Since the late 1990s, outbreaks of community-acquired methicillin-resistant Staphylococcus aureus (CA-MRSA) infection have occurred in select populations such as rural Native Americans,1 sports teams,2,3 incarcerated persons,4,5 and military personnel,6,7 among others.8 Outbreaks have been described among men who have sex with men (MSM)-many of whom have been infected with HIV-in Los Angeles,9 San Francisco,10,11 and New York.12
A similar pattern has been noted at the Owen (HIV) Clinic of the University of California, San Diego (UCSD). We investigated the epidemiology of MRSA infections in this clinic population because MRSA infection in outpatients has implications for the potential prevention and initial management of clinically significant (CS) staphylococcal infections.
Our study had 4 aims: (1) to develop a cohort-based estimate of trends in the incidence of CS-MRSA events in HIV-infected adults, (2) to characterize cases according to likely source of acquisition (nosocomial or community), (3) to characterize cases by antimicrobial susceptibility patterns, and (4) to identify risk factors for CA-MRSA events.
We performed a retrospective cohort analysis of patients with documented HIV infection having 1 or more visits to the UCSD Owen Clinic for the period January 1, 2000 through December 31, 2003. For this cohort, the number of clinical isolates from all body sites (except stool) yielding MRSA or methicillin-sensitive S aureus (MSSA) was summed by half year of the study period, and frequencies of MRSA and MSSA isolates were normalized per 100 patients under care in each half year. The ratio of normalized MSSA/MRSA isolate frequencies was estimated by half year. Sensitivity testing using antibiotic gradient plates was performed using the method of Braude et al13 for cefazolin, oxacillin, cotrimoxazole, gentamicin, erythromycin, ciprofloxacin, tetracycline, rifampin, and vancomycin. MRSA is defined as >4 μg/mL for oxacillin and >10 μg/mL for cefazolin.
The primary study outcome was an initial episode of CS-MRSA infection during the study period. Medical record review was conducted by one of the authors (WCM) to validate the clinical significance of the potential MRSA event and to categorize the primary site of infection. The clinical significance of each MRSA episode was validated by structured implicit review (SIR)14 of medical records, and patients with a CS culture were included in the analysis. In our application of SIR, a reviewer judged the clinical significance of an MRSA isolate against internalized standards; however, in forming that judgment, the reviewer was directed to examine the presence or absence of specific supporting evidence in the medical record.15 Factors abstracted from the medical record for SIR included documentation of (1) prescription of appropriate antibiotics, (2) surgical drainage or catheter removal, (3) clinician diagnosis of MRSA infection as opposed to colonization, (4) culture of MRSA from a normally sterile body site, and (5) the number of body sites yielding MRSA within 7 days of the index date.
A multiple logistic regression model was fit to identify the relative importance of factors identified during structured medical record review in the final implicit judgment regarding the clinical significance of MRSA isolates. The independent variables in this model were binary (yes vs. no or unknown). Model discrimination was evaluated as the area under the receiver operating characteristic curve (ROC). Episodes were categorized by primary site of infection as (1) skin or soft tissue, (2) blood, (3) respiratory, and (4) other sites. Acquisition was defined as nosocomial if (1) hospitalization was documented within 365 days before the initial MRSA event, (2) peritoneal or hemodialysis was administered within 12 months before the infection, or (3) an indwelling vascular device was present at the time of onset of MRSA infection. Otherwise, acquisition was defined as CA. This definition is based on the similarity of MRSA incidence among patients whose last hospitalization was more than 1 year ago and never-hospitalized patients16 and the demonstration of prolonged carriage of MRSA after hospitalization,17 and it is consistent with criteria proposed for “health care-associated” MRSA infection.18,19
For each patient, an initial or index MRSA isolate was defined as the first MRSA isolate (excluding source as stool) occurring during the risk period. Isolates coded with the source “stool” in the laboratory information system were excluded from further analysis. For each patient, the initial and all subsequent isolates of MRSA collected within 7 days of the initial isolate were categorized by source as (1) skin or soft tissue, (2) blood, (3) respiratory, or (4) all other sources.
