A Prospective Study of Outcomes, Healthcare Resource Utilization, and Costs Associated With Postoperative Nosocomial Infections
Objective. We evaluated 4 important outcomes associated with postoperative nosocomial infection: costs, mortality, excess length of stay, and utilization of healthcare resources.
Design. The outcomes for patients who underwent general, cardiothoracic, and neurosurgical operations were recorded during a previous clinical trial. Multivariable analyses including significant covariates were conducted to determine whether nosocomial infection significantly affected the outcomes.
Setting. A large tertiary care medical center and an affiliated Veterans Affairs Medical Center.
Patients. A total of 3,864 surgical patients.
Results. The overall nosocomial infection rate was 11.3%. Important covariates included age, Karnofsky score, McCabe and Jackson classification of the severity of underlying disease, National Nosocomial Infection Surveillance system risk index, and number of comorbidities. After accounting for covariates, nosocomial infection was associated with increased postoperative length of stay, increased costs, increased hospital readmission rate, and increased use of antimicrobial agents in the outpatient setting. Nosocomial infection was not associated independently with a significantly increased risk of death in this surgical population.
Conclusion. Postoperative nosocomial infection was associated with increased costs of care and with increased utilization of medical resources. To accurately assess the effects of nosocomial infections, one must take into account important covariates. Surgeons seeking to decrease the cost of care and resource utilization must identify ways to decrease the rate of postoperative nosocomial infection.
Received June 9, 2005; accepted June 14, 2006; electronically published November 17, 2006.
Every year, approximately 2.1 million patients in the United States acquire infections during medical care. Of patients who undergo surgery, 9%‐30% acquire nosocomial infections.1 Overall, nosocomial infections increase mortality and morbidity above that expected on that basis of the patients’ underlying illnesses and increase the cost of care by 5‐10 billion dollars.2
Employers, third‐party payers, the government, and managed care organizations now scrutinize the cost of medical care and deny payment for specific tests, treatments, hospital days, and complications of care. Concurrently, changes in treatment protocols, patient populations, nurse‐to‐patient ratios, and healthcare delivery may be increasing the risk of complications such as nosocomial infections.3,4 As a result, costs and utilization of healthcare resources may increase despite efforts and demands to decrease both.
Several investigators have demonstrated that nosocomial infections are more expensive than other serious surgical complications.5,6 Taylor et al.5 found that sternal wound infections were the most expensive complication of coronary artery bypass graft procedures, costing a mean of $41,559. The next most expensive complication, respiratory failure, added a mean of $28,756 to the cost of care.5 Weintraub et al.6 documented that surgical site infections after coronary artery bypass graft procedures increased the length of hospitalization from a mean (±SD) of 
days to 32.3 ± 25.8 days. Thus, surgeons who want to provide high‐quality, cost‐effective care must know the costs associated with postoperative nosocomial infections and must find ways to prevent these infections. This prospective study was designed to assess important outcomes, including costs, mortality, and resource utilization, associated with nosocomial infections after operative procedures.
Methods
Study Population and Data Collection
From April 1995 through January 1998, a total of 3,864 adult patients undergoing general, cardiothoracic, or neurological operative procedures at the University of Iowa Hospitals and Clinics (UIHC) or the Veterans Affairs Medical Center (VAMC) in Iowa City, Iowa, were enrolled in a clinical trial of intranasal mupirocin therapy to prevent Staphylococcus aureus infection.7 Research assistants collected pertinent clinical data as they enrolled patients. Infection control professionals and research assistants used standard definitions to identify nosocomial infections.7 Data from the previous clinical trial were used in the current study to assess outcomes associated with nosocomial infection. Patients were observed for a mean of 30 days (range, 25‐35 days) after their operation to determine whether they acquired infection.
The institutional review boards at the University of Iowa College of Medicine and the VAMC approved the involvement of human subjects in this study.
Data Analysis
Direct inpatient costs were calculated as described elsewhere.8 Costs and length of stay data were expressed as median values; the frequency of antibiotic use among outpatients and the rate of hospital readmission were expressed as percentages. Statistical analyses were conducted using SAS, version 9.1 (SAS Institute).
