Quality of life and utility decrement associated with Clostridium difficile infection in a French hospital setting

This is an observational prospective study performed in 7 French acute-care facilities located in Paris, Lyon, Lille and Valenciennes, 4 of which were large University Hospitals. The recruitment period was over the calendar year of 2016. Patient selection criteria were the following: adult (age > 18 years) inpatients presenting with a bacteriologically confirmed diagnosis of CDI during a stay in one of the participating hospitals for any clinical causes, accepting to participate and having given an informed consent. Patients presenting with chronic inflammatory bowel disease, involved in an on-going clinical study or unable to fill out a questionnaire for any linguistic or cognitive reasons were excluded. All cases were identified by the bacteriologists that established the CDI diagnosis in each participating facility.

The bacteriologist informed the interviewer in charge of the study of each case. The interviewer then met with patients to explain the study and ask for their informed consent in case they met the selection criteria. Patients were asked to fill out the EQ-5D-3 L [19] to describe their present state of health at the moment of the CDI episode (7 ± 2 days after diagnosis). Due to difficulties in arranging a new appointment with the patient after resolution of the current CDI episode because of hospital discharge or patient’s death, a retrospective assessment of the patients’ overall perception of health before the episode was also performed. Patients were asked to use the same instrument to describe retrospectively their state of health immediately before the present CDI episode (baseline). In the EQ-5D-3 L, the health status is described by ticking off one of three levels of functioning (“no problems”, “some problems” and “extreme problems”) on five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Using a set of weights (value set or so called tariff) for each level of functioning in each dimension, the descriptive information can be converted into a single utility score. Since this study was conducted in France, the French value set was used to calculate utility scores [20]. Patients were also asked to rate their overall perception of health on the EQ-5D-3 L Visual Analog Scale (EQ-VAS), which ranges from 0 (worst imaginable health state) to 100 (best imaginable health state). All relevant clinical data were collected from the patients’ records: primary and secondary diagnoses as coded in the Hospital Discharge Database, length of stay, stay in ICU, immunosuppression, presence of any CDI episode within the previous 2-year period and all clinical data needed to characterize the severity of CDI episode. A classical measure of CDI severity is the Zar score which ranges from 0 to 6 (6 indicating the most important severity) [21]. This score is calculated by attributing one point to each of the following items: age > 60 years, temperature > 38.3 °C, albumin level < 2.5 mg/dL, or peripheral White Blood Cell count > 15,000 cells/mm3 within 48 h of enrollment. Two points are given for endoscopic evidence of pseudomembranous colitis or treatment in the intensive care unit. Patients with 2 points or more are considered by clinicians to have severe CDI.

Regarding comorbidity, a revised version of the Charlson index [22, 23], based on the ICD-10 coding of primary and secondary diagnoses recorded in the hospital discharge database was used. This index is calculated on the basis of the presence or absence of 17 comorbidities weighted according to their potential influence on mortality. Immunosuppression was considered present for patients with a concomitant immunosuppressive treatment; chemotherapy or radiotherapy; corticosteroid therapy for 30 days or more or at high doses (> 5 mg/kg prednisolone > 5 days); and/or presence of blood disease, metastatic cancer, or HIV infection with CD4 < 500/mm.

Mean and SD for utility scores and EQ-VAS were calculated at the time of the CDI episode and at baseline. The decrement of utility scores for each patient was calculated by subtracting the current utility score from the baseline value. No adjustment for missing data in the EQ-5D was performed. It was tested whether the decrements of score were respectively dependent on the following categorical factors (Student t-test): age (< 65 years versus ≥65 years), gender (M/F), severity of the episode (Zar score ≥ 3 versus < 3), presence of previous CDI episodes (yes/no), Charlson index (< 3 versus ≥3), EQ-5D at baseline (negative versus positive), presence of immunosuppression (yes/no). A multivariate linear regression analysis was performed using the SAS (SAS V9.3) procedure PROC GLM to analyze the overall effect of these categorical variables. All variables were included simultaneously in the model without any option of automatic selection. The number of Quality Adjusted Days (QALDs) lost due to the CDI episode was calculated by multiplying the number of days from baseline (first symptoms) to the date of the interview by the decrement of utility. Quality Adjusted Years (QALYs) were calculated by dividing QALDs by 365.

A total of 80 patients were included. Three patients died during hospital stay (for causes other than CDI) after having completed the survey and their data were included in the analysis. The population had a mean age of 69.4 years (SD: 14.5) and 55% were women. The characteristics of their hospital stays are presented in Table 1.