Time at risk for a CS-MRSA event began with the date of the first Owen Clinic visit or January 1, 2000 for those already under care. Follow-up time was partitioned into consecutive 6-month calendar periods. Patients with 1 or more encounters at the study institution (clinics, emergency room, or hospital) during a given 6-month interval were considered to be at risk for the entire 6 months. Patients were not considered at risk for any half-year interval for which there were no encounters documented. If no MRSA event was observed within the half-year interval, the observation was right-censored on the closing date of the respective half-year interval. Person-time incidence rates were estimated by half year. Risk factors for those CS-MRSA events that met our definition for CA-MRSA were identified using Cox proportional hazards modeling with time-dependent covariates (Stata version 8.0).20 CD4 lymphocyte count, HIV plasma viral load, antiretroviral (ARV) drugs, and prophylaxis with cotrimoxazole and azithromycin were modeled as time-dependent covariates but assumed to be constant within each half-year period (median absolute CD4 count; median log10 plasma HIV viral load; documentation in the medical record of exposure to ARV therapy, coded as the number of ARV drug classes (1, 2, or 3-4) prescribed at any time during a given half year; and exposure to cotrimoxazole and azithromycin prophylaxis, recorded as the number of days of exposure within each half year). Assumptions of the multivariate model were assessed using the goodness-of-fit test by Moreau et al21 and May and Hosmer.22 The study was approved by the UCSD Human Subjects Committee.
During the 4-year period from 2000 through 2003, we found 3455 patients who received care at the Owen Clinic who met cohort eligibility criteria. For this cohort, the normalized frequencies of all clinical isolates (omitting isolates with a source labeled “stool”) yielding MSSA declined over the study period from 9 to 6 per 100 patients under care, whereas over the same period, the normalized frequencies of isolates yielding MRSA rose from 2 to 5 per 100 patients. The ratio (MSSA/MRSA) of normalized isolate frequencies declined from 5.5 in the first half of 2001 to 1.1 in the second half of 2003 (Fig. 1).
Of the 3455 eligible patients, 126 had 1 or more cultures positive for MRSA. Based on medical record review, 94 (3%) had potential initial MRSA events that were classified as CS. The 3455 cohort members contributed 7003 person-years of follow-up, with a median follow-up per patient of 2 years.
The relative importance of factors in the final implicit judgment regarding the clinical significance of MRSA isolates was as follows (adjusted odds ratios [aOR]): (1) appropriate antibiotics were prescribed (aOR = 8.6; P = 0.003), (2) surgical drainage or catheter removal was documented (aOR = 25.4; P = 0.003), (3) clinician-documented diagnosis of MRSA infection (aOR = 3.4; P = 0.077), (4) MRSA culture from a normally sterile body site (aOR = 0.7; P = 0.648), and (5) more than 1 body site positive for MRSA (aOR = 3.9; P = 0.165). The model ROC area was 0.90.
Primary sources of the 94 initial CS-MRSA events were distributed as follows: (1) 78 (83.0%) skin or soft tissue, (2) 9 (9.6%) blood, (3) 6 (6.4%) respiratory, and (4) 1 (1.0%) all other sources. Initial MRSA isolates were more commonly CA (60%) than nosocomial (40%). Potential risk factors for MRSA infection among the 94 cases were documented in the medical record as follows (overall percent [percent nosocomial, percent CA]): injection drug use (IDU), 23% (32%, 18%); amphetamine abuse, 32% (29%, 34%); cocaine abuse, 10% (10%, 9%); antecedent dermatitis, 9% (16%, 4%); and diabetes, 4% (11%, 0%). Of these potential risk factors, only antecedent dermatitis and diabetes were significantly (P < 0.05) less common among the CA cases. Figure 2 presents the distribution of initial CS-MRSA events by half year and by nosocomial or community acquisition. Antimicrobial susceptibility (expressed as percent resistant) of the 94 initial MRSA isolates was distributed as follows (overall percent (percent CA, percent nosocomial]): ciprofloxacin, 76% (73%, 82%); erythromycin, 91% (91%, 92%); rifampin, 1% (0%, 3%); trimethoprim-sulfamethoxazole, 9% (2%, 18%); tetracycline, 24% (31%, 16%); and vancomycin, 0% (0%, 0%). Only trimethoprim-sulfamethoxazole resistance was significantly (P < 0.05) associated with nosocomial acquisition (OR = 12.4; P = 0.021).
