The last 2 authors contributed equally to this article.
Health care-associated infections after major cancer surgery
Temporal trends, patterns of care, and effect on mortality
Version of Record online: 19 MAR 2013
Copyright © 2013 American Cancer Society
Volume 119, Issue 12, pages 2317–2324, 15 June 2013
How to Cite
Sammon, J., Trinh, V. Q., Ravi, P., Sukumar, S., Gervais, M.-K., Shariat, S. F., Larouche, A., Tian, Z., Kim, S. P., Kowalczyk, K. J., Hu, J. C., Menon, M., Karakiewicz, P. I., Trinh, Q.-D. and Sun, M. (2013), Health care-associated infections after major cancer surgery. Cancer, 119: 2317–2324. doi: 10.1002/cncr.28027
The funding sources had no impact on the design, collection, data analysis, or interpretation of data. All researchers are independent from the funding sources. The last 2 authors contributed equally to this article.
- Issue online: 4 JUN 2013
- Version of Record online: 19 MAR 2013
- Manuscript Accepted: 24 JAN 2013
- Manuscript Revised: 20 JAN 2013
- Manuscript Received: 26 NOV 2012
- health care-associated infection;
- cancer surgery;
- urinary tract infection;
- surgical site infection;
- blood stream infection
Approximately 1.7 million individuals per year are affected with health care-associated infections (HAIs) in the United States. The authors examined trends in the incidence of HAI after major cancer surgery (MCS) and risk factors for HAI to describe the effects of HAI on mortality after MCS.
Patients undergoing 1 of 8 MCS procedures within the Nationwide Inpatient Sample between 1999 and 2009 were identified (n = 2,502,686). Generalized linear regression models were used to estimate the impact of the primary predictors (procedure type, age, sex, race, insurance status, Charlson comorbidity index, hospital volume, and hospital bed size) on the odds of HAI and in-hospital mortality. Trends in incidence were evaluated with linear regression.
Overall, MCS-associated HAI incidence increased 2.7% per year (P < .001), whereas mortality decreased 1.3% per year (P < .001). Male gender (odds ratio [OR], 1.12, 95% confidence interval [CI], 1.10-1.14), advancing age (OR, 1.02; 95% CI, 1.02-1.02), black race (OR, 1.26; 95% CI, 1.21-1.31), ≥1 comorbidities (OR, from 1.08 [95% CI, 1.04-1.13] to 1.31 [95% CI, 1.27-1.35]), and nonprivate insurance (OR, from 1.18 [95% CI, 1.15-1.22] to 1.67 [95% CI, 1.59-1.76]) were associated with an increased odds of HAI on multivariable analysis. Conversely, increasing hospital volume was associated with lower odds of HAI (OR, 0.999; 95% CI, 0.99-0.99). Patients with MCS-associated HAI had increased odds of mortality (OR, 8.66; 95% CI, 8.51-8.82).
Between 1999 and 2009, the incidence of MCS-associated HAI events increased; however, HAI-associated mortality decreased. That said, significant disparities exist in the hospital and demographic attributes associated with MCS-associated HAI, with attendant health policy implications. Moreover, HAI remains detrimentally linked to mortality during hospitalization. Cancer 2013;119:2317–2324. © 2013 American Cancer Society.
Approximately 1.7 million individuals per year are affected with health care-associated infections (HAIs) in the United States, which makes it 1 of the most prevalent adverse events during hospitalization. Whereas HAIs—predominantly blood stream infections (BSIs), surgical site infections (SSIs), ventilator associated pneumonia (VAP), and urinary tract infections (UTIs)—impose significant morbidity, mortality, and economic burden, they are also largely preventable.[3, 4] In response to increased public awareness, many state legislatures have made HAI reporting mandatory; as of 2010, at least 27 states have enacted laws requiring health systems to report HAI rates,[5, 6]
In the setting of oncologic patients, susceptibility to HAIs may even be greater.[7, 8] For most types of cancer, surgery remains a likely intervention when there is curative intent. Ironically, surgical intervention represents a risk factor for developing HAIs, with variations according to age, baseline conditions, case complexity, and subsequent care.[10, 11] Unfortunately, in studies that investigate HAIs after surgery, cancer patients often are excluded because of their higher risk for HAI.[12, 13]
Given the increasing burden of oncologic care in the United States as well as the relative paucity of data focusing on cancer surgery patients, we sought to evaluate the incidence of HAIs in the last decade in patients undergoing surgery for 1 of 8 solid organ malignancies. In addition, we sought to examine patient and hospital characteristics that may predispose patients to HAIs. Finally, we focused on the associations of HAIs and mortality during hospitalization.
