J Clin EpidemiolVol. 45, No. 4, pp. 357-364, 1992 Printed in Great Britain. All rights reserved
RISK FACTORS
0895s4356/92$5.00+ 0.00 Copyright 0 1992Pergamon Press plc
FOR SURGICAL WOUND AMONG THE ELDERLY
INFECTION
L. E. NICOLLE,’ S. A. HUCHCROPT* and P. J. E. CRUS$ ‘Infection Control Unit, Health Sciences Centre, Winnipeg, Manitoba, ‘Division of Epidemiology and Preventive Oncology, Alberta Cancer Board, Calgary, Alberta and 3Department of Surgery, University of Calgary, Alberta, Canada (Received in revised form
I8 November 1991)
Abstract-One hundred fifty-seven elderly patients with surgical wound infection were matched on wound classification and date of surgery to non-infected control patients. Factors examined for their association with wound infection included medical history, functional status, behaviour (e.g. smoking), factors predisposing to infection (e.g. results of CBC and urinalysis) and operative factors such as preparation, duration and type of operation. Conditional logistic regression analysis identified factors already known to be risk factors for wound infection at all ages (e.g. type and duration of operation), as well as factors unique to the elderly (e.g. age > 70 years and limited mobility). Wound infection
The elderly
Surgery
INTRODUCTION
Risk factors
associated with the increased frequency of infection at older ages [8]. Some chronic diseases, such as diabetes mellitus and peripheral vascular disease are directly associated with increased infection. Others, such as Alzheimer’s disease, result in functional impairment (e.g. immobility or incontinence) which leads to the development of lesions such as decubitus ulcers. It is well documented that infections and asymptomatic bacteriuria occur with increased frequency in the more functionally impaired elderly [lo, 111. Surgical wound infection is second to urinary tract infection as the most common hospital acquired infection and is associated with significant morbidity and mortality as well as increased costs of hospitalization [3, 12, 131. A wound infection lengthens a patient’s hospital stay, on average, by 10.2 days with its associated hospital costs and loss of earnings [14]. As with all nosocomial infections, the surgical wound infection rate increases with advancing age [ 15, 161. For example, the clean wound infection rate at the Foothills Hospital, Calgary, Alberta (n = 62,939) has been observed to increase
all developed countries, the population is aging. For example, in Canada, the senior population is growing at twice the rate of the population as a whole, By the year 2021, six million Canadians, or 21% of the population, will be 65 or older. As a consequence, the health problems of seniors are becoming an urgent issue [l]. One of the health problems of the elderly is increased susceptibility to infection [2-51. Aging, particularly after the sixth decade of life, is associated with a decline in many functions of the immune system. It has been postulated that immunologic changes may permit infection, although an association between immunologic changes and infection in elderly individuals has not been documented [6-81. In addition, physiologic changes in aging organ systems such as decreased elasticity of the lungs, and impaired healing of skin wounds, are likely important determinants of infection in the elderly [6,8]. It has been suggested that chronic diseases, which are more prevalent among the elderly, are
In
357
L. E. NICOLLE et al.
358
steadily from 0.6% for persons 1-14 years of age to 3.2% for those over 65 years [15]. In particular, age 70 has been identified as an important cutpoint in terms of risk of surgical wound infection [17]. What is not known is whether the risk of surgical wound infection among the elderly remains high when confounding factors such as comorbidity are considered. In sum, the increased risk of surgical wound infection among the elderly may be explained by: (a) greater frequency among the elderly of the known risk factors for surgical wound infection at all ages, interactions between these known risk fac(b) tors and increasing age. (i.e. the known risk factors for wound infection exert a greater impact on the elderly than on the young), (c) additional risk factors for surgical wound infection which are unique to the elderly and have not been documented for younger populations, and/or (d) an age effect independent of all other factors. The present study was concerned mainly with hypothesis (c)-the possibility of risk factors for wound infection which are largely unique to the elderly.
METHODS
Since 1967, the Department of Surgery, Foothills Hospital has conducted a wound study and maintained a database of information on all operations (with the exception of rectal and vaginal operations, burns and circumcisions) so that the incidence of wound infection is made available on a monthly and yearly basis as the measure of quality assurance in the department. Surgical wounds are considered uninfected if they heal per primum without purulent discharge. A wound is classified as infected if pus discharges, even if organisms are not cultured from the purulent materials. Wounds that are inflamed without discharge and wounds that drain culture-positive serous fluid are considered possibly infected. Such wounds are inspected daily until they discharge pus (infected) or resolve, at which time they are classified as not infected. A surgical nurse observes all wounds in hospital and completes a 28-day follow-up by telephoning the offices of the surgeons.