Figure 3 shows the cohort incidence rates and 95% confidence intervals of initial CS-MRSA events by half year. The MRSA incidence increased 6.2-fold from the first half year (early 2000) to the last half year (late 2003) of the study period.
Baseline characteristics of the 94 patients with and 3361 without MRSA infection are compared in Table 1. At cohort entry, eventual patients were more likely (P < 0.05) to be male and white, to have acquired HIV by IDU, to report MSM contacts, to have a higher median plasma HIV viral load, and to be off ARV therapy during the initial half year of follow-up.
To identify potential risk factors for the 56 initial CA-MRSA events, bivariate or unadjusted proportional hazards models were fit. The following variables were examined: sex (male or female), age quartile (19-33, 34-38, 39-44, or 45-80 years), race or ethnicity (black, Hispanic, white, or other/unknown), HIV transmission risk factor (MSM but not IDU, MSM plus IDU, IDU but not MSM, heterosexual, or other/unknown), CD4 count group (0-49, 50-199, 200-349, or ≥350 cells/μL or data missing), number of ARV classes (0, 1, 2, or 3-4), and number of days of exposure to cotrimoxazole and to azithromycin prophylaxis. In these unadjusted models, the following variables were identified as risk factors (P < 0.05) for further analysis: white race (hazard rate [HR] = 2.3), risk factor for HIV transmission (MSM, IDU, or both: HR = 5.9), low absolute CD4 count (<50 cells/μL: HR = 3.1), higher plasma viral load, no exposure to ARV therapy (HR = 0.5), and cotrimoxazole prophylaxis for <120 days in the current and preceding half years combined (HR = 0.4). Those variables found to be risk factors (P < 0.05) in unadjusted models were entered simultaneously in a multipredictor model (Table 2). Based on examination of the unadjusted models, race, HIV transmission risk factors, CD4 cell count category, ARV therapy, and cotrimoxazole prophylaxis were dichotomously recoded. Significant (P < 0.05) independent predictors of initial CA-MRSA events included (1) MSM, IDU, or both as risk factors for HIV transmission; (2) higher HIV plasma viral load (with an apparent “dose response” effect); (3) CD4 count <50 cells/μL; and (4) cotrimoxazole prophylaxis for <120 days. White race was of borderline significance. In additional models (not shown), evidence (P < 0.10) was not found for a statistical interaction between ARV therapy and CD4 cell count or HIV plasma viral load or between CD4 cell count and HIV plasma viral load. The method of using the median CD4 cell count and HIV plasma viral load and an indicator for any ARV therapy as a constant for each half year at risk represents a simple approximation of underlying distribution of these factors. Even with this simple approximation, significant effects of these factors are apparent.
Confirming previous case series, we documented a 6.2-fold increase over a 4-year period in rates of initial CS-MRSA events in a clinic-based cohort of HIV-infected adults. In addition, we documented that the risk of CA-MRSA infection increases with HIV-induced immunosuppression (low CD4 lymphocyte count), high plasma HIV-1 viral load, and lack of ARV therapy.
