From November 2005 through June 2008, 35,639 blood cultures from 29,883 patients met our criteria for analysis; 6,114 (17%) cultures came from patients who had 2 or more cultures separated by ≥ 7 days. Among patients with > 1 culture, the median time between cultures was 106 days (interquartile range: 33–242 days). Antibiotic use in the 72 hours before blood culture was reported for 9,726 (27%) of 35,639 patient cultures. Similarly, 5,782 (24% of 24,538) blood cultures were from patients with serum antimicrobial activity. Antibiotic use, both reported and by serum antimicrobial activity, differed slightly by province and hospital size (). Among patients with serum testing, the prevalence of antimicrobial activity increased with age. Serum antimicrobial activity testing was often not performed for patients with small blood volumes collected (e.g., very young children); as a result, patients with serum disc testing results were older on average than those without testing. According to both measures, antibiotic use before blood culture was highest in 2005 and significantly lower each year afterward, (P < 0.05).
Antibiotic use before blood culture defined by reported use* and by serum antimicrobial activity; Sa Kaeo and Nakhon Phanom, Thailand, November 2005 to June 2008
Detailed information was available for 19,838 patients included in the pneumonia surveillance system. Among these patients, reported antibiotic use prevalence was 28%, and the median age was 29 years. Compared with patients without reported pre-culture antibiotic use, those with antibiotic use were more likely to have fever (38% versus 36%, P = 0.03), respiratory symptoms (83% versus 80%, P < 0.01), and require supplemental oxygen or intubation (31% versus 29%, P < 0.01). The case-fatality proportion was also higher among patients with antibiotic use (4.6% versus 3.5%, P < 0.01). Human immunodeficiency virus (HIV) status was available for 5,523 (15.5%) of 35,642 patient cultures; HIV-positive patients had similar rates of pre-culture antibiotic use as HIV-negative patients (28% versus 29%, P = 0.33).
Antibiotic names were available for 82% (8,004) of patients with reported antibiotic use. The most commonly reported drugs used were third generation cephalosporins (38%) and penicillin/penicillin derivatives (34%), both in monotherapy or in combination; 1,742 (22%) received ≥ 2 antibiotics. We could determine the likely mode of antibiotic administration for 6,368 (84% of 9,746 with reported antibiotic use); of those, 50% (3,148) had only antibiotics typically administered orally and 50% (3,218) had at least one IM or IV antibiotic.
Correlation between serum antimicrobial activity and reported antibiotic use was moderate. After excluding patients missing antibiotic use data by either measure, 16% of patients were positive for antibiotic by both measures, 62% negative by both, and 23% discordant (Kappa = 0.42). Correlation increased slightly when reported antibiotics were limited to those administered IM or IV (Kappa = 0.48). Correlation was higher for patients aged 5 years and older (0.45) than for children less than 5 years of age (0.27). The major source of discordance, particularly in young children, was negative serum antimicrobial disc testing among patients with reported antibiotic use ().
Correlation between reported antibiotic use and serum antimicrobial activity by disc testing among patients < 5 years of age*
Blood culture yields were significantly lower among patients with pre-culture antibiotic use compared with those without antibiotic use (). This was true when the culture endpoint was alarm positivity, organism isolation, or pathogen isolation, and regardless of whether pre-culture antibiotic use was determined by reported use, serum disc testing, or a combination of the two measures. Overall, E. coli was isolated more commonly from those without antibiotic use compared with those with pre-culture antibiotics. In contrast, B. pseudomallei yield was higher among patients with pre-culture antibiotic use compared with those without antibiotic use. Streptococcus pneumoniae was isolated more often in patients without antibiotics regardless of whether antibiotic use was determined by reported use, serum disc testing, or a combination of the two measures.
Blood culture outcomes by reported antibiotic use, serum antimicrobial activity, and a combination of measures, Sa Kaeo and Nakhon Phanom, Thailand, November 2005 to June 2008
When analyses were limited to children < 5 years of age, trends in culture yields for both antibiotic use measures were similar but not statistically significant for pathogen isolation. None of the 554 children < 5 years of age with serum antimicrobial activity had a positive blood culture for S. pneumoniae, compared with 12 (0.4%) of 3,291 without serum antimicrobial activity (P = 0.09). Unexpectedly, E. coli isolation rates were higher among children < 5 years of age with positive serum disc testing (0.9%, N = 5; all < 1 year of age) compared with negative serum disc tests (0.2%; N = 5).
Culture yield, as measured by alarm positivity, organism isolation, and pathogen isolation, including S. pneumoniae, was significantly lower among children < 5 years of age compared with older patients (P < 0.05 for all), which may have been caused by lower blood volumes available for culture in children. Less than half (45%) of children < 5 years of age had the target 4 mL available for blood culture and 27.7% had less than 2 mL. Streptococcus pneumoniae yield was lowest among children < 5 years of age with < 2 mL collected compared with those with 2–3.99 mL collected and 4 mL or more: 0.1% (N = 4), 0.2% (N = 7), and 0.3% (N = 13), respectively. Streptococcus pneumoniae yield within blood volume strata among children < 5 years of age indicated non-significant reductions associated with pre-culture antibiotics, although these analyses are limited by small numbers: no S. pneumoniae isolated for both positive and negative serum antimicrobial activity with < 2 mL; 0 (serum antimicrobial activity) versus 0.4% (no serum antimicrobial activity) for 2–3.99 mL; and 0 versus 0.4% for 4 mL or more.
Figure 1 illustrates the potential impact of pre-culture antibiotic use on age-specific incidence rates of hospitalized pneumococcal bacteremia. The adjusted incidence rates were significantly greater than the overall observed rate (3.2 per 100,000 person-years; 95% CI = 2.7, 3.9), 60% higher when adjusted for antibiotic use defined as reported use or serum antimicrobial activity (5.1 per 100,000). The adjusted rates within each age category were similar across methods of defining antibiotic exposure, except for young children. Among children < 5 years of age, the adjusted incidence was 63% higher when antibiotic use was defined by serum antimicrobial activity (17.9 versus 11.0 per 100,000 person-years [95% CI for observed incidence = 7.0, 16.3]), 9.6% higher when adjusted for reported antibiotic use, and 18% higher when adjusted for antibiotic exposure defined as either reported use or serum antimicrobial activity. When limited to children < 5 years of age with the targeted 4 mL blood volume, incidence was 40% higher after adjustment for serum antimicrobial activity: 13.0 versus 18.2 per 100,000 person-years.