ANTIMICROBIAL AGENTS AND CHEMOTHERAPY, Nov. 1989, p. 1855-1859 0066-4804/89/111855-05$02.00/0 Copyright 1989, American Society for Microbiology Vol. 33, No. 11 Antimicrobial Susceptibility in Gram-Negative Bacteremia: Are Nosocomial Isolates Really More Resistant? JOHN E. McGOWAN, JR.,1,2* E. C. HALL,3 AND PATRICIA L. PARROTT4 Department of Pathology and Laboratory Medicine' and Department of Epidemiology and Biostatistics, Emory University School of Medicine, Atlanta, Georgia 30322, and Clinical Microbiology Laboratory2 * and Epidemiology Department,4 Grady Memorial Hospital, Atlanta, Georgia 30335 Received 3 April 1989/Accepted 9 August 1989 Bloodstream isolates of gram-negative aerobic bacilli from nosocomial infections are more likely to be resistant to antimicrobial agents than isolates from community-acquired cases are. It is not clear, however, how much this is due to the markedly different distribution of organisms in the two groups. We compared the susceptibilities of organisms of a given species which caused community-acquired bacteremia with the susceptibilities of isolates from nosocomial cases. Nine antimicrobial agents were tested against 1,077 isolates which were obtained during a 4-year nonepidemic period. Marked differences in crude rates of resistance were noted for all isolates from nosocomial cases versus all isolates from cases acquired in the community. When results were adjusted for the different organism distributions in the two groups, statistically significant differences were found for only six drug-organism pairs; in each of these, resistance rates were higher in nosocomial isolates. However, when results were further adjusted for the effect of multiple analyses, no significant differences were seen. The major factor leading to the greater prevalence of antimicrobial resistance in our hospital organisms was the markedly different distribution of organisms in the nosocomial and community-acquired groups. For individual organisms, greater resistance in nosocomial strains was confined to certain drugs. Factors that influence differences in organism distribution may not be solely the result of antimicrobial use. Bloodstream isolates of gram-negative aerobic bacilli from nosocomial infections are generally more likely to be resistant to antimicrobial agents than are those from communityacquired cases of bacteremia. It is not clear, however, how much this is due to the markedly different distribution of organisms in the two groups. Within a given species, it is uncertain whether resistance is more prevalent in hospital organisms than in community strains. A greater prevalence of resistance is found in nosocomial bloodstream isolates of a given species during hospital outbreaks (18). Outside epidemic periods, data to establish increased resistance in nosocomial strains are scarce and conflicting (1, 18). Some studies suggest that there is a greater resistance in nosocomial blood isolates. At the Bronx Municipal Hospital, New York, greater resistance in hospital isolates from blood was noted for four organisms (Escherichia coli and Klebsiella, Enterobacter, and Proteus species) (10). Increased resistance in some nosocomial bacteremic strains (E. coli, Klebsiella-Enterobacter group, and Proteus species) was noted in 1972 at Boston City Hospital (17). At hospitals in Boston and Paris from 1973 to 1974, community-acquired isolates from blood (those isolated 3 days or less after admission) were less resistant than nosocomial strains (those obtained later in hospitalization) (22). These studies of blood isolates are paralleled by data for organisms recovered from other sites. At the Jewish Memorial Hospital, New York, hospital-acquired E. coli strains were significantly (P < 0.05) more resistant to ampicillin, sulfisoxazole, and kanamycin than community-acquired isolates were. Nosocomial Pseudomonas aeruginosa isolates were significantly more resistant to streptomycin (3). Nosocomial bacteremic strains of E. coli and Klebsiella and Proteus spp. from intensive