MATERIALS AND METHODS
Relying on the Nationwide Inpatient Sample (NIS), hospital discharges in the United States between January 1, 1999 and December 30, 2009 were abstracted. The NIS is a set of longitudinal hospital inpatient databases included in the Healthcare Cost and Utilization Project family, created by the Agency for Healthcare Research and Quality through a Federal-state partnership. The database includes discharge abstracts from 8 million hospital stays and is the sole hospital database in the United States with charge information on all patients regardless of payer, including individuals covered by Medicare, Medicaid, and private insurance along with the uninsured. Each discharge includes up to 15 inpatient diagnostic and 15 procedural codes. All procedures and diagnoses are coded using the International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM).
In total, 8 major surgical oncologic procedures were selected for to evaluate HAI: colectomy, cystectomy, esophagectomy, gastrectomy, hysterectomy, pneumonectomy, pancreatectomy, and prostatectomy. These operations include a group of commonly performed, complex procedures, which are either associated with the most common cancers or carry a significant risk of morbidity and mortality, as previously reported.[16, 17] Relying on specific ICD-9-CM procedure codes, each surgical procedure was assessed independently, and analyses were restricted to cancer diagnoses only (Table 1).
|Procedure||ICD-9 Procedure Codes||ICD-9 Diagnosis Codes|
|Colectomy||457, 4571, 4572, 4573, 4574, 4575, 4576, 4579, 458, 4581, 4582, 4583||153.x|
|Cystectomy||577, 5771, 5779||188.x|
|Esohagectomy||424, 4240, 4241, 4242||150.x|
|Gastrectomy||435, 436, 437, 439, 4391, 4399||151.x|
|Hysterectomy||683, 6831, 684, 6841, 685, 6851, 687, 6871||182.x|
|Pneumonectomy||323, 3230, 3239, 324, 3241, 3249, 325, 3250, 3259||162.x|
|Pancreatectomy||526, 527, 525, 5251, 5252, 5253, 5259||157.x, 156.x|
Patient and Hospital Characteristics
For all patients, the following variables were available: age, race (white, black, Hispanic, Asian/Pacific Islander, Native American, or other unspecified), insurance status, and Charlson comorbidity index (CCI). Baseline CCI was calculated according to Charlson et al, as adapted by Deyo and colleagues. Insurance categories are combined into general groups, namely, private insurance, Medicare, Medicaid, and other (self-pay). Hospital characteristics include hospital volume and the number of beds, categorized as small, medium, and large, specific to the hospital's region and teaching status.
All demographic characteristics were weighted according to discharge level estimates provided by the Healthcare Cost and Utilization Project. First, descriptive statistics were generated on frequencies and proportions of categorical variables (sex, race, insurance status, CCI, annual hospital volume, hospital location, hospital region, hospital bed size, and hospital teaching status), stratified according to HAI occurrence. Medians and interquartile ranges were reported for continuously coded variables (age). Chi-square tests and Kruskal-Wallis tests were used to compare the statistical significance of differences within categorical and continuous variables, respectively.
Second, temporal trends in rates were analyzed using the linear regression methodology, as previously suggested. Third, multivariable generalized linear regression models were used to assess independent predictors of HAI after major cancer surgery (MCS). The covariates analyzed were age, sex, race, CCI, insurance status, number of hospital beds, and hospital volume. Fixed and random effects for hospital clustering, year of hospitalization, and length of stay also were accounted for. Finally, in separate generalized linear regression models, we assessed the correlation between any HAI, UTI, VAP, SSI, or BSI on perioperative mortality during hospitalization within the entire cohort and within each surgery. Additional analyses were performed to test the potential interactive effect between hospital bed size and hospital volume throughout all models. In that regard, none of the product terms (hospitalvolume*hospitalbedsize) were statistically significant (all Pinteraction ≥ .05).