Of the sociodemographic characteristics associated with the outcome in unadjusted models (race or ethnicity and HIV transmission risk factors), after controlling for CD4+ lymphocyte count, plasma HIV RNA levels, ARV therapy, and cotrimoxazole prophylaxis, only the HIV transmission risk factors category was independently predictive of CA-MRSA events. Relative to other risk categories, those reporting MSM, IDU, or both as transmission risk factors were at higher risk. No CA-MRSA cases were observed among those reporting heterosexual transmission of HIV, even though they represented 15% of the non-cases (see Table 1). The risks for IDU and MSM as HIV transmission risk factors were similar. Interestingly, previous investigations conducted before MRSA prevalence was high in communities found that MSM were not at higher risk for S aureus colonization compared with non-MSM.23 Therefore the increase in MRSA infection in this population may be unrelated to colonization (eg, acquired via sexual activity and skin-to-skin contact), as suggested by the outbreak investigation in Los Angeles, in which high-risk sexual behavior was a risk factor for MRSA infection among HIV-infected MSM.24
In the second half of 2002, nosocomial and CA-MRSA infection seem to have increased, but the number of CA cases rose more dramatically. Because some of our nosocomial cases may have been misclassified, we cannot exclude the possibility that the entire increase is driven by CA-MRSA infection. Recognizing the limitations of a time-based approach to classification of source of acquisition, Salgado et al25 have recommended that MRSA isolates be classified as “community onset” (with or without risk factors) or nosocomial based on the patient's location at the time of identification of MRSA. Because our study was retrospective, the S aureus isolates were not available and classification by chromosomal staphylococcal cassette chromosome (SCC) mec or toxin genotyping was not possible as a means of distinguishing nosocomial strains from CA strains.16,26 Our finding that trimethoprim-sulfamethoxazole resistance was highly associated (OR = 12.4; P = 0.005) with presumed nosocomial acquisition is consistent with several previous reports that susceptibility to trimethoprim-sulfamethoxazole, fluoroquinolones, and amino glycosides is characteristic of CA-MRSA, however.8,16,27-29
The association between indicators of HIV disease stage (absolute CD4+ lymphocyte count and HIV plasma viral load) and risk of CA-MRSA infection is relatively novel. Previous studies have largely explored the relation between CD4 cell count and MRSA colonization but have rarely considered infection. One investigation of HIV-infected patients in Taiwan found an association between lower CD4 cell count and higher likelihood of nasal MRSA colonization,30 but this investigation did not examine MRSA infection per se. Another investigation found that increasing severity of HIV disease was associated with increased likelihood of MRSA colonization; however, interestingly, these investigators found no relation between CD4 cell count and MRSA colonization.31 Finally, 1 case-control study of MRSA infection among HIV-infected MSM failed to demonstrate an association between lower CD4 cell count and MRSA infection.24 Why advanced immunosuppression indicated by a low CD4 cell count predisposes to S aureus colonization or infection is not clear, because T-cell-mediated immunity has not been considered to play a major protective role against S aureus infection.32
In unadjusted analyses, ARV therapy seemed to be a factor protective against CA-MRSA. This effect was attenuated when CD4 cell count and viral load were both taken into account, however. We believe this attenuation was attributable to the fact that the effects of ARV therapy are mediated through effects on CD4 cell count and viral load. Therefore, we cannot conclude that ARV therapy is not protective against CA-MRSA events. Neither CD4-mediated effects of HIV on immunity to S aureus nor those that are independent of CD4 cell counts are well understood.32 High levels of plasma HIV seem to have an antiproliferative effect on CD4 cells stimulated in vitro with HIV antigens.33 This direct effect of HIV on lymphocytes could also influence defenses against staphylococci, but this has not been demonstrated. Alternatively, a high viral load and failure to receive ARV therapy may be markers for unmeasured covariates that promote MRSA infection such as IDU use, non-IDU, sharing needles, high-risk sexual behavior, poor hygiene, and/or crowded living conditions.34 The “dose-response” relation35 between higher viral load and greater risk of MRSA infection suggests a mechanism involving decreased resistance rather than greater colonization. Confirmation of these relations should be sought in other cohorts.
Our study has several potential limitations. First, the increased incidence in MRSA infection could have reflected ascertainment bias as clinicians became more aware of MRSA infection and therefore submitted more cultures. This explanation is unlikely, however, because the frequency of normalized MRSA isolates increased over the study period by approximately 3-fold, whereas the comparable frequency of normalized MSSA isolates declined approximately 1.5-fold. In addition, increasing rates of CS-MRSA infection in our clinic are consistent with other reports of increases in MRSA infections in other HIV-infected populations.36,37
Second, we may have underestimated the rate of MRSA infections because some patients sought care at other clinics outside our medical system. We believe these episodes are relatively few, because the Owen Clinic provides primary and comprehensive care for its patients. Furthermore, because health care delivery for HIV remained fairly constant in our locale, care outside our system was probably unchanged over the study period. To account for patients who may have left the catchment area of the study clinic temporarily, as described in the Methods section, we did not consider them to be at risk for any half year during which no encounters were documented.