care * Corresponding author. unit patients were more frequently resistant to gentamicin or cefuroxime than were the community-acquired strains, but the differences were not statistically significant (6). On the other hand, data from the Comprehensive Hospital Infections Project of the Centers for Disease Control (4) suggested few, if any, differences in susceptibility among community and hospital strains of the same genus, regardless of culture source. In the only cases from this study in which a difference in prevalence was seen, the nosocomial strains were more susceptible than those acquired in the community. Likewise, strains recovered from several sites of infection at a Veterans Administration Hospital in Buffalo, N.Y. (9), showed little difference in susceptibility between community and hospital isolates of the same species. In a group of hospitals in Israel, community-acquired isolates of E. coli, Klebsiella spp., and P. aeruginosa varied little in susceptibility to several antimicrobial agents, compared with nosocomial isolates of the same organisms from urine, wound, and lower respiratory sites (24). Another Israeli study showed no statistically significant difference in resistance to eight drugs between nosocomial and community-acquired bacteremic strains (26). Lynch et al. (11) found gentamicin resistance associated only with nosocomial cases of gram-negative bacteremia, but no statistically significant differences were noted when studies were controlled by type of organism. E. coli isolates with aminoglycoside-modifying enzymes were acquired both in the community and in the hospital in a study of aminoglycoside resistance in nine United States hospitals (8). In many other studies comparing the susceptibilities of nosocomial and community organisms, results have been combined for all organisms, so that whether the data are confounded by the different distributions of organism type in the two groups cannot be determined. This study compared the susceptibilities of gram-negative 1855
1856 McGOWAN ET AL. ANTIMICROB. AGENTS CHEMOTHER. TABLE 1. Susceptibility of gram-negative aerobic bacilli isolated from blood culture, by type of infectiona % Susceptible strains Drug Community acquired Nosocomial 1983 1984 1985 1986 1983 1984 1985 1986 (n = 199) (n = 176) (n = 196) (n = 269) (n = 122) (n = 96) (n = 94) (n = 102) Ampicillin 51 57 49 50 23 27 20 22 Carbenicillin 60 65 63 65 46 52 50 54 Cefamandole 87 87 82 85 79 71 66 63 Cefotaxime-ceftizoximeb 94 96 91 95 88 84 84 82 Cefoxitin 85 85 82 86 69 65 57 56 Cephalothin 71 68 61 68 56 49 44 44 Gentamicin 98 96 94 93 95 92 87 81 Tobramycin 98 94 94 95 95 91 87 86 Trimethoprim-sulfamethoxazole 86 88 82 85 84 78 78 72 a When multiple isolates of a given strain with same susceptibility pattern were obtained from the same patient, only the first isolate was included. Isolates were obtained at Grady Memorial Hospital from 1983 to 1986. b Cefotaxime tested January 1983 to June 1985; ceftizoxime tested July 1985 to December 1986. bacillary organisms of a given species causing communityacquired bacteremia with the susceptibilities of nosocomial isolates of the same species to determine whether differences in prevalence of resistance exist. Results for nine antimicrobial agents tested against 1,177 isolates recovered from blood during a 4-year period are reported. MATERIALS AND METHODS Grady Memorial Hospital is a large, metropolitan teaching hospital serving the indigent population of Atlanta, Ga. Demographic features of our patients have been described elsewhere (14, 15). Community-acquired infections account for 73% of all infections detected at this institution (16). Patterns of occurrence have been reported for communityacquired cases of bacteremia in 1975 (14) and for nosocomial cases in 1975 and 1983 (15). Surveillance methods and definitions that we used for investigation of bacteremia have been described in detail elsewhere (14, 15). Identification of a case as nosocomial was made by members of the Epidemiology Department, Grady