Search strategy

This systematic review was carried out in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement (S1 File) [34]. According to the PRISMA guidelines, a flow diagram was used throughout this study to illustrate the step-by-step study selection process, and rationale was provided for those studies that were excluded [34].

Three different search strategies were employed- database search, grey literature, and backward snowballing. Namely, PubMed, Web of Science, and Embase were systematically searched for potentially eligible studies during week 25, 2023. All studies that quantified DALYs were included, regardless of whether their primary focus was on assessing the BoD or other aspects (e.g., cost-effectiveness and cost of illness studies). The search included the use of the following terms: Campylobacter, Campylobacteraceae, burden of disease, disability-adjusted life year, years of life lost, years lived with disability, cost effectiveness and cost of illness. The search strategy employed Boolean operators and Medical Subject Headings (S2 File). Additionally, a grey literature search was conducted using Google Scholar and OAIster. Moreover, the search was supplemented using the backward snowballing search strategy, involving the manual screening of reference lists from included publications to identify additional eligible studies missed during the initial search.

Eligibility criteria

Peer-reviewed journal articles and grey literature published between 1990 and 2023 were considered for inclusion. Studies published prior to 1990 were excluded due to the introduction of the DALY metric in the early 1990s [29]. All studies utilizing the BoD methodology to calculate DALYs, regardless of their primary focus, were included. Only studies assessing Campylobacter BoD in human populations were eligible. There were no language or study type restrictions; however, abstracts, conference proceedings, reviews, letters to editors, and studies without full-text availability were excluded from the search.

Study selection and screening

Titles and abstracts of publications retrieved from preselected databases (PubMed, Web of Science, and Embase), as well as grey literature sources, were organised and managed using the Zotero reference manager, with duplicates identified and removed. The publications were then imported into Rayyan.ai [35] to further manage duplicates and ensure accuracy. Duplicates were identified based on author(s), publication year, and title, with abstracts compared if needed. Subsequently, articles underwent eligibility screening, including evaluation based on title, abstract, and full text, conducted by an independent reviewer (MT), with CDB reviewing. A sample of articles was pilot screened by the study supervisors (CDB and ZK) to assess conformity. Any uncertainties or discrepancies were discussed (MT, CDB, ZK), and documentation of reasons for article rejection was provided in a PRISMA flow diagram (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow diagram illustrating the literature search and study selection, which has been adapted from the PRISMA 2020 guidelines for systematic reviews [

39].

https://doi.org/10.1371/journal.pntd.0012681.g001

Data abstraction

Data abstraction for all eligible studies was conducted by an independent reviewer (MT) through a full text assessment. This process utilized a standardized data abstraction form (S3 File), which was piloted with study supervisors and adapted from systematic literature reviews by Charalampous et al., [36] and Di Bari et al., [28] to align with the objectives of this review. The data abstraction form facilitated the collection of study characteristics, including data sources for epidemiological data, DALYs results, and methodological considerations for calculating YLDs and YLLs. YLDs in DALY calculations can be derived from either an incidence- or prevalence-based approach. The incidence perspective, used by BCoDE and WHO/FERG, projects future BoD based on current exposures and focuses on pathogen-related outcomes, including sequelae [28,37,38]. On the other hand, the prevalence-based approach, employed by GBD, captures the populations health status at a specific point in time [25,28,37]. GBD also adopts an outcome-based approach to allocate BoD across health conditions and estimate major infectious disease-related outcomes [38]. Following data abstraction, a second reviewer (CDB) reviewed the finalized form. Any revisions or discrepancies were resolved through discussion and consensus.

At present, there is a lack of established tools specifically designed to evaluate the quality of studies estimating DALYs. Consequently, as the primary objective of this study is to evaluate the methodological approaches and assumptions employed in Campylobacter BoD studies, a comprehensive assessment of overall publication quality was conducted as part of this systematic review. This renders the use of specific quality appraisal tools unnecessary. Furthermore, a meta-analysis was not conducted in this review. This decision was primarily driven by the study’s focus on exploring various methodological choices in Campylobacter BoD studies, rather than estimating pooled DALY rates.

Literature search

Fig 1 illustrates a flow diagram detailing the selection process for studies in this systematic review. A total of 292 publications were initially identified through specified databases, grey literature sources, and citation searches. Following the removal of duplicates (n=16), 50 full-text articles were assessed for eligibility. Subsequently, 28 articles were excluded as they utilized health metrics other than YLD, YLL, or DALYs in the BoD assessment. Furthermore, an additional study focusing on GBD data was identified. Finally, data were extracted from 23 BoD studies for analysis.