Wounds are classified as: clean, clean contaminated, contaminated or dirty based on a clinical estimate of contamination made by the circulating nurse in the operating room. The Foothills Hospital criteria are slightly modified from those of the National Research Council. Clean wounds include those in which the gastrointestinal tract or respiratory tract was not entered, no apparent inflammation was encountered, and no break in aseptic technique occurred. Cholecystectomy, incidental appendectomy and hysterectomy are included in this category if no acute inflammation was present. Clean contaminated includes clean operations that entered the gastrointestinal tract or respiratory tract but in which there was no significant spillage. Contaminated includes operations in which acute inflammation (without pus formation) was encountered or in which gross spillage from a hollow viscus occurred. Fresh traumatic wounds and operations in which a major break in aseptic technique occurred are included in this category. Cholecystectomy, incidental appendectomy and hysterectomy are also included in this category if inflammation was present. Dirty includes operations in which pus was encountered or in which a perforated viscus was found. Old traumatic wounds are also included in this group. Operations are coded on the Wound Study Data Base using an adaptation of the International Classification of Diseases [18]. Patients aged 65 and over with wound infection were identified from the Wound Study Data Base. Since patients aged 65 and over represented 5-6% of surgical patients at Foothills Hospital during the time period of observation, and since wound infections occurred in fewer than 5% of surgical procedures, the cases for this study were selected from approximately 80,000 procedures performed over a 6-year period from 1980 to 1985. For each case a non-infected control was chosen whose operation most closely followed the case’s and whose wound was within the same classification (i.e. clean, clean-contaminated, etc.). Cases and controls were matched on wound classification because it is known to be the single strongest predictor of wound infection, with dirty wounds having 20 times the rate of infection as clean wounds [16]. Matching on date of surgery was performed for both pragmatic and scientific reasons-the former to facilitate economic processing of such a large data file, and the latter to control for any trend in
Surgical
Wound
Infection
wound infection rates during the study period. A declining incidence of wound infection had been noted between 1968 and 1977 [16]. A sufficient pool of potential controls permitted matching of cases with controls whose operation was no more than 3 days after the cases’. Excluded from the study were subjects with primary bacteraemia on admission, those who lived less than 28 days after the date of operation, those whose wound closure was delayed, and cases for which no matched control could be found (Table 1). When a potential control was excluded, a replacement, whose date of surgery was the next closest, was selected from the remaining potential controls. A sample size of at least 136 per group would ensure 80% power in detecting an odds ratio as small as 2.0 for variables whose frequency of occurrence in the control group was 0.50 or less (a = 0.05). Variables retained for analysis from the Foothills Hospital wound study database were operation type, wound classification, break in aseptic technique, types of skin preparation and drape, bowel preparation, suture material, and endpoint (e.g. died or discharged alive). Break in aseptic surgical technique was the unexpected exogenous contamination of the operative field by glove perforation or tears, contact of gloves or gowns of the operating team with any nonsterile object (e.g. the O.R. lamp), wet drapes which would allow skin bacteria access to the wound by means of capillary action, or nonsterile foreign bodies falling into the wound. Any member of the operating team could declare a break in technique. Variables abstracted from hospital charts included date of birth, domicile prior to hospital admission (community, seniors’ apartment, nursing home), associated diseases (e.g. diabetes, peripheral vascular disease, renal disease, liver disease, neurologic disease), medications (e.g. antacids, H2 blockers), functional status (including mental status, mobility, incontinence of bladder or bowel), previous surgery within 10 years, smoking history, associated infections or factors predisposing to infection (e.g. decubiti), results of CBC and urinalysis, as well as any prior cultures, and presence of invasive devices (e.g. intravascular devices, urinary catheters, tracheal intubation). For patients admitted more than 7 days before surgery, only those lab results or conditions present within 7 days prior to the operation were considered. Diseases were coded according to the International Classification of Diseases [19]. A data collection
Among
the Elderly
359
form was developed and pre-tested and patient charts were reviewed by two trained technicians. Queries were directed to the principal investigators and a coding manual ensured consistency throughout the data collection phase. Each variable was assessed for its association with wound infection using the x2 test for categorical variables and the t-test for continuous variables [20,21]. Meaningful cut points on continuous variables were determined. Variables found to be associated with wound infections at p < 0.10 unadjusted for multiple comparisons were retained for inclusion as independent variables in a conditional stepwise logistic regression analysis. To create one variable describing operation type, types of operation were grouped into high, average and low risk based on the results of the bivariate analyses and assigned codes of 2, 1 and 0 respectively. Patients who underwent two operations were assigned the higher of the two scores. Operative diagnoses were handled similarly. A composite medical history variable was created by summing the number of chronic diseases. The potential for age as a modifier of risk was assessed by the use of interaction terms in the regression model. The potential utility of the logistic regression model in terms of predicting wound infection among the elderly was assessed. A wound infection risk score was computed for each subject by weighting the presence of each factor in the model by its coefficient, and summing the result. The predictive ability of the model was then assessed using a true-positive vs false-positive plot (ROC curve) [22]. Finally, the association between age (a70 years) and other factors in the model was assessed using a x2 analysis. It should be pointed out that, although available from the database, the effect of antibiotics could not be assessed in this study because of the matched design. Since prophylactic or therapeutic antibiotics are used for all non-clean procedures, only the clean wound category contained subjects who did not receive prophylactic antibiotics. For clean wounds the infection rate is less than 2% resulting in the need for an inordinately large sample size to demonstrate a one percent reduction in the infection rate. RESULTS
Table 1 describes the 86 subjects ineligible for entry into the study. One hundred fifty-seven case-control pairs were available for analysis.
