We examined the epidemiology of Clostridium difficile infection (CDI) for hospitalizations in Texas and estimated the incremental impact of CDI on mortality, length of stay, and costs. For patients hospitalized for other conditions or procedures, CDI may result in higher mortality risk, additional costs, and longer lengths of stay. However, little is known about these incremental impacts from observational data. This study uses publicly available inpatient discharge data from Texas to estimate these impacts. Texas was selected for study owing to the large number of hospitals, geographic and demographic diversity of hospitalized patients, and recent population growth in Texas.

C. difficile is the leading cause of infectious diarrhea in hospitalized patients; the Centers for Disease Control and Prevention reported an almost 10-fold increase in deaths due to CDI between 1999 and 2008.Reference Murphy, Xu and Kochanek ¹ Documented infections have increased since 2001 as an epidemic strain of C. difficile (B1/NAP1) appeared; however, subsequent reports using a national sample of adult hospital discharges indicated the overall rate of CDI in hospitals leveled off between 2008 and 2010.Reference Reveles, Lee, Boyd and Frei ² Although it can be acquired in the community, it has been a known cause of healthcare-associated (nosocomial) infection for approximately 30 years.Reference Kelly and LaMont ³ Associated with use of multiple antibiotics and longer hospital stays, CDI is more likely to impact individuals who are vulnerable to infection, such as older adults and patients transferred from other healthcare settings.Reference Clabots, Johnson, Olson, Peterson and Gerding ^{4 ,} Reference Tartof, Rieg, Wei, Tseng, Jacobsen and Yu ⁵ Because there is a need to treat other conditions with antibiotics, CDI is not always avoidable, and management often includes better antibiotic stewardship along with surveillance to allow for early identification and treatment of cases.

This observational cohort study included most inpatient discharges from Texas hospitals between 2007 and 2011 as reported through the Texas Health Care Information Collection Inpatient Public Use Data Files. The deidentified files contain discharge abstracts from Texas hospitals not exempt from reporting due to rural status, to staffing less than 100 beds, or to not soliciting payments from insurers or the government. ⁶ Other systematic exclusions from the public use data were to protect patient identity through suppression of demographic information, which accounted for approximately 5% of all discharges. ⁶ CDI cases were identified by an International Statistical Classification of Disease, Ninth Revision, Clinical Modification code of 008.45 (C. difficile) in any of the discharge diagnosis fields. CDI discharges were stratified by principal versus secondary diagnosis and those with secondary diagnosis were matched to controls (without CDI) using observed characteristics and one-to-one greedy propensity-score matching methods.Reference Parsons ⁷ In-hospital mortality and length of stay were directly assessed from the discharge records; costs were estimated using facility-specific cost to charge ratios generated from the Centers for Medicare and Medicaid Services cost reports. ⁸ Mortality, mean/median length of stay, and mean/median cost outcomes were compared between CDI cases and matched controls. McNemar’s test was used to assess statistically significant differences in mortality odds ratios because mortality is a dichotomous measure, whereas paired t tests and Wilcoxon signed-rank tests were used to assess statistical significance in differences in length of stay and cost outcomes, respectively.

This study identified 14,723,825 discharge records from Texas hospitals between 2007 and 2011. Of these records, 97,989 (0.67%) had a principal or secondary diagnosis of CDI. The CDI rate per 1,000 discharges across 5 years was 6.66, with a rate of 7.4 cases per 1,000 discharges in 2011, with higher rates for white patients not of Hispanic origin, adults aged 65 or older, and patients arriving from or discharged to other health facilities, including nursing homes (data not shown). For discharges with CDI as a secondary diagnosis, the most common principal diagnoses were septicemia, rehabilitation, acute respiratory failure, pneumonia, pneumonitis, acute renal failure, and pressure ulcer (data not shown).

Hospitalizations with a CDI as a secondary diagnosis had significantly higher in-hospital mortality, longer mean and median lengths of stay, and higher mean and median costs for each data year (see Table 1). Odds ratios for mortality ranged between 1.65 and 1.87 for the overall matched sample comparison of CDI cases compared with controls. CDI hospitalizations were at least 1 week longer, on average, compared with non-CDI hospitalizations; this difference decreased over the study time frame from 9.3 to 7.4 days. Longer median lengths of stay for CDI cases versus controls also decreased from 8 to 6 additional days over the studied years. Median hospitalization costs, which reflect the 50th percentile of discharges and minimize the impact of cost outliers, were approximately $8,000 to $8,350 higher for CDI discharges over the study time frame.

TABLE 1 Estimated Differences in Mortality, Length of Stay, and Cost Between Inpatients With Clostridium difficile Infection (CDI) as a Secondary Diagnosis and Matched Inpatients Without CDI, 2007-2011

NOTE. All odds ratios and differences are significant at P<.001 except for total charges in 2009, which was significant at P<.025. Controls were selected through propensity score logistic regression and one-to-one greedy match algorithms. The matching models evaluated included all available variables and combinations of the following interactions (including none): race and ethnicity, Major Diagnostic Category (MDC) and no. of comorbidities, and payer 1 and payer 2. The final model included all variables and significant interaction terms of payer 1 and payer 2, and MDC and no. of comorbidities. This model yielded the highest C-statistic, had the most complete matches, and had the fewest nonsignificant covariates in the matching model results.

^a Rate is number of deaths per 1,000 patients.

^b Odds ratios were calculated through use of the McNemar’s odds ratio for one-to-one propensity score matching of cases to controls. Significance of differences were measured using McNemar’s test.

^c Mean length of stay and mean costs were calculated through the use of one-to-one propensity score matching of cases to controls. Significance of differences were measured with paired t test.

^d Costs estimated from total charges using cost to charge ratios from the Centers for Medicare and Medicaid Services cost reports.

^e Number of cases differ owing to missing total charges data.

CDI was identified in a small percentage of hospital discharges in Texas, but the rate of CDI per 1,000 discharges increased from 6.02 in 2007 to 7.40 in 2011. Analyses according to year, demographic, geographic, facility, and diagnostic characteristics confirmed that CDI was an increasing problem in Texas hospitals between 2007 and 2011 with disproportionate impacts on older adults, patients in long-term care facilities, and non-Hispanic white patients (data not shown). Even after careful adjustment using propensity score matching, discharges with CDI as a secondary diagnosis had increased in-hospital mortality, longer lengths of stay, and higher costs. The noted decrease in in-hospital mortality rates for CDI hospitalizations over the study time frame was inconsistent with findings from a recent national study but may reflect differences in the underlying hospitalized population in Texas.Reference Reveles, Lee, Boyd and Frei ² Nevertheless, reductions in avoidable CDI cases through improved infection control practices and antibiotic stewardship could improve quality and quantity of life for persons potentially impacted by this infection.