In this editorial, we argue for a shift of attention in the field of mental health epidemiology in conflict-affected settings to include research that explicitly advances the development of mental health prevention and promotion interventions. We purposively use the term mental health epidemiology here as opposed to psychiatric epidemiology, to emphasize our focus on mental health as more than the absence of psychological symptoms – following the definition of health in the World Health Organization's constitution as ‘a state of complete physical, mental and social well-being and not merely the absence of disease’ (World Health Assembly, 2006).

Our focus here is on mental health epidemiology in settings of armed conflict. The Uppsala Conflict Data Program recorded 37 armed conflicts in 2011, defined as a minimum of 25-battle related deaths per year (Themner & Wallensteen, Reference Themner and Wallensteen2012). In their classification, the armed conflict needs to include at least one government, but it is acknowledged that armed conflicts have become more fluid and fragmented, making them difficult to categorize (Machel, Reference Machel2009). Current armed conflicts predominantly occur in low- and middle-income countries (LMIC), particularly countries in Africa (41%), Asia (35%) and the Middle East (16%). The majority of the armed conflicts recorded in 2011 were intra-state, that is, they took place between an internal opposition group and a government (Themner & Wallensteen, Reference Themner and Wallensteen2012). Owing to their changing nature, contemporary armed conflicts disproportionately affect civilians: they are associated with structural human rights violations, including sexual violence, disappearances and torture; displacement; disruption of food supplies; systematic destruction of service and economic infrastructure; and disturbance of community support systems (Pedersen, Reference Pedersen2002). Unfortunately, for many people in areas of armed conflict, conflict-related violence does not represent a sudden rupture in a life previously free of adversity. Rather, conflict-related violence often takes place against a pre-existing background of poverty and systematic marginalization. This creates the potential for vicious cycles in which armed conflicts exacerbate existing inequalities and vice versa (Tol et al., Reference Tol, Kohrt, Jordans, Thapa, Pettigrew, Upadhaya and De Jong2010a).

In settings of armed conflict, mental health epidemiology can have important functions for the promotion of mental health, and the prevention and treatment of mental disorders. We consider three key roles.

First, descriptive epidemiology may add value by detailing the prevalence rates of mental disorders. Such information may be useful for advocating for the necessity of a mental health response, defining priorities and decisions on allocation of resources. The usefulness of descriptive epidemiology with conflict-affected populations has, however, been seriously constrained by a number of factors, both substantive and methodological. An important methodological constraint has been the common use of symptom checklists that have not been validated for use in the population in which they are used (Kohrt et al., Reference Kohrt, Jordans, Tol, Luitel, Maharjan and Upadhaya2011). Using cut-off rates to establish rates of disorder, without knowledge on how this threshold compares with a gold standard (e.g., a psychiatric diagnostic interview), provides questionable prevalence rates. Without such validation, it is easy to mistake normal distress of populations in adverse situations with mental disorder (Rodin & Van Ommeren, Reference Rodin and van Ommeren2009). For example, the cut-off rates for the Hopkins Symptom Checklist and the Self Reporting Questionnaire had to be significantly adjusted for use in Afghanistan (Ventevogel et al., Reference Ventevogel, De Vries, Scholte, Shinwari, Faiz, Nassery, Van Den Brink and Olff2007).

Criterion validation is obviously a crucial task in ensuring useful descriptive epidemiology. However, criterion validation is enormously challenging in areas where the gold standard psychiatric diagnostic interview is not possible, e.g., because there are no trained psychiatrists or clinical psychologists available to conduct interviews. Criterion validation can also be challenged by the lack of a gold standard interview, e.g., in the case of intermittent explosive disorder (Silove et al., Reference Silove, Brooks, Bateman, Steel, Hewage, Rodger and Soosay2009; Al-Hamzawi et al., Reference Al-Hamzawi, Al-Diwan, Al-Hasnawi, Taib, Chatterji, Hwang, Kessler and Mclaughlin2012) or when assessing prevalence of idioms of distress not described in international classification systems. Alternatives are to rely on paraprofessionals or consult with key informants about the absence or presence of disorder (Bolton, Reference Bolton2001; Kohrt et al., Reference Kohrt, Jordans, Tol, Luitel, Maharjan and Upadhaya2011), and to ensure that at least disability associated with symptoms is assessed. Even with these methods, distress and disorder will remain difficult to disentangle, and researchers should be mindful to avoid medicalizing normal distress arising from conditions of adversity.