Third, the observed association between low CD4 cell count and risk of CA-MRSA infection could reflect higher rates of MRSA colonization in advanced patients who are more likely to be in health care settings and to receive antibiotics. Neither of these explanations is likely for several reasons, however: (1) none of the patients with CA-MRSA infection, by definition, were hospitalized in the prior year; (2) others have found no association of MRSA infection with the intensity of outpatient visits16; (3) the CA-MRSA isolates were sensitive (98% of cases) to the most commonly used antibiotic in advanced patients, cotrimoxazole; and (4) colonization of the nasal mucosa by MSSA rather than MRSA is common in MRSA-infected patients, suggesting that asymptomatic nasal carriage is not the reservoir from which MRSA is spread to the skin or lungs.38
Finally, we were limited by the inability to measure some potential confounding factors, such as active substance abuse or short-term antibiotic use, characteristics that are sometimes incompletely documented in our clinical database. A nested case-control study is planned to explore incompletely ascertained potential confounders and effect modifiers.
In summary, we found that MRSA infection was increasing in incidence between 2000 and 2003 among HIV patients in UCSD's Owen Clinic. HIV acquisition by IDU or MSM contact, lower CD4 cell count, higher viral load, and lack of cotrimoxazole prophylaxis were independent risk factors for CA-MRSA infection. Among HIV-infected persons, CA-MRSA infections resemble opportunistic infections with other relatively pathogenic organisms like Mycobacterium tuberculosis, which are capable of affecting persons at any CD4 cell count but are more likely to cause disease in those with low CD4 cell counts. Our observations should be confirmed in other HIV-infected cohorts. Interventions to prevent MRSA infection in HIV-infected patients may be targeted to those with identified risk factors. Empiric treatment of staphylococcal infections in HIV-infected patients should take into account MRSA risk factors.
1. Groom AV, Wolsey DH, Naimi TS, et al. Community-acquired methicillin-resistant Staphylococcus aureus in a rural American Indian community. JAMA
2. Lindenmayer JM, Schoenfeld S, O'Grady R, et al. Methicillin-resistant Staphylococcus aureus in a high school wrestling team and the surrounding community. Arch Intern Med
3. Methicillin-resistant staphylococcus aureus infections among competitive sports participants-Colorado, Indiana, Pennsylvania, and Los Angeles County, 2000-2003. MMWR Morb Mortal Wkly Rep
4. Methicillin-resistant Staphylococcus aureus infections in correctional facilities-Georgia, California, and Texas, 2001-2003. MMWR Morb Mortal Wkly Rep
5. Pan ES, Diep BA, Carleton HA, et al. Increasing prevalence of methicillin-resistant Staphylococcus aureus infection in California jails. Clin Infect Dis
6. Zinderman CE, Conner B, Malakooti MA, et al. Community-acquired methicillin-resistant Staphylococcus aureus among military recruits. Emerg Infect Dis
7. LaMar JE, Carr RB, Zinderman C, et al. Sentinel cases of community-acquired methicillin-resistant Staphylococcus aureus onboard a naval ship. Mil Med
8. Deresinski S. Methicillin-resistant Staphylococcus aureus: an evolutionary, epidemiologic, and therapeutic odyssey. Clin Infect Dis
9. Outbreaks of community-associated methicillin-resistant Staphylococcus aureus skin infections-Los Angeles County, California, 2002-2003. MMWR Morb Mortal Wkly Rep
13. Braude AI, Banister J, Wright M. Use of the gradient plate for routine clinical determinations of bacterial sensitivities to antibiotics. Antibiot Annu
14. Rubenstein LV, Kahn KL, Reinisch EJ, et al. Structured implicit review of the medical record: a method for measuring the quality of inhospital medical care and a summary of quality changes following implementation of Medicare prospective payment system
. N-3033-HCFA. Santa Monica: RAND; 1991.
