Leonard Leibovici, MD Zmira Samra, PhD Hanna Konisberger, MSc Ofra Kalter-Leibovici, MD Silvio D. Pitlik, MD Moshe Drucker, MD
Bacteremia in Adult Diabetic Patients
Objective: To compare the microbiology, sources, complications, and outcome of bacteremia in diabetic and nondiabetic patients. Research Design and Methods: A prospective study was conducted of all episodes of bacteremia in hospitalized diabetic and nondiabetic patients. The study consisted of patients >18 yr of age with bacteremia detected within a 19-mo interval. Results: We compared 124 episodes of bacteremia in 119 diabetic patients to 508 episodes in 480 nondiabetic patients. Diabetic patients were older than nondiabetic patients (median age 74 vs. 68 yr, P = 0.0001). In patients with an indwelling urinary catheter and bacteremic urinary tract infection, the percentage of Klebsiella in diabetic patients was 60% (6 of 10) and in nondiabetic patients was 17% (4 of 23, P = 0.04). In patients without an obvious source of bacteremia, the percentage of staphylococcal isolates in diabetic patients was 29% (10 of 35) and in nondiabetic patients was 14% (24 of 176, P = 0.04). Staphylococci were a common cause of bacteremic infections of the extremities in diabetic patients (12 of 19, 63%) and nondiabetic patients (20 of 50, 40%). Septic shock was the only complication that was more common in diabetic patients. The mortality in diabetic and nondiabetic patients was 28 and 29%, respectively. Conclusions: Our results represent elderly patients with non-insulin-dependent diabetes mellitus. In this group, empirical treatment for suspected bacteremic urinary tract infection in patients with a urinary catheter should include coverage for Klebsiella. Empiric treatment for
From the Department of Internal Medicine B, the Microbiology Laboratory, the Infectious Diseases Unit, and the Institute for Pediatric and Adolescent Endocrinology, Beilinson Medical Center, Petah Tiqva; and the Sackler Faculty of Medicine, Tel Aviv University, Ramat Aviv, Israel. Address correspondence and reprint requests to Leonard Leibovici, MD, Department of Internal Medicine B, Beilinson Medical Center, 49 100, Petah Tiqva, Israel. Received for publication 25 May 1990 and accepted in revised form 24 September 1990.
DIABETES CARE, VOL. 14, N O . 2, FEBRUARY 1991
suspected bacteremia of unknown origin or caused by infection of the extremities should include an antistaphylococcal drug. The prognosis of bacteremia in diabetic and nondiabetic patients was similar. Diabetes Care 14:89-94, 1991
B
acteremia has a mortality rate of 20-30% (1-5). Diabetic patients accounted for a larger percentage of those with bloodstream infections than expected due to their proportion among hospitalized patients (4,5). Thus, the death rates attributable to bacteremia in hospitalized diabetic patients was higher than in nondiabetic patients (4). Early (i.e., empiric) treatment with antibiotics, to which the infecting microorganisms are susceptible, probably reduces mortality caused by bloodstream infections (1-3). A relatively high prevalence of diabetic patients have been found among patients with bacteremic urinary tract infection (UTI; 5), acute pyelonephritis (6), Escherichia coli pneumonia (7), emphysematous cholecystitis (8), and bacteremia caused by Staphylococcus aureus (9,10), Proteus mirabilis (11), group B streptococci (12), and Klebsiella pneumoniae (13). However, it is not clear whether those findings imply that the empiric treatment of suspected bacteremia in diabetic patients ought to be different than that in nondiabetic patients. In this study, we surveyed prospectively all episodes of bacteremia that were detected in diabetic patients during a 19-mo interval and compared them to bacteremia in nondiabetic patients. We looked for significant differences in the prevalence of the isolated microorganisms that would imply the need for a different empiric treatment in diabetic patients. We also compared
89
BACTEREMIA IN DIABETIC PATIENTS
the occurrence of complications and the outcome of infection in the two groups.
RESEARCH DESIGN AND METHODS Beilinson Medical Center is a 900-bed university hospital that serves an urban population of -200,000 as a first-line facility. It is also a referral center for several other hospitals. Included in the study were all episodes of bacteremia that occurred in adult patients ^18 yr of age between March 1988 and September 1989. The episodes were detected by daily surveillance of laboratory records. Ten milliliters of venous blood was aseptically obtained and inoculated into a two-bottle set (6B aerobic and 7D anaerobic tryptic soy broth, Johnston, Towson, MD). The bottles were incubated at 37°C and tested on the Bactec 460 system (Johnston) twice on the 1st day and then daily for 7 days. The median number of sets obtained during one bacteremic episode was three (range 1-9). Identification of microorganisms was performed with standard methods. Organisms that are commonly recovered from the environment or skin were judged as contaminants, unless the clinical findings, results of cultures from other body sites, or the number of positive blood cultures (^2 sets) indicated a high probability for true bloodstream infection. Clinical data were collected prospectively from the patients' charts and the attending physicians. We recorded the patient's age, sex, domicile, functional capacity before hospitalization, cause of admission, medications, and diseases other than the infectious one. Patients were defined as diabetic if they had been previously diagnosed as such or if they received oral hypoglycemic agents or insulin injections before admission and if they had a fasting blood glucose level of >7.8 mM when no longer febrile. In patients who died before defervescence, we used the first fasting blood glucose level obtained. We did not attempt to classify patients as insulin dependent or non-insulin dependent (NIDDM) and did not systematically search for evidence of microvascular or neurological complications. For every episode of bacteremia, we recorded the day and the ward in which it occurred, the source of bacteremia, presence of urinary catheter, central or peripheral intravenous lines and endotracheal tube, the patient's temperature, systolic and diastolic blood pressure, the results of chest X rays, total and differential white blood cells counts, hemoglobin level, and blood chemistry. An episode of bacteremia was defined as hospital acquired if it occurred >48 h postadmission (14). During follow-up, the antibiotic treatment, results of blood and urine cultures, and cultures taken from other sites were recorded. We also recorded complications that occurred during hospitalization (the development of septic shock, adult respiratory distress syndrome,
90
disseminated intravascular coagulopathy, acute kidney failure, coma or other neurological complications, septic emboli, allergic reactions to antibiotic drugs, antibiotic-associated diarrhea or pseudomembranous colitis, and the appearance of bed sores), the duration of the febrile disease and hospital stay, and the outcome of hospitalization, whether discharge or demise. We compared diabetic to nondiabetic patients. Because only 17 diabetic patients were treated with insulin injections, they were included in the diabetic patient group and were not analyzed as a separate group. The deletion of insulin-treated patients from the statistical analysis did not change any of the results. For univariate analysis, we used the x2-test (or the Fisher exact test for small numbers) to look for statistical significance of contingency tables. Comparisons were adjusted for quartiles of age or other confounding variables by the Cochran-Mantel-Haenszel statistic, and the adjusted relative risk with 95% confidence interval (95% Cl) was computed with the Cochran-Mantel-Haenszel statistic (15). The reference category was the group of nondiabetic patients (relative risk 1). Most continuous variables were not normally distributed; therefore, we used Wilcoxon's rank-sum test to look for significance of continuous values compared between two classes. Values of continuous variables are given as the median with range in parentheses. A stepwise logistic regression procedure (LOGIST; 16) was used to perform multivariate regression analysis. To test the influence of multiple factors (including diabetes) on survival, we used Cox's life-table regression model (procedure PHGLM; 17).
