Populations affected by humanitarian emergencies such as those involving natural disasters, conflict or war are often exposed to a wide range of stressors, with elevated levels of distress and mental disorders commonly found in these settings. Reference de Jong, Komproe and van Ommeren1–Reference Steel, Chey, Silove, Marnane, Bryant and van Ommeren5 Historically, there has been a heavy emphasis placed within relevant research and mental health interventions on potentially traumatic events experienced during the emergency (for instance exposure to violence, torture, forced recruitment to fight, loss of home or loved ones, etc.), and the effects these have on emergency-affected people's mental health. Post-traumatic stress disorder (PTSD) has featured most prominently as the mental health outcome indicator of choice in evaluation studies. Reference de Jong, Komproe, van Ommeren, El Masri, Araya and Khaled6–9 Recent reviews and other studies have confirmed that experience of torture, violence, conflict and other potentially traumatic events are strongly associated with mental disorders, such as PTSD and depression. Reference de Jong, Komproe and van Ommeren1,Reference Steel, Chey, Silove, Marnane, Bryant and van Ommeren5,Reference de Jong, Komproe, van Ommeren, El Masri, Araya and Khaled6
In recent years, however, a debate has arisen over the relative value of trauma-focused perspectives in emergency settings. The World Health Organization (WHO) is one of several agencies which, along with others, have called for a refocusing of mental health interventions in humanitarian emergency settings towards a broader framework incorporating mental disorders such as PTSD. Reference de Jong, Komproe and van Ommeren1,2,9–Reference Weiss, Saraceno, Saxena and van Ommeren11 An integrative ‘mental health and psychosocial support’ approach has been advocated in emergency settings, incorporating a range of social (non-clinical) programmes as well as mental health interventions (including trauma-focused ones). 2,Reference Mollica, Lopes Cardozo, Osofsky, Raphael, Ager and Salama3,Reference van Ommeren, Saxena and Saraceno10,Reference Miller and Rasmussen12–Reference Jordans, Tol, Komproe, Susanty, Vallipuram and Ntamatumba14 The broader framework has a bearing on the assessment of needs in humanitarian settings, exemplified by the development of a new scale.
The development and initial validation of the Humanitarian Emergency Settings Perceived Needs Scale (HESPER) has been described in detail elsewhere. Reference Semrau, van Ommeren, Blagescu, Griekspoor, Howard and Jordans15 In short, the HESPER Scale was developed by the WHO and the Institute of Psychiatry at King's College London, to rapidly assess perceived needs of populations in humanitarian settings during conflict or other disasters in low- and middle-income countries. The scale, uniquely, provides a population-level assessment of the prevalence and distribution of needs as perceived by members of the population themselves. In addition to providing descriptive data to inform and monitor humanitarian response, the scale can help answer the following policy question. To what extent do daily stressors (i.e. current needs) experienced in the post-emergency environment (for instance poverty, forced displacement to camps, overcrowding, malnutrition, etc.) account for the impact of traumatic exposure on mental health (for example Miller & Rasmussen, Reference Miller and Rasmussen12,Reference Miller and Rasmussen16 Neuner, Reference Neuner17 Ager Reference Ager18 )? It is not clear how much of the variance in mental health outcome may be predicted by current needs or stressors in the potentially modifiable post-emergency recovery environment v. traumatic events that have already occurred previously. Recently, various models have been proposed that attempt to explain the interaction between these variables. One such model, proposed by Miller & Rasmussen Reference Miller and Rasmussen12,Reference Miller and Rasmussen16 purports that daily stressors mediate the relationship between traumatic experiences and mental health in conflict and post-conflict settings. They argue that the direct traumatising effects of war exposure on affected populations’ mental health are often overemphasised compared with the impact of daily stressors present in post-conflict environments, and advocate an integrative intervention approach, which addresses daily stressors first, and specialised trauma interventions next only for those individuals who are still very distressed. In strong contrast, another model has suggested that poor mental health (which may be caused or exacerbated by past traumatic experiences) may conversely result in a change in the perception of daily stressors, or even the self-generation of daily stressors. Reference Neuner17 This paper aims to further this discussion by assessing the relationship between current perceived needs, past traumatic events and distress in two humanitarian settings. Specifically, the role of perceived needs in explaining the association between traumatic exposure and distress is investigated.
Method
Sample
In Jordan, Iraqi people (n = 269) displaced following the 2003 invasion of Iraq were interviewed in July 2010 (Amman, Zarqa, Irbid, and Madaba). In Nepal, Bhutanese refugees (n = 269) were interviewed in October to November 2010 (Beldangi-II camp in Jhapa district).