Memorial Hospital, using previously published criteria (14). Cases were considered nosocomial if the underlying infection was neither present nor incubating at the time of hospital admission; when this could not be determined, cases with onset more than 72 h after admission were considered nosocomial. Surveillance efforts were conducted by P. L. Parrott during the earlier studies (14, 15) and throughout the current study period. The study organisms were gram-negative aerobic bacilli isolated from blood cultures received in the Clinical Microbiology Laboratory of Grady Memorial Hospital in the period from January 1983 to December 1986. When more than one isolate of the same organism with the same susceptibility pattern was recovered from cultures of a single patient, only the initial isolate was included. During the study period, organisms from blood were tested routinely for susceptibilities to the nine drugs listed in Table 1. During the 3-year period, one change was made in the first-line drugs tested: in July 1985, ceftizoxime was added to the hospital formulary in place of cefotaxime, and the testing of susceptibility to ceftizoxime was substituted for cefotaxime testing at that time. The results of testing for these two drugs were combined for this analysis. Susceptibility testing was performed by a standardized disk agar diffusion method (20) from January 1984 to June 1985. From July 1985 to December 1986, we tested susceptibilities with a commercial automated instrument (Vitek AMS; Vitek Systems, Inc., St. Louis, Mo.) by standard methods recommended by the manufacturer. A comparison of results for the 6-month period before this change of method with data for the same interval afterward showed no systematic pattern of difference. When a susceptibility test result fell in the intermediate or indeterminate category, the organism tested was considered resistant in our analysis. To determine whether the data were from a nonepidemic period, we reviewed minutes of Hospital Infection Control Committee meetings for the study months for mention of clusters (25) of nosocomial infection due to gram-negative aerobic bacilli. Susceptibility rates of community-acquired and nosocomial organisms were compared for each species and drug by using either one-tailed chi-square tests or Fisher exact tests. Differences were considered significant at a nominal or per-comparison (21) significance level of 0.05; in addition, an adjusted or per-species significance level of 0.005 (the nominal level divided by 9, the number of comparisons per species) was used to account for the multiple significance testing on each species (21, 27). Overall susceptibility rates for each drug were compared by pooling data on all organisms for a crude rate and by calculating species-adjusted rates that account for the different number of organisms of each species which make up the two groups. RESULTS The susceptibilities of isolates by year and relative proportion of different organisms in each year for the study period are shown in Tables 1 and 2, respectively. There was little difference in the prevalence of resistance of community-acquired strains to the study drugs, but the prevalence of susceptibility of nosocomial strains to the cephalosporins generally declined during the study period. The relative proportion of bacteremic isolates of various organisms remained steady for the community-acquired strains. Klebsiella strains decreased in relative proportion among nosocomial strains during the period, and Acinetobacter strains rose abruptly during the last study year. Comparison of community-acquired isolates with nosocomial strains was made for seven organism groups, for an eighth group including the remaining organisms, and for all organisms combined (Table 3). For each species, significant