RESULTS During the study, 124 episodes of bacteremia occurred in 119 diabetic patients, and 508 episodes occurred in 480 nondiabetic patients. The median age of diabetic patients was 74 yr (range 34-98 yr) and of nondiabetic patients was 68 yr (range 18-99 yr, P = 0.0001). Fiftyseven percent of diabetic patients and 50% of nondiabetic patients were men (P = 0.2). Twenty-nine percent (36 of 124) of episodes of bacteremia in diabetic patients were hospital acquired vs. 39% (197 of 508) of episodes in nondiabetic patients (P = 0.04). However, this difference was not significant when adjusted for age (relative risk 0.7, 95% Cl 0.51.2). More than one microorganism was isolated in 3 (2.4%) episodes in diabetic patients and in 31 episodes (6%) in nondiabetic patients (P = 0.1). Underlying disorders in diabetic and nondiabetic patients are detailed in Table 1. With stepwise logistic regression analysis, which had as the dependent variable the presence of diabetes and as independent variables age and the underlying disorders, only three factors were found to be significantly and independently associated with diabetes: age (x2 in the model = 11,
DIABETES CARE, VOL. 14, NO. 2, FEBRUARY 1991
L. LEIBOVICI AND ASSOCIATES
TABLE 1 Underlying disorders in diabetic and nondiabetic patients Nondiabetic patients i(n = 480)
Diabetic patients (n == 119)
Hypertension Congestive heart failure Cerebral vascular accident Decubitus ulcer Corticosteroid treatment Chronic kidney failure Hemodialysis
n
%
n
%
P
Relative risk, age adjusted (95% confidence interval)
35 18 14 10 13 11 5
29 15 12 8 11 9 4
50 34 31 19 90 33 9
10 7 6 4 19 7 2
0.0001 0.004 0.04 0.04 0.5 0.35 0.12
3.1 (1.9-5.0) 1.9(1.1-3.6) 1.7 (0.9-3.2) 2.2 (1.0-4.9) 0.6(0.3-1.2) 1.3 (0.7-2.6) 3.5 (1.2-10.0)
P = 0.001), hypertension (x2 = 22, P = 0.0001), and congestive heart failure (\2 = 5, P = 0.03). Sources of bacteremia are detailed in Table 2. The only difference that reached statistical significance on univariate analysis was the preponderance of UTI as a source of bacteremia among diabetic patients (50 of 124 vs. 154 of 508 episodes in nondiabetic patients, P = 0.03), but this difference was explained by the higher age of the diabetic patients (Cochran-MantelHaenszel statistic controlling for age 1.5, P = 0.22, relative risk = 1.3, 95% Cl 0.8-2.0). The overall percentages of microorganisms isolated from blood cultures were similar in diabetic and nondiabetic patients (Table 3). In episodes in which the source of bacteremia was the urinary tract, the percentage of Klebsiella isolates was 25% (13 of 51) in diabetic patients and 12% (20 of 161) in nondiabetic patients (P = 0.03, relative risk adjusted for age 2.4, 95% Cl 1.1-5.2). In patients with a urinary catheter and bacteremic UTI, Klebsiella was isolated in 60% (6 of 10) of diabetic patients and in 17% (4 of 23) of nondiabetic patients (P = 0.04). The percentage of Klebsiella isolates in patients without a urinary catheter was similar in diabetic and nondiabetic patients, but the preponderance of Klebsiella as a cause of bacteremic UTI in diabetic patients was significant even when corrected for the presence of the catheter (relative risk 2.3, 95% Cl 1.0-5.1). This difference was also significant when corrected for the acquisition of bacteremia in the hospital and for the presence of underlying disorders other than diabetes (data not shown). In patients without an obvious source of bacteremia, the percentage of Staphylococcus isolates was 29% (10 of 35) in diabetic patients and 14% (24 of 76) in nondiabetic patients (P = 0.04, relative risk controlling for age 2.7, 95% Cl 1.1-6.7). The higher percentage of staphylococci in diabetic patients without an obvious source of infection remained significant when controlled for the acquisition of bacteremia in the hospital, the presence of peripheral or central intravenous lines, and underlying disorders (data not shown). The percentage of Staphylococcus isolates in diabetic
DIABETES CARE, VOL. 14, N O . 2, FEBRUARY 1991
patients with an infection of the extremities was 63% (12 of 19) and in nondiabetic patients was 40% (20 of 50, P = 0.1). Septic shock was diagnosed in 14% (18 of 124) of bacteremic episodes in diabetic patients and in 7% (37 of 508) of episodes in nondiabetic patients (P = 0.01, relative risk adjusted for age 1.9, 95% Cl 1.0-3.4). All other complications of bacteremia occurred in similar rates in diabetic and nondiabetic patients. Twenty-eight percent (35 of 124) of episodes of bacteremia in diabetic patients ended in death vs. 29% (146 of 504) in nondiabetic patients (P = 0.87, relative risk adjusted for age 1.1, 95% Cl 0.7-1.7). The only subgroup in which diabetic patients fared worse was patients with intra-abdominal infections: 3 of 7 diabetic patients (43%) died vs. 7 of 43 (16%) nondiabetic patients, but the numbers were too small to reach statistical significance (P = 0.1, relative risk adjusted for age 3.9, 95% Cl 0.7-21.1). With stepwise logistic regression analysis, six factors were found to be associated both significantly and independently with a fatal outcome: septic shock, neutropenia, malignancy, hospital-acquired bacteremia, a source of bacteremia other than UTI, and an elevated blood urea nitrogen level on admission. Identical factors
TABLE 2 Sources of bacteremia in diabetic and nondiabetic patients Diabetic patients (n = 124)
Urinary tract Unknown Lower respiratory tract Extremities Intra-abdominal Other
Nondiabetic patients (n = 508 )
n
%
n
%
50 33 11 17 7 6
40* 27 9 14 6 5
154 157 42 46 43 66
30 31 8 9 8 13
*P = 0.03 vs. patients without diabetes.
