Hostname: page-component-6898cfcb55-cbk24 Total loading time: 0 Render date: 2026-11-11T09:46:12.122Z Has data issue: false hasContentIssue false
  • English
  • Français

Electronically Available Comorbid Conditions for Risk Prediction of Healthcare-Associated Clostridium difficile Infection

Published online by Cambridge University Press:  05 February 2018

Anthony D. Harris,
Alyssa N. Sbarra,
Surbhi Leekha,
Sarah S. Jackson,
J. Kristie Johnson,
Lisa Pineles and
Kerri A. Thom
Anthony D. Harris
Affiliation:
Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, Maryland
Alyssa N. Sbarra
Affiliation:
Department of Epidemiology of Microbial Diseases, Yale School of Public Health, New Haven, Connecticut
Surbhi Leekha
Affiliation:
Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, Maryland
Sarah S. Jackson
Affiliation:
Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, Maryland
J. Kristie Johnson
Affiliation:
Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, MarylandDepartment of Pathology, University of Maryland School of Medicine, Baltimore, Maryland
Lisa Pineles
Affiliation:
Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, Maryland
Kerri A. Thom*
Affiliation:
Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, Maryland
*
Address correspondence to Kerri A. Thom, MD, MPH, 655 W Baltimore Street, Bressler Hall M-021, Baltimore, MD 21201 (kthom@som.umaryland.edu).
Rights & Permissions [Opens in a new window]

Abstract

OBJECTIVE

To analyze whether electronically available comorbid conditions are risk factors for Centers for Disease Control and Prevention (CDC)-defined, hospital-onset Clostridium difficile infection (CDI) after controlling for antibiotic and gastric acid suppression therapy use.

PATIENTS

Patients aged ≥18 years admitted to the University of Maryland Medical Center between November 7, 2015, and May 31, 2017.

METHODS

Comorbid conditions were assessed using the Elixhauser comorbidity index. The Elixhauser comorbidity index and the comorbid condition components were calculated using the International Classification of Disease, Tenth Revision, Clinical Modification (ICD-10-CM) codes extracted from electronic medical records. Bivariate associations between CDI and potential covariates for multivariable regression, including antibiotic use, gastric acid suppression therapy use, as well as comorbid conditions, were estimated using log binomial multivariable regression.

RESULTS

After controlling for antibiotic use, age, proton-pump inhibitor use, and histamine-blocker use, the Elixhauser comorbidity index was a significant risk factor for predicting CDI. There was an increased risk of 1.26 (95% CI, 1.19–1.32) of having CDI for each additional Elixhauser point added to the total Elixhauser score.

CONCLUSIONS

An increase in Elixhauser score is associated with CDI. Our study and other studies have shown that comorbid conditions are important risk factors for CDI. Electronically available comorbid conditions and scores like the Elixhauser index should be considered for risk-adjustment of CDC CDI rates.

Infect Control Hosp Epidemiol 2018;39:297–301

Information

Type
Original Articles
Information
Infection Control & Hospital Epidemiology , Volume 39 , Issue 3 , March 2018, pp. 297 - 301
Copyright
© 2018 by The Society for Healthcare Epidemiology of America. All rights reserved 

Although healthcare-associated infection rates have decreased dramatically for a number of infections, this has not been the case for Clostridium difficile infection (CDI). 1 Hospital-onset CDI rates decreased by 8% from 2011 to 2014, 22% short of the US Department of Health and Human Services (HHS) goal of a 30% decrease. 1 Clostridium difficile has been labeled an “urgent threat” by the Centers for Disease Control and Prevention (CDC).Reference Evans and Safdar 2 In an effort to reduce CDI, hospital-onset CDI is a publicly reported, pay-for-performance metric.