Measures
Distress was used as dependent variable, as measured by the General Health Questionnaire (GHQ-12). Reference Goldberg19 The main independent variables were perceived needs (as measured by the newly developed HESPER Scale) 20 and past traumatic events (as measured by the traumatic events list of the PTSD section of the Composite International Diagnostic Interview (CIDI) 3.0). 21
The GHQ-12 provides a distress score based on 12 questions relating to symptoms of depression and anxiety. Reference Goldberg19 An Arabic and Nepali version have been validated. Reference Daradkeh, Ghubash and El-Rufaie22,Reference Koirala, Regmi, Sharma, Khalid and Nepal23 It has been used as a measure of distress in the World Mental Health Survey Initiative in 28 countries. 24 The traumatic events list of the CIDI lists 27 potentially traumatic events, several of which (though not all) directly relate to events common in conflict or other disastrous situations; a total score of traumatic exposure can then be computed by counting the number of events experienced. Participants are also able to list additional events experienced. 21 In the present study, Cronbach's alphas for internal consistency of the GHQ-12 were 0.89 in Jordan and 0.88 in Nepal; for the CIDI traumatic events list, the alphas were 0.71 in Jordan and 0.72 in Nepal.
The HESPER Scale assesses the perceived physical, social, and psychological needs of the general adult population in humanitarian settings. Perceived needs are defined here as needs expressed by members of the affected population themselves. They are thus problem areas for which people would likely want help. Perceived needs are assessed on the HESPER Scale across 26 need items. Examples of need items include ‘Place to live in’ (‘Do you have a serious problem because you do not have an adequate place to live in?’), and ‘Education for your children’ (‘Do you have a serious problem because your children are not in school or are not getting a good enough education?’). Ratings are then made for each need item according to unmet need (or serious problem; ‘1’ rating), no need (or no serious problem; ‘0’ rating), or no answer (i.e. refused, not known, or not applicable; ‘9’ rating). A total score of unmet needs can be calculated by adding up the number of items rated as serious problems.
Validation of the HESPER Scale was conducted with the same data-sets as were used for the present study. Reference Semrau, van Ommeren, Blagescu, Griekspoor, Howard and Jordans15 In brief, the HESPER Scale demonstrated excellent interrater reliability (intraclass correlation coefficients (ICCs, absolute agreement) for total number of unmet needs were 0.998 in Jordan and 0.995 in Nepal; Cohen's kappas for need ratings of the 26 individual need items ranged between 0.66 and 1.0 across sites, with 94.9% of items having kappas over 0.8), and good to excellent test–retest reliability (ICCs, absolute agreement, for total number of unmet needs were 0.961 in Jordan and 0.773 in Nepal; need ratings of individual need items ranged between 0.07 and 1.0, with 55.8% of items having kappas over 0.6). Additionally, acceptable criterion (concurrent) validity was established by comparison with the World Health Organization Quality of Life (WHOQOL-100); 25 total number of unmet needs correlated with total WHOQOL-100 score as predicted in both settings (r = –0.629 in Jordan and r = –0.469 in Nepal), and correlations between individual need items and related WHOQOL-100 questions were mostly as predicted. Reference Semrau, van Ommeren, Blagescu, Griekspoor, Howard and Jordans15 Internal consistency was assessed by calculating Cronbach's alpha, where ‘0’ (‘no serious problem’) and ‘9’ (‘not applicable’) ratings were combined into one rating, as was done for the initial study into the psychometric properties of the HESPER Scale. Reference Semrau, van Ommeren, Blagescu, Griekspoor, Howard and Jordans15 Cronbach's alphas of the HESPER Scale were 0.89 in Jordan and 0.80 in Nepal.
Procedure
Ethics approval for both settings was obtained through the King's College London Psychiatry, Nursing and Midwifery Research Ethics Committee. In Nepal, ethical approval was obtained from the Nepal Health Research Council, and in Jordan permission for the study was obtained from the Ministry of Planning and Ministry of Health. All participants gave their voluntary written or verbal consent to take part. Arabic and Nepali versions of the GHQ-12 and the traumatic events list of the CIDI were obtained from distributors or other research groups before testing commenced; all other project materials (including the HESPER Scale) were translated using back-translation methods. The traumatic events list of the CIDI was adapted to suit the local contexts, i.e. a few items were added or removed based on previous research of what items were locally relevant. In Jordan, an additional need item was added to the HESPER Scale based on previous findings made during pilot-testing. This item pertains to problems related to legal residency and potential resettlement to a third country.
Different sampling methods were employed in the two sites. Iraqi participants in Jordan were recruited through a multistage cluster sampling design, involving 30 clusters of city districts. The sample was geographically representative of Iraqis living in Jordan (around 75% of the sample lived in Amman and 25% outside Amman). In Jordan, the random-walk method was employed to recruit households within clusters or camps; a random-number Kish Table Reference Kish26 was used to select individuals from within chosen households. In Nepal, participants were recruited though simple random sampling methods; a list of randomly selected Bhutanese refugees in the camp was obtained from the United Nation High Commissioner for Refugees.
