
 

In patients with uncontrolled diabetes, three studies of low strength demonstrated increased risk of PJI in both hips and knees, and two in knees alone. Chrastil and colleagues found that hemoglobin A1C (HbA1c) did not perfectly correlate with PJI risk, but perioperative hyperglycemia with maximum preoperative blood glucose   level of 194 or higher had a hazard ratio of 1.44 in their study of 13,372 Veterans Affairs patients. Additionally, patients with history of diabetic complications had an HR of 1.113. Shohat and associates looked at serum fructosamine in evaluating risk of PJI in diabetic patients, identifying that patients with a serum fructosamine of 292 mmol/L or higher had more than 6 times the risk of PJI (odds ratio 6.2) than those under 292 mmol/L. However, there was concern that use of a stepwise regression model without validation could increase the likelihood that there was inadequate control of confounders. The authors note that the benefits of use of fructosamine include its low cost and its reflection of mean glycemic control over a shorter time period than HbA1c. In the final study, reported in 2017, Tarabichi and associates performed a retrospective multicenter study of 1645 diabetic patients undergoing primary THA or TKA, evaluating HbA1c levels as a predictor for adverse events. Their stepwise logistic regression analysis suggested that an HbA1c of 7.7% (ROC with AUC 0.65) was associated with an increased rate of PJI, from 0.8% with HbA1c less than 7.7, to 5.4% with HbA1c >7.7, with AUC of 0.65.

Three additional studies on uncontrolled diabetes found no difference in either PJI or wound complications in   hips or knees, and two found no difference in PJI or deep infection in knees alone.  Five studies demonstrated increased risk of PJI in patients with diagnosis of diabetes alone, in TKA only, and two in THA, with an additional study (Song 2012) indicating increased risk of “deep incisional and/or organ space infection” in THA, and one (Namba 2013) “deep” infection in knee patients. The remaining 20 low strength studies demonstrated no significant differences in occurrence of various extent of infection.

Immunocompromised status other than HIV, including transplant, cancer
There were 12 included low    strength studies evaluating the effect of immunocompromised status (other than HIV, see below) on risk of PJI. Three of these indicated higher risk of PJI in hip and knee patients, and one identified higher PJI incidence in knee patients. According to the findings of Klement et al (Kelment, 2016), kidney, liver, heart or pancreas transplant recipients undergoing hip replacement had higher risk of PJI, at relative risk of 1.56, 1.6, 1.82, and 1.31, respectively. In a review of 2,579,694 patients from the Nationwide Inpatient Sample (NIS) between 1993 and 2011, Cavanaugh et al (Cavanaugh, 2015) identified a higher risk of in-hospital wound healing complications in hip and knee patients with history of heart, lung, or pancreas transplant versus no history of transplant, with an odds ratio of 2.13. The remaining six studies identified no difference in PJI in patients with history of transplant.


Inflammatory arthritis
There were 22 included studies assessing risk of PJI as a function of inflammatory arthritis. In the single moderate strength study by George et al (George, 2017), timing of infliximab cessation was assessed in patient with inflammatory arthritis in over 4288 patients in a Medicare database. While there was no significant difference in PJI risk in patients at various stoppage times before surgery, there was 
increased risk in patients with inflammatory arthritis using glucocorticoids (hazard ratio 2.7, in favor of PJI). Out of the 21 low strength studies, one demonstrated increased risk of PJI in TKA and THA (Bongartz 2008), one for TKA alone (Bozic 2012), two for THA alone (Bozic 2012, Triantafyllopoulos 2016), and eight others demonstrated increased risk of various delineations of infection, including several that indicated higher rates of    late revision for infection (>6 years).

