Burden of Clostridium difficile-associated disease among patients residing in nursing homes: a population-based cohort study

Background Clostridium difficile (C. difficile) infection (CDI) is the leading cause of nosocomial diarrhea in the United States. This study aimed to examine the incidence of CDI and evaluate mortality and economic burden of CDI in an elderly population who reside in nursing homes (NHs). Methods This was a population-based retrospective cohort study focusing on US NHs by linking Medicare 5% sample, Medicaid, Minimum Data Set (MDS) (2008–10). NH residents aged ≥65 years with continuous enrollment in Medicare and/or Medicaid Fee-for-Service plan for ≥12 months and ≥2 quarterly MDS assessments were eligible for the study. The incidence rate was calculated as the number of CDI episodes by 100,000 person-years. A 1:4 propensity score matched sample of cohorts with and without CDI was generated to assess mortality and health care costs following the first CDI. Results Among 32,807 NH residents, 941 residents had ≥1 episode of CDI in 2009, with an incidence of 3359.9 per 100,000 person-years. About 30% CDI episodes occurred in the hospital setting. NH residents with CDI (vs without CDI) were more likely to have congestive heart failure, renal disease, cerebrovascular disease, hospitalizations, and outpatient antibiotic use. During the follow-up period, the 30-day (14.7% vs 4.3%, P < 0.001), 60-day (22.7% vs 7.5%, P < 0.001), 6-month (36.3% vs 18.3%, P < 0.001), and 1-year mortality rates (48.2% vs 31.1%, P < 0.001) were significantly higher among the CDI residents vs non-CDI residents. Total health care costs within 2 months following the first CDI episode were also significantly higher for CDI residents ($28,621 vs $13,644, P < 0.001). Conclusions CDI presents a serious public health issue in NHs. Mortality, health care utilization, and associated costs were significant following incident CDI episodes. Electronic supplementary material The online version of this article (doi:10.1186/s12877-016-0367-2) contains supplementary material, which is available to authorized users.


Background
Clostridium difficile (C. difficile) infection (CDI) is the leading cause of nosocomial diarrhea in the United States and has surpassed the infection rate of other health care-associated infections (HAI) such as methicillin-resistant Staphylococcus aureus [1,2]. In a recent national prevalence survey, C. difficile was the most commonly-reported pathogen, accounting for 12% of HAIs and an estimated 80,400 hospital-onset infections [3]. C. difficile was responsible for almost a halfmillion infections and approximately 28,000 deaths in 2011 [4]. The emergence of the epidemic BI/NAP1/027 strain has been associated with the increased incidence, severity, and mortality of CDI [5]. The increasing incidence of CDI also imposes a large financial burden on the health care system [6]. In 2008, the excess health care costs for CDI in acute care facilities alone were approximately $4.8 billion [7].
The elderly represent the population at greatest risk for CDI, accounting for the highest disease morbidity and mortality. Moreover, residents of nursing homes and other long-term care (LTC) facilities are more prone to CDI due to advanced age, frequent hospitalization, prevalent comorbid illnesses, extended stays, and frequent exposure to antibiotics [1]. In 2013, Garg et al. determined that there was a predominantly higher number of patients diagnosed with LTC-acquired CDI (46.1%) compared to community-acquired (33.3%) or hospital-acquired (20.6%) CDI [8]. Additionally, statewide surveillance in Ohio has found that there were more cases of CDI diagnosed in LTC facilities than in acute-care settings [9]. The increasing trends of CDIincidence in LTC facilities such as nursing homes highlight the need for optimal C. difficile infection control measures [8].
Previous studies have focused on the costs related to CDI in acute-care facilities [10,11]. However, to our knowledge, there are no recent studies assessing the costs of CDI in nursing homes. Therefore, the objectives of this study were to examine the incidence of CDI and evaluate the mortality and economic burden of CDI in an elderly population residing in nursing homes.

