During the 19-year period, 35,580 bacterial pathogens causing nosocomial infections were identified. The hospital’s overall rate of nosocomial infections during the 19-year period ranged from 3.9% to 6.1%. For the four major sites of nosocomial infections, data are presented as numbers of infection per 10,000 patient-days from 1991 to 1999 (Figure 1). BSI ranked first in nosocomial infection sites in 1999, followed by urinary tract, surgical site, and respiratory tract infections.

Data for the catheter- and ventilator-days of the hospitalized patients were not available. However, the mean percentages of patients who stayed in ICUs (six medical ICUs and six surgical ICUs) and used urinary catheters, arterial and central venous catheters, and ventilators were 75.0%, 77.9%, 63.2%, respectively, in 1996 and 79.4%, 81.4%, and 65.2%, respectively, in 1999. The incidence (number of infections/patient-days x 1,000) of urinary catheter-related urinary tract infection, vascular catheter-associated BSI, and ventilator-associated pneumonia in the ICUs was 4.4, 5.6, and 2.7, respectively, in 1996 and 6.0, 7.1, and 3.4, respectively, in 1999.

Gram-negative bacteria remained the predominant pathogens (66.1% in 1981, 51.3% in 1993, and 53.4% in 1999); however, incidences of fungal infections have increased recently (Table 1). In a comparison of data grouped into four time periods (1981-1986, 1987-1992, 1993-1998, and 1999), E. coli decreased from 12.1% of all nosocomial infections in 1981-1986 to 9.5% in 1993-1998; however, infections caused by S. aureus increased from 5.2% (1981-1986) to 12.1% (1993-1998). Candida species showed a considerable increase (from 3.7% in 1981-1986 to 16.2% in 1999) and have ranked first since 1993, followed by S. aureus and P. aeruginosa (Table 1).

Candida species, S. aureus, and Acinetobacter species were also important bloodstream isolates (Table 1), increasing from 1.0%, 5.2%, and 6.1%, respectively, in 1981-1986 to 16.2%, 13.0%, and 7.6%, respectively, in 1999. Although E. coli was also a frequent isolate, it declined in percentage of all BSI from 18.7% in 1981-1986 to 9.0% in 1999. Other pathogens declining in percentage of BSI from 1981-1986 to 1999 included K. pneumoniae, P. aeruginosa, and Enterococcus species.

From 1992 to 1999, 1,065 isolates of Candida species were recovered from patients with nosocomial BSIs. C. glabrata (8.8%) ranked fourth in incidence behind C. albicans (59.3%), C. tropicalis (17.6%), and C. parapsilosis (8.2%) in 1999. Only four blood isolates of C. krusei were identified during the 8-year period.

The most frequent isolates from urinary tract infections in 1993-1998 were Candida species (23.6%), followed by E. coli (18.6%) and P. aeruginosa (11.0%). However, in 1999, E. coli (18.4%) replaced Candida species (14.3%) as the top-ranking pathogen causing urinary tract infections. P. aeruginosa remained the most frequent isolate for respiratory tract and surgical site infections in the past 13 years. Candida species increased in incidence in surgical site infections, from 1.8% in 1981-1986 to 6.4% in 1999. Among pathogens causing respiratory tract infections, Acinetobacter species ranked fifth in 1993-1998 (9.3%) but second in 1999 (13.0%); Candida species accounted for only 2.2% in 1999.

The distributions of selected antimicrobial drug-resistant pathogens causing all nosocomial infections and BSIs of patients hospitalized in intensive care units or general wards is shown in Table 2. Antimicrobial drug-resistant pathogens causing BSIs that increased markedly over the study period were methicillin-resistant S. aureus (4.3% in 1981-1986, 58.9% in 1993-1998, and 69.2% in 1999), cefotaxime-resistant E. coli (0% in 1981-1986, 6.1% in 1993-1998, and 12.5% in 1999), and cefotaxime-resistant Klebsiella pneumoniae (4.0% in 1981-1986, and 25.8% in 1993-1998). The frequencies of these three resistant pathogens were considerably higher in isolates from ICUs than those from general wards (84.6% vs. 48.3% for MRSA, 17.1% vs. 5.1% for cefotaxime-resistant E. coli, and 51.1% vs. 18.3% for cefotaxime-resistant K. pneumoniae in 1993-1998). The incidence of methicillin-resistant coagulase-negative staphylococci (MRCoNS) remained high (72%-90%) during the 19-year period.

Ceftazidime use is associated with trends of several antimicrobial-resistant pathogens during the period 1991 to 1999 (Figure 2). Restriction of third-generation cephalosporins (particularly ceftazidime) was implemented in 1997. In 1999, resistance to cefotaxime in K. pneumoniae diminished; however, resistance to cefotaxime in E. coli and resistance to ceftazidime in P. aeruginosa slightly increased.