All statistical analyses were performed using the R statistical software package (R Foundation for Statistical Computing, Vienna, Austria). The 2-sided significance level was set at P < .05.
A weighted estimate of 2,502,686 patients underwent 1 ofthe 8 examined procedures. Baseline sociodemographic characteristics of the entire cohort are described in Table2.
|Percentage of Sample|
|Variables||Overall||Without HAI||With HAI||P|
|Total percentage of patients||100||89.5||10.5||—|
|Age: Median (IQR), y||66 (58-74)||65 (57-73)||72.0 (62-79)||<.001a|
|Hospital bed size||<.001|
Incidence and Temporal Trends of Health Care-Associated Infections
Over the entire study period, the proportion of patients with ≥1 HAI was 10.5% (Fig. 1). The proportion of patients who had UTI, SSI, VAP, and BSI was 4.2%, 3.2%, 3.5%, and 1.9%, respectively. In temporal trend analyses, the incidence of any HAI increased by 2.7% per year (from 10% in 1999 to 11.4% in 2009; P < .001) (Fig. 2). The highest increase was noted for BSI (+3.9% per year; from 1% to 2.7%; P < .001), and the lowest wasnoted for VAP (+1.3% per year; P< .001).
Major Cancer Surgery-Associated Health Care-Associated Infections
Multivariable linear regression analyses for predicting HAIs are reported in Table 3. Women (odds ratio [OR], 0.89; P < .001), patients of other races (OR, 0.93; P = .009), and those who received treatment at high-volume hospitals (OR, 0.99; P < .001) were less likely to experience HAIs postoperatively. In contrast, older age (OR, 1.02; P < .001), black race (OR, 1.26; P < .001), ≥1 comorbidity (OR, 1.08-1.31; P < .001), nonprivately insured individuals (OR, 1.18-1.67; P < .001), and those who received treatment in more recent years (OR, 1.04; P< .001) were more likely to develop an HAI. Procedure-specific variability of the odds of an HAI also was recorded.
|Variable||OR (95% CI)||P|
|Lung resection||0.890 (0.864-0.916)||<.001|
|Hospital bed size|
|Hospital volume||0.999 (0.998-0.999)||<.001|
Mortality After Health Care-Associated Infections and Temporal Trends
Of all patients who experienced any HAI postoperatively, the rate of in-hospital mortality was 11%. Among those who had an occurrence of UTI, SSI, VAP, and BSI, the in-hospital mortality rate was 6%, 10%, 13%, and 31%, respectively. During the entire study span, mortality decreased by 1.3% per year among patients who had an HAI (from 12.3% to 9.9%; P < .001) (Fig. 3). A similar decrease in mortality was observed among patients with UTIs (−2% per year), SSIs (−3.2% per year), VAP(−2.7% per year), and BSIs (−2.8% per year; all P< .001).
The Effect of Health Care-Associated Infections on Major Cancer Surgery Mortality
In multivariable analyses, the occurrence of any HAI was associated with an 8.7-fold increased odds of in-hospital mortality (P < .001) (Table 4). Patients who experienced UTI, SSI, VAP, and BSI were 2.0-fold, 3.8-fold, 4.9-fold, and 17.3-fold more likely to die during hospitalization in the postoperative setting, respectively (P < .001). When analyses were repeated within each MCS, the results remained similar (Table 4).