VOL. 33, 1989 ANTIBIOTIC RESISTANCE IN NOSOCOMIAL BACTEREMIA 1857 TABLE 2. Frequency of gram-negative aerobic bacilli isolated from blood culture, by type of infection' Frequency (% of strains) Organism Community acquired Nosocomial 1983 1984 1985 1986 1983 1984 1985 1986 (n = 199) (n = 176) (n = 196) (n = 269) (n = 122) (n = 96) (n = 94) (n = 102) Acinetobacter spp. 3 1 3 4 6 3 7 16 Enterobacter spp. 6 6 5 4 17 14 21 13 Escherichia coli 55 55 57 55 29 30 29 25 Klebsiella spp. 16 14 10 14 30 21 21 22 Providencia spp. 1 1 2 3 1 0 0 1 Proteus mirabilis 7 13 12 12 4 4 3 6 Pseudomonas aeruginosa 6 7 6 5 8 12 13 10 Serratia spp. 3 2 2 1 3 11 5 6 Others 3 1 3 3 2 5 1 3 a Isolates were obtained at Grady Memorial Hospital. differences (nominal 0.05 level) between susceptibility rates these differences do not reflect the different distributions of of nosocomial and community-acquired organisms are indi- organisms making up the two groups. cated by a box around the two rates. For example, for When the susceptibility rates of the seven individual Enterobacter species, only the rates of susceptibility to organism groups and the eighth group of other isolates were carbenicillin were significantly different. compared, only six of these species-drug comparisons Overall comparisons for each drug against all organisms showed differences significant at the nominal 0.05 level. In (crude rates) yielded significant differences with both nomi- each of these six statistically significant pairs, the rate of nal and adjusted significance levels. In all cases, the rates of susceptibility was lower for nosocomial organisms than for susceptibility for community-acquired strains were higher community-acquired strains. than rates for nosocomial strains. The largest difference was When the adjusted significance level of 0.005 was used, 22.9 percentage points for cefoxitin, and the smallest differ- only the comparison for gentamicin susceptibility in Acineence was 5.5 percentage points for tobramycin. However, tobacter calcoaceticus subsp. anitratus yielded a significant TABLE 3. Susceptibility of gram-negative aerobic bacilli isolated from blood culture, by infection type and organism' b No. % Susceptible toc: Organism Infection tetd tested Ampi Carb Cman Czxm Cfox Cltn Gent Tobr TmSu Acinetobacter calcoaceticus subsp. CA 21 9.5 71.4 0 71.4 0 0 100 100 95.2 anitratus N 32 0 56.2 0 56.2 0 0 68.8 84.4 71.9 Enterobacter species CA 37 10.8 83.8 83.8 94.6 10.8 8.1 94.6 97.3 89.2 N 67 4.5 65.7 71.6 82.1 7.5 3.0 89.6 91.0 83.6 Escherichia coli CA 430 70.5 71.2 96.3 99.5 99.5 83.0 95.8 96.0 93.3 N 118 62.7 65.3 95.8 99.2 99.2 74.6 97.5 96.6 94.9 Klebsiella species CA 102 0 0 94.1 100 98.0 87.3 93.1 94.1 89.2 N 97 0 0 96.9 100 97.9 90.7 85.6 85.6 84.5 Proteus mirabilis CA 82 96.3 98.8 98.8 100 98.8 97.6 98.8 98.8 97.6 N 14 100 100 100 100 100 100 100 100 100 Pseudomonas aeruginosa CA 42 0 78.6 0 21.4 0 0 90.5 95.2 0 N 42 0 69.0 0 21.4 0 0 76.2 78.6 0 Serratia marcescens CA 12 0 66.7 16.7 100 58.3 0 83.3 66.7 83.3 N 29 0 69.0 24.1 89.7 55.2 0 93.1 89.7 82.8 Other organisms CA 36d 8.3 63.9 77.8 91.7 69.4 8.3 77.8 88.9 58.3 N 16e 0 43.8 81.3 93.8 56.2 25.0 87.5 93.8 81.2 All organisms Crude rates CA 762 513 65.2 85.6 94.0 69.8 94.5 95.4 6.1 N 415 29.1 50.4 69.6 84.6 61.7 47.2 88.4 89.9 78.1 Species-adjusted rates CA 762 42.4 62.6 80.2 92.0 77.0 62.4 94.1 94.8 84.1 N 415 37.8 56.1 79.9 89.8 76.0 59.6 91.6 92.3 83.7 a Statistically significant difference (nominal 0.05 level) between nosocomial and community-acquired susceptibility rates is indicated by a box around the two rates. Isolates were obtained at Grady Memorial Hospital, 1983 to 1986. b CA, Community acquired; N, nosocomial. c Ampi, Ampicillin; Carb, carbenicillin; Cman, cefamandole; Czxm, ceftizoxime-cefotaxime (cefotaxime tested January 1983 to June 1985; ceftizoxime tested July 1985 to December 1986); Cfox, cefoxitin; Cltn, cephalothin; Gent, gentamicin; Tobr, tobramycin; TmSu, trimethoprim-sulfamethoxazole. d l Morganella morganii, 11 P. stuartii, 5 Citrobacterfreundii, 3 Citrobacter diversus, 3 Providencia rettgeri, 2 Acinetobacter calcoaceticus subsp. Iwoffii, 1 Proteus vulgaris. e6 M. morganii, 4 C. diversus, 4 C. freundii, 2 P. stuartii.