L. E. NICOLLE~~al.
360
Table 1. Exclusions Reason
Cases Controls
No match found Primary bacteremia, on admission Died within 28 days after operation Delayed wound closure
23 3 3 30
n/a 4 17 7
Total 23 7 20 37 87*
*One control had primary bacteremia on admission and also died. Therefore the total number of subjects excluded = 86.
Subjects ranged in age from 65 to 99 years with a median of 73 years. Forty-four percent of operations were clean, 7% clean-contaminated, 43% contaminated and 6% dirty. Fifty-five subjects had two operations resulting in a total of 369 procedures. Bearing in mind that nonclean procedures were over-represented because of their increased likelihood of wound infection, the four most frequent operations were on the hip (n = 55), colon (n = 29), gallbladder (n = 28) and urogenital system (n = 26).
Variables identified as possible predictors of wound infection in the elderly and retained for further analysis are presented in Table 2. Other variables examined, but found not to be associated with wound infection, were: number of prior operations within the last 10 years, institutional care, bowel incontinence, use of tobacco (current and ever) alcohol and medications, decubitus ulcer, indwelling catheter, central line, positive urinalysis at time of diagnosis, fatality of the operative diagnosis (non-
Table 2. Variables associated with wound infection (p -C0.10) No. of
No. of Cases controls
xf
Unadjusted p-value
Increased risk Medical history
Neurologic disease >7 Prior admissions within the last 10 years History of >4 medical conditions
55
35
5.95
0.01
10
3 56
3.05 5.22
0.08 0.02
1 11 18 9 21 14 31 70
11.22 6.04 6.65 4.59 4.91 4.12 3.17 4.00 3.26
0.03 0.03 0.04 0.07 0.05 0.07
148 100
13.5 81
5.15 4.23
0.02 0.04
33 15 19 35
15 4 7 20
7.11 5.60 5.07 4.32
0.008 0.02 0.02 0.04
29 18
12 8
7.18 3.40
0.007 0.07
27
56
12.84
0.0003
4 7 2 4
25 21 10 13
15.20 6.63 4.25 3.98
77
A I admission
Limited mobility Age >70 Confused or demented Diagnosis of injury Distant infection WBC > 12,000 Urine incontinence Obesity Male Pre-operative No bowel prep
Admission >2 days before operation
46 120 27 33 22 35 25 43 87
99
0.0008
0.01 0.01
Operative
Duration of operation >2 hours Amputation Uro-genital operation Hip operation Serous cavity or subcutaneous wound drain Catgut sutures Decreased risk At admission
Diagnosis of digestive system disorder Operative
Colon operation Gallbladder operation Heart operation Hernia operation
Surgical Wound Infection Among the Elderly
361
Table 3. Distribution of cases and controls on the composite diagnosis and oneration variables Cases Controls
x2
p
Operative diagnosis
“Low” risk (digestive system disorder) “Average” risk “High’ risk (injury)
18
48
106 33 157
91 18 157
8
53
80 69
53 31
157
157
19.19
0.0001
Operation type
“Low” risk (colon, gall bladder, heart or hernia operation) “Average” risk “High’ risk (amputation or operation on hip or uro-genital tract)
fatal, ultimately fatal, rapidly fatal), operative diagnoses of cancer or circulatory disease, preoperative skin prep, type of skin drape, break in aseptic technique, antibiotics and history of: hypertension, peripheral vascular disease, heart disease, liver disease, pulmonary disease, rheumatologic disease, haematologic disease, malignancy, diabetes, gastric disease, renal disease or other diseases. The distribution of cases and controls on the composite operation and diagnoses variables is presented in Table 3. Nineteen independent variables were entered into the logistic regression model. Regression coefficients for variables with few positive observations (e.g. fewer than 10% of the total sample size) could be unstable. Only one variable“more than 7 prior admissions within the last 10 years” was of concern in this regard. Results of the stepwise conditional logistic regression analysis are presented in Table 4. After type and duration of operation, the most important predictor of surgical wound infection appeared to be age ( 270 years), followed by male sex, limited mobility and elevated WBC. Each of these factors was present in at least 15% of observations. No interaction terms entered
47.96