91
BACTEREMIA IN DIABETIC PATIENTS
TABLE 3 Percentages of microorganisms isolated from blood cultures of diabetic and nondiabetic patients according to source of infection Lower Urinary tract
Isolates (n) Microorganism (%) Staphylococcus aureus Staphylococcus coagulase negative Pneumococcus Enterococcus Other streptococci Total Grampositive bacteria Escherichia coli Klebsiella Pseudomonas Enterobacter Acinetobacter Proteus Other Gramnegative Total Gramnegative bacteria Bacteroides Fungi
re jpiratory tract
Unknown
Abdominal
Other
Extremities
Total
DP
NDP
DP
NDP
DP
NDP
DP
NDP
DP
NDP
DP
NDP
DP
NDP
51
161
35
176
12
48
8
51
19
50
6
68
131
554
0
1
20*
12
8
12
0
6
53
36
0
18
14
14
0 0 0 2
0 0 3
2 2 2 8
0 33 0 0
0 33 2 6
0 0 12 0
0 2 4 6
10 0 5 0
4 2 2 26
0 0 17 67
1 9
1
9 0 6 6
16
4 3 4 5
1 5 3 8
2
4
40
26
42
54
12
18
68
70
83
48
30
28
47 25* 8 2 0 10
60 12 7
15 12 15 5 6
8 2
27 12 25 0 0 0
8 18 8
4
8 9 9 0 0 8
2 6 13
1 9
9 8 8 3 3 0
4 0
0 10 5 5 5 0
0 2 4 2 8 6
0 17 0 0 0 0
6 6 6 3 10 2
23 15 8 2 1 5
26 11 10 5 5 5
4
3
17
9
17
6
25
10
5
6
0
7
10
6
96
96
46
66
50
42
87
28
65
67
2
0
2
0
0
1
2
0
5
8
2
0
2
0
4 0
4
2
3 11
17 0
40
0
80 0
32
0
0
7
5
3
4
4
14
4
DP, diabetic patients; NDP, nondiabetic patients. *P < 0.05 vs. NDP.
were delineated by Cox's life-table regression analysis, which takes into account the time elapsed till death or discharge of the patient. Diabetes was not found to be associated with fatal outcome on either regression procedure. The median duration of hospital stay till discharge was 11 days (range 3-84 days) in diabetic patients and 13 days (range 1-105 days) in nondiabetic patients. The median duration of hospital stay till demise was 13 days (range 2-96 days) in diabetic patients and 15 days (range 1-106 days) in nondiabetic patients. Neither comparison was statistically significant.
CONCLUSIONS
In this study, we compared episodes of bacteremia in diabetic patients to bacteremia in nondiabetic patients. The median age of the diabetic group was 74 yr, and only 14% were treated with insulin. Thus, the data pre-
92
sented herein probably reflect characteristics of elderly NIDDM patients with bacteremia. Patients were designated as diabetic if they had hyperglycemia and had been diagnosed or treated for such in the past. This definition underestimates the number of NIDDM patients in the population (18). However, this is usually the extent of information available to the physician in the emergency ward, whereas the significance of NIDDM discovered by screening is not clear (19). As in previous studies, UTI was a more frequent source of bacteremia in diabetic than nondiabetic patients (4,5). However, when corrected for the higher age of the diabetic patients, this association was not significant (relative risk 1.3, 95% Cl 0.8-2.0). A second question of interest is whether diabetic patients have bacteremia of unknown origin more often than nondiabetic patients. Mellors et al. (20) and Leibovici et al. (21) found an excess of diabetic patients in bacteremic patients without an obvious source of fever. In this study and the study of MacFarlane et al. (5), the percentage
DIABETES CARE, V O L . 14, N O . 2, FEBRUARY 1991
L LEIBOVICI AND ASSOCIATES
of bacteremia without an obvious source was similar in diabetic and nondiabetic patients. Overall, we were unable to demonstrate any significant differences in the sources of bacteremia when comparing diabetic to nondiabetic patients. The overall percentages of microorganisms causing bacteremia were similar in diabetic and nondiabetic patients. However, Klebsiella bacteremia was twice as common in diabetic patients with bacteremic UTI. The difference is explained by the fact that 60% of bacteremic UTIs in diabetic patients with an indwelling urinary catheter were caused by Klebsiella. This finding was not sought for or demonstrated in previous studies on bacteremia in diabetic patients (4,5) or on Klebsiella bacteremia (13,22,23). Klebsiella is generally less susceptible to antibiotics than Escherichia coli, the most common pathogen causing UTI (24-26). Studies on staphylococcal bacteremia mention diabetes as an underlying disease in 12-14% of patients, but because the denominator data (the total number of diabetic and nondiabetic patients with bacteremia) were missing, we cannot conclude from these studies whether diabetic patients are at increased risk for staphylococcal bacteremia. In our study and in two other studies on bacteremia in diabetic patients, staphylococci were isolated in similar rates in diabetic and nondiabetic patients (4,5). Diabetic patients without an obvious source of bacteremia had twice the rate of staphylococcal isolates than nondiabetic patients (29 vs. 14%, P = 0.04). The percentage of staphylococcal bloodstream isolates in infections of the extremities was 63% in diabetic patients. Indeed, staphylococci are the most common pathogens in diabetic foot infections (28) and in osteomyelitis of the feet in diabetic patients (29). The relative risk of diabetic versus nondiabetic patients to develop septic shock was 1.9 (adjusted for age). Patients with septic shock fare worse than other bacteremic patients, but the absolute numbers are too small for the difference in the risk to develop shock to be expressed as a difference in mortality. We found no other complication to be more common in diabetic patients. The overall mortality of diabetic patients in this study was similar to that in nondiabetic patients, as found by Bryan et al. (4) and MacFarlane et al. (5). Diabetes did not correlate with mortality on multivariate regression analysis, and we did not find any subgroup of patients in which diabetic patients were at a higher risk for a fatal outcome. The only exception was the higher mortality of diabetic patients whose source of bacteremia was an intra-abdominal infection, but the number of patients was too small to reach statistical significance. Nevertheless, other studies support this observation (30-32). In summary, we surveyed prospectively 124 episodes of bloodstream infections in diabetic patients and compared them to 508 episodes in nondiabetic patients. Ours was an elderly population representative of
DIABETES CARE, VOL. 14, N O . 2, FEBRUARY 1991
NIDDM patients. Although the overall sources of bacteremia and the rates of the infecting microorganisms were similar in diabetic and nondiabetic patients, two findings justify a different empiric treatment in diabetic patients suspected of bacteremia: a high percentage of Klebsiella isolates in patients with a urinary catheter and a high percentage of staphylococci isolated from the blood of diabetic patients without an obvious source of bacteremia. Empiric antibiotic treatment for bacteremic infection of the extremities should include an antistaphylococcal drug both in diabetic and nondiabetic patients. Diabetic patients had a higher percentage of septic shock and probably fared worse if the source of bacteremia was an intra-abdominal infection. Otherwise, the rates of complications and mortality were similar in diabetic and nondiabetic patients.