1858 McGOWAN ET AL. difference. Of the 72 comparisons (nine drugs tested for each of eight organism groups) between rates for communityacquired and nosocomial strains, 41 differed by less than 5 percentage points, 55 differed by less than 10 percentage points, and 17 differed by more than 10 percentage points. For 35 comparisons, the susceptibility rate was higher for the community-acquired strains, for 23 comparisons the susceptibility rate was higher for the nosocomial isolates, and there were 14 exact ties (usually 0% versus 0%, or 100o versus 100%). None of the comparisons of species-adjusted rates for all of the organisms combined (Table 3) yielded significant differences, although in all comparisons the susceptibility rate was higher for community-acquired strains. The largest difference in species-adjusted rate was 6.5 percentage points for carbenicillin; the smallest was 0.3 percentage points for cefamandole. Thus, while all six of the statistically significant differences (nominal 0.05 level) were in the direction of greater resistance with nosocomial isolates, after appropriate adjustment for multiple significance testing and the calculation of species-adjusted rates, the difference in susceptibility between the community-acquired and nosocomial organisms was slight. Even if the adjusted significance level was increased by deleting the drug-organism comparisons for which no difference was expected because of characteristic resistance (comparisons with 0%o and 0% in Table 3), there were no significant species-adjusted differences between the two groups. If multiple significance testing were adjusted for by using a per combination adjusted significance level of 0.05/72 = 0.0007 (the nominal level of 0.05 divided by the number of combinations [nine drugs tested for each of eight organisms]), the likelihood of a statistically significant difference would be even less than with the per species level presented above. Infection Control Committee minutes for the 36 study months listed 11 clusters of infection due to gram-negative aerobic bacilli (25). These clusters of nosocomial infection were infrequent, they affected small numbers of patients (maximum of 11), and none primarily involved bacteremia. Thus, our study period reflected endemic, not epidemic, nosocomial bacteremia. DISCUSSION Our data suggest that the prevalence of antimicrobial resistance in hospital organisms is largely due to differences between the distribution of organisms which cause nosocomial infections and the distribution of those which cause community-acquired cases. In contrast, hospital strains of a given organism were not likely to be more resistant than community-acquired isolates of that same organism. In the six cases in which nosocomial strains of a given organism were more resistant at the nominal level of significance (0.05) (Table 3), the differences were seen for certain drug-organism pairs and were not a general feature of nosocomial strains against all or most of the drugs tested. Studies of antimicrobial resistance suggest that resistant hospital strains of bacteria have emerged under the selective pressures of antimicrobial use (18). However, the distribution of organisms causing nosocomial infection is not solely the result of antimicrobial use. Some microorganisms have attributes that permit survival in or around medical devices and equipment; these characteristics may be entirely independent of the susceptibility of the organism to antimicrobial agents. For example, the prominence of Providencia stuartii as a pathogen in catheter-associated urinary tract infections ANTIMICROB. AGENTS CHEMOTHER. appears to be related to a unique ability of the organism to survive in catheter materials rather than to its ability to resist several antimicrobial agents (28). Attributes unrelated to antibiotics that enhance hospital survival might confound attempts to measure the relationship between hospital organisms and antibiotics. This confounding should be removed in part by analyzing differences in the rates of resistance by organism, since