A related issue has been the very large variance in prevalence rates found in epidemiological studies: 0–99% for prevalence of post-traumatic stress disorder (PTSD) in 181 studies with conflict-affected populations (Steel et al., Reference Steel, Chey, Silove, Marnane, Bryant and van Ommeren2009). Much of the variance in these prevalence rates could be explained by methodological factors. For example, the unadjusted weighted prevalence rate for PTSD was 30.6%. However, this rate was 15.7% in studies with sample sizes greater than 1000; 26.6% in studies with representative samples; and 24.6% in studies that used diagnostic interviews. Moreover, time since hostilities was a significant predictor of lower rates of PTSD, suggesting reduction in rates over time and the importance of longitudinal studies (Steel et al., Reference Steel, Chey, Silove, Marnane, Bryant and van Ommeren2009). Other methodological issues that may explain variance in prevalence rates concern the quality of translation of standardized instruments and the use of incorrect diagnostic criteria (Rodin & Van Ommeren, Reference Rodin and van Ommeren2009).

Given these methodological issues, we question the utility of a currently common approach to descriptive epidemiology in conflict-affected populations, i.e., the undertaking of small studies with non-validated symptom checklists for PTSD and depression in convenience samples. These studies may have an important role to play in advocating for mental health programmes. However, such advocacy could also be achieved through different means, including generalizing from the existing knowledge on prevalence rates of mental disorders as established by robust studies (Charlson et al., Reference Charlson, Steel, Degenhardt, Chey, Silove, Marnane and Whiteford2012); mixed methods participatory approaches that document perspectives on needs and resources as expressed by conflict-affected populations; and assessing the prevalence of individual symptoms without making claims regarding the prevalence of mental disorder (e.g., reporting that a certain percentage of people experienced extreme hopelessness, anxiety, or anger) – an approach taken by recent guidelines on assessment in humanitarian settings (WHO & UNHCR, 2012).

Overall, an essential ethical principle should be to conduct research that minimizes the research burden in a population under distress while maximizing benefits in accruing vital knowledge that will guide priorities and approaches to intervention. A balance may be achieved by coordinating research with the existing health and community actors (to avoid duplicating efforts) and conducting thorough desk reviews to identify what is already known (for an example of such a review, see Kirmayer, Reference Kirmayer2010). From an ethical perspective, collecting new data in a separate study may represent an unreasonable burden if the same information is already available. Research burden can also be diminished by integrating questions on mental health in ongoing needs assessments by humanitarian actors. An additional ethical consideration concerns timeliness of information, as needs on the ground change quickly in acute conflict-affected settings. If descriptive epidemiology is to aid in the development of interventions, it is crucial that the results are provided in a time-frame that allows a response before needs have evolved (WHO & UNHCR, 2012). Useful descriptive epidemiology requires continuous reflection on the balance between the excellence (e.g., the certainty that all aspects of the problem under study have been fully investigated) and relevance of research (e.g., the usefulness of findings to policy and practice) (Tol et al., Reference Tol, Patel, Tomlinson, Baingana, Galappatti, Silove, Sondorp, van Ommeren, Wessells and Panter-Brick2012).