Study characteristics

Of the 23 eligible BoD studies included for review, 19 were performed at a single-country scale, while 4 were multi-country analyses, as depicted in Fig 2. Within the multi-country studies, Mangen et al., [40] and Cassini et al., [31] undertook regional assessments across countries within the EU/EEA. Additionally, 2 studies addressed the burden of campylobacteriosis on a global level, with one conducted by Kirk et al., [15] under the WHO/FERG study. They utilized a pathogen-based approach to calculate DALYs, considering estimated cases, sequelae, and fatalities. Conversely, Vos et al., [41] summarised GBD methodologies, presenting results from an outcome-perspective. Although GBD provides specific estimates for campylobacteriosis, these are often categorised under broader infectious disease groups rather than as a discrete entity. For instance, Campylobacter is integrated within overall estimates for diarrhoeal diseases. The study by Vos et. al., [41] did not mention campylobacteriosis independently but included it within the collective analysis of diarrhoeal diseases. The appendix of the article details that during data input, diarrhoeal aetiologies, including Campylobacter, were extracted from scientific literature reporting the proportion of diarrhoea cases testing positive for each pathogen [41]. This methodology enables comprehensive estimates of diarrhoeal disease burden but obscures discrete data on specific pathogens like Campylobacter.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Existing Campylobacter BoD studies compiled from published literature based on year of publication.

https://doi.org/10.1371/journal.pntd.0012681.g002

Over the 2000-2023 period, the largest number of national-level studies originated from the Netherlands (n=7) [42–48], followed by Denmark (n=2) [49,50]. Additional individual studies spanned various countries, including the United States [26], Belgium [51], Taiwan [52], Australia [53], Switzerland [54], Japan [55], Sweden [56], Greece [57], Germany [58], with the most recent publication stemming from Rwanda in 2023 [59].

Methodological choices

Approaches for estimating YLL: Choice of life expectancy table

Most studies (n=21) included YLL due to campylobacteriosis. However, one study by Haagsma et al., [45] explicitly omitted the YLL component in the DALY calculation, as they did not associate IBS with increased mortality. Among the single-country independent BoD studies (n=8), national life tables were predominantly utilized to compute YLLs [26,47,50,52–54,56,58]. Moreover, four of the independent single-country studies employed the aspirational global Coale-Demeny (West level 25 or 26) model life tables [42,44,48].

Notably, GBD data for campylobacteriosis does not exist as a standalone disease; rather, it is classified as an aetiology of overall infectious diseases. Although Vos et al., [41] utilized GBD life tables in their analysis, it is important to clarify that these tables do not provide specific YLL data for campylobacteriosis. The remaining single- or multi-country studies employed various life expectancy tables, including aspirational model life tables akin to those used in the GBD study [31,46,55], standard life model tables from WHO [15,49,59], or a combination of the aforementioned tables [ 40,51,57].

Social weighting: Age weighting and time discounting

In total, five studies applied either age weighting or time discounting rates, but none of the studies implemented both social weighting parameters (Table 1). Among the studies employing these variables, time discounting was more prevalent, as evidenced by four studies [40,42,44,47]. Notably, the discounting rates varied significantly across studies. For instance, Mangen et al., [42] and Havelaar et al., [44] presented disease burden data both undiscounted and discounted at a rate of 1.5%. In contrast, an earlier study by Havelaar et al., [47] in 2005 presented disease burden data using different discounting rates, including undiscounted data and data discounted at a rate of 4%. Similarly, the regional study by Mangen et al., [40], which estimated the burden of campylobacteriosis across six EU/EEA countries, presented undiscounted data alongside different discounting rates, including 3%, 6%, and country-specific rates. Although Gibney et al., [53] did not utilize time discounting, they did apply age weighting rates for acute GE to the reference population.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summarises these methodological differences, as extracted from the studies included in this review (for the full data abstraction form, refer to S3 File).

https://doi.org/10.1371/journal.pntd.0012681.t001

Approaches for estimating YLD: Incidence- vs. Prevalence-based approach

Most studies (n=22) utilised the incidence-based approach to compute YLDs associated with campylobacteriosis [15,26,31,40,42–59], while the GBD study by Vos et al., [41] employed the prevalence-based approach. Some studies addressed comorbidity to enhance burden estimates. For instance, Mangen et al., [47] adjusted for comorbidity by applying a correction factor of 1.86 across all ages. This factor was derived from a Danish study [60] that estimated the relative risk of mortality for patients with laboratory-confirmed campylobacteriosis within the first year after symptom onset. The estimation utilized an index based on relative mortality rates from different diagnostic groups. This adjustment revealed a relative risk of 1.86 (90% CI 1.56–2.20), accounting for other health conditions. Similarly, Kirk et al., [15] included a comorbidity adjustment in their global assessment of 22 diseases, however, their study did not specifically address the burden associated with campylobacteriosis.