Risk factors for CDI have been previously described in the literature.Reference Garey, Sethi, Yadav and DuPont 3 Important risk factors that have been identified in many studies include the use of both antibiotics and gastric acid suppression therapy, which includes proton pump inhibitors and histamine blockers.Reference Dial, Alrasadi, Manoukian, Huang and Menzies 4 , Reference Donskey 5 These are potentially preventable risk factors where interventions targeting a decrease in unnecessary use of these medications could lead to decreased C. difficile rates. Other risk factors for CDI such as age and comorbid conditions are not usually preventable.Reference Evans and Safdar 2 Comorbid conditions identified as risk factors for CDI in various studies include liver disease, organ transplant, weight loss, inflammatory bowel disease, recent bowel surgery, congestive heart failure, renal failure, coagulopathy, and malignant tumors.Reference Evans and Safdar 2 , Reference Khanafer, Vanhems, Barbut and Luxemburger 6 – Reference Rodemann, Dubberke, Reske, Seo and Stone 9 However, no US studies have assessed whether these comorbidities as captured by International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes are risk factors for CDI. If comorbid conditions captured by ICD-10-CM codes are identified as CDI risk factors, these could eventually be used for risk adjustment and prediction rules. The aim of the current study was to investigate whether comorbid conditions captured by ICD-10-CM codes are risk factors for CDI after controlling for important risk factors such as antibiotic and gastric acid suppression therapy use and thus determine whether electronically available comorbid conditions may be candidates for risk adjustment.

METHODS

We performed a retrospective cohort study among all adult patients admitted to the University of Maryland Medical Center (UMMC), a 755-bed tertiary-care hospital located in Baltimore, Maryland. The University of Maryland Institutional Review Board approved this study.

The cohort was created using the hospital’s central data repository, which is a relational database containing administrative, pharmacy, clinical, and laboratory data from electronic medical records. The cohort included all adult patients (aged ≥18 years) admitted to UMMC between November 7, 2015, and May 31, 2017. Data extracted included age, gender, race or ethnicity, ICD-10-CM codes, medication administration, length of stay, C. difficile test date, and C. difficile test result. Validation of a random sample of the fields used was performed by chart review. In this study and in prior studies, the positive and negative predictive values of the fields in the electronic medical record or central data repository were >99%.Reference Harris, Jackson and Robinson 10 – Reference Harris, Fleming and Bromberg 13

Per CDC NHSN case definition, a patient was considered to have hospital-onset CDI if the patient tested positive for C. difficile at least 3 days after hospital admission and had no prior positive in the preceding 14 days. 14 Throughout the study period, testing for toxigenic C. difficile in liquid stool samples was performed using a nucleic acid amplification test (NAAT, Illumigene C. difficile assay, Meridian Bioscience, Cincinnati, OH). The laboratory did not test stool specimens that were formed. Best practice alerts recommending against C. difficile in the presence of diarrhea due to laxative use and against repeat testing went into effect in the electronic medical record on December 20, 2016.

The ICD-10-CM codes were mapped to appropriate comorbid components of the Elixhauser comorbidity index as outlined by Quan et al.Reference Quan, Sundararajan and Halfon 15 The Elixhauser comorbidity index is a sum of 30 unweighted comorbid conditions in which possible scores range from 0 to 30.Reference Elixhauser, Steiner, Harris and Coffey 16 Multiple ICD-10-CM codes were used in defining each comorbidity component, but the components were operationalized as binary variables. For example, a patient was considered to be positive for the lymphoma component binary variable if his or her medical record had any of the following ICD-10-CM codes: 200.x-202.3x, 202.5-203.0, 203.8, 238.6, 273.3, V10.71, V10.72, or V10.79.Reference Quan, Sundararajan and Halfon 15

Patients were considered to have received antibiotic or gastric acid suppression therapy if they had received at least one dose from the time of hospital admission until the time of a positive C. difficile test or until discharge for patients who did not develop the outcome. Antibiotic use was analyzed in individual classes in the bivariate analysis and in 1 of the multivariable models. Histamine blocker and proton pump inhibitors were analyzed as separate classes. In the primary multivariable model, antibiotic use was analyzed as a binary variable defined as whether patients received or did not receive any systemic antibiotic to analyze whether comorbid conditions were risk factors beyond all antibiotic use. Antibiotic therapy and gastric acid suppression therapy use were analyzed because of their association with C. difficile in prior studies.Reference Dial, Alrasadi, Manoukian, Huang and Menzies 4 , Reference Donskey 5