On a more conceptual level, a debate has focused on the application of psychiatric diagnostic categories from international classification systems in conflict-affected populations. Scholars have debated whether these psychiatric categories – originating in socio-cultural settings different from settings in which most armed conflicts occur – are relevant to these populations. Particularly the importance of PTSD for conflict-affected populations remains debated (Weiss et al., Reference Weiss, Saraceno, Saxena and van Ommeren2003). While the controversy regarding PTSD will not likely be resolved anytime soon, we would like to point to the importance of rapid ethnographic approaches in planning descriptive mental health epidemiology with conflict-affected populations (de Jong and van Ommeren, Reference De Jong and van Ommeren2002; Bolton et al., Reference Bolton, Tol and Bass2009). Techniques such as free listing and pile sorting, semi-structured interviews with the affected population, key informant interviews, group interviews and participant observation can aid in the selection and development of instruments that measure mental health concepts with local relevance and validity (Miller et al., Reference Miller, Kulkarni and Kushner2006; Tol et al., Reference Tol, Reis, Susanty and De Jong2010b). Such techniques can also be employed in developing measures of functional impairment (Bolton & Tang, Reference Bolton and Tang2002; Tol et al., Reference Tol, Komproe, Jordans, Susanty and De Jong2011b), in order to establish which symptoms or symptom groups are most strongly associated with impairment in daily activities of conflict-affected populations (Tol et al., Reference Tol, Komproe, Thapa, Jordans, Sharma and De Jong2007; Silove et al., Reference Silove, Bateman, Brooks, Fonseca, Steel, Rodger, Soosay, Fox, Patel and Bauman2008). If these techniques are implemented in collaboration with national universities, relevant governmental agencies and non-governmental organizations, they may contribute to mental health research capacity building, and ultimately contribute to the development of a sustainable mental health infrastructure in conflict-affected LMIC (Silove et al., Reference Silove, Rees, Tam, Liddell and Zwi2011). An important challenge here lies in how to connect knowledge of local experience of mental health on the one hand with the need for international comparison and exchange of knowledge on the other hand. One way of addressing this problem may consist of local adaptation of standardized measures using the same mixed methods procedures across socio-cultural settings: content of the measures may then vary, but overall comparison remains possible.

In addition, we would like to stress the importance of attention to mental health beyond PTSD, including neuropsychiatric disorders such as psychosis, substance use disorders and dissociative disorders. For example, an epidemiological study in conflict-affected Timor–Leste found high rates of psychotic-like symptoms (predicted by traumatic events) which were strongly associated with disability (Silove et al., Reference Silove, Bateman, Brooks, Fonseca, Steel, Rodger, Soosay, Fox, Patel and Bauman2008; Soosay et al., Reference Soosay, Silove, Bateman-Steel, Steel, Bebbington, Jones, Chey, Ivancic and Marnane2012). Similarly, rapid assessments in a variety of refugee populations have identified alcohol and substance use disorders as an important mental health problems (Ezard et al., Reference Ezard, Oppenheimer, Burton, Schilperoord, Macdonald, Adelekan, Sakarati and van Ommeren2011). Very little research has focused on dissociative disorders in conflict-affected settings, even though they have been shown to be associated with exposure to traumatic events and may be a common idiom of distress in conflict-affected populations (van Duijl et al., Reference Van Duijl, Nijenhuis, Komproe, Gernaat and De Jong2010).

Second, analytic epidemiology may contribute through the identification of risk and protective factors for mental health outcomes. A recent study aimed at identifying research priorities in humanitarian settings for the next decade placed this area of epidemiology at the top of the priority list. Overall, strong consensus was found in this study for a research agenda that generates practical knowledge that could be translated to immediate tangible benefits for programming in humanitarian settings, rather than addressing the key debates that have dominated the academic literature so far. The top 10 of most highly prioritized research questions focused on problem analysis: ‘What are the stressors faced by populations in humanitarian settings?’ (#1); ‘How do affected populations themselves describe and perceive mental health and psychosocial problems in humanitarian settings?’ (#3); ‘What are the major protective factors (including individual [e.g., coping, hope] and contextual [e.g., justice mechanisms, religious practices]) for mental health and psychosocial problems in humanitarian settings?’ (#7) and ‘To what extent do current mental health and psychosocial supports address locally perceived needs?’ (#9) (Tol et al., Reference Tol, Patel, Tomlinson, Baingana, Galappatti, Panter-Brick, Silove, Sondorp, Wessells and van Ommeren2011c, Reference Tol, Patel, Tomlinson, Baingana, Galappatti, Silove, Sondorp, van Ommeren, Wessells and Panter-Brick2012).