nonantibiotic factors which permit survival in the hospital would also be present in community-acquired strains of the same organism. When we did this, we saw little difference between the two groups of strains. Prospective reimbursement has set a premium on early discharge from the hospital. This has increased the interchange of patients and organisms between hospital and community. Many organisms involved in community-acquired infections are more resistant to antimicrobial agents than in the past (13). In particular, resistant organisms now are found more frequently in nursing homes (29). When nursing-home patients who have acquired a resistant strain enter the hospital, or when patients with nosocomial infections more frequently require readmission shortly after discharge, the distinction between community and hospital strains may be blurred (23). The degree to which these trends influence our findings is not clear. Our data fail to find statistically significant differences between the two groups; however, it is possible that small differences that we could not detect by studying a population of only 1,077 isolates exist between groups. Whether detection of these smaller differences by using a larger data base would be clinically meaningful would probably vary for each hospital, according to the level of resistance for each organism-drug pair and the antimicrobial agents available for therapy at that institution (5, 12). We studied gram-negative aerobic bacilli isolated from blood cultures; whether our findings can be extrapolated to other organisms or to other sites of infection is not known. Likewise, the results depend critically on the ability to distinguish community-acquired cases of bacteremia from nosocomial cases; it is likely that the ease with which this can be done varies from center to center (7, 23). Because of this, our findings may not apply in other settings (2, 5, 11). However, our results from a municipal hospital from 1983 to 1986 agree with those of Dixon from eight community hospitals from 1972 to 1973 (4) and with other results, from a Veterans Administration Hospital, published in 1982 (9). Thus, similar rates of susceptibility in nosocomial and community-acquired strains can be found in several different types of hospitals in different time periods. The organismspecific nature of antimicrobial resistance in hospital infections has been noted by several authors (1, 10, 18, 19, 29); perhaps the characteristics of the microbe are more important than the type of hospital in which the organisms are found. ACKNOWLEDGMENTS We are indebted to C. Perlino, C. Job, C. Parrish, L. Frawley, and E. Whitworth of the Epidemiology Department, Grady Memorial Hospital, for their help in identifying nosocomial cases. LITERATURE CITED 1. Alford, R. H., and A. Hall. 1987. Epidemiology of infections caused by gentamicin-resistant Enterobacteriaceae and Pseudomonas aeruginosa over 15 years at the Nashville Veterans Administration Medical Center. Rev. Infect. Dis. 9:1079-1086. 2. Atkinson, B. A., and V. Lorian. 1984. Antimicrobial agent
VOL. 33, 1989 ANTIBIOTIC RESISTANCE IN NOSOCOMIAL BACTEREMIA 1859 susceptibility patterns of bacteria in hospitals from 1971 to 1982. J. Clin. Microbiol. 20:791-796. 3. Cohen, J. 1976. Variations in sensitivities to antibiotics. Nosocomial versus community-acquired infections caused by same organism. N. Y. State J. Med. 76:391-393. 4. Dixon, R. E. 1975. Epidemiology of drug resistance in hospitals, p. 349-360. In S. Mitsuhashi, L. Rosival, and V. Krcmery (ed.), Drug-inactivating enzymes and antibiotic resistance. Springer- Verlag, New York. 5. EUner, P. D., D. J. Fink, H. C. Neu, and M. F. Parry. 1987. Epidemiologic factors affecting antimicrobial resistance of common bacterial isolates. J. Clin. Microbiol. 25:1668-1674. 6. Forgacs, I. C., S. J. Eykyn, and R. D. Bradley. 1986. Serious infection in the intensive therapy unit: a 15-year study of bacteraemia. Q. J. Med. 232:773-779. 7. Garner, J. S., W. R. Jarvis, T. G. Emori, T. C. Horan, and J. M. Hughes. 