Interviews were conducted during one-to-one assessments in participants’ homes in Arabic in Jordan and Nepali in Nepal; these were conducted by 12 and 6 local interviewers respectively, who had previously been trained for 3.5 to 4 days. Interviewers were supervised by a local team leader. Participants were first administered the HESPER Scale, followed by the GHQ-12 and then the traumatic events list of the CIDI.
Statistical analyses
To identify the mechanism that underlies the relationship between trauma exposure and distress in humanitarian settings, a series of mediation analyses were conducted. A mediator model hypothesises that the effect of the independent variable on the dependent variable is transformed by the mediator variable. Reference Frazier, Tix and Barron27,Reference Baron and Kenny28 The mediator variable thus clarifies the nature of the relationship between the independent and dependent variables and is the mechanism through which a predictor influences an outcome variable. Traditional approaches to mediator testing have been criticised, and alternative approaches have been developed to test direct v. indirect effects and using bootstrapping procedures. Reference Preacher and Hayes29 Direct effects refer to the relationship between a predictor variable (X) and outcome variable (Y). Indirect effects refer to the effect of (X)on(Y) via a mediator variable (M). Predictor variables may have both direct and indirect effects on an outcome variable. The hypothesised mediator between trauma exposure and distress in the current study was perceived needs. Therefore, both the direct effect of trauma exposure on distress and the indirect effect of trauma exposure on distress via perceived needs were examined.
A regression-based approach to mediation analysis was used with bootstrap estimation of indirect effects, with data being resampled 5000 times. Many mediation studies use regression analysis-based approaches. Reference MacKinnon, Lockwood, Hoffman, West and Sheets30 Bootstrapping is a non-parametric technique for obtaining parameter estimates by numerous resamplings of the data, and is used for testing the likelihood of the mediation by producing robust confidence intervals for the estimates. A Preacher & Hayes’ ‘indirect’ macro for SPSS was used in the present study, which calculates bootstrap estimates of indirect effect for models of one or more mediators Reference Preacher and Hayes31 (see www.afhayes.com/spss-sas-and-mplus-macros-and-code.html for the macro). Simulation studies have shown that bootstrap estimation compares favourably with distribution-based estimation or significance tests for simple mediation models (i.e. one mediator). Reference MacKinnon, Lockwood and Williams32 A further strength of the ‘indirect’ macro is that it allows for the inclusion of covariates. In case the problem of confounding (of associations) is present, a model that does not control for the effect of the confounding variable produces a biased estimate of the exposure effect. The present study includes covariates of age and gender, as these variables have been significant covariates in prior multiple mediation models in Nepal. Reference Kohrt, Speckman, Kunz, Baldwin, Upadhaya and Acharya33 All statistical analyses were performed in SPSS v.17.0 on Windows. Standardised scores were used for analyses of mediator models. Standardised variables were employed (obtained by subtracting the mean and dividing by the standard deviation) in the mediator analyses, in order to reduce the problem of multicollinearity that partly (but not exclusively) arises when interaction terms are introduced in the analysis.
Within the employed statistical procedure, the c path refers to a significant relation between the predictor (X) and the outcome (Y), when not accounting for indirect effects (i.e. total effect). The a path refers to the relationship of the predictor (X) with the mediator (M). The b path refers to the relationship of the mediator (M) with the outcome variable (Y). When M is included in the model, the indirect effect refers to the ab pathway, and c′ refers to the direct effect of X on Y. The indirect effect also can be conceptualised as the difference between the total effect and the direct effect (c–c′). The present study investigated the theoretical model in which the ‘current perceived needs’ variable was the mediator (M) for the association between the predictor ‘trauma exposure’ (X) and the outcome ‘distress’ (Y). Analyses per country sample were conducted and no between-country comparisons were performed.
Results
Sociodemographic information for respondents in both study sites is presented in Table 1. Table 2 presents the mean scores of the hypothesised predictor (CIDI total score), outcome (GHQ total score) and mediator (HESPER total score) variables.
In both samples a significant total effect was found for trauma exposure on distress when covariates were taken into account, but possible mediation was not (Jordan: 0.42, Nepal: 0.32). In the Jordan sample, there was a significant indirect effect via perceived needs. Analyses confirmed a significant association between the predictor and the putative mediator (a path). Furthermore, perceived needs had a statistically significant association with distress (b path). The association between trauma exposure and distress was no longer statistically significant (c′ 0.11) when accounting for the indirect effect of perceived needs (0.31). Age and gender as covariates played no role in predicting distress in this model. The adjusted R 2 was 0.44 for the distress variable in the tested model (i.e. 44% of the distress variance was explained by the tested meditational model) (Fig.1).
