Epidemiology and Healthcare Costs of Incident Clostridium difficile Infections Identified in the Outpatient Healthcare Setting

Jennifer L. Kuntz, PhD; Eric S. Johnson, PhD; Marsha A. Raebel, PharmD; Amanda F. Petrik, MS; Xiuhai Yang, MS; Micah L. Thorp, DO, MPH; Steven J. Spindel, MD; Nancy Neil, PhD; David H. Smith, PhD

Disclosures

Infect Control Hosp Epidemiol. 2012;33(10):1031-1038. 

In This Article

Methods

We conducted a population-based, retrospective cohort study among Kaiser Permanente Colorado and Kaiser Permanente Northwest members between June 1, 2005, and September 30, 2008. During the study time period, Kaiser Permanente Northwest and Kaiser Permanente Colorado collectively had a membership of approximately 900,000 on any given day. Data on patient membership, pharmacy dispensings, demographics, and clinical measures were collected from regional electronic databases. The study was reviewed and approved by the Institutional Review Boards at both health plans.

Identification and Categorization of Clostridium Difficile Infections and Follow-Up

In the outpatient setting, we identified CDIs through (1) a diagnosis of International Classification of Diseases, Ninth Revision (ICD-9) code 008.45 (intestinal infection due to C. difficile) or (2) a positive C. difficile toxin test. We further required that positive toxin tests be associated with metronidazole or vancomycin dispensed in the outpatient pharmacy in the 7 days before or after a positive test. All CDIs in the inpatient setting were identified by ICD-9 code 008.45; that code's sensitivity and specificity has been shown to be high for inpatients.[13,14] The index date of CDI was defined as the date on which the first indication of CDI occurred (eg, date of C. difficile diagnosis, positive toxin test result, or metronidazole or vancomycin dispensing). Both health plans consistently utilized the Meridian Premier toxin A/B enzyme immunoassay (Meridian Bioscience) during the study time period. For inpatient-identified CDIs, the index date was the admission date for the hospitalization during which the infection was diagnosed. We then categorized CDIs by the setting (inpatient or outpatient) in which they were first identified.

Patients were required to have continuous membership in the health plan and prescription drug coverage for at least 1 year before the CDI index date. Exclusion criteria were a recorded history of a prior CDI in the 180 days before the index date, as evidenced by a C. difficile diagnosis, a positive C. difficile toxin test, or an outpatient prescription fill for vancomycin. For patients with more than 1 CDI during the study time period, we used the first incident CDI for data collection and analysis.

Measurement of Potential Risk Factors for CDI Identified in the Outpatient Setting

We assessed patient demographic characteristics and comorbidity in the 365 days before the CDI index date. Healthcare utilization and outpatient prescription medication use were gathered for the preceding 180 days to ensure incident events; however, only 60 days of history for these exposures were used in the model to predict the setting of CDI identification.

To measure underlying comorbidity, we determined whether persons with CDI were diagnosed with cardiovascular disease, chronic pulmonary disease, diabetes, inflammatory bowel disease, liver disease, malignancy and metastatic tumors, or rheumatologic disease, as identified by ICD-9 diagnosis codes (codes available on request). We used estimated glomerular filtration rates to evaluate renal function[15] and hemoglobin laboratory values to identify anemia. A history of immunosuppression or chemotherapy was identified through diagnosis and procedure codes or medication utilization. Healthcare utilization was measured at baseline by identifying inpatient and outpatient healthcare encounters or an admission to a nonacute healthcare institution.

We identified prescription medications filled at outpatient pharmacies in the preceding 60 days, specifically, antimicrobials, gastric acid suppressants (including proton pump inhibitors and histamine-2 receptor antagonists), statins, and chronic oral corticosteroids. We did not evaluate exposure to medications during hospital admissions. We examined receipt of selected antimicrobials, number of unique antimicrobials received, and timing of the receipt of antimicrobials in relation to the CDI index date. Antimicrobials were categorized by class, although we also created a category for other antimicrobials that included clindamycin, daptomycin, linezolid, metronidazole, rifampin, telithromycin, synercid, and tigecycline. Use of gastric acid suppressants and statins was categorized as never received or ever received. Chronic corticosteroid use was defined as at least a 90-day supply dispensed in the previous 180 days.

Measurement of Healthcare Utilization and Adverse Events Following Outpatient-identified CDI

We measured the occurrence of outpatient visits, emergency department visits, and hospitalizations listing a diagnosis of ICD-9 code 008.45 in the 180-day time period including and following the index date of CDI. For patients with outpatient-identified CDI who were subsequently hospitalized, we reported time from outpatient diagnosis of CDI to hospital admission. In addition, we collected information about all-cause mortality among patients with CDI. Patients were included in this analysis if they had 180 days of complete follow-up or complete follow-up until death during the 180-day time period after CDI. It should be noted that we did not include a non-CDI comparison group, so these follow-up findings should not be construed to imply attribution to CDI.

Statistical Analyses

We calculated summary statistics for demographic characteristics, healthcare utilization, comorbid conditions, and medication use. We utilized a logistic regression model to determine how strongly baseline characteristics predict outpatient-identified versus inpatient-identified CDIs. Patients were excluded from modeling if they had missing values for covariates. Covariates were not selected on the basis of statistical significance.[16] Instead, we initially included all covariates and excluded only those that measured a concept similar to another variable in the model (eg, number of antimicrobial agents). Thus, our model is a full model, and, as recommended by Harrell,[16] we minimized potential overfitting by allowing at least 20 CDI events per degree of freedom. All odds ratios (ORs) were simultaneously adjusted for other characteristics in the logistic regression model.

Calculation of Healthcare Costs Associated With Clostridium Difficile Infection

We calculated healthcare costs in the 180-day time period including and following the first occurrence of CDI among patients with complete follow-up during that time. We based our costing method on previously developed procedures.[17] For outpatient costs, standard prices were created for office visits by specialty, department, and type of clinician (eg, physician, physician assistant). The number of visits (per department and clinician type) for each patient was then multiplied by the appropriate unit price. Medication costs approximate retail prices within the local community and were based on Kaiser Permanente Northwest only. Hospitalizations were classified into diagnosis-related groups, and the average daily rate per diagnosis-related groups was then multiplied by the length of stay. Laboratory testing costs were derived from the 2009 Centers for Medicare and Medicaid Services Medicare fee schedule. All costs are reported in 2009 US dollars, using year-specific inflation factors from the US Bureau of Labor and Statistics.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....