ABCs conducts active laboratory- and population-based surveillance for invasive infections due to GAS and other bacterial pathogens of public health importance. We reviewed ABCs reports of invasive GAS cases among persons >65 years of age occurring from January 1, 1998, through December 31, 2003, in the following sites: San Francisco, California (3 counties); Baltimore, Maryland (6 counties); Albany and Rochester, New York (15 counties); Portland, Oregon (3 counties); Chattanooga, Knoxville, Memphis, and Nashville, Tennessee (11 counties); and the entire states of Connecticut, Georgia, and Minnesota. Five counties in the Denver, Colorado, metropolitan area were added in 2000. The total surveillance area encompassed a 2000 Census population of 3,446,404 persons >65 years of age (10% of the total US population in this age group).

ABCs methodology has been published previously (2,17). Briefly, ABCs sites maintain active contact with clinical laboratories to identify all cases and perform audits of laboratory records at least every 6 months to ensure complete reporting. Surveillance officers review case-patient medical records to obtain information on demographic characteristics, clinical syndrome, underlying disease, and illness outcome. Case-patients with GAS-positive blood cultures but without an identifiable clinical syndrome are categorized as having bacteremia without focus. Otherwise, multiple clinical syndromes—including cases of pneumonia, cellulitis, osteomyelitis, non-skin abscesses, and other syndromes (18) when accompanied by a sterile site isolate—may be reported for each case. Underlying illness information (18) was not consistently collected in Georgia from 1998–1999, Maryland from 1998–2000, or Tennessee in 1998. Information on smoking status was collected beginning in 2000 and history of cerebrovascular accident (CVA) in 2001.

ABCs defines a case of invasive GAS infection as isolation of GAS from a normally sterile site (e.g., blood, cerebrospinal fluid) or from a wound when accompanied by STSS or NF in a resident of an ABCs surveillance area. ABCs defines an LTCF as a skilled nursing facility, nursing home, rehabilitation hospital, or other chronic-care facility in which the patient has been living for at least 30 days before GAS infection. The definition did not include facilities in which the patient receives daily outpatient therapy or prisons, group homes, and assisted living facilities.

To determine whether outbreaks contributed significantly to GAS disease among LTCF residents, we looked for clusters within LTCFs. We defined a GAS LTCF cluster as >2 invasive infections with the same emm type occurring within 12 months (duration of some previously reported GAS outbreaks [10]) among residents >65 years of age living at the same facility. Surveillance staff confirmed the residence of case-patients within each cluster.

To describe incidence trends for persons >65 years of age (regardless of residence type) from 1998 to 2003, we analyzed GAS cases and deaths reported from 54 ABCs counties that conducted GAS surveillance during the entire 6-year period (1998 population: 1,981,662 persons >65 years of age). For annual rate calculations, we used national census and postcensus population estimates for these counties as the annual population denominators.

To calculate incidence of invasive GAS infection among persons >65 years of age stratified by residence type, we included ABCs GAS case-patients during the year 2000 and imputed cases with missing residence information on the basis of distribution of cases with known residence. For the denominator we used residence type–specific population estimates from the US Census 2000 Summary File 1 for ABCs counties (19); census data on residence type were only available for the year 2000. To calculate national estimates of disease, we applied age- and race-specific GAS rates from the ABCs surveillance area to the age and racial distribution of the US population in 2000; we redistributed those of unknown race on the basis of the reported distribution for known cases.

For residence-specific analyses, we excluded cases of invasive GAS infection if residence was missing or unknown. To calculate case-fatality ratios (CFRs) we included only case-patients with known outcomes.

ABCs sites forwarded all available GAS isolates to CDC’s Streptococcal Genetics Laboratory. GAS isolates underwent T typing and amplicon restriction profiling of the emm gene as described at www.cdc.gov/ncidod/biotech/strep/protocol_emm-type.htm (20). Using a reference database containing ≈180 group A streptococcal emm sequence types, we categorized an isolate as a given emm type if it had >92% identity over the first 30 codons encoding the processed M protein with one of the reference emm types (21).

Antimicrobial drug susceptibility testing of available GAS isolates in 1999, 2001, and 2003 was performed at CDC by using broth microdilution. To report antimicrobial susceptibility, we used established Clinical and Laboratory Standards Institute breakpoints for MICs and defined isolates with intermediate or high-level resistance as nonsusceptible (22).

We used SAS version 9.1 (SAS Institute Inc., Cary, NC, USA) for all analyses. To analyze incidence trends, we used Cochran-Armitage calculations for linearity and trend. In univariate analysis, we used Cochran-Mantel-Haenszel statistics to compare case-patient and GAS isolate characteristics stratified by case-patient residence; we also analyzed factors associated with death among LTCF residents and community-based case-patients separately.

We used logistic regression to characterize factors associated with death, checking for 2-way interactions and collinearity. We included in our model all variables associated with death on univariate analysis (p<0.15) controlling for age group, race, and sex. We stratified emm type into each of the 10 most common emm types and an 11th category including all remaining emm types (“other”). We classified case-patients with multiple clinical syndromes in the category with the highest CFR. The model was restricted to cases for which information on all variables was available. We considered p values <0.05 statistically significant.