Bivariate associations were performed using log binomial regression. Age was considered a continuous covariate in bivariate analysis due to its previously established linear associations with CDI.Reference Jump 17 Candidacy for covariates in a multivariable analysis was determined by the significance of association in the bivariate approach as indicated by P<0.10. The linearity of association between CDI and total Elixhauser score was assessed, and due to its linear association, the Elixhauser index was analyzed as a continuous variable. Age was assessed for linear association with CDI and was analyzed as a continuous variable. Tests for effect modification and confounding between antibiotic use and Elixhauser score were conducted. Multivariable log binomial regression was used to assess the risk of CDI. We conducted a likelihood ratio test to assess model fit for models with and without the Elixhauser Index. A second multivariable logistic regression model was conducted to analyze individual antibiotics, individual comorbid conditions, age, and gastric acid suppression. Stepwise regression was used for variable selection. Because the outcome was rare, we interpreted the odds ratios as risk ratios. Analyses were performed using SAS version 9.4 software (SAS Institute, Cary, NC).

RESULTS

Of 37,246 patient admissions that met entry criteria, 3,001 had a C. difficile test performed, of whom 257 (8.5%) were positive and met the CDC definition for hospital-onset C. difficile infection (CDI). The CDI incidence rate for our hospital was 9.5 per 10,000 patient days in calendar year 2015 and 9.1 per 10,000 patient days in calendar year 2016. Among patients with CDI, the average length of hospital stay was 10.8 days (standard deviation [SD], 10 days) before a positive result. Among all patients in the cohort, the mean age was 52.7 years (SD, 18 years) (Table 1). Among patients with CDI, 80% had gastric acid suppression therapy compared to 54% of those without CDI (P<.001), and 88% of persons with CDI had antibiotic therapy compared to 58% of persons without CDI (P<.001). The mean Elixhauser score among all patients in the cohort was 3.4 (SD, 2). Patients with CDI had a mean Elixhauser score of 5.3 (SD, 2), and those without CDI had a mean score of 3.4 (SD, 2; P<.001).

TABLE 1

Selected Characteristics of Patients Admitted to University of Maryland Medical Center in Baltimore, Maryland, Between November 7, 2015, and May 31, 2017 by Clostridium difficile Infection Status

NOTE. PCR, polymerase chain reaction assay; SD, standard deviation; LOS, length of hospital stay.

a P value is for the t test or χ2 test as appropriate for continuous and categorical variables, respectively.

In the bivariate analysis, age, antibiotic use for all antibiotic classes, gastric acid suppression therapy use, and Elixhauser score were significantly related to CDI (Table 2). After controlling for any antibiotic use, age, and gastric acid suppression in a multivariable log binomial regression, for every one-point increase in Elixhauser score, there was 1.26 times the odds of having CDI (95% CI, 1.21–1.33) in model 1 (Table 3). We compared this model to one without the Elixhauser Index and found that the model with the index was a better fit (P<.0001). The individual antibiotic and individual comorbid condition multivariable model is presented in model 2 (Table 3). Carbapenem, vancomycin, piperacillin-tazobactam, and cefepime were individual antibiotics statistically associated with CDI. Malnutrition, coagulopathy, lymphoma, renal failure, and neurological conditions were statistically associated with CDI. Complicated hypertension was statistically protective.

TABLE 2

Risk Ratios and 95% Confidence Intervals (CI) for the Associations Between Selected Covariates and Clostridium difficile Infection

a Estimates are obtained from a log binomial regression.

b Wald 95% confidence intervals.

TABLE 3

Multivariable Regression to Assess Risk of Clostridium difficile Infection

NOTE. CI, confidence interval.

a Estimates are obtained from a logistic regression.

b Wald 95% CI.

c For each additional comorbidity, the risk of CDI increases by 26%.

d For each 1 year in age, the risk of CDI increases by 1%.