The statistical analyses were done to determine whether there was an association between nosocomial infection and the postoperative outcomes of death, outpatient antibiotic use, hospital readmission, postoperative length of stay, and 30‐day postoperative total hospital costs after adjusting for the effect of covariates. Nosocomial infection was categorized as either surgical site infection (SSI), other nosocomial infection, or no nosocomial infection. The covariates used were type of surgery (general, cardiothoracic, or neurosurgery), age, body mass index, Karnofsky score, McCabe and Jackson classification of underlying disease as ultimately fatal, number of comorbid conditions, diabetes, obesity, preoperative infection, preoperative length of stay, National Nosocomial Infection Surveillance system (NNIS) risk index, and pulmonary, cardiovascular, or rheumatologic disease. Logistic regression was used in the analysis of death, postoperative outpatient antibiotic use, and hospital readmission. Multiple linear regression using the natural log transformation of 
(to normalize the cost distribution) was done to analyze 30‐day postoperative total hospital costs. Fitted Cox proportional hazard regression analysis was used to evaluate postoperative length of stay. Data from patients who died before discharge were censored at the time of death.
For each outcome variable, the effect of nosocomial infections was tested with all covariates in the regression model. A second model was fitted, with nonsignificant covariates (
) removed from the regression model by backward elimination. This model was expanded to evaluate effects involving 2‐factor interaction between nosocomial infection and each significant covariate. The backward elimination procedure was also used when selecting which 2‐factor interactions to include in the final model.
For logistic regression analysis, significant covariates modeled as continuous variables were examined to determine whether a linear relationship existed between the continuous variable and the logit. If an increasing or decreasing linear trend in the estimated logistic regression coefficients was not observed, the continuous variable was categorized into quartiles. Adjacent categories having coefficient estimates of similar magnitude and greatly overlapping confidence intervals (CIs) were combined; the lowest or highest quartile served as the reference category.
Results
Of the 3,864 patients enrolled in the mupirocin trial, 2,408 (62.3%) underwent general surgical procedures, 732 (18.9%) underwent neurosurgical procedures, and 724 (18.7%) underwent cardiothoracic surgical procedures. A total of 438 patients (11.3%) acquired at least 1 nosocomial infection, of which 316 were surgical site infections. One hundred six patients had 122 infections at sites other than the surgical wound, including 54 urinary tract infections, 43 respiratory tract infections, and 18 bloodstream infections. Staphylococci, the most common etiologic agents, caused 40% of nosocomial infections. One hundred patients (2.6%) were infected with S. aureus, and 76 (2.0%) were infected with coagulase‐negative staphylococci.
Patients who had nosocomial infections were significantly older and had larger body mass indexes, longer preoperative lengths of stay, higher NNIS risk indexes, higher Karnofsky scores, worse McCabe and Jackson scores (ie, ultimately fatal underlying diseases), and more comorbid conditions, including diabetes, obesity, preoperative infection, and pulmonary, cardiovascular, and rheumatologic diseases. These covariates were included in the regression models for each outcome variable.
Eleven (2.5%) of 438 patients with nosocomial infections and 45 (1.3%) of 3,425 patients without nosocomial infections died (
). When all covariates were included in the model, there was no significant association between nosocomial infection and death (odds ratio [OR], 1.14 [95% confidence interval {CI}, 0.54‐2.38]; 
). When only statistically significant covariates were included in the model, the NNIS risk index, the Karnofsky score, preoperative infection, and age were independently associated with death (Table 1).
With all covariates in the model, nosocomial infection was significantly associated with postoperative length of stay (P < .001). After adjusting for the significant covariates (Karnofsky score, preoperative length of stay, and age), SSI and other nosocomial infections were significantly associated with postoperative length of stay. The NNIS risk index score and preoperative infections were also significant covariates and had significant interactions with surgical site infection. Likewise, McCabe and Jackson classification of underlying disease as ultimately fatal, body mass index, and preoperative infection had significant interactions with other nosocomial infections. Table 2 provides examples of how combinations of covariates with a significant interaction with nosocomial infections (including the highest and lowest levels) affect the excess length of stay due to SSI or other nosocomial infection. These examples are estimated from the fitted Cox proportional hazard regression model; they demonstrate that SSI and other nosocomial infections significantly increase length of stay at both the lowest and the highest values of the covariates and that the magnitude of the increase is dependent on the values of the covariates.