|HAI Effect on Mortality: OR (95% CI)|
|Cancer Type||Urinary Tract Infection||Pneumonia||Surgical Site Infection||Blood Stream Infection||Any HAI|
|Overall||1.97 (1.85-2.10)||4.90 (4.64-5.17)||3.78 (3.56-4.02)||17.29 (16.33-18.31)||8.66 (8.51-8.82)|
|Colectomy||2.04 (1.88-2.20)||4.44 (4.13-4.78)||3.59 (3.32-3.89)||16.42 (15.27-17.67)||5.82 (5.50-6.15)|
|Cystectomy||2.31 (1.80-2.98)||3.92 (2.84-5.42)||5.73 (4.57-7.18)||16.00 (12.63-20.26)||7.03 (5.66-8.73)|
|Esophagetomy||1.61 (1.02-2.55)||2.65 (1.96-3.58)||2.05 (1.49-2.81)||8.08 (5.98-10.92)||4.38 (3.36-5.71)|
|Gastrectomy||1.47 (1.17-1.85)||2.98 (2.51-3.53)||3.31 (2.78-3.93)||13.16 (11.20-15.45)||4.99 (4.34-5.72)|
|Hysterectomy||3.35 (2.19 -5.11)||13.20 (8.19-21.28)||8.51 (5.53-13.10)||59.12 (35.32-98.96)||7.81 (5.60-10.89)|
|Lung||1.73 (1.60-1.88)||7.28 (6.97-7.60)||5.92 (5.49-6.39)||31.35 (29.43 -33.40)||8.22 (7.88-8.58)|
|Pancreatectomy||1.66 (1.45-1.90)||2.99 (2.64-3.38)||2.59 (2.35-2.85)||9.98 (9.04-11.01)||4.60 (4.22-5.00)|
|Prostatectomy||5.62 (2.34-13.49)||9.30 (3.29-26.28)||14.36 (5.48-37.61)||66.50 (27.78-159.16)||10.64 (6.11-18.55)|
Our results demonstrate that the overall rate of in-hospital HAI after MCS (10.5%) in the current cohort was much higher than the rates previously recorded for elective surgeries (3%). These differential rates suggest that patients with cancer who undergo surgery are indeed at greater risk of HAI. Most important, our results indicate that, although the incidence of any HAI increased during the study period by approximately 2.7% per year, mortality among those with HAI decreased by approximately 1.3% per year. These results also were modulated significantly by sociodemographic components, such as race and insurance status. Black patients, compared with white patients, have a 26% increase in HAIs. Furthermore, if payment was through either Medicare or Medicaid or if the patient was uninsured, then odds of an HAI event increased 18 to 67%.
The use of the NIS for assessing HAI allows the examination of large populations that are nationally representative and permits the generation of trends over an extensive time frame. The database also contains information on all types of payment as well as multiple inpatient diagnostic, complication, and procedural codes. Despite the benefits of population-level data, several studies have suggested that discharge diagnosis codes provide inadequate positive predictive value to facilitate hospital surveillance with ICD-9 data exclusively. Nonetheless, billing codes are a component in most automated systems for HAI surveillance, and it has been demonstrated that billing codes increase sensitivity and decrease the costs of surveillance. Electronic surveillance algorithms using coding data are more likely to be applied consistently over time and suffer from less interobserver variability between reviewers and hospitals. In our analyses, we address broad trends in HAIs and characterize risk. We are not seeking to measure 1 hospital against another. The biases associated with data acquisition were likely unchanged over the study period and evenly distributed amongst studied hospitals.
After the publication of a landmark report from the Institute of Medicine (To Err is Human), numerous guidelines and programs have been advanced for better detection and prevention of HAIs. Without a doubt, HAIs place a considerable burden on the delivery of medical care[1, 3] and have a negative impact on hospital stay, health care costs, and mortality. To date, noncancer hospital inpatient HAIs have received the greatest attention, and the majority of epidemiologic studies that have describe the risk factors and outcomes associated with HAIs largely have avoided cancer patients.[25, 26] However, the risk of HAIs in cancer patients is widespread, and the challenges surrounding the detection and prevention of HAIs are equally (if not more) important in these patients given their increased susceptibility to acquiring HAIs.[7, 8] In that context, the treatment of cancer patients with surgical intervention is associated with a particularly high incidence of HAI.[27, 28]
In addition, our analyses demonstrated that black patients consistently had greater odds of HAIs despite simultaneous adjustment for other confounding patient characteristics. Similarly, nonprivately insured individuals appeared to be predisposed to more postoperative HAIs in adjusted analyses. In the light of previous studies linking insurance type to health care access, it is entirely possible that patients with public insurance may be diagnosed at a more advanced stage and grade than their privately insured counterparts. These cases may be more complex, requiring longer operative time and subsequent postoperative care, thus increasing the odds of HAI. Moreover, decreased access to care presupposes an implicit disparity in access to high-quality hospitals with subspecialized practice profiles. Certain processes of care related to the prevention of HAIs are likely implemented better at high-volume institutions, and there is some evidence to support this hypothesis.