Study design and data sources
This population-based retrospective cohort study included adults aged ≥65 years enrolled in Medicare or with dual eligibility with the Medicaid fee-for-service (FFS) health plan during 2008-2010. A Medicare 5% random sample was used, and all eligible enrollees were required to have both Part A (hospital insurance) and Part B (supplementary medical insurance) during the study period [12]. Because some Medicare enrollees, may also be covered by Medicaid due to socioeconomic factors, we linked the Medicare 5% sample with the 100% Medicaid claims data.
To identify Medicare enrollees residing in nursing homes, we linked the 5% Medicare sample and 100% Medicaid with the 100% Minimum Data Set (MDS) and required ≥2 consecutive MDS quarterly assessments in the 12 months prior to the CDI diagnosis. The MDS is a federally-mandated nursing home clinical assessment performed on all residents at admission, discharge, annually, and quarterly, regardless of the payer.

Identification of CDI episodes
The primary outcome was the incidence of CDI, identified as residents with ≥1 medical claim for CDI (International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM] code: 008.45) in Medicare or Medicaid data, or identified CDI (1 = yes) in MDS data from January 1, 2009 through December 31, 2009. An incident CDI episode was defined as a patient with no occurrences of CDI within 60 days prior to the first CDI claim. Among patients with multiple episodes of CDI, each episode was considered as incident if there was at least a 60-day gap between the 2 episodes. We also identified the setting (eg, inpatient, nursing home) of infection onset using the first CDI claim setting.

Case-control analysis
As a secondary objective, we examined mortality at 30, 60, and 180 days, and at 1 year following the onset of the index CDI episode. Since mortality information is derived from the Medicare enrollment file which is updated by the Social Security Administration, we were able to capture all deaths. We also analyzed all-cause health care costs and utilization stratified by health plan (eg, Medicare and Medicaid) within 2 months following the first incident episode among those with CDI (cases) and those without CDI (controls). For this analysis, in contrast to the incidence computation, residents were required to have no diagnosis of CDI within the 12 months prior to the first incident episode (index date) in 2009. The 12 months prior to this index date was used as an observation period to allow sufficient time to determine baseline demographic and clinical characteristics and health care utilization parameters. Residents with CDI episodes identified in the MDS typically lacked complete Medicare and Medicaid data; therefore, these residents were excluded from this portion of the analysis. Eligible residents enrolled in Medicare or Medicaid without a CDI diagnosis between January 1, 2008 and December 31, 2010, served as the control group and were randomly assigned an index date in 2009.

Statistical analyses
The incidence rates of CDI were calculated as the number of residents with a CDI episode per 100,000 personyears and stratified by age, gender, race, and region. For the case-control analysis, each CDI case resident was matched to 4 non-CDI control residents using a 'greedy' match method [13]; CDI cases were matched to non-CDI controls within 0.001 units of the propensity score [14]. The propensity score was calculated via a logistic regression model of factors associated with the risk of CDI, including the residents' age, gender, race, US census region, comorbidities using the Charlson Comorbidity Index (CCI) score, prior hospitalizations, nursing home stays, antibiotic exposure, gastric acid suppressant use within 90 days and 1 year prior to the incident CDI, individual comorbidities, and comorbid conditions included the MDS data.
Chi-square tests were used to evaluate differences in categorical variables. Student t-tests were used for evaluating differences in continuous variables. In addition to p-values, standardized differences were calculated for each variable. A standardized difference of >10% was used to assess significant practical differences in the case-control comparison [15].
Statistical significance was set a priori at the alpha <0.05. All analyses were performed using Statistical Analysis System (SAS) version 9.3 (SAS Institute Inc., Cary, NC).

CDI incidence rates
We identified 1,051 incident episodes of CDI among 941 residents with ≥1 CDI episode in a cohort of 32,807 residents who met the eligibility criteria. The annual incidence density of CDI was 3,752.7 episodes per 100,000 person-years and 3,359.9 residents per 100,000 person-years (Table 1). Across the age strata, an inverse linear trend was observed: the incidence ranged from a low of 3,414.9 episodes per 100,000 person-years among residents aged ≥85 years to a high of 4,247.5 episodes/ per 100,000 person-years among those aged 65-74 years. Incidence was slightly higher among men than women and lower among whites as compared to blacks or other race categories. Geographically, the incidence density of CDI was highest in the Northeast (4,641.1 per 100,000 person-years), and lowest in the Midwest (3,306.8 episodes per 100,000 person-years). The majority of CDI episodes occurred in a nursing home (71.7%), and 28.3% occurred in an inpatient hospital setting.