REFERENCES 1. Kreger BE, Craven DE, McCabe WR: Gram-negative bacteremia. IV. Re-evaluation of clinical features and treatment in 612 patients. Am I Med 68:344-54, 1980 2. Weinstein R, Murphy JR, Reller LB, Lichtenstein KA: The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. II. Clinical observations, with special reference to factors influencing prognosis. Rev Infect Dis 5:54-70, 1983 3. Rayner BL, Willcox PA: Community-acquired bacteremia: a prospective survey of 239 cases. Q / Med 69:907-19, 1988 4. Bryan CS, Reynolds KL, Metzger WT: Bacteremia in diabetic patients: comparison of incidence and mortality with nondiabetic patients. Diabetes Care 8:244-49, 1985 5. MacFarlane IA, Brown RM, Smyth RW, Burdon DW, FitzGerald MG: Bacteremia in diabetics. / Infect Dis 12:213-19, 1986 6. Seymour A, Phear D: The causes of death in diabetes mellitus: a study of diabetic mortality in the Royal Adelaide Hospital from 1956 to 1960. Med J Aust 50:89094, 1963 7. Tillotson JR, Lerner AM: Characteristics of pneumonia caused by Escherichia coli. N Engl I Med 277:115-22, 1967 8. Mentzer RM, Golden GT, Chandler JG, Horsley JS III: A comparative appraisal of emphysematous cholecystitis. Am I Surg 129:10-15, 1975 9. Musher DM, McKenzie SO: Infections due to Staphylococcus aureus. Medicine 56:383-409, 1977 10. Cluff LE, Reynolds RC, Page DL, Breckinridge JC: Staphylococcal bacteremia and altered host resistance. Ann Intern Med 69:859-73, 1968 11. Adler JL, Burke JP, Martin DF, Finland M: Proteus infections in a general hospital. II. Some clinical and epidemiological characteristics with an analysis of 71 cases of Proteus bacteremia. Ann Intern Med 75:531-36, 1971 12. Opal SM, Cross A, Palmer M, Almazan R: Group B streptococcal sepsis in adults and infants: contrasts and comparisons. Arch Intern Med 148:641-45, 1988
93
BACTEREMIA IN DIABETIC PATIENTS
13.
14.
15. 16. 17.
18.
19. 20.
21. 22. 23.
94
Haddy Rl, Lee M, Sangal SP, Walbroehl GS, Hambrick CS, Sarti GM: Klebsiella pneumoniae bacteremia in the community hospital. / Earn Pract 28:686-90, 1989 Centers for Disease Control: National Nosocomial Infections Study Site Definitions Manual. Atlanta, GA, Centers for Disease Control, 1972, p. 882-84 SAS Institute Inc: SAS User's Guide: Statistics, Version 5 Edition. Cary, NC, SAS Inst., 1985, p. 420-23 Harrell FE: The LOGIST procedure. In SAS Institute Inc. SUGI Supplemental Library User's Guide, Version 5 Edition. Cary, NC, SAS Inst., 1986, p. 269-93 Harrell FE: The PHGLM Procedure. In SAS Institute Inc. SUGI Supplemental Library User's Guide, Version 5 Edition. Cary, NC, SAS Inst., 1986, p. 437-66 Harris Ml, Hadden WC, Knowler WC, Bennett PH: Prevalence of diabetes and impaired glucose tolerance and plasma glucose levels in U.S. population aged 20-74 yr. Diabetes 36:523-34, 1987 Singer DE, Samet JH, Coley CM, Nathan DM: Screening for diabetes mellitus. Ann Intern Med 109:639-49, 1988 Mellors JW, Horwitz Rl, Harvey MR, Horwitz SM: A simple index to identify occult bacterial infection in adults with acute unexplained fever. Arch Intern Med 147:66671, 1987 Leibovici L, Cohen O, Wysenbeek AJ: Occult bacterial infection in adults with unexplained fever: validation of a diagnostic index. Arch Intern Med 150:1270-72, 1990 Montgomerie JZ, Ota JK: Klebsiella bacteremia. Arch Intern Med 140:525-27, 1980 Garcia de la Torre M, Romera-Vives J, Martinez-Beltran J, Guerrero A, Meseguer M, Bouja E: Klebsiella bacteremia: an analysis of 100 episodes. Rev Infect Dis 7:14350, 1985
24. Atkinson BA, Lorian V: Antimicrobial susceptibility patterns of bacteria in hospitals from 1971 to 1982. ) Clin M/crob/o/230:791-96, 1984 25. European Study Group in Antibiotic Resistance: Susceptibility to beta-lactam antibiotics in septicemia isolates from twenty-nine European laboratories, fur / Glin Microb/'o/6:515-20, 1987 26. Martin MA, Pfaller MA, Rojas PB, Woolson RF, Wenzen RP: In-vitro susceptibility of nosocomial gram-negative bloodstream pathogens to quinolones and other antibiotics: a statistical approach. / Antimicrob Chemother 23:353-61, 1989 27. Mylotte JM, McDermott C, Spooner JA: Prospective study of 114 consecutive episodes of Staphylococcus aureus bacteremia. Rev Infect Dis 9:891-907, 1987 28. Wheat LJ, Allen SD, Henry M, Kernik CB, Siders JA, Kuebler T, Fineberg N, Norton J: Diabetic foot infections: bacteriologic analysis. Arch Intern Med 146:1935-40, 1986 29. Baumberger DM, Daus GP, Gerding DN: Osteomyelitis in the feet of diabetic patients: long-term results, prognostic factors, and the role of antimicrobial and surgical therapy. Am ) Med 83:653-60, 1987 30. DiPalo S, Ferrari G, Castoldi R, Fiacco E, Cristallo M, Staudacher C, Chiesa R, DiCarlo V: Surgical septic complications in diabetic patients. Ada Diabetol Lat 25:4954, 1988 31. Hickman MS, Schwesinger WH, Page CP: Acute cholecystitis in the diabetic: a case-control study of outcome. Arch Surg 123:409-11, 1988 32. Kalfarentzos FE, Dougenis DV, Cristopoulos DC, Spiliotis JD, Williams M, Androulakis J: Prognostic criteria in intraabdominal sepsis. Int Surg 72:185-87, 1987
DIABETES CARE, VOL. 14, N O . 2, FEBRUARY 1991
Bacteremia in Adult Diabetic Patients
Objective: To compare the microbiology, sources, complications, and outcome of bacteremia in diabetic and nondiabetic patients. Research Design and Methods: A prospective study was conducted of all episodes of bacteremia in hospitalized diabetic and nondiabetic patients. The study consisted of patients >18 yr of age with bacteremia detected within a 19-mo interval. Results: We compared 124 episodes of bacteremia in 119 diabetic patients to 508 episodes in 480 nondiabetic patients. Diabetic patients were older than nondiabetic patients (median age 74 vs. 68 yr, P = 0.0001). In patients with an indwelling urinary catheter and bacteremic urinary tract infection, the percentage of Klebsiella in diabetic patients was 60% (6 of 10) and in nondiabetic patients was 17% (4 of 23, P = 0.04). In patients without an obvious source of bacteremia, the percentage of staphylococcal isolates in diabetic patients was 29% (10 of 35) and in nondiabetic patients was 14% (24 of 176, P = 0.04). Staphylococci were a common cause of bacteremic infections of the extremities in diabetic patients (12 of 19, 63%) and nondiabetic patients (20 of 50, 40%). Septic shock was the only complication that was more common in diabetic patients. The mortality in diabetic and nondiabetic patients was 28 and 29%, respectively. Conclusions: Our results represent elderly patients with non-insulin-dependent diabetes mellitus. In this group, empirical treatment for suspected bacteremic urinary tract infection in patients with a urinary catheter should include coverage for Klebsiella. Empiric treatment for