the model, suggesting that age did not modify the effects of other risk factors. The stability of the coefficients was tested by repeating the analysis with subsets of the 5 statistically significant variables. As the number of distinct covariate patterns was reduced from 80 to 3, the maximum percent change in the coefficients was 8 for operation type, 14 for duration of operation, 17 for age and 44 for sex and mobility. Only for sex did the percent change exceed the standard error. Also, in 4 of the 7 models, which included sex, it was of borderline statistical significance (0.05-O. 11). The ROC plot is presented in Fig. 1. It illustrates that over and above operation type and duration, neither age nor the remaining factors in the full model contributed substantial additional precision to the ability to predict wound infection. For example, at a sensitivity of 80% (true positive fraction) the false positive fraction ranged from approximately 50% when only operation type and duration were used, to approximately 42% when age was added. This approximately 8% reduction in the false positive rate could be considered the degree of improvement in precision afforded by the
Table 4. Results of stepwise conditional logistic regression analysis Variable 1. Operation type (“high” vs “low” risk) 2. Duration of operation (>2 hr) 3. Age ( > 70 yr) 4. Male sex 5. Limited mobility 6. WBC >12,000
Coefficient
SE
Adjusted odds ratio
1.4482
0.2810
4.3
(2.5,7.4)
40.04
1.1259
0.4647
3.1
(1.2,7.7)
7.45
0.006
0.8292
0.3469
2.3
(1.1,4.5)
6.74
0.009
0.7696 0.7918 0.7945
0.3222 0.3808 0.4565
2.2 2.2 2.2
(1.1,4.7) (1.1,4.7) (0.9,5.3)
5.97 4.74 3.24
0.015 0.029 0.072
95% CI Improvement x2
p
362
L. E. NICOLLEer al.
90 %
T r U e
8070605040-
-
3020-
-
Full
Model
Operation
.---.-. Oper’n
0
10
I
I
20
30
+ Duration
+ Dur’n
+ Age
I
I
I
I
I
I
40
50
60
70
80
90
100
% False Positive Fig. 1. The addition of age and other factors to type and duration of operation contribute modestly to the ability to predict wound infection using a risk score derived from the logistic regression analysis.
addition of age into the model after operation type and duration. Finally, age 270 years was observed to be associated with high risk operations. Thirtyseven percent of subjects aged 270 years had a high risk operation compared to 2 1% of those aged ~70 years (x2 = 7.52, df = 2, p = 0.02). However, age was not associated with prolonged duration of operation. DISCUSSION
Two groups of factors determine the likelihood of a wound infection-the dose of bacterial contamination and the host’s resistance to pathogens. Wound classification is based on the likelihood of endogenous contamination by pathogens at operation. The incidence of wound infection is therefore related directly to the wound classification, ranging from a wound infection rate of less than 2% for clean wounds to approximately 40% for dirty wounds [16]. Fortunately, the majority of surgical wounds (approximately 72%) fall in the clean category, whereas dirty wounds comprise 3% of the total. Other risk factors for the development of surgical wound infections have been described, such as the type of pre-operative preparation, the duration of pre-operative hospitalization, infection at a remote body site, the duration of the operation, use of steroids, and the presence
of diabetes, severe malnutrition, foreign bodies in the wound, vascular compromise and advanced age [13, 15, 16,23-251. Multivariate analyses have been used to predict the likelihood of surgical wound infection [26,27]. The results of the present study of wound infection among the elderly are consistent with those of Haley et al.[26] with respect to risk factors at all ages. These investigators also found type of operation, duration of operation and wound classification to be the three major predictors of surgical wound infection. A fourth factor identified by Haley et al. as a risk factor for all ages was three or more underlying diagnoses. In our study, neither a history of individual diseases nor the composite history measure predicted wound infection. However, our observation that limited mobility was associated with wound infection might be considered a proxy for comorbidity. One of the weaknesses of restrospective studies such as ours is the inability to precisely measure some variables such as the presence and severity of comorbidity. The design of the present study precluded the incorporation of the few scales of severity of illness that exist such as the APACHE score [28] or the New York Heart Association classification for illness [29]. These measures were not in use at Foothills Hospital during the time period of our observations and cannot be used validly and reliably on data
Surgical Wound Infection Among the Elderly
