Incidence, microbiology, and patient characteristics of skin and soft-tissue infections in a U.S. population: a retrospective population-based study

Kaiser Permanente of Northern California (KPNC) is a nonprofit, integrated health care delivery system providing care to over 3 million members. The member population reflects the general population in the Northern California region, although, as an insured population, it under-represents persons with very low levels of education and income [32]. Clinical diagnostic and laboratory results data from hospital discharges and ambulatory settings, including emergency departments, are archived in electronic medical record and administrative databases. KPNC databases contain individual patient records and are readily linked using the patient’s unique medical record number which remains with the patient for life. In this retrospective, observational, study, clinical decisions regarding diagnosing and testing reflect usual care by KPNC providers [29]. The study was approved by the KPNC Institutional Review Board.

Using electronic databases, we identified all clinic visits, Emergency Department visits, and hospital-based encounters occurring between January 1, 2009 and December 31, 2011, with a diagnosis (International Classification of Diseases, 9th Revision, Clinical Modification [ICD9-CM]) relating to SSTIs. We selected this three-year period in order to reflect current practice patterns, to reflect a period when MRSA incidence appears to have stabilized [29], and because it was the most recent period for which we had complete diagnostic and laboratory data available for this population. For hospital-based encounters, we only included SSTIs that were listed as the primary diagnosis, thus likely excluding hospital-onset cases – although some cases with a second co-primary diagnoses such as sepsis will also have been excluded. SSTIs were defined by ICD9-CM code as the following: Cellulitis and abscess, comprising abscess of anal and rectal region (566×), abscess of breast (6751×, 6752×), cellulitis and abscess of finger and toe (681×), other cellulitis and abscess (682×), and pilonidal cyst with abscess (6850); Carbuncle and furuncle (680×); Impetigo (684); Other infections of skin and subcutaneous tissue (686×); Folliculitis (7048×); and Other skin and soft tissue infections, comprising erysipelas (035), other mastitis (6752×), acute lymphadenitis (683), hydradenitis (70583), and necrotizing fasciitis (72886). These codes include most of those used by Edelsberg et al [25]. in their analysis of hospitalizations caused by SSTIs, plus erysipelas, abscess of breast, nonpurulent mastitis, folliculitis and hydradenitis. Because our focus was on acute, community-onset SSTIs, and not on nosocomial or chronic SSTIs, we did not include in these analyses chronic/decubitus ulcer, surgical site infection, post-operative wound infection, infection due to device, gangrene, or amputation stump infection.

SSTI episodes were defined as beginning on the date of the first SSTI diagnosis and ending with the last SSTI or complication diagnosis, not followed by another SSTI or complication diagnosis within 42 days. A six week period was considered short enough as to be unlikely to include multiple different episodes, but long enough to extend beyond the time that a patient might be asked to return for a follow-up visit. The start of an episode was defined solely by one of the diagnoses noted above. However, receipt of one or more complications diagnoses could prolong the episode. The following ICD-9-CM codes were considered potential SSTI complications: 038 (sepsis/septicemia), 04082 (toxic shock syndrome), 7280 (myositis), 730× (osteomyelitis), 7854 (gangrene), 78552 (septic shock), 7970× (bacteremia), 99590–99592 (systemic inflammatory response syndrome.) We retained only those episodes where the patient was a KPNC member at the start of the episode.

Patient age and gender were extracted from KPNC membership databases. Although race/ethnicity is not systematically captured for all members, the race/ethnicity of a majority of patients (about 94%) was identified via a search of electronic medical records, membership databases and other administrative databases. Patients were classified as Hispanic if their ethnicity was identified as “Hispanic” in any of these sources. For patients who did not have race/ethnicity recorded in administrative databases, we imputed race/ethnicity using the Bayesian Improved Surname and Geocoding algorithm, which uses census data and surname to impute the probabilities that the patient belonged to each of six different racial/ethnic groups [33].