The analysis of total postoperative costs at day 30 after the procedure included only patients whose operations and postoperative stays were at the UIHC. The median crude total hospital costs for patients without infection was $3,343 (range, $0‐$85,712), compared with $6,364 (range, $0‐$79,148) for patients with surgical site infection only, $9,562 (range, $1,570‐$40,845) for patients with urinary tract infection only, and $26,863 (range, $10,989‐$103,375) for patients with bloodstream infection only.
When all covariates were included in the model, nosocomial infection was significantly associated with total hospital cost (
). The significant covariates were the Karnofsky score, NNIS risk index, number of comorbid illnesses, obesity, preoperative length of stay, and age. The covariates with significant interactions with nosocomial infection were type of surgery and the McCabe and Jackson classification. When only the significant covariates and the covariates with significant interactions were included in the model, SSI and other nosocomial infections were independently associated with postoperative hospital costs at 30 days for all patients who underwent general surgical procedures, regardless of the McCabe and Jackson classification (Table 3). In contrast, only other nosocomial infections increased costs for patients who had cardiothoracic surgical procedures and who had a McCabe and Jackson classification of underlying disease as ultimately fatal. Other nosocomial infections increased costs for patients who had neurosurgical procedures, regardless of the McCabe and Jackson classification.
Only patients who survived to be discharged were evaluated to identify risk factors for readmission to the hospital. One hundred six (24.2%) of 438 patients with nosocomial infections were readmitted, compared with 269 (8.0%) of 3,370 patients (data for 56 patients were missing) who did not have nosocomial infections (
). When all covariates were included in the model, nosocomial infections were significantly associated with hospital readmission (SSI vs no infection: OR, 3.61 [95% CI, 2.68‐4.87]; other nosocomial infection vs no infection: OR, 1.91 [95% CI, 1.14‐3.22]; 
for both comparisons). The significant covariates were McCabe and Jackson classification of underlying disease as ultimately fatal, NNIS risk index, number of comorbid illnesses, rheumatologic disease, and preoperative length of stay, and the covariate with a significant interaction with nosocomial infection was surgical service. When only these variables were included in the model, SSI was associated with hospital readmission for all 3 surgical services (Table 4). The odds of hospital readmission were also increased for patients who acquired other nosocomial infections after general surgical procedures (
) and cardiothoracic surgical procedures (
).
Outpatient antimicrobial use within 30 days after discharge was evaluated for patients who survived hospitalization. With all covariates in the model, nosocomial infection was significantly associated with postoperative outpatient antimicrobial use (SSI vs no nosocomial infection: OR, 8.32 [95% CI, 6.33‐10.93]; other nosocomial infection vs no nosocomial infection: OR, 2.56 [95% CI, 1.72‐3.79]; 
for both comparisons). The significant covariates were McCabe and Jackson classification of underlying disease as ultimately fatal, Karnofsky score, cardiovascular disease, surgical service, and age. The covariates that had significant interactions with nosocomial infection were number of comorbid illnesses and preoperative infection. When only the significant covariates and the covariates with significant interactions were included in the model, SSI was significantly associated with outpatient antimicrobial use, regardless of the number of comorbid conditions or the presence of preoperative infection (Table 5). For patients without preoperative infection, other nosocomial infections were also associated with an increased odds of outpatient antibiotic use. For patients with preoperative infection, other nosocomial infections were associated with an increased odds of outpatient antimicrobial use, but the difference did not reach statistical significance (Table 5).
Discussion
Green and Wenzel9 were the first investigators to tightly match groups of patients with and patients without surgical site infection to quantify the associated increased cost and length of stay. They documented that surgical site infection doubled the expected length of stay, even after patients were matched for variables that might increase cost (ie, age, underlying disease, surgeons, and pathological findings).9
Twenty years later, we evaluated data from a large clinical trial and demonstrated that nosocomial infection in postoperative patients was associated with significantly increased costs, hospital readmission rates, length of hospital stay, and frequency of outpatient antimicrobial use. In contrast to some studies, we did not find an association between postoperative nosocomial infection and death.10‐12 Similar to our study, the study by Kollef et al.13 found an association between nosocomial infection and death in their univariate analysis but not in their multivariable analysis.