DISCUSSION

Patients with comorbid conditions identified by ICD-10-CM codes had a greater risk of hospital-onset C. difficile infection. Specifically, patients with higher Elixhauser scores (representing greater number of comorbidities) had a greater risk of CDI. This association remained strong even when controlling for important confounding variables of antibiotic therapy, age, and gastric acid suppression therapy.

A nested case-control study among French patients using ICD-10 codes found that chronic renal or liver disease and malnutrition were comorbid conditions that were risk factors for CDI while controlling for other confounding variables.Reference Khanafer, Vanhems, Barbut and Luxemburger 6 Among these comorbid conditions, 2 were found in our study as well. The French study measured these comorbid conditions by chart review, whereas in our study, electronically available ICD-10 codes were used, which has practical implications for broader application of such risk adjustment. In a study in Boston, a secondary analysis of an existing dataset was performed specifically to assess the risk of gastric acid suppression therapy for nosocomial CDI.Reference Howell, Novack and Grgurich 7 Using ICD-9 codes mapped to Elixhauser index components to control for possible confounding and similarly to our study, investigators found that an increase in the number of comorbid conditions was associated with an increased risk of CDI.

Current CDC NHSN hospital-onset CDI models adjust for the following variables: (1) academic or nonacademic affiliation, (2) bed size, (3) type of C. difficile testing, (4) community-onset prevalence, and (5) number of ICU beds. However, none of these variables are patient-level risk factors; therefore, they may not completely account for the patient case mix and associated nonmodifiable CDI risk at different hospitals.Reference Dudeck, Weiner, Malpiedi, Edwards, Peterson and Sievert 18 Our group has demonstrated that comorbid conditions identified by ICD codes are significant risk factors for surgical site infections and central-line–associated bloodstream infections.Reference Jackson, Leekha and Magder 19 , Reference Jackson, Leekha and Magder 20 We have also shown that risk adjustment models using ICD codes and accounting for comorbidity had better predictive discrimination and calibration than the current CDC models for these infections.Reference Jackson, Leekha and Pineles 21 A study in California showed substantial variation in hospital rankings when adjusting for group-level hospital characteristics including comorbid conditions.Reference Datta, Kazerouni and Rosenberg 22 These issues lead many hospital epidemiologists to desire better risk adjustment modeling and analysis before national reporting of outcomes occurs.Reference Datta, Kazerouni and Rosenberg 22 – Reference McGregor and Harris 24

The major limitation of this is study is that it is a single-center study, which impacts the generalizability. This work needs to be repeated at multiple institutions, including hospitals with distinct characteristics such as nonacademic and nonurban hospitals. Another limitation is that our work did not adjust for the severity of illness of patients. Further work is needed in this area as evidenced by a study that showed that a hospital’s case-mix index, reflecting severity of illness, was a risk factor for CDI.Reference Thompson, Edwards, Dudeck, Fridkin and Magill 25 A criticism of the use of ICD codes in research is that they fail to capture all patient comorbidities or could reflect codes that maximize reimbursement.Reference Quan, Parsons and Ghali 26 , Reference Schneeweiss and Maclure 27 Research comparing the Charlson and Elixhauser comorbidity indices derived from ICD-9-CM codes to those same scores extracted from chart review has found that the sensitivity of the individual components varies greatly but the specificity is nearly 100%.Reference Quan, Li and Saunders 28 , Reference Leal and Laupland 29 Future work could further elucidate the association between comorbid condition scores, such as the Elixhauser comorbidity index and the Charlson comorbidity index, and their association with CDI. In addition, different results could be obtained if antibiotics were based on duration received and not just as a categorical variable. Another limitation is that we did not control for hospital admission source because it was not available in our database. It has been shown that patients who transfer from long-term care facilities are at increased risk of CDI. However, the prevalence of community-onset CDI (ie, CDI cases diagnosed within the first 3 days of hospital admission) did not change during the study period.

In conclusion, if our work is repeated at multiple institutions and shows that electronically available comorbid conditions are risk factors for CDI, the CDC and the Centers for Medicare and Medicaid Services (CMS) should strongly consider adding comorbidity-based risk adjustment to publicly reported C. difficile infection rates.

