Risk Factors for Bloodstream Infections Among an Urban Population with Skin and Soft Tissue Infections: A Retrospective Unmatched Case-Control Study

This retrospective, case-control study was conducted at two major medical centers located in downtown Detroit, Michigan: the Detroit Medical Center (DMC) and the Henry Ford Health System (HFHS). Both medical centers have centrally located campuses and therefore service a large section of the inner-city Detroit patient population. The population of interest consisted of patients hospitalized with CO-ABSSSI diagnosed within 48 h of admission, as previously defined [16] between January 2010 and December 2015. Patients were excluded if they had any of the following concomitant diagnoses because these infections tend to have a higher likelihood of poor patient outcomes irrespective of concomitant BSI: [17] osteomyelitis, septic arthritis, prosthetic device-related infection, an animal or human bite, burn, necrotizing fasciitis, pyomyositis, or ABSSSI secondary to recent (within 2 weeks) surgery. Cases were defined as inpatients with at least one positive blood culture meeting the Center for Disease Control and Prevention (CDC) BSI criteria [18] collected within 48 h of admission and in whom ABSSSI was documented as the primary source. Study controls consisted of inpatients with CO-ABSSSI and blood cultures collected within 48 h of admission demonstrating no microbiologic evidence of BSI.

Both case and control patients were identified using International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] diagnostic codes via the DMC and HFHS electronic medical records (EMRs). Data elements including demographics, comorbid conditions, prior hospitalization (180 days), history of S. aureus infection (60 days), prior ABSSSI (60 days), recent antibiotic exposure (30 days), vital signs (using the worst physiologic values within 24 h of admission), laboratory values, microbiology and antibiotic treatment were extracted from the EMR. All cases were screened and entered in the secure database, REDCap (Research Electronic Data Capture) [19], an electronic data capture tool hosted by Wayne State University by a trained research assistant. Relevant data in the EMR were used to determine whether the patient met the criteria for ABSSSI according to FDA definitions.

Income and education levels were derived from the Census Bureau of Statistics by the patient’s zip code. Severe ABSSSI was defined as need for incision and drainage or wound debridement in the operating room, amputation, admission to the intensive care unit (ICU) or severe sepsis, as previously defined [20]. Complications secondary to ABSSSI infection were defined as the development of endocarditis, osteomyelitis, and/or other sites of metastatic infection not present on admission. Thirty-day post-discharge vital status was ascertained from in- and outpatient records and categorized as deceased or alive with or without readmission. Patients with no follow-up visits or readmissions within 30 days of discharge were assumed to be alive without readmission. Readmission was considered ABSSSI-related based on physician documentation. This study was approved by the institutional review boards (IRBs) at Walden University (IRB# 0303767) and Wayne State University (IRB# 010616M1E), with the latter serving as coordinating IRB for the DMC review committee and HFHS IRB. A waiver of informed consent was granted. The study was performed in accordance with the 1964 Helskinki Declaration and its later amendments.

The a priori minimum sample size for this unmatched 1:1 case-control study was calculated using the statistical test for differences in proportions [21]. A sample size of 187 case and 187 control subjects was required to achieve 80% power at a 0.05 significance level to detect an adjusted odds ratio (aOR) of ≥ 2 for a risk factor with a known prevalence of 10–20% (diabetes, renal failure, prior antibiotics, recurrent infection) [17, 22‐24].

Descriptive statistics were used to characterize cases and controls. Continuous variables were compared by Student's t test if normally distributed and Mann-Whitney U test if the normality assumption was not met. Categorical variables were evaluated using Pearson’s chi-square test or a two-tailed Fisher’s exact test, as appropriate. Explanatory variables that were associated with case status (P ≤ 0.1) in the unadjusted analyses were included in a multivariable logistic regression model. A backwards-stepwise elimination approach was used to eliminate the variable with the largest P value at each step. The Hosmer-Lemeshow statistic was used to evaluate the predicted and observed probability to test for evidence of a lack of a model fit [25]. Relationships between the explanatory variables and BSI were evaluated using the odds ratio (OR) and corresponding 95% confidence interval (CI). An identical procedure was used to identify a set of risk factors independently associated with increased risk of severe ABSSSI. All calculations were performed using IBM SPSS Statistics software, version 22.0 (IBM SPSS, Armonk, NY).

This study had a retrospective study design and therefore was subject to all of the potential biases associated with this type of investigation. For example, it was possible that there were missing data or incomplete data recorded in the EMR. While it appeared that incomplete or missing data were minimal during the data collection, it is possible that some data were not recorded or were recorded incorrectly. Another potential limitation was the assumption that in patients with the diagnosis of cellulitis and BSI, the S. aureus identified in the bloodstream originated from the patient’s cellulitis. Cellulitis is a deep-seated infection of the subcutaneous tissues of the skin, and therefore it is difficult to culture the pathogen causing disease. While S. aureus is a pathogen associated with cellulitis, other bacterial pathogens such as streptococci are also a frequent cause of cellulitis [2, 10]. Therefore, although unlikely, cases of BSI where cellulitis without abscess was identified as the infection type may have been misclassified. Additionally, we evaluated the impact of race and ethnicity on severity of disease. The composite definition for severity included amputation. However, no amputations were recorded in the population cohort. It was possible that the exclusion of diabetic foot infections from this study is the most likely reason for the lack of amputations because surgical intervention including amputation is common among this population. Further, one of the objectives of this investigation was to determine whether sociodemographic characteristics were associated with ABSSSI + BSI. However, this information was derived by zip code that was provided through Federal census data. Information on education and level of income was not available on an individual basis; therefore, the information was limited. We observed that patient outcomes were poorer among those with bacteremia. We did not further investigate the impact of treatment factors, such as drug choice or time to effective therapy, because it was beyond the scope of our research objective; however, future studies investigating the impact of management strategies on outcomes specifically in ABSSI +BSI would be valuable. Unfortunately, the time from symptom onset to presentation was not consistently documented, and therefore we were unable to analyze this factor. Lastly, the study was conducted on a population of mostly African American individuals who resided in Detroit or its immediate surrounding areas. Therefore, it is possible that the results of this investigation cannot be extrapolated to other metropolitan populations.