Pathogen- vs. Outcome-based approach

The majority of studies (91%) utilised the pathogen-based approach, many of which also employed the incidence-based approach. Notably, two recent studies by Lai et al., [52] and van Lier et al., [43] reported adopting the WHO/FERG and BCoDE protocols, respectively. It is important to note that the WHO/FERG [37] study was not directly included in the data abstraction process because it did not provide specific BoD estimates. Instead, it focused on demonstrating how the methodology can be used to convert epidemiological data into BoD estimates [37]. In contrast, two studies followed an outcome-based approach; one of these was the GBD study [41]. Diverging from the GBD methodology, Haagsma et al., [45] applied both an incidence and outcome-based approach, specifically examining the disease burden of post-infectious IBS following campylobacteriosis by calculating the number of IBS cases attributable to Campylobacter infection.

Health states

Fig 3 presents the perceived outcome tree for campylobacteriosis, adapted from Mangen et al., [48] and Lackner et al., [58]. However, there was notable variation in the health states associated with campylobacteriosis, leading to the application of seven different disease models across studies (S4 File). Among studies calculating DALYs for campylobacteriosis, seven studies reported various health states including GE, IBS, IBD, ReA, GBS, and mortality (disease model 1) [26,42–44,54,57,58]. Nonetheless, Martins et al., [54] considered cases of GE in conjunction with diarrhoea. Conversely, some studies [15,49,50,52,59] treated diarrhoea as a distinct health state from GE, including individual-level national (n=4) and global studies (n=1). While most studies reported all health states and their sequelae, some omitted IBD [31,40,48–50,53] or IBS [47,55], or both [46,51,56]. Haagsma et al., [45] exclusively considered post-infectious IBS following Campylobacter infection and did not compute YLLs since IBS was not associated with increased mortality (disease model 7).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Illustration of the perceived outcome tree for Campylobacter infection in humans (adapted from Mangen et al., [

48] and Lackner et al., [58]).

https://doi.org/10.1371/journal.pntd.0012681.g003

Choice of disability weights

Based on the 23 studies, the GBD DW was the most commonly used (39%) [15,40,41,46,48–50,54,59]. Among these, the DWs largely stemmed from the GBD 2010 study [15,54,59]. Notably, Martins et al., [54] and Kirk et al.,[15] computed DWs based on the GBD 2010 framework, albeit with variances in the classification of health states. For instance, Kirk et al., [15] categorized certain clinical outcomes for campylobacteriosis into acute Campylobacter diarrhoea (mild), acute Campylobacter diarrhoea (moderate), and acute Campylobacter diarrhoea (severe), assigning DWs of 0.061, 0.202, and 0.281, respectively. Similarly, Martins et al., [54] adopted these DWs as proposed by GBD 2010 but considered GE alongside diarrhoea, analysing them as mild, moderate, or severe, following the same pattern. In contrast, the European DW set, recommended and implemented by the BCoDE study, conducted by Cassini et al., [31] were also included in three other studies. These national-level studies conducted in Japan [55], Belgium [51], and Germany [58], were published in 2015, 2017, and 2019 respectively, after the elicitation of the European DWs in 2015. Several studies opted for alternative DWs, primarily the Dutch DWs [26,42,44,47,56]. The majority of studies (n=3) employing the Dutch DWs were conducted at a national level within the Netherlands [42,44,47]. Conversely, two national studies, in Sweden [56] and the United States [26] also used the Dutch DWs. The remaining studies either used specific DWs for sequelae [45], or amalgamated existing DW sets [43,52,53,57].

Duration

Data concerning the duration of disease were obtained from various sources, including existing literature, data analysis, and expert opinion (Table 1). However, one study by Scallan et al., [26] lacked clarity regarding this process. Additionally, as campylobacteriosis is not considered a standalone cause of morbidity and mortality in the GBD study [41], data extraction was not feasible. Among the studies reviewed, the majority (n=13) relied solely on existing literature to define health states and duration data [15,31,40,44–48,51,53,55,57,58]. However, significant variations emerged regarding the duration of sequelae associated with campylobacteriosis [44–46,48,53]. For example, Gibney et al., [53] reported durations of 0.6 years, 5 years, and lifelong for ReA, IBS, and GBS, respectively. Mangen et al., [48] similarly concluded that IBS could persist for 5 years as well as severe cases of GBS lasting for an individual’s remaining life expectancy. They did, however, state that GE could only persevere for 0.017 (years) [48]. In contrast, Havelaar et al., [46] suggested a shorter duration for GE (5.08 days) and ReA (6 weeks), but surprisingly, did not consider GBS as a lifelong disability, estimating its duration at 1 year. Conversely, another study by Havelaar et al., [44] assumed IBD to be lifelong, while Haagsma et al., [45] indicated that symptoms of post-infectious IBS could last for 5 years.