Patient addresses were extracted from membership databases and were used to determine the patient’s census block group. Using the American Community Survey 2006–2010 Summary File [34], we determined the median block group income for each patient, and classified their SSTIs based on the quintile of the block group median income. Persons whose address information did not map to a valid census block group were classified into a sixth category “unknown”.

Because persons with diabetes are widely considered to be at increased risk of SSTIs [35‐40], we classified SSTI episodes by the diabetes status of the patient. An SSTI episode was considered to be for a patient with diabetes if the patient was on the KPNC Diabetes Registry and was considered to have diabetes prior to the start of the episode. The KPNC Diabetes Registry was established in 1993 to monitor the quality of care and health outcomes for plan members with diabetes and is updated annually by identifying all plan members with diabetes from automated databases of pharmacy data, laboratory data, hospitalization records, and outpatient diagnoses [41‐43].

Microbiology test results were extracted from KPNC’s Laboratory Utilization and Reporting System. We retained for analysis all blood, body fluid, tissue, and other miscellaneous bacterial specimens (such as those taken from abscesses, pustules, boils, wounds, etc.) obtained from patients with SSTIs in the period from 7 days prior to the start of their SSTI episode to 7 days after the end of their SSTI episode. These tests were considered to have been performed to determine SSTI etiology. We excluded respiratory, cerebrospinal fluid and urine specimens. For those SSTI episodes where a microbiology test was ordered, we defined the “index specimen” as being the specimen from which the first clinically-relevant organism was isolated, or the first specimen obtained during the relevant time period if no organism was isolated. The isolate from the index specimen was defined as the “index isolate”, and we considered this organism to be the cause of the infection. Episodes could be associated with more than one type of pathogen if multiple pathogens were identified from the index specimen.

Organisms were considered clinically-relevant pathogens only if the laboratory performed antibiotic susceptibility testing on the isolate, or if it was one of the following organisms which do not routinely receive antibiotic susceptibility testing: BHS, Streptococcus pneumoniae, Bacteroides spp., Fusobacterium spp., Haemophilus spp., Pasteurella spp., Peptococcus, Porphyromonas spp., Prevotella spp., Peptostreptococcus spp., and S. aureus. The following organisms were considered contaminants and were not counted as clinically-relevant pathogens: coagulase negative staphylococcus, Ochrobactrum, Corynebacterium, Micrococcus spp., Bacillus spp., gram positive cocci, gram positive rods, and streptococcus species other than BHS, S. pneumoniae, Streptococcus milleri, Streptococcus anginosus, Streptococcus constellatus, and Streptococcus intermedius. In addition, some uncommon organisms that could not be definitively speciated by the laboratory were also excluded.

The denominator used for calculating incidence rates was KPNC member years. We first identified all persons who were members of KPNC at any time between January 1, 2009 and December 31, 2011. KPNC membership systems track membership on a monthly basis and persons could enter and leave the health plan throughout this time. We summarized member months into strata by calendar year (2009, 2010 and 2011), age (in one-year increments), gender, race/ethnicity and diabetes status. SSTI episodes were also summarized by year, age group, gender, race/ethnicity and diabetes status. Persons could contribute more than one episode to the numerators.

Incidence rates were calculated by dividing the number of episodes in each stratum by the person-years in that stratum. For reporting, one-year age strata were summarized into five age groups (less than 5, 5- < 18, 18- < 50, 50- < 65, and 65+ years of age). Ninety-five percent confidence intervals for incidence rates were calculated using the following formulas:

Rate ratios were estimated using a single multivariate negative binomial regression in which episodes from all years were included. Each record in the analytic dataset was one of the age, gender, race/ethnicity, and diabetes strata noted above. The dependent variable was the number of SSTI episodes for all patients in the stratum. Independent variables were gender, age group (less than 5, 5- < 18, 18- < 50, 50- < 65, and 65+ years of age), race/ethnicity, and diabetes status. The log of person-years in each stratum was used as an offset term. In a sensitivity analysis, we ran the same model but included only the first SSTI per patient.