Our study documented that although nosocomial infections significantly increased cost, length of stay, readmission rate, and use of antimicrobial agents in the outpatient setting, other factors, such as age, number of comorbid illnesses, specific underlying diseases, McCabe and Jackson classification, NNIS risk index, functional status of the patient (as measured by the Karnofsky score), and preoperative length of stay significantly influenced the size of the effect caused by infections. Moreover, we documented that surgical site infection and other nosocomial infections have different effects on the length of stay, the risk of readmission, and the frequency of outpatient antimicrobial use. Most of the other nosocomial infections would be identified while patients are hospitalized, and thus these infections are likely to increase the length of stay but are less likely to increase the incidence of readmission or outpatient antimicrobial use. Conversely, many surgical site infections are identified after patients are discharged. Thus, surgical site infection is likely to increase the incidence of readmission and use of antimicrobial agents by outpatients.
Previous studies of surgical populations often oversimplified the effect of infection on outcomes such as cost, length of stay, and mortality by not taking covariates into account or by addressing them through inadequate matching. In addition, most reports in the literature give a single “attributable cost” or a single “attributable length of stay” resulting from nosocomial infections. Our study demonstrates that the effect of a nosocomial infection is not the same for every patient and that outcomes vary significantly with patient characteristics.
Numerous investigators have evaluated outcomes of either surgical site infection alone or all nosocomial infections after surgical procedures.10‐21 However, the study populations, methods, and analyses varied substantially. Some studies included data on patients who were readmitted, whereas other studies did not. Most studies included only hospital‐based data, but ours and that by Reilly et al.14 address some outcomes in the community. Some studies evaluated all patients in a cohort, whereas others considered only matched pairs; some used only univariate analyses, whereas others used multivariable analysis. Some studies calculated outcomes by comparing patients who where infected with those who were not, whereas other studies reviewed the patients’ medical records and bills and assessed whether costs and days were attributable to infection. Given these differences, it is difficult to compare the results of these studies. However, the data suggest that some nosocomial infections increase mortality rates and that most increase costs, length of postoperative hospital stay, and readmission rate. The extent to which infections affected these outcomes remains elusive because investigators used different methodologies.
Recently, several groups have done studies to determine the best methodology for assessing postoperative outcomes related to infection. Erbaydar et al.17 found that the increased length of stay among surgical patients with nosocomial infections in a matched study was 6.4 days shorter than that found in an unmatched study of the same population. Asensio and Torres18 did multivariable analysis on data from a cohort of 701 patients who had open‐heart surgery and compared these results with those from a nested case‐control study in which 31 patients with deep surgical site infection were matched with 31 uninfected patients by the variables most closely associated with increased length of stay in the multivariable analysis and by other variables. The mean excess postoperative length of stay determined by multivariable analysis was 20.8 days (95% CI, 16.7‐24.9). However, in the matched analysis, the mean increased length of stay was 14.3 days (95% CI, 3.2‐25.4; ie, 6.5 days shorter), and the median excess length of stay was 26.5 days (5.8 days longer).
Merle et al.19 compared the appropriateness evaluation protocol method with multivariable analysis in a study involving patients undergoing digestive tract surgical procedures and found that for all infections the excess length of stay calculated by the appropriateness evaluation protocol method was half (3.5 days) that from the multivariable analysis (7.2 days). However, the 2 methods gave almost identical results for deep surgical site infection. This discrepancy may have occurred because this type of infection was the primary cause of increased length of stay among patients with deep infections, compared with patients with less serious infections, for whom other factors, such as underlying illnesses, also affected length of stay.