The increasing incidence reflects the growing awareness and sensitization in the medical community and the broader public of HAIs, whereas the decreasing mortality rates reflect the positive impact of programs, risk scores, and guidelines that were proposed in the hope of reducing the risk of mortality after the occurrence of HAIs. Such trends may suggest progress against postoperative HAIs among oncology patients. That said, after controlling for all other covariates, HAIs retained a detrimental association with a nearly 9-fold increase of in-hospital mortality. Thus, much work remains to be done with respect to reducing a serious adverse event (ie, mortality) after an HAI. With respect to specific events, our results demonstrated that BSI-related in-hospital mortality was the highest, at 31%. In that regard, BSI was associated with a 17.3-fold increased rate of in-hospital mortality after adjusting for other covariates. In 2011, the Centers for Disease Control and Prevention released updated guidelines to specifically target the prevention and detection of BSIs. The current findings support the importance of adherence to recommendations made by such guidelines.
There are several limitations to our study, and our findings must be interpreted in this context. One of the significant limitations of the NIS is that, as with most population-based analyses, complications and HAIs were captured through claims data, a process with well described limitations, as mentioned above. Another very important limitation of the NIS data includes a lack of consistent surgeon identification, precluding adjustment for the effect of surgeon volume or individual surgeon practice patterns. Furthermore, in the current analysis, we focused exclusively on postoperative infections in cancer patients, and we were unable to assess the role infections play in the causality of cancer. Finally, in this study, we examined race as a risk factor for HAIs and controlled for it in assessing for the odds of postoperative mortality; however, we were unable to assess the effect of race on cancer incidence.
In summary, we were able to provide new insight into the implications of HAIs on surgical oncology. We observed consistently increasing HAI rates over the last decade, yet decreasing mortality rates, within the context of MCS. This likely reflects improved processes of care in the treatment of critically ill patients as well as progress against HAI-related mortality. However, several patient and hospital attributes were associated with the occurrence of HAIs after MCS. Disparities in HAI incidence and mortality, based on race and insurance coverage, highlight the need for improved access to quality health care to avoid potentially devastating outcomes because of preventable HAIs.
Pierre I Karakiewicz is partially supported by the University of Montreal Health Centre Urology Specialists, Fonds de la Recherche en Santé du Québec, the University of Montreal Department of Surgery, and the University of Montreal Health Centre (CHUM) Foundation.
CONFLICT OF INTEREST DISCLOSURE
The authors made no disclosures.
- 5National Conference of State Legislatures. Lessons from the Pioneers: Reporting Healthcare-Associated Infections. Available at: http://www.ncsl.org/documents/health/haireport.pdf. Accessed July 2,2012.
- 15Healthcare Cost and Utilization Project (HCUP). HCUP Nationwide Inpatient Sample (NIS), 2007-2009. Available at URL: http://www.hcup-us.ahrq.gov/nisoverview.jsp. Accessed September 2012.
- 20Healthcare Cost and Utilization Project (HCUP).Description of Data Elements: Nationwide Inpatient Sample (NIS).Washington, DC:H-CUP, Agency for Healthcare Research and Quality;2011.
- 24To err is human: an interview with the Institute of Medicine's Linda Kohn.Jt Comm J Qual Improv.2000;26:227–234..
- 31HELICS-ICU.The ICU-HELICS programme: towards European surveillance of hospital-acquired infections in intensive care units.Euro Surveill.2002;7:127–128., , , ;