Risk for Asymptomatic Household Transmission of Clostridioides Difficile Infection Associated With Recently Hospitalized Family Members

Aaron C. Miller; Alan T. Arakkal; Daniel K. Sewell; Alberto M. Segre; Sriram V. Pemmaraju; Philip M. Polgreen

Disclosures

Emerging Infectious Diseases. 2022;28(5):932-939. 

In This Article

Methods

Data Source

We constructed our study population from the US Commercial Claims and Medicare Supplemental datasets of IBM MarketScan Research Databases (https://www.ibm.com) from 2001–2017. These databases contain employer-sponsored commercial insurance claims and Medicare supplemental claims for >195 million enrollees during the 17-year study period. This dataset represents one of the largest longitudinal administrative databases in the United States. The databases provided insurance claims for inpatient, outpatient, and emergency department encounters, along with outpatient medications, demographic characteristics, employment, and enrollment characteristics. We were able to link claims from multiple family members in the same enrollment plan by using a family identifier along with a variable indicating each enrollee's relationship to the primary enrollee, which indicated spouse, child, or dependent.

Study Population

We restricted our study population to enrolled households in which ≥2 family members could be identified on the same insurance plan. Our analysis was based on monthly CDI incidence, so we restricted our study population to those enrollees that were continuously enrolled for an entire month. We used code 008.45 from International Classification of Diseases, 9th Revision (ICD-9), and codes A04.7, A04.71, and A04.72 from the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM), to identify CDI cases in outpatient and inpatient settings. To eliminate recurrent infections or subsequent care for the same infection, we focused on CDI cases in which the patient had no prior CDI diagnosis ≤60 days prior to the index month.

To isolate the potential effect of asymptomatic household transmission attributable to a prior hospitalization, we applied 2 additional restrictions to remove potential symptomatic exposures that might confound our results. First, we restricted our analysis to only enrollees that did not have a family member with CDI diagnosed in the period ≤60 days prior to the index month. Second, we restricted our analysis to those enrollees who were not hospitalized themselves ≤60 days prior to the index month.

Analysis

We compared the monthly incidence of CDI between persons in households where another family member had been recently hospitalized and discharged, ≤60 days prior to the index month, to those without recently hospitalized family members. We used a regression model to stratify enrollees into monthly enrollment strata based on the year and month, along with other demographic and patient characteristics, such as age, sex, prior antimicrobial drug use, PPI use, presence of an infant ≤2 years of age in the household, and exposure to a recently hospitalized family member. We then estimated the CDI incidence within each monthly enrollment strata, as a function of these various characteristics (Appendix, https://wwwnc.cdc.gov/EID/article/28/5/21-2023-App1.pdf).

We separated enrollees into categories for ages 0–17, 18–40, 41–65, and >65 years. We also categorized antimicrobial drugs into separate risk strata for high-CDI-risk antibiotics (clindamycin, fluoroquinolones, cephalosporins, carbapenems, ampicillin/sulbactam, pipercillin/tazobactam, and later-generation cephalosporins) or low-CDI-risk antibiotics (penicillin, macrolides, sulfonamides, trimethoprim, tetracyclines, and first-generation cephalosporins). We identified patients taking 1 of the following PPIs within 30 days before the CDI index date: omeprazole, esomeprazole, lansoprazole, rabeprazole, pantoprazole, dexlansoprazole, and omeprazole with sodium bicarbonate. We included an indicator for the presence of an infant ≤2 years old in the household because higher colonization rates have been found in infants.[17,21]

Quantifying Exposure to Recently Hospitalized Family Members. We evaluated the effect of exposure to a recently hospitalized family member in 2 ways. First, we defined a single dichotomous stratification based on whether any other family member spent time in the hospital ≤60 days prior to the index month. We then analyzed the incidence rate ratio (IRR) of CDI associated with exposure to a recently hospitalized family member. Second, we investigated whether a dose-response relationship existed between risk for CDI and the total amount of time that recently hospitalized family members spent in the hospital ≤60 days prior to the index month by computing the total days of within-family hospitalization. Specifically, we summed the lengths of stay across recently hospitalized family members' inpatient stays that overlapped the previous 60-day exposure window. For example, a case-patient with 2 family members discharged in the prior 60 days, 1 with a length of stay of 2 days and the other 3 days, would have 5 total days of within-family hospitalization (Appendix Figure 1). Finally, we sorted total days of within-family hospitalization into categories of 0, 1–3, 4–10, 11–20, 21–30, and >30 days by using 0 days (i.e., no hospitalization or a hospitalization of <1 day) of prior exposure as the reference.

Statistical Approach. We started by computing monthly CDI incidence for each of the patient characteristics used to define the various strata we described. We then estimated IRRs for the various patient strata while accounting for potential confounding effects by using a log-linear regression model, along with a quasi-Poisson distribution to account for overdispersion. Specifically, we estimated the mean CDI incidence in each monthly enrollment strata as a function of the binary criteria that define a stratum (Appendix). Of note, this approach and study population previously have been used to estimate the risk for secondary CDI infections among family members in household settings.[11]

Sensitivity Analyses. We conducted 2 sensitivity analyses. First, we evaluated whether underlying susceptibility at a household level might confound our results. For example, households with family members more susceptible to CDI also could be more likely to have longer or more frequent hospitalizations (Appendix Figure 2). To evaluate this effect, we analyzed 2 models in which we reversed the temporal order and evaluated whether CDI risk is associated with future hospitalizations in a family (Appendix).

Second, we explored the time window used to define prior exposures. Specifically, we considered a 90-day exposure window before index CDI events to compute total days of within household exposure and prior exposure to antimicrobial drugs.

processing....