We examined the relationship between demographic and clinical characteristics and the risk of S. aureus and MRSA. We created an analytic dataset that included all SSTI episodes from which a culture was obtained. Odds ratios were estimated using a multivariate logistic regression where the dependent variable was whether or not the culture was positive for S. aureus and the independent variables were gender, age (treated as a categorical variable with five levels: less than 5, 5- < 18, 18- < 50, 50- < 65, and 65+ years of age), race/ethnicity, diabetes status, quintile of census block income, and the type of SSTI. Because some patients had more than one SSTI between 2009 and 2011, we used a generalized estimating equation approach with an exchangeable covariance structure to account for correlation among records for the same person. The resulting odds ratios represent the odds of testing positive for S. aureus given that microbiological testing was performed. A second model was run to identify the risk factors for testing positive for MRSA among those SSTI episodes that were positive for S. aureus. The dependent variable in this model was whether or not the SSTI was positive for MRSA and the independent variables were the same as in the first model. The resulting odds ratios represent the odds of testing positive for MRSA given that the SSTI was culture-confirmed S. aureus.

During the three-year study period, 376,262 unique individuals experienced 471,550 SSTI episodes (Table 1). The majority of these SSTIs were coded as “cellulitis and abscess” (63%), and the average age of patients was 41 years. The racial profile of SSTI patients was similar to that of the KPNC membership except that Asians were under-represented (12% of SSTIs versus about 18% of KPNC members) and Whites were over-represented (54% versus about 49%). Twenty-three percent of SSTI episodes were cultured, and 12% of all episodes had a potentially clinically-relevant pathogen identified.

The rate of clinically-diagnosed SSTIs was 496 per 10,000 person-years, with the highest crude incidence rates being in persons 65 years of age and older, Native Americans, African-Americans, persons identified as Multiracial, and persons with diabetes (Table 2). After adjusting for patient demographics and diabetes (and including only main effects), children under 5 years of age had higher rates of SSTI than persons aged 65 years and older (Rate Ratio (RR): 1.21, Confidence Interval (CI): 1.17 to 1.26), while persons in other age groups had lower rates of SSTIs compared to persons aged 65 years and older. Asians (RR: 0.51, CI: 0.50 to 0.52), African-Americans (RR: 0.94, CI: 0.92 to 0.96), and Hispanics (RR:0.81, CI: 0.80 to 0.83 all had lower rates of SSTIs than Whites. Persons with diabetes had nearly double the rate of SSTI compared to persons without diabetes (RR: 1.93, CI: 1.90 to 1.96). When we restricted the model to include only the first episode per individual, the RR associated with diabetes decreased from 1.98 to 1.69 (CI: 1.66 to 1.72), but all other results remained substantially the same. When diabetes was excluded from the main model (the one allowing multiple episodes per person), the results were substantially the same as when diabetes was included, with the exception that children under 5 years of age had a lower RR (RR: 0.901, CI: 0.85 to 0.95) compared to persons aged 65 years and older, and Native Americans and persons identified as Multiracial join the other race/ethnicities in having lower risk of SSTI compared to Whites.

In a post-hoc analysis to investigate differences among the racial/ethnic groups in the relationship between age and the risk of SSTI, we re-ran the model but included an interaction term for race/ethnicity by age group. In this model, within each racial/ethnic group, the rates of SSTI in the younger age groups were compared to the rate in persons age 65 years and older (Table 3). This model indicated that children less than 5 years of age had a significantly higher rate of SSTI compared to persons 65 years of age and older in Asians (RR: 1.86, CI: 1.74 to 2.00), African-Americans (RR: 1.51, CI 1.40 to 1.62), and Hispanics (RR: 1.22, CI: 1.14 to 1.30), but not in Native Americans, persons classified as multiracial, or Whites. Also notable was that African-Americans aged 18 to <50 and 50 to <65 had substantially higher rates of SSTI compared to African-Americans 65 years of age and older (RR: 1.59, CI: 1.52 to 1.66, and, RR: 1.19, CI: 1.14 to 1.25, respectively). When the same model was rerun, but parameterized in such a way that within each age group the rate in Whites was the reference group, Asians were found, at every age, to have lower rates of SSTI than Whites of the same age, and Hispanics were found to have lower rates of SSTI at every age group except children under 5 years, compared to Whites of the same age group (not shown). African Americans aged 18 to <50 had a higher rate of SSTI (RR: 1.27, CI: 1.23 to 1.31) compared to whites of the same age, but at other ages had similar, or lower rates, than Whites.