Malnutrition – One study on the effect of malnutrition on risk of PJI was of moderate strength, and demonstrated no increased risk of deep infection in TKA patients. In this study, Wagner and associates (2016) reviewed data from 22,289 consecutive knees from their institutional joint registry, finding that TKA patients with BMI less than 18 had nearly twice the risk of deep infection (hazard ratio 1.96[95% CI .42 to 9.14]). Four low strength studies evaluated malnutrition in the primary joint arthroplasty population. Manrique et al (Manrique, 2017) found TKA patients with BMI less than 18.5 at an over 23 times higher risk of deep infection (odds ratio 23.3), but also note a higher incidence of rheumatoid arthritis in their study population than the general population, potentially confounding the result. One by Gramatico-Guillon et al (Gramatico-Guillon, 2015) identified increased risk of PJI in TKA and THA patients. In a study of 4551 patients undergoing right TKA in the NSQIP database, Kamath and associates (Kamath, 2016) identified that patients with preoperative albumin levels below 3.5 mg/dL were at an increased risk of deep incisional (odds ratio 2.3) and deep organ space infection (odds ratio 3.79) than those with albumin 3.5 or higher. Zorrilla and colleagues (Zorrilla,2006) identified that patients with low serum zinc had a higher incidence of delayed wound healing after total hip arthroplasty. Huang and associates (Huang, 2013) evaluated hip and knee patients and found that those with low albumin or transferrin levels had greater odds of acute infection within 3 months of surgery (odds ratio of 2.37[95% CI .73 to 7.76]), which was statistically insignificant, but the study was likely underpowered due to a low event rate.

Mental health disorders, including depression
Three included low strength studies evaluated at least one  component of mental health. Bozic and colleagues reported on 587 unilateral THA at 5 clinical sites, identifying an adjusted hazard ratio of 1.96 for PJI in patients with diagnosis of depression (Bozic 2014). Using the Medicare 5% sample claims database,40,919 patient who underwent THA between 1998 and 2007 were evaluated by Bozic et al (Bozic 2012) reporting a hazard ratio of 1.38 with concomitant diagnosis of depression and 1.48 for psychosis.  Finally, using similar methods and the same database, this time for 83,0111 TKA performed between 1998 and 2007, they reported a 1.28 hazard ratio for depression, and a similar 1.26 for psychosis.


Liver disease (hepatitis, cirrhosis, other)
Fourteen low strength studies assessed the effect of liver disease on PJI, and one moderate strength study specifically evaluated the risk of deep PJI in patients with cirrhosis. Most seemed to indicate a roughly two-fold higher rate of PJI in patients with liver disease. The moderate strength study by Deleuran et al (Deleuran, 2015) compared PJI rates for TKA and THA in 363 patients with cirrhosis versus 109,159 patients without cirrhosis in the Danish healthcare registries, identifying twice the rate of infection (hazard ratio 2.1) in patients with cirrhosis. As is a frequent issue with registry data, comorbidities were measured retrospectively through diagnosis codes, whose validity is unclear. Additionally, alcohol
 consumption was not a measured variable and not controlled for, though authors identify that cirrhotic patients currently drinking would not have been offered joint replacement in their setting. Nonetheless, the remaining   low strength studies also favored a higher risk of PJI in patients with liver disease, with higher rates of hip and knee PJI in patients with presence of viral hepatitis in two studies (Jiang et al 2014, Kildow 2017), as well as knee PJI in another (Kuo, 2016). Three studies predicted higher risk of PJI with presence of liver disease in both hip and knee patients (Grammatico-Guillon 2015, Kao 2017, and Cai 2014). Grammatico-Guillon et al found a hazard ratio of 2.88 in patients with diagnosis code of liver disease based on a French database of 32,678 patients. A low strength study of all 255,568 Taiwanese residents who underwent TKA or THA between 1997 and 2009 by Kao and associates identified a similar 2.09 adjusted hazard ratio in favor of PJI in patients with diagnosis of chronic liver disease. In a case-control study of 903 patients undergoing TKA or THA, Cai and associates identified an odds ratio of 7.03 for PJI in patients with liver disease versus those without. The remaining 8 low strength studies found no significant difference for patients with liver disease, and no study identified a lower risk of PJI.

Peripheral vascular disease
Six low strength studies evaluated peripheral vascular disease (PVD) as a risk factor for PJI. One demonstrated increased risk in THA and TKA patients (Jiang 2014), one for TKA patients alone (Bozic 2012), and one for THA patients alone (Bozic 2012). The remaining three showed no difference.


Prior joint infection
There was one moderate strength study and one low strength study identifying increased risk of PJI in patients with history of prior prosthetic joint infection. In the moderate quality study Bedair and   colleagues reported a 21 times higher relative risk of PJI in subsequent joint replacements when a patient had a history of PJI in a previously replaced hip or knee (Bedair, 2015). Intuitively, Mortazavi and associates
 demonstrated a higher risk of infection after revision for infection than for those with aseptic revisions (Mortazavi, 2010).