Case-control analysis
After accounting for the inclusion and exclusion criteria, a sample of 789 residents with an incident CDI episode and 30,258 without were included for this analysis. The mean age of the residents with and without CDI was 83.1 years and 83.6 years, respectively. Geographically, a higher proportion of CDI residents resided in the Northeast region (36.1% vs 28.5%, P < 0.001), whereas a higher proportion of non-CDI residents resided in the Midwest (34.3% vs 27.8%, p < 0.001; Table 2). Approximately two-thirds of residents had Medicaid coverage along with their Medicare insurance.
Residents with CDI were more often diagnosed with comorbidities than those without CDI (CCI scores: 4.6 ± 2.8 in CDI vs 3.2 ± 2.4 in non-CDI groups, P < 0.001; Additional file 1: Table S1). The hospitalization burden among those with CDI within the preceding 1 year (64.8% vs 32.4%, P < 0.001) or 90 days (43.2% vs 10.9%, P < 0.001) was significantly higher compared to residents without CDI. After propensity score matching (PSM), 653 residents with an incident CDI episode and 2,612 residents without a CDI episode were compared for the secondary objectives.

Mortality
During the 12-month follow-up period, the 30-day (14.7% vs 4.3%, P < 0.001) mortality rate was significantly higher among the propensity score (PS)-matched CDI residents as compared to the PS-matched non-CDI residents, so are mortality rates at 60-day, 6-month and 1year follow up (Fig. 1).

Health care utilization
Among Medicare-covered claims, during the 2-month follow-up period, the number of inpatient claims was significantly higher among PS-matched CDI residents compared to PS-matched non-CDI residents (0.66 vs 0.13, P < 0.001). Similarly, the number of skilled nursing facility (SNF) and carrier claims was significantly higher among PS-matched CDI residents. Among the Medicaid covered claims, the number of inpatient claims was significantly higher among PS-matched CDI residents as compared to PS-matched non-CDI residents (0.23 vs 0.08, P < 0.001); however, the number of pharmacy claims was significantly higher in the PS-matched non-CDI cohort as compared to the PS-matched CDI cohort (2.37 vs 1.74, p < 0.001; Table 3).