With regard to risk factors, the existing research has confirmed that past exposure to conflict-related events (particularly torture) is an important predictor of worse mental health outcomes (Steel et al., Reference Steel, Chey, Silove, Marnane, Bryant and van Ommeren2009). In addition to exposure to traumatic events, new research is confirming that exposure to current adversity (e.g., poverty, lack of access to basic services) is highly relevant to mental health in conflict-affected populations (Brooks et al., Reference Brooks, Silove, Steel, Steel and Rees2011). For example, a recent study found that current perceived needs mediated the relation between traumatic events and psychological distress in both Iraqi and Bhutanese refugees (Jordans et al., Reference Jordans, Semrau, Thornicroft and van Ommeren2012). Similar findings have been reported with refugees in Darfur (Rasmussen et al., Reference Rasmussen, Nguyen, Wilkinson, Vundla, Raghavan, Miller and Keller2010) and Sri Lankan youth (Miller et al., Reference Miller, Fernando and Berger2009). In children and adolescents affected by armed conflict, longitudinal research has shown the importance of post-conflict social and economic hardship (Betancourt et al., Reference Betancourt, Brennan, Rubin-Smith, Fitzmaurice and Gilman2010), as well as family violence (Panter-Brick et al., Reference Panter-Brick, Goodman, Tol and Eggerman2011) as risk factors for psychological symptoms over time. Taken together, these research findings suggest that epidemiology in conflict-affected settings should be inclusive of the determinants of mental health that have been identified in LMIC more generally, i.e., poverty, marginalization and diverse forms of violence (Patel et al., Reference Patel, Araya, De Lima, Ludermir and Todd1999; Rees et al., Reference Rees, Silove, Chey, Ivancic, Steel, Creamer, Teesson, Bryant, Mcfarlane, Mills, Slade, Carragher, O'Donnell and Forbes2011).

The study of protective factors has been a growing area of interest, particularly in the area of resilience, that is, good mental health outcomes despite exposure to adversity. A recent systematic review of studies in LMIC on promotive and protective factors in children affected by armed conflict identified 53 studies, 38 of which were quantitative and 15 of which applied qualitative or mixed methods. This review found important methodological shortcomings in this area of research: most studies were cross-sectional in nature; few studies measured positive outcomes in addition to psychological symptoms; and more consistent effort to measure predictors and outcomes in culturally sensitive ways is necessary (Tol et al., Reference Tol, Song and Jordans2013). The review also identified only two studies focused on the early childhood period, even though this area understandably is regarded as the most cost-effective period for preventive efforts (Heckman, Reference Heckman2006; Shonkoff & Garner, Reference Shonkoff and Garner2012). Authors of similar reviews have all emphasized that an important future research direction among conflict-affected children and adolescents concerns contextually sensitive longitudinal epidemiological research focused on identifying modifiable risk and protective factors at multiple levels of the social ecology (e.g., individual-, family-, school-, community-levels) (Reed et al., Reference Reed, Fazel, Jones, Panter-Brick and Stein2012; Betancourt et al., Reference Betancourt, Borisova, Williams, Meyers-Ohki, Rubin-Smith, Annan and Kohrt2013). Such a multi-level model research could, for instance, focus on family-level variables by sampling family units rather than individuals within households (Farhood et al., Reference Farhood, Zurayk, Chaya, Saadeh, Meshefedjian and Sidani1993), which is the current norm.

The importance of heeding socio-cultural context is reinforced by findings on the complexity of resilience. Determinants of mental health have been shown to have different impacts depending on the phase of conflict, gender and developmental stage. Protective factors in one setting may be risk factors in other settings. For example, political affiliation appeared protective among Nepali former child soldiers (Kohrt et al., Reference Kohrt, Jordans, Tol, Perera, Karki, Koirala and Upadhaya2010), but the reverse was observed in Bosnian adolescents (Jones, Reference Jones2002). Such complexity warns against pre-packaged resilience promotion approaches in conflict-affected populations, but rather suggests the need for careful assessment of risk – and protective factors in new settings to inform development of mental health prevention and promotion interventions that are tailored to context (Tol et al., Reference Tol, Song and Jordans2013).