Renal disease
Data from one moderate and fourteen low strength studies were conflicting as it applied to 
renal disease, but three of the low strength studies (Bozic 2014, Grammatico-Guillon 2015, Bozic 2012) indicated higher risk of PJI in hip, hip & knee, and knee patients, respectively. An additional low strength study by Tan and associates (Tan, 2016) identified a higher risk of revision for PJI in hip and knee patients. In the one included moderate strength study, Miric and associates (Miric, 2014) reported on the results of 20,720 patients who underwent hip replacement from an integrated healthcare system database. Defining deep SSI as attributable to index THA up to 365 days post-operatively, they found no statistically significant difference between those patients with chronic kidney disease (CKD) or end stage renal disease (ESRD), and those without renal disease, but were unable to adjust for confounding variables.  This was true of many of the   studies, and many patients with more advanced stages of renal disease seemed to have higher incidences of other comorbidities. There were no studies that predicted a lower risk of PJI in patients with renal disease.

Tobacco
 All twelve studies that met inclusion criteria for tobacco and associated risk of PJI were considered low strength. One study indicated higher risk of “deep soft tissues or any part of the anatomy,” including organ space in hips, but not knees (Sahota, 2018).Two studies demonstrated increased risk of “deep infection” (not specifically PJI) in hip and knee patients who use tobacco, and another found increased risk of wound healing complications, including dehiscence and deep wound infection (Duchman, 2015).


POSSIBLE RISKS AND HARMS OF IMPLEMENTATION
Candidacy for surgical intervention is at once, one of the most essential and complex decisions in surgical practice, ethically balancing the degree of pathology and risk of the operation, with the positive benefits to the patient and society at large, in a shared decision-making process between patient and surgeon. The positive effect of indicated lower extremity arthroplasty on quality of life and reduced morbidity is well established. However, its cost to the health care system has been the subject of increased scrutiny. Surgeons are under increasing pressure from payors, health care systems, and peers alike to provide highest value care, balancing the overall cost of care with excellence in patient outcomes. As such, delay or denial of surgical care based on any one or multiple factors in order to avoid one of the most devastating complications of one of the best value surgeries ever practiced should not be taken lightly. Given that all risk factors listed – with the exception of obesity – are supported with only low strength or conflicting evidence, the decision to proceed is individual, and based on myriad other factors, including but not limited to the ability of the system in which the surgeon practices to handle varying degrees of complexity, volume and experience of the surgeon and system, and other confounding factors not herein assessed. Additionally, at this time it is unclear based on the literature if modification of any risk factor, including obesity, actually reduces the risk of PJI. Payors and healthcare systems alike should understand that though tactics to reduce cost may include delaying or avoiding operating on patients with these risk factors, such practice may deny surgery to a much larger proportion of patients who may otherwise significantly benefit and not endure PJI.

FUTURE RESEARCH
Despite the volume of literature addressing risk factors for periprosthetic joint infection, there is a paucity of moderate-quality studies, and complete absence of high-quality studies. Future research must attempt to better control for individual confounding variables prospectively, with better delineation of disease states. For example, though BMI may not be the best measure of obesity overall, its stratification in many studies has helped allow for better comparison between groups, improving the quality of data available. Simply identifying whether or not a disease process is present based off an individual entry of a diagnostic code from the patient’s potentially remote past medical history does not ensure best quality data. Unfortunately, the relatively low incidence of PJI requires large numbers for appropriate statistical power, making registries and large healthcare databases an optimal target for research. Better quality abstraction for such databases is therefore necessary to help de-confound. Additional assessments of markers of disease status and their associated thresholds may also help the clinician further and more accurately stratify risk. Finally, identification of risk associated with a condition or stage of comorbidity does not by itself afford the provider the ability to proselytize for change, as the effect of modification and optimization of the status of a listed condition is still unclear. Future research endeavors should specifically be designed to determine if risk factor modification truly results in a reduction in the risk for PJI after hip or knee arthroplasty surgery. Given frequently conflicting conclusions among studies, the individual system and even provider-specific management of comorbidities – which was typically not delineated – may account for such discrepancies. Prospective, appropriately controlled studies incorporating these considerations will better afford surgeon and patient the ability to predict and potentially minimize risk of periprosthetic joint infection.