Search strategy

We identified guidelines for research ethics in disaster situations by a three-tiered search strategy: 1) searching two databases (PubMed and Google Scholar), 2) an Internet search (Google), and 3) a search of the references in the documents included from the first two searches. We used the following search terms: (guidelines AND “research ethics” AND (disaster OR emergency OR crisis)). Assessment of eligibility was limited to the first 200 hits retrieved in Google and to the first 250 hits in Google Scholar ordered by relevance in accordance with the methods used in numerous similar systematic reviews. Limits were not placed on the PubMed search.

The screening process is summarized in Fig. 1. At the first screening stage, one researcher reviewed the document titles. Only documents written in English or translated into English by the guideline developers were included. Titles clearly not related to the topic, as well as scientific and popular articles, books, presentations, and opinion pieces which were clearly not guidelines were excluded. This gave 110 documents which were further screened. The second eligibility screening was performed independently by two researchers. Each researcher evaluated the documents against the inclusion criteria and screened the document’s reference list for additional disaster research ethics guidelines. Independent results were compared between the two researchers. When discrepancies existed, a third researcher was involved to resolve any eligibility disagreements.

Inclusion/exclusion criteria

Qualitative analysis

We used the constant comparative method (CCM) for analysis of included guidelines [6, 7]. This method of qualitative analysis combines inductive category coding with a simultaneous comparison of all obtained units of meaning. A unit of meaning is defined as a part of the text (e.g., phrase, sentence, several sentences) that “must be understandable without additional information, except for knowledge of the researcher’s focus of inquiry” [7]. Open coding was applied as a first step in the coding process to identify units of meaning and to allow categories to emerge from the data. According to CCM, each new unit of meaning is “compared to all other units and subsequently grouped (categorized and coded) with similar units of meaning” [7]. In the process of open coding and comparison, initial categories were changed, merged and omitted when necessary. The second step involved axial coding to explore connections between categories and subcategories. Selective coding as a third step involved selecting the core themes. As a result of selective coding, we identified two core themes emerging from our analysis: research ethics review process and vulnerability.

Issuing organizations

The guidelines identified during the search showed a diversity in geography and types of issuing organizations. Seven out of 14 guidelines are applicable internationally, five are national level guidelines (two each from Canada and the US, and one from India), and two apply to a particular organization (Médecins Sans Frontières (MSF) or Doctors Without Borders) or institution (US Centers for Disease Control and Prevention). However, it should be taken into account that only guidelines written in English were included, which is a limitation of this study.

International organizations issuing guidelines included the World Health Organization (WHO) [8], MSF [9] and the International NGO Training and Research Centre [10]. Several national organizations issued guidelines, including the Indian Council of Medical Research [11] and the Canadian Medical Association [12]. Three guidelines arose from particular meetings or specific working groups including: a meeting entitled “Ethical Issues Pertaining to Research in the Aftermath of Disaster,” organized by the New York Academy of Medicine and the National Institute of Mental Health [13]; the Harvard Humanitarian Action Summit [14]; and the Working Group on Disaster Research and Ethics [15].

Content of the guidelines

Qualitative analysis of the 14 included guidelines revealed two core themes: vulnerability and REC review process. Much of the material addressed in the guidelines could be grouped under one of these themes. At the same time, the themes overlap and a clear distinction between the two is not possible. Within each of the two core themes, various categories were identified, and within each of these, further subcategories were identified. These are summarized in Additional file 2: Table S2 and Additional file 3: Table S3, and described in detail in the sections below.

Vulnerability as a concept

We identified three subcategories within this category: definitions of vulnerability (in four guidelines), reasons for vulnerability (in six guidelines), and gaps in the existing guidelines (in two guidelines).

Four out of 14 guidelines included a definition of vulnerability or analysed the concept of vulnerability. For example, Collogan et al. discussed different approaches to the definition of vulnerability and contrasted vulnerability as “a characteristic of the group” with “certain individual characteristics” [13]. The authors of this guideline criticized very broad definitions of vulnerability because these can be applied “to cover almost any person, group, or situation” and often serve to stereotype and disempower research participants [13].