1988. CDC definitions for nosocomial infections, 1988. Am. J. Infect. Control 16:128-140. 8. Gaynes, R. P., R. Cooksey, C. Thornsberry, J. M. Swenson, and J. M. Hughes. 1987. Mechanisms of aminoglycoside resistance among beta-lactam-resistant Escherichia coli in the United States. Diagn. Microbiol. Infect. Dis. 7:45-50. 9. Klingman, K., and J. M. Mylotte. 1982. Ampicillin and cephalothin susceptibility of community-acquired Enterobacteriaceae. N. Y. State J. Med. 82:1801-1804. 10. Lorian, V., and B. Topf. 1972. Microbiology of nosocomial infections. Arch. Intern. Med. 130:104-110. 11. Lynch, J. M., G. R. Hodges, G. M. Clark, and D. L. Dworzack. 1981. Gram-negative bacteremia. Analysis of factors for clinical assessment of gentamicin resistance. Arch. Intern. Med. 141: 582-586. 12. Magnussen, C. R., and M. T. Jacobson. 1984. Longitudinal analysis of endemic gentamicin- and tobramycin-resistant gramnegative bacilli in a community hospital. Infect. Control 5: 88-92. 13. McCue, J. D. 1985. Cefoxitin resistance in community-acquired gram-negative bacillary bacteremia. Arch. Intern. Med. 145: 834-836. 14. McGowan, J. E., Jr. 1983. Antimicrobial resistance in hospital organisms and its relation to antibiotic use. Rev. Infect. Dis. 5:1033-1048. 15. McGowan, J. E., Jr. 1985. Changing etiology of nosocomial bacteremia and fungemia and other hospital-acquired infections. Rev. Infect. Dis. 7(Suppl. 3):S357-S370. 16. McGowan, J. E., Jr. 1987. Is antimicrobial resistance in hospital microorganisms related to antibiotic use? Bull. N. Y. Acad. Med. 63:253-268. 17. McGowan, J. E., Jr., C. Garner, C. Wilcox, and M. Finland. 1974. Antibiotic susceptibility of gram-negative bacilli isolated from blood cultures. Results of tests with 35 agents and strains from 169 patients at Boston City Hospital during 1972. Am. J. Med. 57:225-238. 18. McGowan, J. E., Jr., P. L. Parrott, and V. P. Duty. 1977. Nosocomial bacteremia: potential for prevention of procedurerelated cases. J. Am. Med. Assoc. 237:2727-2729. 19. Murthy, S. K., A. L. Baltch, R. P. Smith, E. K. Desjardin, M. C. Hammer, J. V. Conroy, and P. B. Michelsen. 1989. Oropharyngeal and fecal carriage of Pseudomonas aeruginosa in hospital patients. J. Clin. Microbiol. 27:35-40. 20. National Committee for Clinical Laboratory Standards. 1984. Approved standard M2-A3. Performance standards for antimicrobial disk susceptibility tests, 3rd ed. National Committee for Clinical Laboratory Standards, Villanova, Pa. 21. O'Brien, P. C., and M. A. Shampo. 1988. Statistical considerations for performing multiple tests in a single experiment. 1. Introduction. Mayo Clin. Proc. 63:813-815. 22. O'Brien, T. F., J. F. Acar, A. A. Medeiros, R. A. Norton, F. Goldstein, and R. L. Kent. 1978. International comparison of prevalence of resistance to antibiotics. J. Am. Med. Assoc. 239:1518-1523. 23. Rhame, F. S. 1987. Surveillance objectives: descriptive epidemiology. Infect. Control 8:454-458. 24. Rubinstein, E., Z. Mark, G. Keren, M. Alkan, S. Berger, and B. Bogokowski. 1986. Comparative activity of ofloxacin with reference to bacterial strains isolated in in-patients and out-patients. Infection 14(Suppl. 1):S20-S25. 25. Schaberg, D. R., R. W. Haley, A. K. Highsmith, R. L. Anderson, and J. E. McGowan, Jr. 1980. Nosocomial bacteriuria: a prospective study of case clustering and antimicrobial resistance. Ann. Intern. Med. 93:420-424. 26. Siegman-Igra, Y., D. Schwartz, I. Gonen, and N. Konforti. 1987. Bacteremic infections in granulocytopenic patients in Israel. Isr. J. Med. Sci. 23:1214-1216. 27. Smith, D. G., J. Clemens, W. Crede, M. Harvey, and E. J. Gracely. 1987. Impact of multiple comparisons in randomized clinical trials. Am. J. Med. 83:545-550. 28. Warren, J. W. 1986. Providencia stuartii: a common cause of antibiotic-resistant bacteriuria in patients with long-term indwelling catheters. Rev. Infect. Dis. 8:61-67. 29. Weinstein, R. A. 1987. Resistant bacteria and infection control in the nursing home and hospital. Bull. N. Y. Acad. Med. 63:337-344.