From the Department of Internal Medicine B, the Microbiology Laboratory, the Infectious Diseases Unit, and the Institute for Pediatric and Adolescent Endocrinology, Beilinson Medical Center, Petah Tiqva; and the Sackler Faculty of Medicine, Tel Aviv University, Ramat Aviv, Israel. Address correspondence and reprint requests to Leonard Leibovici, MD, Department of Internal Medicine B, Beilinson Medical Center, 49 100, Petah Tiqva, Israel. Received for publication 25 May 1990 and accepted in revised form 24 September 1990.
DIABETES CARE, VOL. 14, N O . 2, FEBRUARY 1991
suspected bacteremia of unknown origin or caused by infection of the extremities should include an antistaphylococcal drug. The prognosis of bacteremia in diabetic and nondiabetic patients was similar. Diabetes Care 14:89-94, 1991
B
acteremia has a mortality rate of 20-30% (1-5). Diabetic patients accounted for a larger percentage of those with bloodstream infections than expected due to their proportion among hospitalized patients (4,5). Thus, the death rates attributable to bacteremia in hospitalized diabetic patients was higher than in nondiabetic patients (4). Early (i.e., empiric) treatment with antibiotics, to which the infecting microorganisms are susceptible, probably reduces mortality caused by bloodstream infections (1-3). A relatively high prevalence of diabetic patients have been found among patients with bacteremic urinary tract infection (UTI; 5), acute pyelonephritis (6), Escherichia coli pneumonia (7), emphysematous cholecystitis (8), and bacteremia caused by Staphylococcus aureus (9,10), Proteus mirabilis (11), group B streptococci (12), and Klebsiella pneumoniae (13). However, it is not clear whether those findings imply that the empiric treatment of suspected bacteremia in diabetic patients ought to be different than that in nondiabetic patients. In this study, we surveyed prospectively all episodes of bacteremia that were detected in diabetic patients during a 19-mo interval and compared them to bacteremia in nondiabetic patients. We looked for significant differences in the prevalence of the isolated microorganisms that would imply the need for a different empiric treatment in diabetic patients. We also compared
89
BACTEREMIA IN DIABETIC PATIENTS
the occurrence of complications and the outcome of infection in the two groups.
RESEARCH DESIGN AND METHODS Beilinson Medical Center is a 900-bed university hospital that serves an urban population of -200,000 as a first-line facility. It is also a referral center for several other hospitals. Included in the study were all episodes of bacteremia that occurred in adult patients ^18 yr of age between March 1988 and September 1989. The episodes were detected by daily surveillance of laboratory records. Ten milliliters of venous blood was aseptically obtained and inoculated into a two-bottle set (6B aerobic and 7D anaerobic tryptic soy broth, Johnston, Towson, MD). The bottles were incubated at 37°C and tested on the Bactec 460 system (Johnston) twice on the 1st day and then daily for 7 days. The median number of sets obtained during one bacteremic episode was three (range 1-9). Identification of microorganisms was performed with standard methods. Organisms that are commonly recovered from the environment or skin were judged as contaminants, unless the clinical findings, results of cultures from other body sites, or the number of positive blood cultures (^2 sets) indicated a high probability for true bloodstream infection. Clinical data were collected prospectively from the patients' charts and the attending physicians. We recorded the patient's age, sex, domicile, functional capacity before hospitalization, cause of admission, medications, and diseases other than the infectious one. Patients were defined as diabetic if they had been previously diagnosed as such or if they received oral hypoglycemic agents or insulin injections before admission and if they had a fasting blood glucose level of >7.8 mM when no longer febrile. In patients who died before defervescence, we used the first fasting blood glucose level obtained. We did not attempt to classify patients as insulin dependent or non-insulin dependent (NIDDM) and did not systematically search for evidence of microvascular or neurological complications. For every episode of bacteremia, we recorded the day and the ward in which it occurred, the source of bacteremia, presence of urinary catheter, central or peripheral intravenous lines and endotracheal tube, the patient's temperature, systolic and diastolic blood pressure, the results of chest X rays, total and differential white blood cells counts, hemoglobin level, and blood chemistry. An episode of bacteremia was defined as hospital acquired if it occurred >48 h postadmission (14). During follow-up, the antibiotic treatment, results of blood and urine cultures, and cultures taken from other sites were recorded. We also recorded complications that occurred during hospitalization (the development of septic shock, adult respiratory distress syndrome,
90
disseminated intravascular coagulopathy, acute kidney failure, coma or other neurological complications, septic emboli, allergic reactions to antibiotic drugs, antibiotic-associated diarrhea or pseudomembranous colitis, and the appearance of bed sores), the duration of the febrile disease and hospital stay, and the outcome of hospitalization, whether discharge or demise. We compared diabetic to nondiabetic patients. Because only 17 diabetic patients were treated with insulin injections, they were included in the diabetic patient group and were not analyzed as a separate group. The deletion of insulin-treated patients from the statistical analysis did not change any of the results. For univariate analysis, we used the x2-test (or the Fisher exact test for small numbers) to look for statistical significance of contingency tables. Comparisons were adjusted for quartiles of age or other confounding variables by the Cochran-Mantel-Haenszel statistic, and the adjusted relative risk with 95% confidence interval (95% Cl) was computed with the Cochran-Mantel-Haenszel statistic (15). The reference category was the group of nondiabetic patients (relative risk 1). Most continuous variables were not normally distributed; therefore, we used Wilcoxon's rank-sum test to look for significance of continuous values compared between two classes. Values of continuous variables are given as the median with range in parentheses. A stepwise logistic regression procedure (LOGIST; 16) was used to perform multivariate regression analysis. To test the influence of multiple factors (including diabetes) on survival, we used Cox's life-table regression model (procedure PHGLM; 17).