Six guidelines mentioned and analysed the reasons for vulnerability in disaster settings. These arose from specific disaster situations and included, for example, political status and human rights abuses in refugee populations [16]; young and old age of research subjects, social vulnerability, physical injuries, and experience of violent and traumatic events leading to mental health problems in the aftermath of disaster [13]; psychological and physical consequences of a disaster, as well as poverty in pre-disaster settings and disempowerment in post-disaster settings [17], increased public risks and devastation [18] and substantial psychological stress in humanitarian settings [19].

Two guidelines issued in the early 2000s pointed to gaps in other general research ethics guidelines (such as the Declaration of Helsinki, Council for International Organizations of Medical Sciences (CIOMS) guidelines, etc.) and noted that these existing guidelines look at vulnerability very generally and do not address the special circumstances of disaster settings [17] or do not provide an applicable definition of vulnerability [13, 16].

Risks and burdens

This category covered different types of risks and burdens that research subjects might face during disaster research. The category included six subcategories: physical harm (in seven guidelines), re-traumatization (in five guidelines), manipulation (in two guidelines), exploitation (in eight guidelines), unrealistic expectations (in eight guidelines) and stigmatization (in two guidelines).

Various guidelines mentioned different sources of possible direct physical harm to research subjects, depending of the type of research study and the authors’ experience. Goodhand mentioned a possibility that research interventions might “affect the incentives system and structures driving violent conflict or impact upon the coping strategies and safety of communities” [10]. Leaning referred to situations where “sampling techniques may confer unintended negative attention or focus on particularly vulnerable subpopulations […] and make conditions worse” [16]. Allden et al. mentioned several examples of physical risks, e.g., the fact that “the presence of western researchers in itself, could present a danger in targeting of local civilians” [14]. The authors gave an example where “researchers gave great attention to ethics and staff behaviour only to learn subsequently that the research group’s driver, who stayed with the research staff in the village where research was being conducted, engaged in transactional sex with girls <18 years of age” [14]. Tansey et al. paid attention to physical risks posed by research involving novel interventions and suggested that, “Early detection of toxicities or other harmful effects of a research intervention could help to reduce research-related harms by leading to adjustments in protocol or, if necessary, termination of the intervention” [12]. Collogan et al. and Curry et al. provided the general term “physical risks” and did not elaborate on particular types of risks or reasons for them [13, 19].

Guidelines were more detailed in their analysis of different types of non-physical risks. The most often mentioned risks were exploitation of research subjects and unrealistic expectations, as well as manipulation. While some guidelines just mentioned exploitation in general, others gave more detailed examples, e.g., Schopper et al. mentioned the collection, export, and analysis of tissues as a potential source of “exploitation of communities from which tissues have been taken” [20]. Allden et al. paid attention to children as a vulnerable group “who could be targeted for recruitment by groups that could exploit them as a result of research and program activities” [14]. Sumathipala et al. expressed deep concerns regarding possible exploitation and referred specifically to “exploitation through undue inducement and compensation, and through an understandable confusion regarding the researchers’ objectives” [15]. This type of exploitation leads into the next subcategory: unrealistic expectations. This was analysed broadly, with the general advice that researchers should be “particularly vigilant in ensuring that prospective study participants do not confuse research procedures with clinical care and evaluation and thus fall prey to the so-called therapeutic misconception” [13] and to “take great care to inform potential research participants […] that some interventions to which they are subjected might be undertaken primarily for the benefit of the research” [21].

The risk of stigmatization was mentioned less often, being named in two guidelines. Allden et al. emphasized that vulnerable groups might be stigmatized after participation in research, but individual research participants might be stigmatized by researchers who lack cultural competencies and knowledge of specific socio-cultural contexts [14]. Sumathipala et al. paid attention to researchers’ duty to prevent stigmatization [15].

Another risk, mentioned in five guidelines, was re-traumatization of research participants. Goodhand explained that by involving subjects in a research study and by asking sensitive questions “researchers may inadvertently re-open wounds by probing into areas respondents may not wish to talk about” [10]. Therefore, RECs should assess the risk of re-traumatization [20]. Specific ways of addressing re-traumatisation were mentioned in some guidelines, such as Sumathipala et al. describing the campaign “Prevent Re-traumatisation of the Traumatised” as an example of good practice [15]. Some guidelines noted the importance of recognising that disaster settings offer “limited opportunities for therapeutic interventions to handle adverse psychological reactions” [17]. At the same time, some authors argued that “an individual noted to be upset during participation in research might not necessarily regret participation” [13].