Schulgen et al.21 evaluated 5 methods of estimating excess length of stay related to nosocomial infections, including surgical site infection. Two‐group comparisons (infected vs uninfected) yielded the highest estimates of extra hospital stay (20.7 days; 95% CI, 18.4‐23.0). Matching for confounders alone yielded an estimate of 16.9 days (95% CI, 12.9‐20.9), and matching for confounders and time to infection yielded an estimate of 11.4 days (95% CI, 7.1‐15.7). The 2 methods developed by Schulgen et al.21 to take into account the heterogeneity of the patient population and the time to infection yielded estimates of 9.8 days (95% CI, 5.7‐13.8) and 11.5 days (95% CI, 8.9‐14.0), respectively. They concluded that methods to estimate excess length of stay that do not take into account both of these factors substantially overestimate the effect of nosocomial infection on postoperative length of stay.21
The extent to which nosocomial infections increase costs and length of stay has differed among studies for other reasons, including differences in periods studied (eg, costs in older studies may be lower than those in recent studies because of inflation) and in the way costs are calculated. For example, numerous investigators used hospital charges rather than true costs. However, a charge, the price a hospital places on its services, is not the same as a cost, the value of resources consumed in the production of an item or the delivery of a service.22 Thus, the “costs” reported by different investigators may vary because different hospitals charge different amounts for the same service and not because the services used were truly different. Because hospital charges are not surrogates for costs,23 we used a cost‐to‐charge ratio to estimate costs associated with nosocomial infection.
We quantified some outcomes of postsurgical nosocomial infection that occur after discharge. Both readmission rate and use of antimicrobial agents in the outpatient setting were increased by nosocomial infection. Similarly, Reilly et al.14 found that patients with surgical site infection had a median of 4 “readmission days,” 1 visit to a general practitioner, 1 home visit by a nurse, and 1 prescription for an antimicrobial agent. Most studies, including ours, assessed only direct costs of nosocomial infection (ie, hospital expenditures and resources consumed). Most studies have not assessed indirect costs, such as the cost of time lost from work, expenses for traveling and child care, and the cost to society of the patient’s diminished productivity. The patients’ quality of life may also diminish due to nosocomial infection. Our study demonstrated that patients who have nosocomial infection after an operation have a prolonged hospital stay and are frequently readmitted to the hospital. Consequently, these patients are away from their families, homes, and jobs longer than patients who do not acquire this type of infection. Thus, essentially all studies substantially underestimate the total socioeconomic burden caused by nosocomial infection.
Our study had several limitations. We could not get data on the cost and length of stay for the patients who were discharged to the VAMC. In addition, we could not obtain the costs for care of infections that were treated primarily in the outpatient setting. Moreover, we did not have the exact costs of the services but used a cost‐to‐charge ratio to estimate the costs from the charges. Furthermore, we could analyze only covariates that were collected during the prior clinical trial. Thus, we may have missed important covariates or confounders.
In summary, our study is unique because we evaluated a large group of patients, we evaluated several outcomes and numerous significant covariates by multivariable analyses, and we included time to infection in our analyses of excess postoperative length of stay. Our study demonstrated that nosocomial infection was significantly associated with increased costs and use of healthcare resources by surgical patients. In addition, our study demonstrated that the relationship between infection and outcomes, such as cost and length of stay, is not simple and is affected substantially by numerous patient characteristics. Despite this complexity, surgeons who desire to provide high‐quality care in a cost‐effective manner must find ways to decrease the risk of infections among their patients. Infection control programs can help surgeons by identifying specific remediable risk factors for postsurgical infection in specific populations, by collaborating with surgeons in programs to implement current guidelines for preventing nosocomial infection, and by providing surgeons with rates of infection and estimates of the associated cost, length of stay, and other outcomes.24,25 Together, surgeons and infection control personnel can improve patient safety and the economic viability of surgical programs.
Acknowledgments
This study was funded in part by grants from GlaxoSmithKline (GSK) and the Prevention Epicenters Program of the Centers for Disease Control and Prevention (CDC). GSK funded the clinical trial7 but did not fund the study of outcomes reported here. A Prevention Epicenters grant from the CDC funded the data analysis.
References
- 1. Horan TC, Culver DH, Gaynes RP, et al. Nosocomial infections in surgical patients in the United States, January 1986‐ June 1992. Infect Control Hosp Epidemiol 1993; 14:73‐80.
- 2. Wenzel RP, Pfaller MA. Feasible and desirable future targets for reducing the costs of hospital infections. J Hosp Infect 1991; 18(Suppl A):94‐98.
- 3. Jarvis WR. Selected aspects of the socioeconomic impact of nosocomial infections: morbidity, mortality, cost and prevention. Infect Control Hosp Epidemiol 1996; 17:552‐557.