RESULTS During the study, 124 episodes of bacteremia occurred in 119 diabetic patients, and 508 episodes occurred in 480 nondiabetic patients. The median age of diabetic patients was 74 yr (range 34-98 yr) and of nondiabetic patients was 68 yr (range 18-99 yr, P = 0.0001). Fiftyseven percent of diabetic patients and 50% of nondiabetic patients were men (P = 0.2). Twenty-nine percent (36 of 124) of episodes of bacteremia in diabetic patients were hospital acquired vs. 39% (197 of 508) of episodes in nondiabetic patients (P = 0.04). However, this difference was not significant when adjusted for age (relative risk 0.7, 95% Cl 0.51.2). More than one microorganism was isolated in 3 (2.4%) episodes in diabetic patients and in 31 episodes (6%) in nondiabetic patients (P = 0.1). Underlying disorders in diabetic and nondiabetic patients are detailed in Table 1. With stepwise logistic regression analysis, which had as the dependent variable the presence of diabetes and as independent variables age and the underlying disorders, only three factors were found to be significantly and independently associated with diabetes: age (x2 in the model = 11,
DIABETES CARE, VOL. 14, NO. 2, FEBRUARY 1991
L. LEIBOVICI AND ASSOCIATES
TABLE 1 Underlying disorders in diabetic and nondiabetic patients Nondiabetic patients i(n = 480)
Diabetic patients (n == 119)
Hypertension Congestive heart failure Cerebral vascular accident Decubitus ulcer Corticosteroid treatment Chronic kidney failure Hemodialysis
n
%
n
%
P
Relative risk, age adjusted (95% confidence interval)
35 18 14 10 13 11 5
29 15 12 8 11 9 4
50 34 31 19 90 33 9
10 7 6 4 19 7 2
0.0001 0.004 0.04 0.04 0.5 0.35 0.12
3.1 (1.9-5.0) 1.9(1.1-3.6) 1.7 (0.9-3.2) 2.2 (1.0-4.9) 0.6(0.3-1.2) 1.3 (0.7-2.6) 3.5 (1.2-10.0)
P = 0.001), hypertension (x2 = 22, P = 0.0001), and congestive heart failure (\2 = 5, P = 0.03). Sources of bacteremia are detailed in Table 2. The only difference that reached statistical significance on univariate analysis was the preponderance of UTI as a source of bacteremia among diabetic patients (50 of 124 vs. 154 of 508 episodes in nondiabetic patients, P = 0.03), but this difference was explained by the higher age of the diabetic patients (Cochran-MantelHaenszel statistic controlling for age 1.5, P = 0.22, relative risk = 1.3, 95% Cl 0.8-2.0). The overall percentages of microorganisms isolated from blood cultures were similar in diabetic and nondiabetic patients (Table 3). In episodes in which the source of bacteremia was the urinary tract, the percentage of Klebsiella isolates was 25% (13 of 51) in diabetic patients and 12% (20 of 161) in nondiabetic patients (P = 0.03, relative risk adjusted for age 2.4, 95% Cl 1.1-5.2). In patients with a urinary catheter and bacteremic UTI, Klebsiella was isolated in 60% (6 of 10) of diabetic patients and in 17% (4 of 23) of nondiabetic patients (P = 0.04). The percentage of Klebsiella isolates in patients without a urinary catheter was similar in diabetic and nondiabetic patients, but the preponderance of Klebsiella as a cause of bacteremic UTI in diabetic patients was significant even when corrected for the presence of the catheter (relative risk 2.3, 95% Cl 1.0-5.1). This difference was also significant when corrected for the acquisition of bacteremia in the hospital and for the presence of underlying disorders other than diabetes (data not shown). In patients without an obvious source of bacteremia, the percentage of Staphylococcus isolates was 29% (10 of 35) in diabetic patients and 14% (24 of 76) in nondiabetic patients (P = 0.04, relative risk controlling for age 2.7, 95% Cl 1.1-6.7). The higher percentage of staphylococci in diabetic patients without an obvious source of infection remained significant when controlled for the acquisition of bacteremia in the hospital, the presence of peripheral or central intravenous lines, and underlying disorders (data not shown). The percentage of Staphylococcus isolates in diabetic
DIABETES CARE, VOL. 14, N O . 2, FEBRUARY 1991
patients with an infection of the extremities was 63% (12 of 19) and in nondiabetic patients was 40% (20 of 50, P = 0.1). Septic shock was diagnosed in 14% (18 of 124) of bacteremic episodes in diabetic patients and in 7% (37 of 508) of episodes in nondiabetic patients (P = 0.01, relative risk adjusted for age 1.9, 95% Cl 1.0-3.4). All other complications of bacteremia occurred in similar rates in diabetic and nondiabetic patients. Twenty-eight percent (35 of 124) of episodes of bacteremia in diabetic patients ended in death vs. 29% (146 of 504) in nondiabetic patients (P = 0.87, relative risk adjusted for age 1.1, 95% Cl 0.7-1.7). The only subgroup in which diabetic patients fared worse was patients with intra-abdominal infections: 3 of 7 diabetic patients (43%) died vs. 7 of 43 (16%) nondiabetic patients, but the numbers were too small to reach statistical significance (P = 0.1, relative risk adjusted for age 3.9, 95% Cl 0.7-21.1). With stepwise logistic regression analysis, six factors were found to be associated both significantly and independently with a fatal outcome: septic shock, neutropenia, malignancy, hospital-acquired bacteremia, a source of bacteremia other than UTI, and an elevated blood urea nitrogen level on admission. Identical factors
TABLE 2 Sources of bacteremia in diabetic and nondiabetic patients Diabetic patients (n = 124)
Urinary tract Unknown Lower respiratory tract Extremities Intra-abdominal Other
Nondiabetic patients (n = 508 )
n
%
n
%
50 33 11 17 7 6
40* 27 9 14 6 5
154 157 42 46 43 66
30 31 8 9 8 13
*P = 0.03 vs. patients without diabetes.