The first clinical isolate of VRE was recognized in 1995 25,26. Since then, 80 isolates of VRE (49 of E. faecalis and 31 of E. faecium) have been recovered from hospitalized patients. The incidence of VRE in isolates causing nosocomial infection increased from 1.8% in 1995 to 6.7% in 1997 and 25.2% in 1999 (Figure 3). Among these VRE isolates, 4 were from blood, 15 from urine, and the rest were pus or drainage fluid. Only 12 (15%) of these isolates were from patients admitted to ICUs, and 6 of these 12 isolates were recovered in 1999. The incidence of VRE in enterococci causing nosocomial infections in ICUs was 7.0%. The relationship of increasing vancomycin use and the increase in vancomycin resistance in enterococci is shown in Figure 3.

The incidences of imipenem-resistant P. aeruginosa (1993 to 1998) and imipenem-resistant Acinetobacter baumannii (1999) isolated from ICUs were five- to ten-fold higher than isolates recovered from non-ICU settings. However, this was not the case with imipenem-resistant P. aeruginosa in 1999 or imipenem-resistant A. baumannii in 1993-1998 (Figure 4).

All 150 isolates of Candida species were inhibited by 1 μg/mL of amphotericin B (MIC range 0.03 to 1 μg/mL). The MIC₅₀ and MIC₉₀ of C. glabrata were 16 μg/mL and >32 μg/mL, respectively. Twenty-one isolates (70%) of C. glabrata were nonsusceptible to fluconazole (MICs >8 μg/mL). Four (15.4%) of the C. parapsilosis isolates had MICs >8 μg/mL (2 had MICs 16 μg/mL and the other 2 had MICs 64 μg/mL).

Hospitals worldwide are continuing to face the crisis of the upsurge and dissemination of antimicrobial-resistant bacteria, particularly those causing nosocomial infections in ICU patients 1,27–29. Among resistant bacteria, MRSA, MRCoNS, VRE, third-generation cephalosporin-resistant Enterobacteriaceae, and imipenem- or ciprofloxacin-resistant P. aeruginosa and A. baumannii are of great concern because these bacteria have spread worldwide and ultimately will compromise the antimicrobial therapy of infections caused by these organisms 2,25–28,30.

This report describes trends in major nosocomial pathogens and shifts in antimicrobial resistance during a 19-year period in a large teaching hospital in Taiwan. In a comparison of data from a recent NNIS study and other surveillance systems 2,4,9,31, our results suggest four conclusions. First, Candida species, rather than P. aeruginosa, E. coli, or staphylococci, now are the most frequent pathogens causing overall nosocomial infections and BSIs in this hospital. The upward trend in coagulase-negative staphylococci, which was the leading cause of BSI in the recent NNIS study, was not confirmed in our study. Second, we observed a remarkably high incidence of MRSA, especially in ICUs, contrasted with a relatively low incidence of VRE. Third, we report an increase in incidence of cefotaxime-resistant Enterobacteriaceae, especially in the incidence of cefotaxime-resistant K. pneumoniae in ICUs. Fourth, although the overall incidence of imipenem resistance among P. aeruginosa and A. baumannii in recent years has remained stable (4% to 11%), higher incidences of imipenem-resistant P. aeruginosa or A. baumannii were found in ICUs than in general wards in most time periods.

Since 1990, Candida species have continued to be an important cause of nosocomial BSI in the United States, and the proportion (40%-50%) of these infections due to species of Candida other than C. albicans may be increasing (12,19,20,31,32). Among the species of Candida other than C. albicans, C. glabrata (prone to be resistant to fluconazole) and C. krusei (intrinsically resistant to fluconazole) are of clinical importance 31,33. Although the proportion (40%) of candidemia due to non-albicans Candida species in 1999 in our hospital was similar to that (48%) reported in the United States from April 1995 to June 1996, the incidences of C. glabrata (8.8%) and C. krusei (0%) in our hospital were lower than those (20% and 5%, respectively) in the United States 31.

Although an upsurge in the incidences of K. pneumoniae and E. coli isolates resistant to cefotaxime was noted in our ICUs, an investigation is under way into the mechanisms of resistance and potential outbreaks (clonal dissemination or gene transfer) 34. The abrupt increase in the proportion of A. baumannii isolates resistant to imipenem in 1999 resulted from wide dissemination of several multidrug-resistant clones in ICUs and many general wards in the hospital (data not shown).

In summary, surveillance of the microbial etiology of nososcomial infections over prolonged time periods not only can provide important information for day-to-day decision making in antimicrobial therapy in individual hospitals but also can reflect local trends and shifts in etiology and antimicrobial drug resistance. Nosocomial pathogens have shifted away from easily treated bacteria toward more resistant bacteria and even to Candida species with fewer options for therapy. These shifts continue to present challenges for nosocomial infection control and prevention.