Alternatively, two independent studies in Denmark estimated campylobacteriosis duration through data analysis [49,50]. Pires et al., [49] based their findings on life expectancy data provided by Denmark Statistics. They determined that symptoms of diarrhoea typically lasted for 3 days, while ReA persisted for 222 days. In contrast, both IBS and GBS were identified as severe, long-lasting conditions, with IBS potentially lasting for 1825 days and GBS extending for the duration of an individual’s remaining life expectancy [49]. Mangen et al., [42] based their duration data on expert elicitation. The remaining studies (n=5) employed a combination of methods, drawing from published literature, data analysis, and expert opinion [43,52,54,56,59]. Toljander et al., [56], for instance, synthesized duration data from literature, national notification systems, and hospitalization records, establishing that reported GE typically lasted for 8.37 days, ReA for 0.71 years, short-term (≤1 year) GBS for 0.685 years, and long-term (>1 year) GBS for 15.2 years.

Epidemiological data and data adjustments

Based on the included studies, several epidemiological data sources were utilized to calculate DALYs. The majority of studies (n=9) drew information from published literature [15,31,45,47,48,52,53,56,59]. Moreover, three studies combined literature findings with national surveillance data [15,47,56], one study relied on sentinel surveillance [48], and another cited the WHO/FERG database [59]. Notably, two studies incorporated data from multiple sources, with Gibney et al., [53] integrating national estimates from administrative data and additional data from OzFoodNet, while Lai et al., [52] gathered information from various Taiwanese authorities, national databases, and nationwide surveys. In contrast, five studies predominantly obtained incidence data from laboratory-confirmed cases [40,43,46,50,51]. Each of these studies enhanced their findings by incorporating supplementary data from various sources, including sentinel surveillance, statutory notification data, and reports from outbreaks, population-based surveys, and general practitioner surveillance [40,43,46,50,51]. Notably, the studies from Scallan et al., [26], Pires et al., [49], Mangen et al., [42], Havelaar et al., [44], and Gkogka et al., [57], sourced data from multiple inputs, including published literature, laboratory cases, active surveillance, and national medical registers. Furthermore, three studies explicitly utilized alternative data sources, such as those from the Federal Office of Public Health or local government statistics, registration records, and national patient surveys [54,55,58].

To mitigate underestimation of incidence, fourteen studies adopted data adjustment techniques. Five studies carried out data adjustments through a data-driven approach [26,46,52,55,57]. Lai et al., [52] adjusted for underestimation by selecting the highest estimate as the most accurate reflection of the actual incidence of diagnoses, while Gkogka et al., [57] employed a PERT distribution derived from previously published literature as multipliers to account for food attribution. Alternatively, others relied on expert opinions to manage underestimation [40,43,48,49,51,53,54]. Some of these studies [40,43,48,51] based their adjustments on guidelines outlined by Havelaar et al., [61]. The remaining studies also adopted an expert-driven approach, albeit with differing methodologies [49,53,54]. Gibney et al., [53] utilized regional data to estimate the number of notifiable cases of Campylobacter in Australia, complemented by case fatality ratios from a study on FBDs in the United States [26]. Similarly, Pires et al., [49] tailored the surveillance pyramid outlined by Haagsma et al., [62] to adjust for under-reporting and underdiagnosis of diseases. Only one study [50] combined both approaches, incorporating expert opinions and non-pathogen-specific and pathogen-specific parameters to adjust for under-reporting and underdiagnosis. Additionally, several studies (n=13) employed Monte Carlo simulation to model uncertainty [15,31,42–45,48,51,53,55,56,58,59].

Limitations

This review has several limitations. Firstly, a risk of bias assessment was not performed for the included studies, as it is not deemed suitable for evaluating BoD studies. Additionally, despite efforts to retrieve relevant literature from various databases and platforms, there is a possibility of overlooking certain grey literature sources. Notably, only a single study was identified from a LMIC, where the burden of campylobacteriosis may be higher but remains unaccounted for in this review. Furthermore, the database and grey literature searches were last conducted in June 2023, which may have resulted in the exclusion of more recent studies. Although this systematic review can be updated in the future as additional eligible studies become available, logistical constraints prevented us from updating the search for this review. Additionally, the focus on the GBD 2019 iteration, rather than including all GBD iterations, was a methodological decision, and the time frame of the search has been explicitly acknowledged to provide transparency regarding the study’s scope and methodology.