ACKNOWLEDGMENTS

We thank Colleen Reilly and Yuan Wang for database maintenance and abstraction. The opinions expressed here are those of the authors and do not reflect the official position of the Agency for Healthcare Research and Quality, the National Institute of Allergy and Infectious Diseases, or the US Department of Health and Human Services.

Financial support: This project was partially supported by the AHRQ (grant no. R01HS022291), by the NIAID (grant no. 2K24AI079040-06), by the US Department of Health and Human Services, and by the National Institutes of Health (NIH grant no. 5K24AI079040).

Potential conflicts of interest: All authors report no conflicts of interest relevant to this article.

References

REFERENCES

1. HAI progress report FAQ. Centers for Disease Control and Prevention website. https://www.cdc.gov/hai/surveillance/progress-report/faq.html. Updated 2016. Accessed July 31, 2017.Google Scholar
2. Evans, CT, Safdar, N. Current trends in the epidemiology and outcomes of Clostridium difficile infection. Clin Infect Dis 2015;60(Suppl 2):S66–S71.Google Scholar
3. Garey, KW, Sethi, S, Yadav, Y, DuPont, HL. Meta-analysis to assess risk factors for recurrent clostridium difficile infection. J Hosp Infect 2008;70:298–304.CrossRefGoogle ScholarPubMed
4. Dial, S, Alrasadi, K, Manoukian, C, Huang, A, Menzies, D. Risk of Clostridium difficile diarrhea among hospital inpatients prescribed proton pump inhibitors: cohort and case-control studies. CMAJ 2004;171:33–38.Google Scholar
5. Donskey, CJ. The role of the intestinal tract as a reservoir and source for transmission of nosocomial pathogens. Clin Infect Dis 2004;39:219–226.Google Scholar
6. Khanafer, N, Vanhems, P, Barbut, F, Luxemburger, C, CDI01 Study Group. Factors associated with Clostridium difficile infection: a nested case-control study in a three-year prospective cohort. Anaerobe 2017;44:117–123.Google Scholar
7. Howell, MD, Novack, V, Grgurich, P, et al. Iatrogenic gastric acid suppression and the risk of nosocomial Clostridium difficile infection. Arch Intern Med 2010;170:784–790.CrossRefGoogle ScholarPubMed
8. Dubberke, ER, Reske, KA, Olsen, MA, et al. Risk for Clostridium difficile infection after allogeneic hematopoietic cell transplant remains elevated in the postengraftment period. Transplant Direct 2017;3:e145.Google Scholar
9. Rodemann, JF, Dubberke, ER, Reske, KA, Seo, DH, Stone, CD. Incidence of Clostridium difficile infection in inflammatory bowel disease. Clin Gastroenterol Hepatol 2007;5:339–344.CrossRefGoogle ScholarPubMed
10. Harris, AD, Jackson, SS, Robinson, G, et al. Pseudomonas aeruginosa colonization in the intensive care unit: prevalence, risk factors, and clinical outcomes. Infect Control Hosp Epidemiol 2016;37:544–548.Google Scholar
11. Pepin, CS, Thom, KA, Sorkin, JD, et al. Risk factors for central-line-associated bloodstream infections: a focus on comorbid conditions. Infect Control Hosp Epidemiol 2015;36:479–481.CrossRefGoogle ScholarPubMed
12. Harris, AD, Johnson, JK, Thom, KA, et al. Risk factors for development of intestinal colonization with imipenem-resistant Pseudomonas aeruginosa in the intensive care unit setting. Infect Control Hosp Epidemiol 2011;32:719–722.Google Scholar
13. Harris, AD, Fleming, B, Bromberg, JS, et al. Surgical site infection after renal transplantation. Infect Control Hosp Epidemiol 2015;36:417–423.Google Scholar
14. Multidrug-resistant organism and Clostridium difficile infection (MDRO/CDI) module. Centers for Disease Control and Prevention website. https://www.cdc.gov/nhsn/pdfs/pscmanual/12pscmdro_cdadcurrent.pdf. Updated 2017. Accessed July 31, 2017.Google Scholar