Among episodes with a culture, 54% had a pathogen identified (Table 4). S. aureus was by far the most common pathogen, isolated in 81% of pathogen-positive index specimens. Among S. aureus isolates, 46% were MRSA (37% of all pathogen-positive index specimens). Other important pathogens were BHS (10% of pathogen-positive index specimens) and gram negative bacteria (14%). Multiple pathogens were identified in 6% of index specimens with a pathogen.

Among SSTI episodes with a culture, there was a consistent trend of younger persons being more likely to test positive for S. aureus (Table 5). Episodes in children under 5 years of age had 3.89 (CI: 3.64 to 4.14) times the odds of being S. aureus compared to episodes in persons 65 years and older. Hispanics were somewhat more likely to test positive for S. aureus than were whites, and persons living in less affluent census blocks were more likely to test positive for S. aureus than persons in more affluent census blocks. Carbuncle/furuncle and cellulitis and abscess were more likely to be culture-positive for S. aureus than were other types of SSTIs.

Among SSTI episodes with a positive S. aureus culture, isolates from females were slightly more likely to be MRSA (Odds Ratio (OR): 1.04, CI: 1.00 to 1.09, compared to males). Compared to Whites, isolates from African-Americans were more likely to be MRSA (OR: 1.79, CI: 1.67 to 1.92) as were those of Hispanics (OR: 1.24, CI: 1.18 to 1.31). Isolates from Asians, on the other hand, were substantially less likely to be MRSA than those of whites (OR: 0.73, CI: 0.68 to 0.78) or other racial/ethnic groups. Isolates from persons living in less affluent census blocks were more likely to be MRSA compared to isolates from persons living in more affluent census blocks. Isolates from diabetics were less likely to be MRSA than isolates from non-diabetics (OR: 0.75, CI: 0.71 to 0.80), and carbuncle/furuncle and cellulitis and abscess isolates were more likely to be MRSA than were isolates from other types of SSTI (OR: 1.22, CI: 1.14 to 1.30).

Our selection criteria explicitly excluded SSTIs diagnoses that we thought likely be represent nosocomial or chronic infections, and thus our incidence rates and microbiology results do not reflect inclusion of those types of infections. In addition, our findings reflect the routine practice of providers in this health plan with respect to diagnosing and microbiologic testing, and these practices may differ from those of other settings. The differences we found in risk of SSTIs among the different gender, race/ethnicity, and age groups, may to some extent reflect differences in the propensity of persons in these groups to seek treatment. As an observational study, our analysis of risk factors could be affected by unmeasured confounders. We did not include, for example, nursing home status, family size, injection drug use, or medical conditions other than diabetes. We included diabetes as a risk factor due to the high prevalence of the condition, and because prior literature suggests its relationship to increased risk of SSTI [35‐37, 39, 40]. However, other conditions – in particular, immuno-compromising conditions such as HIV – might also be associated with increased risk of SSTIs. In this study we did not assess severity or health services utilization. Therefore, higher risk of clinically-diagnosed SSTIs does not necessarily imply more severity or greater overall cost of care. Cultures were included based on their temporal proximity to a patient’s SSTI diagnosis. It is possible that some patients may have multiple infections, and that the pathogen we associated with the SSTI infection may not have been the cause of that infection.