Design

An integrative review design was used. An integrative design was chosen because it summarises empirical and theoretical literature and allows for the synthesis of results when study designs and methodologies vary between studies. As a result, it provides a more comprehensive understanding of a particular issue[11].

Search Methods

The literature was accessed through searches on two electronic databases—Medline and Pubmed—limited to the period 1 January 2005 to 30 April 2011. This limitation reflects the recently changing epidemiology of CDI. Other limits included only searching literature published in English, and studies involving humans. Key search words used were ‘Clostridium difficile and mortality’ and ‘Clostridium difficile and death’. These searches were combined, with duplicate studies removed.

The next step of the search strategy process involved a preliminary review of the articles. Only prospective or retrospective articles that were case controlled, cohort or reviews were included. Additionally, articles were only included if they examined the mortality of hospitalised patients and were not limited to CDI in a specific patient group, for example, a person with cancer. Only articles that did not use CDI as a comparison group were included, for example, mild versus severe CDI and related mortality. The rationale for excluding such articles was to gain a better understanding of CDI in the general hospitalised population rather than those with a specific illness.

A second review of the articles was undertaken, and this step excluded articles that did not examine mortality at fixed intervals, for example, 30 or 90 days. The inclusion of studies that documented mortality at fixed intervals was chosen in an attempt to assist the pooling of data during data analysis.

Search Outcome

The initial search yielded 362 articles. Following the first review, 303 articles were excluded and 59 articles remained. The second review, which excluded articles that did not examine CDI mortality at fixed intervals, resulted in a further exclusion of 23 articles. A total of 26 articles remained after the second review. Figure1 provides a summary of the search strategy process.

Quality Appraisal

All papers included in this review were published in peer-reviewed journals.

Data abstraction and synthesis

Study Characteristics

Of the 24 articles included, 12 were prospective, 11 retrospective and one was a review. Eighteen studies reported mortality at 30 days or less, with 12 studies reporting mortality at further endpoints, predominantly 90 or 180 days. One study that reported 30-day mortality used 30-day post-discharge as the definition[12]. The choice of 30 days post-discharge as the method used to examine mortality was made due to limitations in available data, which should be noted when considering results from this study. Four studies examined in-hospital mortality in addition to 30-day mortality. The reported mortality in each of these studies is detailed in Table1 and is explored in more detail later.

The search strategy used to identify articles for this review did not identify the same articles in the review published by Karas, Enoch & Aliyu (2010). Seven articles included in the review by Karas, Enoch and Aliyu[13] were not include in our review. These studies did not meet our inclusion criteria and were excluded if the participants were from a selected group. For example, they had a specific strain of CDI or had severe CDI. Conversely, our study did identify and included 14 studies not used by Karas, Enoch and Aliyu[13]. The primary reason for this discrepancy was the recent publications of studies, with 11 of these 14 published in the past two years. The review by Karas, Enoch and Aliyu[13], also focussed on attributable mortality.

Reported mortality

As reported in the included studies, mortality varied considerably, both in the type of mortality calculated and the figures represented. The study design, exclusion criteria of participants and the data collected in the studies varied, making the pooling of data for analysis impossible. All-cause mortality at 30 days varied from 9% to 38%, as shown in Table1. Three studies reported all-cause mortality at 90 days, with a range of 27% to 30%[14–16]. Similarly, three studies report attributable mortality at 30 days, varying from 5.7% to 6.9%[17–19]. In-hospital mortality was reported in four studies and ranged from 8% to 37%[20–23].

Eighteen studies documented all-cause mortality at 30 days, or 28 days in the case of the study undertaken by Sundram et al.[24], which examined two strains of C.difficile (027 and 106) and their respective mortality. The total cohort size was larger than the two groups combined; therefore, it was unclear whether participants with other strains of C.difficile died. Acknowledging the limitations of different study designs, and excluding the study undertaken by Sundram et al.[24], 2,041 of 7,774 persons (26.3%) died within 30 days of CDI diagnosis. These figures include people identified with CDI in these studies who died within 30 days (all-cause mortality)[14, 15, 17–20, 22, 25–34].

The manner in which mortality was analysed and presented in the studies differed; however, there were some similarities. In several studies, participants were divided into those with CDI and those without[19, 22, 27, 35]. Other studies divided participants into two groups: survivors and non-survivors[18, 26, 33]. Regardless of the methods used, similar data analysis involved using mortality at a fixed period. Commonly, X², or Fisher’s exact test, was used for categorical data analysis, while a t test was used for continuous variables[14, 20]. Kaplan–Meier survival analysis and regression model were used in some studies[18, 23, 28, 30, 32, 33]. Not all studies undertook adjustment for variables when determining mortality rates. Despite collecting basic demographic data on cases, Charlson co-morbidity data and exposure to recent medication such as antimicrobials, the study undertaken by Chung et al.[20] did not adjust for any variables in the calculation of hospital mortality and all-cause mortality.

In addition to studies comparing survivors and non-survivors, or those with and without CDI, some studies examined mortality by age group[17, 31, 34]. In the retrospective cohort study undertaken by McGowan et al.[31], age groups were divided into decades, with those under the age of 40 years grouped together. A significant difference in 30-day mortality was found for the three groups aged over 60 years, using <60 yeardsas the reference group (X² = 65.82; df: 2; P < 0.05). The study did not examine the effect of co-morbidity on outcome[31]. Similar to the work of McGowan et al.[31], a prospective multi-centred study of 1,430 participants demonstrated that age-specific mortality increased sharply after 60 years of age[17]. These two studies demonstrate the importance of collecting and analysing age-related data.

Some studies presented data that compared groups based on the causative strain of C.difficile[24, 30, 31]. The retrospective cohort study undertaken by Labbe et al.[30] found that patients affected by C.difficile ribotype 027 were twice as likely to die within 30 days of diagnosis than patients infected with other ribotypes. Similar results were demonstrated in a prospective case-controlled study comparing persons with CDI caused by ribotype 027, 106 and all other ribotypes[24].