Risk management

Taking into account the vulnerability of research participants leads to a need to manage risks. The risk management category included six subcategories: accountability and monitoring of research (in nine guidelines); avoiding over- or underestimation of risks (in three guidelines); the need for empirical evidence on risk (in three guidelines); providing psychological support to research subjects (in four guidelines); quality of informed consent (in six guidelines); and evaluation of power relationships between researchers and subjects (in two guidelines).

Accountability and monitoring of research was covered by eight guidelines. Authors of these guidelines, especially after 2008, emphasized that risks in disaster research can be diminished by monitoring and control [13] and mentioned corresponding values, e.g., accountability and transparency [8]. The most detailed description of monitoring was developed by Curry et al. [19]. However, most guidelines did not clearly describe what monitoring and control should include and what institutions should oversee the research studies. Direct oversight of the whole process of research implementation was mentioned in some guidelines, but seen as practically impossible [20]; however, close monitoring of particular parts of the process, e.g., informed consent procedures [12], was viewed as necessary. Some guidelines mentioned that new information arising during the research should be carefully monitored, e.g., protocol amendments, side effects, adverse effects and early stopping of a study. A different approach was to increase monitoring of specific types of research “where risk is high or uncertain” [12]. Various monitoring bodies were proposed, with some guidelines suggesting that monitoring could be done “by the central IRB or a separate data and safety monitoring board” [17].

Some guidelines noted that preliminary risk assessment should avoid both over- and underestimation of risks, e.g., overestimation by labelling all research participants as ‘vulnerable’ [21] or misusing the concept of ‘re-traumatization’ [13], and underestimation of risks by denying that research might add additional risks to those posed by the disaster [17]. The same guidelines dealing with estimation of risks mentioned the need for empirical evidence to evaluate risks posed by research [13, 17, 21].

One possible additional approach to risk management mentioned in four guidelines was ensuring psychological support to research subjects, including training of researchers and development of procedures to provide psychological support [21], as well as “explicit mechanisms available for timely referral of subjects in need of mental health consultation” [13]. These support mechanisms should be culturally and politically acceptable [14]. Other guidelines emphasized the need to identify available local services and to help research participants access these services when needed [17].

Another subcategory of risk management was the quality of informed consent, evaluated by some authors as one of the major ethical challenges in disaster research [20]. Examples of shortcomings in informed consent included “incomplete information given to the participants about objectives, risks, adverse effects, and planned house visits; information too detailed and complicated; formulation of the text biased to induce a positive answer; overestimation of the benefit for participants and community; and lack of procedures to ensure that the information provided is understood” [20]. To ensure higher quality informed consent, the guidelines included criteria usually mentioned by general research ethics guidelines, as well as criteria specific to disaster settings. These specific criteria emphasized the impact of vulnerability on research subjects and the safety of the setting where informed consent procedures take place [13]. Allden et al. suggested that researchers should “take consent at multiple times during the research process, including at the end of data collection” and “take consent from multiple agencies including community, parents, and partners as appropriate” [14].

An additional subcategory of risk management was mentioned in two guidelines: the evaluation of power relationships between researchers and subjects [10, 14]. Allden et al. stated that in disaster settings researchers hold more power than participants and therefore researchers should “be aware of power differentials between the researcher and respondent that may increase their likelihood of participation” [14].

Decisional capacity of research subjects

We identified three subcategories for this category: factors diminishing decisional capacity (in four guidelines); underestimation of decisional capacity (in two guidelines); and the need for a specific procedure for informed consent (in five guidelines). Two guidelines mentioned the traumatic experiences of research participants [13, 21] and one guideline referred to “inherent tensions and pressures” in situations of public emergencies [18] as a possible reason for impaired decision-making capacity. To take account of this, a specific procedure for informed consent was proposed that would address diminished decisional capacity of research participants in disaster settings. This specific procedure should involve “a time lag between an initial contact and eventual interview” [21], “language specific to the unique situation of protection of victims of disasters as well as their communities” [17], and use of oral instead of written consent in cases when participants decline to sign anything [20]. At the same time, authors of several guidelines warned about possible underestimation of decisional capacity of participants by stating that it would “be inaccurate and potentially stigmatizing to assume that all persons who have experienced terror or other disasters are decisionally impaired and unable to make choices for themselves” [13].