This links to the third area where epidemiology may contribute to mental health of conflict-affected populations; that is intervention research. The current emphasis in global mental health research in general is on the scaling up of evidence-based treatments for mental disorders, by making them available in primary care settings (World Health Organization, 2010; Patel, Reference Patel2011). Although crucial, there is also consensus that a key ‘grand challenge’ rests in advancing prevention and implementation of early interventions for mental disorders (Collins et al., Reference Collins, Patel, Joestl, March, Insel, Daar, Anderson, Dhansay, Phillips, Shurin, Walport, Ewart, Savill, Bordin, Costello, Durkin, Fairburn, Glass, Hall, Huang, Hyman, Jamison, Kaaya, Kapur, Kleinman, Ogunniyi, Otero-Ojeda, Poo, Ravindranath, Sahakian, Saxena, Singer and Stein2011). A systematic review of rigorous evaluations of mental health and psychosocial support interventions in humanitarian settings found a large gap between research and practice in this regard. Whereas a survey of practice showed the popularity of preventive approaches (e.g., community-based social supports, structured social activities including child-friendly spaces, raising awareness and psycho-education), the most rigorous evaluations had focused on specialized interventions, such as psychological therapies for PTSD (Tol et al., Reference Tol, Barbui, Galappatti, Silove, Betancourt, Souza, Golaz and van Ommeren2011a).

We realize that pursuing a prevention agenda is easier said than done. A key challenge in pursuing such an agenda is that many of the systemic determinants of mental health are highly entrenched (e.g., systematic marginalization of ethnic or religious groups, violence against women) and outside the influence of mental health practitioners. Nevertheless, it does not seem sensible to treat the mental health consequences of living in areas of armed conflict, without attention to what causes mental health concerns – and this is where epidemiology has an essential role to play. Pursuing preventive efforts will require mental health professionals and researchers to step outside the confines of clinics and disciplinary confines and engage with advocacy groups, community leaders, policy makers and other key stakeholders that affect social change. It will also require the development of an additional skill-set among mental health professionals in integrating their concerns in for example human rights protection efforts, social welfare and the education system. And it demands from mental health researchers that they collaborate with a wide variety of specialists, e.g., in the fields of agriculture, nutrition, poverty alleviation, political science and human rights, to have their research concerns addressed in multi-disciplinary efforts. Throughout this process, it is crucial that mental health practitioners and researchers first listen to the needs and possibilities for social action as expressed by these stakeholders, in order to avoid the uni-directional imposition of cultural values from the outside.

A particularly attractive research strategy concerns preventive interventions in early childhood, given evidence that stress during that developmental period has adverse consequences across the life course. As brain development proceeds sequentially, with later development building on earlier achievements, prevention efforts in this period can have multiplicative impacts across the life span (Shonkoff & Garner, Reference Shonkoff and Garner2012). This could be achieved through, for example, treatment of maternal mental disorders, developing parenting skills and enhancing the quality of the home environment, including the prevention of domestic violence. Although evidence-based strategies exist to improve early childhood outcomes in LMICs (e.g., breast-feeding, parenting and pre-school programmes), there is a need for the development of integrated interventions that can address the multiple stressors that children are commonly exposed to (Engle et al., Reference Engle, Fernald, Alderman, Behrman, O’gara, Yousafzai, De Mello, Hidrobo, Ulkuer, Ertem and Iltus2011). More evidence is required to determine whether the treatment of maternal mental disorders will improve developmental outcomes, or how integrated nutrition and early stimulation can best be applied in conflict-affected settings (Morris et al., Reference Morris, Jones, Berrino, Jordans, Okema and Crow2012).