- 4. Fridkin SK, Pear SM, Williamson TH, et al. The role of understaffing in central venous catheter–associated bloodstream infection. Infect Control Hosp Epidemiol 1996; 17:150‐158.
- 5. Taylor GJ, Mikell FL, Moses HW, et al. Determinants of hospital charges for coronary artery bypass surgery: the economic consequences of postoperative complications. Am J Cardiol 1990; 65:309‐313.
- 6. Weintraub WS, Jones EL, Craver J, et al. Determinants of prolonged length of stay after coronary bypass surgery. Circulation 1989; 80:276‐284.
- 7. Perl TM, Cullen JJ, Wenzel RP, et al. Intranasal mupirocin to prevent postoperative Staphylococcus aureus infections. The Mupirocin and the Risk of Staphylococcus aureus Study Team. New Engl J Med 2002; 346:1871‐1877.
- 8. Herwaldt LA, Swartzendruber SK, Zimmerman MB, et al. Hemorrhage after coronary artery bypass graft procedures. Infect Control Hosp Epidemiol 1998; 19:9‐16.
- 9. Green JW, Wenzel RP. Postoperative wound infections: a controlled study of the increased duration of hospital stay and direct cost of hospitalization. Ann Surg 1977; 185:264‐268.
- 10. Kollef MH, Sharpless L, Vlasnik J, et al. The impact of nosocomial infection on patient outcomes following cardiac surgery. Chest 1997; 112:666‐675.
- 11. Hollenbeak CS, Murphy DM, Koenig S, et al. The clinical and economic impact of deep chest surgical site infections following coronary artery bypass graft surgery. Chest 2000; 118:397‐402.
- 12. Rebollo MH, Bernal JM, Llorca J, et al. Nosocomial infections in patients having cardiovascular operations: a multivariate analysis of risk factors. J Thorac Cardiovasc Surg 1996; 112:908‐913.
- 13. Kirkland KB, Briggs JP, Trivette SL, et al. The impact of surgical site infections in the 1990s: attributable mortality, excess length of hospitalization, and extra costs. Infect Control Hosp Epidemiol 1999; 20:725‐730.
- 14. Reilly J, Twaddle S, McIntosh J, et al. An economic analysis of surgical wound infection. J Hosp Infect 2001; 49:245‐249.
- 15. Coello R, Glenister H, Fereres J, et al. The cost of infection in surgical patients: a case‐control study. J Hosp Infect 1993; 25:239‐250.
- 16. Jenney AW, Harrington GA, Russo PL, et al. Cost of surgical site infections following coronary artery bypass surgery. Aust NZ J Surg 2001; 71:662‐664.
- 17. Erbaydar S, Akgun A, Eksik A, et al. Estimation of increased hospital stay due to nosocomial infections in surgical patients: comparison of matched groups. J Hosp Infect 1995; 30:149‐154.
- 18. Asensio A, Torres J. Quantifying excess length of postoperative stay attributable to infections: a comparison of methods. J Clin Epidemiol 1999; 52:1249‐1256.
- 19. Merle V, Germain J‐M, Chamouni P, et al. Assessment of prolonged hospital stay attributable to surgical site infections using appropriateness evaluation protocol. Am J Infect Control 2000; 28:109‐115.
- 20. Zoutman D, McDonald S, Vethanayangan D. Total and attributable costs of surgical‐wound infections at a Canadian tertiary‐care center. Infect Control Hosp Epidemiol 1998; 19:254‐259.
- 21. Schulgen G, Kropec A, Kappstein I, et al. Estimation of extra hospital stay attributable to nosocomial infections: heterogeneity and timing of events. J Clin Epidemiol 2000; 53:409‐417.
- 22. Larson L. Cost determination and analysis. In: Bootman JL, Townsend RJ, McGhan WF, eds. Principles of Pharmacoeconomics. 2nd ed. Cincinnati: Harvey Whitney Book Company; 1996:44‐59.
- 23. Finkler SA. The distinction between costs and charges. Ann Intern Med 1982; 96:102‐109.
- 24. Cruse PJE, Foord R. The epidemiology of wound infection: a 10‐year prospective study of 62,939 wounds. Surg Clin North Am 1980; 60:27‐40.
- 25. Wenzel RP. The economics of nosocomial infections. J Hosp Infect 1995; 31:79‐87.