abstracted from chart review. Consequently, our measure of comorbidity may not have been sensitive enough to permit detection of any association between comorbidity and surgical wound infection, Age appeared to be a risk factor for surgical wound infection independent of the measures of comorbidity available to us. Our results indicating a 2.3 fold risk of surgical wound infection among subjects aged z 70 years relative to those aged 65-69 are also consistent with those of Sandfast et al. who found rates of nosocomial infection among those aged over 70 to be twice those among subjects aged 40-70. However, our study provided no evidence that age modifies the effects of other risk factors (hypothesis b). The design of our study precluded a measure of immune response which has been postulated as declining with age and which might have explained the association of surgical wound infection with age. WBC > 12,000 has not been documented by others as a risk factor for surgical wound infection. However, WBC > 12,000 may be an indicator of concomitant infection which has been suggested as a risk factor for surgical wound infection [ 131.Our observation of an association between male sex and surgical wound infection may be unique to a study of the elderly in that it may reflect a greater age-specific susceptibility to disease and the foreshortened life expectancy of males relative to females. However, sex appeared to be the least stable of all the predictors and hence is of questionable utility as an indicator of risk of surgical wound infection among the elderly. The observation that colon surgery was associated with reduced risk of surgical wound infection appears paradoxical in light of the fact that colon surgery appears to carry a higher likelihood of endogenous contamination than other procedures. However, the results of this study are independent of wound classification and therefore, colon surgery carries a lower risk of surgical wound infection relative to other types of surgery within the same classiJication. This may be because much colon surgery is elective and permits precaution against wound infection, whereas the same category of wounds includes emergency operations for which there is often insufficient time for preoperative prophylaxis. Our study suggests that most of the risk of surgical wound infection among the elderly is explained by operative risk factors common
363
to all ages but which may occur with greater frequency among the elderly (hypothesis a). Nevertheless, the elderly appear to be at modestly increased risk of surgical wound infection because of additional factors independent of comorbidity (hypotheses c and d). However, because of the instability of the coefficients for sex and limited mobility, these latter factors are speculative. Unfortunately we were unable to identify risk factors among the aged which could be altered to reduce the risk of surgical wound infection. At best we can suggest that interventions which might improve host resistance to pathogens be considered. Also, an area for further investigation may be to investigate the relationship between immune status among the elderly and the likelihood of developing a surgical wound infection. Acknowledgements-The authors thank Grace Williams and Shirley Pedersen who reviewed the charts, Randy Lentjes who programmed the data entry routine, Jennine Chomack and Susan Price who prepared the manuscript and Dr. Penny Brasher who offered statistical advice. This project was funded by Foothills Hospital Research and Development Fund, Calgary and by the Centre on Aging, Winnipeg. This work was presented, in part at the 30th Inferscience Congress
on Antimicrobial
Agents
and
Chemotherapy,
Atlanta, 21-24 October 1990. Finally, without the existence of the Wound Study Data Base, our study would have required 6 years’ worth of prospectively gathered data. We are therefore grateful for the research opportunity afforded by the Wound Study Data Base at Foothills Hospital.
REFERENCES 1. Chambers LW. Research on health problems of seniors its time has come. Chron Dis Canada 1988; 9(5): 78-79.
Haley RW, Hooton IN, Culver DH ef a/. Nosocomial infections in U.S. hospitals, 1975-1976. Estimated frequency by selected characteristics of patients. Am J Med 1981; 70: 9477959. 3. Allen JR, Hightower AW, Martin AM, Dixon RE. Secular trends in nosocomial infections: 197&1979. Am J Med 1981; 70: 389-392. 4. Gross PA, Rapuano C, Andrignolo A, Shaw B. Nosocomial infections: Decade-specific risk. Infect Control 1983; 4: 145-147. 5. Saviteer SM, Samsa GP, Rutala WA. Nosocomial infection in the elderly: Decade-specific rates per hospital day. Twenty-fourth Interscience Conference on 2.