91
BACTEREMIA IN DIABETIC PATIENTS
TABLE 3 Percentages of microorganisms isolated from blood cultures of diabetic and nondiabetic patients according to source of infection Lower Urinary tract
Isolates (n) Microorganism (%) Staphylococcus aureus Staphylococcus coagulase negative Pneumococcus Enterococcus Other streptococci Total Grampositive bacteria Escherichia coli Klebsiella Pseudomonas Enterobacter Acinetobacter Proteus Other Gramnegative Total Gramnegative bacteria Bacteroides Fungi
re jpiratory tract
Unknown
Abdominal
Other
Extremities
Total
DP
NDP
DP
NDP
DP
NDP
DP
NDP
DP
NDP
DP
NDP
DP
NDP
51
161
35
176
12
48
8
51
19
50
6
68
131
554
0
1
20*
12
8
12
0
6
53
36
0
18
14
14
0 0 0 2
0 0 3
2 2 2 8
0 33 0 0
0 33 2 6
0 0 12 0
0 2 4 6
10 0 5 0
4 2 2 26
0 0 17 67
1 9
1
9 0 6 6
16
4 3 4 5
1 5 3 8
2
4
40
26
42
54
12
18
68
70
83
48
30
28
47 25* 8 2 0 10
60 12 7
15 12 15 5 6
8 2
27 12 25 0 0 0
8 18 8
4
8 9 9 0 0 8
2 6 13
1 9
9 8 8 3 3 0
4 0
0 10 5 5 5 0
0 2 4 2 8 6
0 17 0 0 0 0
6 6 6 3 10 2
23 15 8 2 1 5
26 11 10 5 5 5
4
3
17
9
17
6
25
10
5
6
0
7
10
6
96
96
46
66
50
42
87
28
65
67
2
0
2
0
0
1
2
0
5
8
2
0
2
0
4 0
4
2
3 11
17 0
40
0
80 0
32
0
0
7
5
3
4
4
14
4
DP, diabetic patients; NDP, nondiabetic patients. *P < 0.05 vs. NDP.
were delineated by Cox's life-table regression analysis, which takes into account the time elapsed till death or discharge of the patient. Diabetes was not found to be associated with fatal outcome on either regression procedure. The median duration of hospital stay till discharge was 11 days (range 3-84 days) in diabetic patients and 13 days (range 1-105 days) in nondiabetic patients. The median duration of hospital stay till demise was 13 days (range 2-96 days) in diabetic patients and 15 days (range 1-106 days) in nondiabetic patients. Neither comparison was statistically significant.
CONCLUSIONS
In this study, we compared episodes of bacteremia in diabetic patients to bacteremia in nondiabetic patients. The median age of the diabetic group was 74 yr, and only 14% were treated with insulin. Thus, the data pre-
92
sented herein probably reflect characteristics of elderly NIDDM patients with bacteremia. Patients were designated as diabetic if they had hyperglycemia and had been diagnosed or treated for such in the past. This definition underestimates the number of NIDDM patients in the population (18). However, this is usually the extent of information available to the physician in the emergency ward, whereas the significance of NIDDM discovered by screening is not clear (19). As in previous studies, UTI was a more frequent source of bacteremia in diabetic than nondiabetic patients (4,5). However, when corrected for the higher age of the diabetic patients, this association was not significant (relative risk 1.3, 95% Cl 0.8-2.0). A second question of interest is whether diabetic patients have bacteremia of unknown origin more often than nondiabetic patients. Mellors et al. (20) and Leibovici et al. (21) found an excess of diabetic patients in bacteremic patients without an obvious source of fever. In this study and the study of MacFarlane et al. (5), the percentage
DIABETES CARE, V O L . 14, N O . 2, FEBRUARY 1991
L LEIBOVICI AND ASSOCIATES
of bacteremia without an obvious source was similar in diabetic and nondiabetic patients. Overall, we were unable to demonstrate any significant differences in the sources of bacteremia when comparing diabetic to nondiabetic patients. The overall percentages of microorganisms causing bacteremia were similar in diabetic and nondiabetic patients. However, Klebsiella bacteremia was twice as common in diabetic patients with bacteremic UTI. The difference is explained by the fact that 60% of bacteremic UTIs in diabetic patients with an indwelling urinary catheter were caused by Klebsiella. This finding was not sought for or demonstrated in previous studies on bacteremia in diabetic patients (4,5) or on Klebsiella bacteremia (13,22,23). Klebsiella is generally less susceptible to antibiotics than Escherichia coli, the most common pathogen causing UTI (24-26). Studies on staphylococcal bacteremia mention diabetes as an underlying disease in 12-14% of patients, but because the denominator data (the total number of diabetic and nondiabetic patients with bacteremia) were missing, we cannot conclude from these studies whether diabetic patients are at increased risk for staphylococcal bacteremia. In our study and in two other studies on bacteremia in diabetic patients, staphylococci were isolated in similar rates in diabetic and nondiabetic patients (4,5). Diabetic patients without an obvious source of bacteremia had twice the rate of staphylococcal isolates than nondiabetic patients (29 vs. 14%, P = 0.04). The percentage of staphylococcal bloodstream isolates in infections of the extremities was 63% in diabetic patients. Indeed, staphylococci are the most common pathogens in diabetic foot infections (28) and in osteomyelitis of the feet in diabetic patients (29). The relative risk of diabetic versus nondiabetic patients to develop septic shock was 1.9 (adjusted for age). Patients with septic shock fare worse than other bacteremic patients, but the absolute numbers are too small for the difference in the risk to develop shock to be expressed as a difference in mortality. We found no other complication to be more common in diabetic patients. The overall mortality of diabetic patients in this study was similar to that in nondiabetic patients, as found by Bryan et al. (4) and MacFarlane et al. (5). Diabetes did not correlate with mortality on multivariate regression analysis, and we did not find any subgroup of patients in which diabetic patients were at a higher risk for a fatal outcome. The only exception was the higher mortality of diabetic patients whose source of bacteremia was an intra-abdominal infection, but the number of patients was too small to reach statistical significance. Nevertheless, other studies support this observation (30-32). In summary, we surveyed prospectively 124 episodes of bloodstream infections in diabetic patients and compared them to 508 episodes in nondiabetic patients. Ours was an elderly population representative of
DIABETES CARE, VOL. 14, N O . 2, FEBRUARY 1991
NIDDM patients. Although the overall sources of bacteremia and the rates of the infecting microorganisms were similar in diabetic and nondiabetic patients, two findings justify a different empiric treatment in diabetic patients suspected of bacteremia: a high percentage of Klebsiella isolates in patients with a urinary catheter and a high percentage of staphylococci isolated from the blood of diabetic patients without an obvious source of bacteremia. Empiric antibiotic treatment for bacteremic infection of the extremities should include an antistaphylococcal drug both in diabetic and nondiabetic patients. Diabetic patients had a higher percentage of septic shock and probably fared worse if the source of bacteremia was an intra-abdominal infection. Otherwise, the rates of complications and mortality were similar in diabetic and nondiabetic patients.