Discussion
This study estimated the burden of CDI among nursing home residents. To our knowledge, this is the first study to estimate the health care costs and utilization associated with each CDI episode in this elderly population.
We reported an incidence rate of 3,753 cases per 100,000 person-years (1.03 cases per 10,000 resident days), and 3,360 CDI patients per 100,000 person-years among nursing home residents. A study conducted by Kim et al. reported that the yearly incidence rate of LTC-associated CDI increased from 0.4 cases per 10,000 person-years in 2008 to 0.8 cases per 10,000 personyears in 2009 [1]. According to the results from another study conducted in 2010, the incidence rate of CDI in nursing homes was 2.3 cases per 10,000 resident days [16]. These results suggest a persistent increasing trend of CDI incidence among the nursing home residents in recent years. The emergence of new, severe, and hypervirulent epidemic strains of C. difficile over the past years has particularly affected older adults and played a role in the increasing CDI rates [17]. Previous literature has noted the increased incidence of CDI; however, the majority of these studies focused on acute care settings [7,10,18]. Besides hospitalized patients, a trend of increased CDI incidence has been observed in nursing homes and communities [8,19]. CDI outbreaks have been reported in nursing facilities since 1986 [5]. An influx of CDI in nursing homes from acute care may lead to subsequent transmission within the facility, leading to an overall increase in nursing home CDI rates. Residents may also acquire CDI during their stay at the nursing home, adding to the increased incidence rate [1]. Several factors may have contributed to the shift from primarily a hospital-acquired infection to a nursing home-acquired infection, such as an increase in the elderly population admitted to nursing home facilities, an increase in use of multiple medications (especially antibiotics and proton pump inhibitors [PPIs]), and multiple comorbidities among the elderly population [8]. Because nursing home residents are one of the groups most vulnerable to CDI, preventative measures are crucial to effectively prevent the disease. These include additional nursing care, meticulous hand hygiene, contact precautions, appropriate use of antimicrobials and future interventions, such as vaccines. Accompanied by a high increase in the incidence rate, high mortality associated with CDI was also observed. Although there is limited data on long-term mortality, a few studies have assessed 30-day mortality [4,19] and our 30-day mortality result (10.4%) is similar to these studies. As the incidence of CDI continues to rise, health care costs and utilization related to CDI will have a tremendous impact on the health care system. Previous studies reported costs attributable to CDI ranging from $8,000-$30,000 that vary with coexisting conditions [20]. However to our knowledge, no study has systematically assessed healthcare costs and utilization associated with CDI in nursing homes where incidence of CDI is quickly growing. Our study is unique since it used a large data sample by linking Medicare, Medicaid, and MDS; examined total health care costs including inpatient, outpatient, SNF, and pharmacy costs; and provided an overall picture of the costs among residents in a nursing home facility. Our study included both Medicare and Medicaid covered claims and expenses. Medicare covers relatively few LTC services, such as SNF and home health care [21,22]. Over 60% of nursing home resident care, and an even higher percentage of longterm resident care, is paid for by state Medicaid programs [23]. Thus, providing costs using both claims data offers more generalizable results, helps to better understand the true overall resource utilization, and indicates that CDI is associated with high health care costs, especially among patients residing in nursing home facilities.
A few limitations should be noted for this study. We included both Medicare and Medicaid claims data, but a certain proportion of residents may have supplemental long-term care insurance such as Medigap plans to fill the 'gaps' that are not covered by Medicare plans. These direct medical costs were not captured in this study. The indirect costs, such as those associated with additional nursing care, cleaning, isolation, and possible transmission to other patients, were also not accounted for in this study. Therefore, this study underestimates the true total burden of CDI among nursing home residents. Given that Medicare and Medicaid data contain administrative information from multiple inpatient and outpatient sources, under-reporting or misclassification of health outcomes of interest may occur. Since we required patients to have at least 2 MDS quarterly assessments, the sample consisted entirely of long-term stay residents. Patients who were in the nursing home for a short amount of time and were discharged were excluded from the study.
Although the Medicare, Medicaid, and MDS datasets contain uniquely integrated data, its beneficiaries are considered a vulnerable population since they are older and have a higher disease burden. Our results demonstrated an inverse distribution of CDI incidence as the residents grew older. Few previous studies have also demonstrated such an inverse trend among nursing home residents; however, no explanations have been provided [24,25]. To verify our findings, we did a sensitivity analysis to analyze the baseline CCI score and medical costs across the age strata. We found a similar inverse trend in which residents ≥85 years residing in nursing homes had a lower comorbidity score, health care utilization, and costs prior to the incident CDI compared to patients aged 65-74 years and 75-84 years. An explanation may be that older adults (≥85 years) are more likely to reside in a nursing home facility due to comorbid conditions and frailty, and younger seniors (65-74) are more likely to reside in a nursing home facility due to poor health and may experience increased health care exposures, putting them at a higher risk for CDI.

Conclusions
Results from our study showed 3,753 cases per 100,000 person-years, 10.4% 30-day mortality, and $15,000 in expenditures per case. With 1.4 million people residing in nursing homes in the United States, [26,27] we estimated that approximately 53,000 annual CDI cases were associated with 5,500 deaths and $800 million in costs among this population. These numbers suggest that C. difficile is endemic in nursing home facilities, and more effective disease prevention and infection control measures are warranted to prevent CDI and thus improve patient outcomes and reduce overall health care costs.

Additional file
Additional file 1: Table S1. Clinical Characteristics and Health Care Utilizations Between Non-CDI and CDI Residents Before and After PSM. Table S1. includes the baseline clinical characteristics such as hospitalization, antibiotic use, and heart, musculoskeletal, neurological, and psychiatric conditions between the CDI and non-CDI residents before and after propensity score matching.