15. Quan, H, Sundararajan, V, Halfon, P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43:1130–1139.Google Scholar
16. Elixhauser, A, Steiner, C, Harris, DR, Coffey, RM. Comorbidity measures for use with administrative data. Med Care 1998;36:8–27.CrossRefGoogle ScholarPubMed
17. Jump, RL. Clostridium difficile infection in older adults. Aging Health 2013;9:403–414.Google Scholar
18. Dudeck, MA, Weiner, LM, Malpiedi, PJ, Edwards, JR, Peterson, KD, Sievert, DM. Risk adjustment for healthcare facility-onset C. difficile and MRSA bacteremia laboratory-identified event reporting in NHSN. Centers for Disease Control and Prevention website. https://www.cdc.gov/nhsn/pdfs/datastat/riskadjustment-mrsa-cdi.pdf. Published 2013. Accessed July 31, 2017.Google Scholar
19. Jackson, SS, Leekha, S, Magder, LS, et al. Electronically available comorbidities should be used in surgical site infection risk adjustment. Clin Infect Dis 2017;65:803–810.Google Scholar
20. Jackson, SS, Leekha, S, Magder, LS, et al. The effect of adding comorbidities to current centers for disease control and prevention central-line-associated bloodstream infection risk-adjustment methodology. Infect Control Hosp Epidemiol 2017:1–6.Google Scholar
21. Jackson, SS, Leekha, S, Pineles, L, et al. Improving risk adjustment above current centers for disease control and prevention methodology using electronically available comorbid conditions. Infect Control Hosp Epidemiol 2016;37:1173–1178.Google Scholar
22. Datta, R, Kazerouni, NN, Rosenberg, J, et al. Substantial variation in hospital rankings after adjusting for hospital-level predictors of publicly reported hospital-associated clostridium difficile infection rates. Infect Control Hosp Epidemiol 2015;36:464–466.Google Scholar
23. Sexton, DJ, Chen, LF, Moehring, R, Thacker, PA, Anderson, DJ. Casablanca redux: We are shocked that public reporting of rates of central line-associated bloodstream infections are inaccurate. Infect Control Hosp Epidemiol 2012;33:932–935.Google Scholar
24. McGregor, JC, Harris, AD. The need for advancements in the field of risk adjustment for healthcare-associated infections. Infect Control Hosp Epidemiol 2014;35:8–9.Google Scholar
25. Thompson, ND, Edwards, JR, Dudeck, MA, Fridkin, SK, Magill, SS. Evaluating the use of the case mix index for risk adjustment of healthcare-associated infection data: an illustration using Clostridium difficile infection data from the national healthcare safety network. Infect Control Hosp Epidemiol 2016;37:19–25.Google Scholar
26. Quan, H, Parsons, GA, Ghali, WA. Validity of information on comorbidity derived rom ICD-9-CCM administrative data. Med Care 2002;40:675–685.Google Scholar
27. Schneeweiss, S, Maclure, M. Use of comorbidity scores for control of confounding in studies using administrative databases. Int J Epidemiol 2000;29:891–898.Google Scholar
28. Quan, H, Li, B, Saunders, LD, et al. Assessing validity of ICD-9-CM and ICD-10 administrative data in recording clinical conditions in a unique dually coded database. Health Serv Res 2008;43:1424–1441.CrossRefGoogle Scholar
29. Leal, JR, Laupland, KB. Validity of ascertainment of co-morbid illness using administrative databases: a systematic review. Clin Microbiol Infect 2010;16:715–721.Google Scholar
Figure 0

TABLE 1Selected Characteristics of Patients Admitted to University of Maryland Medical Center in Baltimore, Maryland, Between November 7, 2015, and May 31, 2017 by Clostridium difficile Infection Status

Figure 1

TABLE 2Risk Ratios and 95% Confidence Intervals (CI) for the Associations Between Selected Covariates and Clostridium difficile Infection

Figure 2

TABLE 3Multivariable Regression to Assess Risk of Clostridium difficile Infection