REFERENCES 1. Kreger BE, Craven DE, McCabe WR: Gram-negative bacteremia. IV. Re-evaluation of clinical features and treatment in 612 patients. Am I Med 68:344-54, 1980 2. Weinstein R, Murphy JR, Reller LB, Lichtenstein KA: The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. II. Clinical observations, with special reference to factors influencing prognosis. Rev Infect Dis 5:54-70, 1983 3. Rayner BL, Willcox PA: Community-acquired bacteremia: a prospective survey of 239 cases. Q / Med 69:907-19, 1988 4. Bryan CS, Reynolds KL, Metzger WT: Bacteremia in diabetic patients: comparison of incidence and mortality with nondiabetic patients. Diabetes Care 8:244-49, 1985 5. MacFarlane IA, Brown RM, Smyth RW, Burdon DW, FitzGerald MG: Bacteremia in diabetics. / Infect Dis 12:213-19, 1986 6. Seymour A, Phear D: The causes of death in diabetes mellitus: a study of diabetic mortality in the Royal Adelaide Hospital from 1956 to 1960. Med J Aust 50:89094, 1963 7. Tillotson JR, Lerner AM: Characteristics of pneumonia caused by Escherichia coli. N Engl I Med 277:115-22, 1967 8. Mentzer RM, Golden GT, Chandler JG, Horsley JS III: A comparative appraisal of emphysematous cholecystitis. Am I Surg 129:10-15, 1975 9. Musher DM, McKenzie SO: Infections due to Staphylococcus aureus. Medicine 56:383-409, 1977 10. Cluff LE, Reynolds RC, Page DL, Breckinridge JC: Staphylococcal bacteremia and altered host resistance. Ann Intern Med 69:859-73, 1968 11. Adler JL, Burke JP, Martin DF, Finland M: Proteus infections in a general hospital. II. Some clinical and epidemiological characteristics with an analysis of 71 cases of Proteus bacteremia. Ann Intern Med 75:531-36, 1971 12. Opal SM, Cross A, Palmer M, Almazan R: Group B streptococcal sepsis in adults and infants: contrasts and comparisons. Arch Intern Med 148:641-45, 1988
93
BACTEREMIA IN DIABETIC PATIENTS
13.
14.
15. 16. 17.
18.
19. 20.
21. 22. 23.
94
Haddy Rl, Lee M, Sangal SP, Walbroehl GS, Hambrick CS, Sarti GM: Klebsiella pneumoniae bacteremia in the community hospital. / Earn Pract 28:686-90, 1989 Centers for Disease Control: National Nosocomial Infections Study Site Definitions Manual. Atlanta, GA, Centers for Disease Control, 1972, p. 882-84 SAS Institute Inc: SAS User's Guide: Statistics, Version 5 Edition. Cary, NC, SAS Inst., 1985, p. 420-23 Harrell FE: The LOGIST procedure. In SAS Institute Inc. SUGI Supplemental Library User's Guide, Version 5 Edition. Cary, NC, SAS Inst., 1986, p. 269-93 Harrell FE: The PHGLM Procedure. In SAS Institute Inc. SUGI Supplemental Library User's Guide, Version 5 Edition. Cary, NC, SAS Inst., 1986, p. 437-66 Harris Ml, Hadden WC, Knowler WC, Bennett PH: Prevalence of diabetes and impaired glucose tolerance and plasma glucose levels in U.S. population aged 20-74 yr. Diabetes 36:523-34, 1987 Singer DE, Samet JH, Coley CM, Nathan DM: Screening for diabetes mellitus. Ann Intern Med 109:639-49, 1988 Mellors JW, Horwitz Rl, Harvey MR, Horwitz SM: A simple index to identify occult bacterial infection in adults with acute unexplained fever. Arch Intern Med 147:66671, 1987 Leibovici L, Cohen O, Wysenbeek AJ: Occult bacterial infection in adults with unexplained fever: validation of a diagnostic index. Arch Intern Med 150:1270-72, 1990 Montgomerie JZ, Ota JK: Klebsiella bacteremia. Arch Intern Med 140:525-27, 1980 Garcia de la Torre M, Romera-Vives J, Martinez-Beltran J, Guerrero A, Meseguer M, Bouja E: Klebsiella bacteremia: an analysis of 100 episodes. Rev Infect Dis 7:14350, 1985
24. Atkinson BA, Lorian V: Antimicrobial susceptibility patterns of bacteria in hospitals from 1971 to 1982. ) Clin M/crob/o/230:791-96, 1984 25. European Study Group in Antibiotic Resistance: Susceptibility to beta-lactam antibiotics in septicemia isolates from twenty-nine European laboratories, fur / Glin Microb/'o/6:515-20, 1987 26. Martin MA, Pfaller MA, Rojas PB, Woolson RF, Wenzen RP: In-vitro susceptibility of nosocomial gram-negative bloodstream pathogens to quinolones and other antibiotics: a statistical approach. / Antimicrob Chemother 23:353-61, 1989 27. Mylotte JM, McDermott C, Spooner JA: Prospective study of 114 consecutive episodes of Staphylococcus aureus bacteremia. Rev Infect Dis 9:891-907, 1987 28. Wheat LJ, Allen SD, Henry M, Kernik CB, Siders JA, Kuebler T, Fineberg N, Norton J: Diabetic foot infections: bacteriologic analysis. Arch Intern Med 146:1935-40, 1986 29. Baumberger DM, Daus GP, Gerding DN: Osteomyelitis in the feet of diabetic patients: long-term results, prognostic factors, and the role of antimicrobial and surgical therapy. Am ) Med 83:653-60, 1987 30. DiPalo S, Ferrari G, Castoldi R, Fiacco E, Cristallo M, Staudacher C, Chiesa R, DiCarlo V: Surgical septic complications in diabetic patients. Ada Diabetol Lat 25:4954, 1988 31. Hickman MS, Schwesinger WH, Page CP: Acute cholecystitis in the diabetic: a case-control study of outcome. Arch Surg 123:409-11, 1988 32. Kalfarentzos FE, Dougenis DV, Cristopoulos DC, Spiliotis JD, Williams M, Androulakis J: Prognostic criteria in intraabdominal sepsis. Int Surg 72:185-87, 1987
DIABETES CARE, VOL. 14, N O . 2, FEBRUARY 1991
![[Occult bacteremia or bacteremia in adult patients discharged from the Emergency Department].](/assets/img/hold.png)