15. Ashton CM, Kuykendall DH, Johnson ML, et al. An empirical assessment of the validity of explicit and implicit process-of-care criteria for quality assessment. Med Care
16. Charlebois ED, Perdreau-Remington F, Kreiswirth B, et al. Origins of community strains of methicillin-resistant Staphylococcus aureus. Clin Infect Dis
17. Scanvic A, Denic L, Gaillon S, et al. Duration of colonization by methicillin-resistant Staphylococcus aureus after hospital discharge and risk factors
for prolonged carriage. Clin Infect Dis
18. Morin CA, Hadler JL. Population-based incidence
and characteristics of community-onset Staphylococcus aureus infections with bacteremia in 4 metropolitan Connecticut areas, 1998. J Infect Dis
19. Community-associated methicillin-resistant Staphylococcus aureus in Minnesota. Minnesota Department of Health Disease Control Newsletter. 2004;32:61-67.
20. Cleves MA, Gould WW, Gutierrez RG. An Introduction to Survival Analysis Using Stata
. College Station, TX: Stata Press; 2002:157-177.
21. Moreau T, O'Quigley J, Lellouch J. On Schoenfeld's approach for testing the proportional hazards assumption. Biometrika
22. May S, Hosmer DW. Hosmer and Lemeshow type goodness-of-fit statistics for the Cox proportional hazards model. In: Handbook of Statistics 23: Advances in Survival Analysis
. Amsterdam, Netherlands: Elsevier B. V.; 2004:383-394.
23. Battan R, Raviglione MC, Wallace J, et al. S. aureus nasal carriage among homosexual men with and without HIV infection
. Am J Infect Control
24. Lee NE, Taylor MM, Bancroft E, et al. Risk factors
for community-associated methicillin-resistant Staphylococcus aureus skin infections among HIV-positive men who have sex with men. Clin Infect Dis
25. Salgado CD, Farr BM, Calfee DP. Community-acquired methicillin-resistant Staphylococcus aureus: a meta-analysis of prevalence and risk factors
. Clin Infect Dis
26. Roberts RB, de Lencastre A, Eisner W, et al. Molecular epidemiology of methicillin-resistant Staphylococcus aureus in 12 New York hospitals. MRSA Collaborative Study Group. J Infect Dis
27. Baggett HC, Hennessy TW, Rudolph K, et al. Community-onset methicillin-resistant Staphylococcus aureus associated with antibiotic use and the cytotoxin Panton-Valentine leukocidin during a furunculosis outbreak in rural Alaska. J Infect Dis
28. Fey PD, Said-Salim B, Rupp ME, et al. Comparative molecular analysis of community- or hospital-acquired methicillin-resistant Staphylococcus aureus. Antimicrob Agents Chemother
29. Naimi TS, LeDell KH, Como-Sabetti K, et al. Comparison of community- and health care-associated methicillin-resistant Staphylococcus aureus infection. JAMA
30. McDonald LC, Lauderdale TL, Lo HJ, et al. Colonization of HIV-infected outpatients in Taiwan with methicillin-resistant and methicillin-susceptible Staphylococcus aureus. Int J STD AIDS
31. Ganesh R, Castle D, McGibbon D, et al. Staphylococcal carriage and HIV infection
32. Lowy FD. Staphylococcus aureus infections. N Engl J Med
33. McNeil AC, Shupert WL, Iyasere CA, et al. High-level HIV-1 viremia suppresses viral antigen-specific CD4+
T cell proliferation. Proc Natl Acad Sci USA
35. Rothman KJ, Greenland S. Modern Epidemiology
. 2nd ed. Philadelphia: Lippincott William & Wilkins; 1998:7-28.
36. Bar A, Hantschke D, Mirmohammadsadegh A, et al. Spectrum of bacterial isolates in HIV-positive patients with skin and soft tissue infections: emergence of methicillin-resistant Staphylococci. AIDS
37. Sattler CA, Mason EO Jr, Kaplan SL. Prospective comparison of risk factors
and demographic and clinical characteristics of community-acquired, methicillin-resistant versus methicillin-susceptible Staphylococcus aureus infection in children. Pediatr Infect Dis J
38. Rieg G, Daar E, Witt M, et al. Community-acquired methicillin-resistant Staphylococcus aureus colonization among HIV-infected men who have sex with men: a point prevalence survey [abstract 877]. Presented at: 12th Conference on Retroviruses and Opportunistic Infections; 2005; Boston.