Antimicrobial
Agents and Chemotherapy;
1984.
6.
Gardner ID. The effect of aging - - on susceptibility to infection. Rev Infect Dis 1980; 2: 801-810._ 7. Schwab R. Walters CA. Weksler ME. Host defense mechanisms and aging: Semin Oncol 1989; 16(l): 20-27. 8. Finkelstein MS. Unusual features of infections in the aging. Geriatrics 1982; 37(4): 65-78. 9. Nicolle LE, McIntyre M, Zacharias H, MaxDonell JA. Twelve-month surveillance of infections in institutionalized elderly men. J Am Ceriab Sot 1984; 2: 513-519.
364
L. E. NICOLLEet al.
10. Boscia JA, Kobasa WD, Knight RA, Abrutyn E, Kevison ME, Kaye D. Epidemiology of bacteriuria in an elderly ambulatory population. Am J Med 1986; 80: 208-214. 11. Nicolle LE, Henderson E, Bjornson J, MacIntyre M, Harding G, MacDonell JA. Bacteriuria and survival in elderly institutionalized men: A five-year follow-up. bast No. 1100. Twenth-fifth Interscience Conference on Antimicrobial Agents and Chemotherapy; 1985. 12. Haley RW, Schaberg DR, Crossley KB, von Allmen SD, McGowan JE Jr. Extra charges and prolongation of stay attributable to nosocomial infection: A prospective inter-hospital comparison. Am J Med 1981; 70: 51-58. 13. Simmons BP. Guideline for prevention of surgical wound infections. Am J Infect Control 1983; ll(4): 133-143. 14. Cruse PJE. Wound study reduces infection, enhances care. Quest (Foothills Hospital Newsletter) 1986 July; l-2. 15. Cruse P. Wound infection surveillance. Rev Infect Dis 1981; 3: 734-737. 16. Cruse PJE, Foord R. The epidemiology of wound infections. A IO-year prospective study of 62,939 wounds. Surg Clin North Am 1980; 60: 2740. 17. Sandfast SJ, Michelsen PB, Baltch AL, Smith RP, Latham EK, Spellacy AB, Venezia RA, Andritz MH. A prevalence survey of infections in a combined acute and long-term care hospital, Infect Control 1984; 5: 177-184. 18. National center for Health Statistics Public Health Services Publication # 719: International Classification of Diseases Adapted for Indexing Hospital Records by Diseases and Operations. Washington, DC: U.S. Government Printing Office; March 1965.
19. Committee on Professional Hospital Activities. Inter?atlooal ? Classification of Diseases, Version 9, clinical Modification (ICD9CM). Ann Arbor, MI: Annual Updates; 1980-1985. 20. Rosner B, Hennekens CH. Analytic methods in matched pair epidemiological studies. Int J Epidemiol 1978; 7: 367-372. 21. Feinstein AR. Quantitative ambiguities in matched versus unmatched analyses of the 2 x 2 table for a case-control study. Int J Epidemiol 1987; 16: 128-134. 22. Metz C. Basic principles of ROC analysis. Semin Nucl Med 1978; 8: 283-298. 23. Ehrenkranz NJ. Surgical wound infection in clean operations. Risk stratification for interhospital comparisons. Am J Med 1981; 70: 909-914. 24. Coppa GF, Eng K. Factors involved in antibiotic selection on elective colon and rectal surgery. Surgery 1988; 104: 8533858. 25. Galandiuk S, Polk HC Jr, Jagelman DG, Fazio VW. Re-emphasis of priorities in surgical antibiotic prophylaxis. Surg Gynecol O&et 1989; 169: 219-222. 26. Haley RW, Culver DH, Morgan WM, White JW, Emori TG, Hooton TN. Identifying patients at high risk of surgical wound infection. A simple multivariate index of patient susceptibility and wound contamination. Am J Epidemiol 1985; 121: 206215. 27. Simchen E, Shapiro M, Sacks TG, Michel J, Durst A, Eyal Z. Determinants of wound infection after colon surgery. Ann Surg 1984; 199: 26&265. 28. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: A severity of disease classification system. Crlt Care Med 1985; 13: 818-829. 29. Goldman L, Hashimoto B, Cook F, Loscalzo A. Comparative reproducibility and validity of systems where assessing cardiovascular function class: Advantages of a new specific activity scale. Circulation 1981; 64: 1227.