From 1998 to 2003, a total of 5,889 cases of invasive GAS infection of all ages were reported, including 1,762 (30%) among persons >65 years of age. Incidence of invasive GAS infection in this elderly age group increased from 10.0 cases per 100,000 population in 1998 to 10.9 cases per 100,000 population in 2003 (Table 1). Type of residence was available for 1,662 elderly case-patients (94%). Of these, 383 case-patients resided in LTCFs, accounting for 23% of cases in those >65 years of age. In 2000 (the only year with reliable US Census population estimates for residence type), the incidence of invasive GAS among LTCF residents was almost 6 times higher than among community-based residents (41.0 vs. 6.9 cases per 100,000 persons, p<0.01). Projecting to the US population, we estimate that 650 cases among LTCF-residents and 2,250 cases among community-based residents >65 years of age occurred nationwide in 2000. Among both LTCF- and community-based residents, GAS incidence was highest among black men (78.9 and 13.8 cases per 100,000 persons, respectively) and lowest among white women (35.1 and 4.9 cases per 100,000 persons, respectively).

In comparison to community-based case-patients, LTCF case-patients were older (median 83 years vs. 75 years for community case-patients, p<0.01) and more frequently female (Table 2). Underlying illness information was available for 1,538 (93%) case-patients. Congestive heart failure (CHF), diabetes mellitus, chronic obstructive pulmonary disease, and atherosclerotic cardiovascular disease were common in both groups. However, LTCF case-patients more frequently had CHF and a history of cerebrovascular accident but less commonly had diabetes mellitus or were current smokers than community-based case-patients. In addition, LTCF residents were less likely to have penetrating trauma preceding the infection (0.8% vs. 2.7%, p<0.05). Compared to community-based case-patients, LTCF case-patients more commonly had bacteremia without focus and pneumonia but less frequently had cutaneous or soft tissue infections as the possible source of the invasive GAS isolate identified (Table 3).

GAS was identified from blood cultures in 1,491 (90%) of the 1,662 elderly case-patients with known residence. Of the remaining 171 nonbacteremic patients, GAS was most commonly isolated from joint fluid (n = 57) and surgical specimens (n = 51). GAS was identified from multiple body sites in 125 (8%) case-patients.

GAS isolates were available in 1,414 (85%) of the 1,662 case-patients. From a total of 63 emm types identified, 5 (emm1, emm3, emm12, emm28, and emm89) accounted for most infections (57% among LTCF residents; 62% among community-based residents) (Table 4). Antimicrobial susceptibility testing was performed on 781 GAS isolates including 187 isolates from LTCF case-patients. Fourteen (7%) isolates from LTCF case-patients and 34 (6%) from community-based case-patients were not susceptible to erythromycin (p = 0.38). Three isolates from LTCF case-patients and 5 from community case-patients were not susceptible to levofloxacin; 2 from community case-patients were not susceptible to clindamycin. No isolates were resistant to penicillin, ampicillin, cefazolin, vancomycin, or cefotaxime.

The CFR among case-patients >65 years of age was 24%. CFR increased with age among both LTCF- and community-based case-patients. However, when compared to the CFR for the 65- to 74-year-old group, the CFR among 75- to 84-year-old persons and those >85 years of age was significantly greater only among community-based case-patients (Figure). LTCF case-patients were 1.5 times as likely to die from the infection as community-based GAS case-patients (33% vs. 21%, p<0.01); however, this group was less often hospitalized (90% vs. 95%, p<0.01). CFRs among hospitalized and nonhospitalized case-patients were comparable in both LTCF (33% vs. 33%, p = 0.92) and community case-patients (21% vs. 25%, p = 0.44).

Univariate analysis of LTCF case-patients showed that those with CHF had significantly higher CFR (42% with CHF died vs. 27% without CHF, p<0.01) as did those with infections caused by emm1 (51% vs. 28%, p<0.01) or emm3 (45% vs. 30%, p<0.05) when compared to other emm types. We also observed higher CFR among LTCF case-patients with STSS (73% vs. 31%, p<0.01), NF (64% vs. 31%, p<0.05), or pneumonia (42% vs. 30%, p<0.05) than those with other syndromes. Sex, race, and hospitalization of LTCF case-patients were not significantly associated with death. These same variables were associated with significantly higher case-fatality rates among community-based case-patients.

In the final multivariate logistic regression model, independent predictors of death included LTCF residence; lack of hospitalization; infection due to emm1, emm3, or emm12; disease manifesting as STSS, NF, pneumonia, or bacteremia without focus; and interaction between female sex and presence of congestive heart failure (Table 5). Age was not a significant risk factor associated with death.

We identified 18 GAS clusters comprising a total of 40 cases (10% of LTCF cases). Fourteen clusters consisted of only 2 cases; the other 4 clusters had 3 cases each. The median interval between the first and second cases was 2.5 months (range 0.2–9.2 months). The most common emm types identified were emm28 and emm89, which caused 4 and 3 clusters, respectively. Case-patients in clusters were of similar age (median 85.5 years), sex (68% female), and race (75% white) to overall LTCF GAS case-patients >65 years of age. The most common syndromes of clustered patients were cellulitis (40%) and bacteremia without focus (38%). Fifteen case-patients died (CFR 38%).