The CDI was coded as primary diagnosis in 22.8% of cases indicating that it was the reason for the patient’s admission to the hospital. Other primary diagnoses were very diverse, the three most frequent being congestive heart failure (19 patients, 23.8%), chronic pulmonary disease (16 patients, 20.0%) and diabetes mellitus (14 patients, 17.5%). The mean Charlson index was 5.1 (SD: 2.8) with a percentage of 60% of patients being ≥5 indicating a high risk of mortality associated with severe comorbidities. Three patients died after a mean of 34 days (SD: 22.1) spent in hospital and 2 of them were treated in the ICU for their CDI.

The characteristics of the CDI episodes are displayed in Table 1. The CDI episode was considered as clinically severe for 32.5% of patients according to the Zar score (score ≥ 2). The number of stools per day may be also used as a severity indicator. At diagnosis, patients declared a mean number of 6.3 stools/day (up to 20 for one patient) and 88.6% of patients had more than 3 stools/day. At the date of EQ-5D-3 L completion, these figures were reduced to 2.4 stools/day in average and 38.5% of patients had > 3 stools/day.

The EQ-5D-3 L instrument, including the associated EQ-VAS, was administered on average 6.6 days (SD: 1.6) after CDI bacteriological diagnosis but 19.1 days (SD: 21.3) after the onset of symptoms (according to patient declaration), with a median of 12.0 days (min 3 days / max 124 days), among the 74 patients with no missing data. Additional file 1: Table S1 presents the answers to the 5 sub-scales as completed by the participants with the left column corresponding to the retrospective assessment of their health state before the occurrence of the current CDI episode (baseline) and the right column corresponding to their assessment at the day of interview.

The answers to the EQ-5D-3 L at baseline suggest that patients were already presenting with a severe condition having a substantial impact on their quality of life. A proportion of 34.5% of them reported at least one important or extreme problem (level 3) among the 5 dimensions of the questionnaire, especially anxiety/depression (12.5%), inability to perform usual activities (11.4%) and pain/discomfort (11.3%). This proportion increased to 68.7% during the CDI episode, with 5 patients mentioning the most extreme impact on all 5 dimensions. Pain/discomfort was rated extreme in almost half of the population (48.8%), inability to perform usual activities and confinement to bed was also very frequently reported (43 and 40% respectively).

Table 2 shows the utility measures obtained from the EQ-5D-3 L at baseline and during the current episode. Utility scores dropped from 0.542 (SD: 0.391) at baseline to 0.050 (SD: 0.404) during the CDI episode. Utility scores at baseline were negative in 14 patients (18%) with a mean of − 0.086 (SD: 0.075) and the number of patients that considered themselves with a perfect state of health (utility 1) was only 11 (13.9% of all). During the episode, utility scores were negative in 43 patients (54.4%) with a mean of − 0.266 (SD: 0.162) and positive in the rest of the population with a mean utility score of 0.427 (SD: 0.249).

The mean utility decrement associated with the CDI episode measured in the 78 patients with no missing data was 0.492 points (SD: 0.398) with a maximum value of 0.750 and a minimum of − 1.416 in one patient (utility gain). It may be worth noticing that such a gain was observed in 5 patients indicating an improvement of their score at the moment of completion of the self-questionnaire compared to baseline.

At baseline, patients rated their state of health on the EQ-VAS instrument at a mean score of 61.0 (SD: 23.0), that dropped to 35.2 (SD: 19.8) during the episode i.e. a decrement of 25.8 (SD: 23.0) points.

The utility decrement from baseline showed statistically significant increases according to the following characteristics of patients: Zar score of severity ≥3 (p = 0.001), a positive utility score at baseline (p = 0.032), being a woman (p = 0.041) and having an age > 65 years (p = 0.041) (Table 3). Conversely, the Charlson index, the presence of immuno-suppression or existence of a recurrence did not significantly modify this decrement. The same analyses performed with the EQ-VAS score were only significant for the Zar score of severity (p = 0.0189). The SAS procedure PROC GLM was used with the same variables on the utility decrement with a global statistical significance (p = 0.0025) (Table 4). It confirmed the previously documented results with the exception of the effect of gender which was no longer significant.

Assuming that the health-related quality of life as measured by the EQ-5D-3 L was stable during the CDI episode from symptoms onset until the end of treatment, it is possible to derive from these results an estimate of the utility loss associated with a CDI episode. The result obtained in the 74 patients with no missing data was 10.22 QALD (Quality-Adjusted Life Days) (SD: 19.44) or 0.028 QALY (SD: 0.053). By restricting the calculation to patients with less than 60 days of diarrhea (n = 72), the results are 7.2 QALD (SD: 7.41) or 0.020 QALY. These estimates did not include the QALY loss due to the excess mortality associated with CDI which should be estimated from external sources [5].