TABLE 1 Demographic characteristics of respondents at both study sites
| Jordan (n = 269) |
Nepal (n = 269) |
|
|---|---|---|
| Gender, n (%) | ||
| Men | 116 (43.1) | 139 (51.7) |
| Women | 153 (56.9) | 139 (51.7) |
| Age, years: mean (s.d.) | 40.24 (13.36) | 36.92 (14.15) |
| Marital status, n (%) | ||
| Married | 191 (71.0) | 217 (80.7) |
| Unmarried | 56 (20.8) | 50 (18.6) |
| Widowed | 16 (5.9) | 0 |
| Divorced/separated | 6 (2.2) | 2 (0.7) |
| Cohabiting | 0 | 0 |
| Number of children, mean (s.d.) | 2.11 (1.95) | 2.39 (2.37) |
| Level of education, n (%) | ||
| Illiterate/no formal education | 7 (2.6) | 108 (40.1) |
| Primary school (grades 1–6) | 29 (10.8) | 63 (23.4) |
| Secondary school (grades 7–12) | 104 (38.7) | 89 (33.1) |
| University | 129 (48.0) | 9 (3.3) |
| Religion, n (%) | ||
| Christian | 45 (16.7) | 16 (5.9) |
| Muslim | 221 (82.2) | 0 |
| Hindu | 0 | 178 (66.2) |
| Buddhist | 0 | 52 (19.3) |
| Other religionFootnote a | 3 (1.1) | 23 (8.6) |
| No religion | 0 | 0 |
| Time displaced, year: mean (s.d.) | 3.84 (2.18) | 18.95 (0.93) |
a. Other religions include Kirat, Sanatan, Biswasi, Manab, Nastak (Nepal), Haba'i, Sa'aebiya (Jordan).
TABLE 2 Mean scores for trauma exposure, distress and perceived needs for both study sites
| Mean (s.d.) | ||
|---|---|---|
| Jordan | Nepal | |
| GHQ total score | 16.50 (7.30) | 13.77 (7.01) |
| CIDI total score | 7.53 (3.62) | 4.51 (3.23) |
| HESPER total score | 10.61 (5.88) | 8.10 (4.64) |
GHQ, General Health Questionnaire; CIDI, Composite International Diagnostic Schedule, past traumatic life events as measured by the traumatic events list (28 items were included in Nepal and 16 in Jordan); HESPER, Humanitarian Emergency Settings Perceived Needs Scale (27 items were included in Jordan and 26 in Nepal).
In the Nepal sample similarly there was an indirect effect via perceived needs of trauma exposure on distress. Analyses confirmed a statistically significant association between all of the variables in the model. When the indirect effect of perceived needs was included in the model, the direct effect regression coefficient between trauma exposure and distress was reduced from a total effect (c path) of 0.32 to an adjusted effect (c′ path) of 0.20. Thus, there remained a direct effect of traumatic exposure on distress when accounting for the perceived needs indirect effect. Trauma exposure in Nepal therefore predicted distress directly as well as indirectly through perceived needs. The significant association of age (older) and gender (female) meant that these were independent predictors of distress in addition to the earlier mentioned direct and indirect pathways of trauma exposure. The adjusted R 2 was 0.34 for the distress variable in the tested model (Fig. 2). Table 3 presents all confidence intervals for the above findings.
FIG. 1 Mediation model Jordan (n = 269).
* P<0.05, ** P<0.001; bootstrap resamples: 5000.
Discussion
Main findings
This study tested the hypothesis that there are indirect effects of trauma exposure on distress via current perceived unmet needs. This question informs mental health policy and programmes in humanitarian settings. The study's hypothesis was confirmed for both the Jordan and Nepal samples. It was found that the tested model had a good fit with the Jordan data, where the direct association between traumatic exposure and distress was no longer significant in the model with indirect effects of unmet perceived needs. Similarly, an indirect effect for current needs was found in Nepal. Nepal differed from Jordan in that the Nepal sample showed a significant direct effect of trauma on distress when accounting for unmet needs. Age and gender were significant covariates in Nepal (being female and being older respectively).
FIG. 2 Mediation model Nepal (n = 269).
* P<0.05, ** P<0.001; bootstrap resamples: 5000.
TABLE 3 Direct and indirect effects for trauma exposure on distress
| Effect (95% CI) | ||
|---|---|---|
| Jordan | Nepal | |
| Model 1: total trauma exposure effect (without mediator)Footnote a | ||
| Trauma exposure | 0.42 (0.30 to 0.54) | 0.32 (0.22 to 0.42) |
| Age | –0.02 (–0.13 to 0.09) | 0.34 (0.24 to 0.45) |
| Gender | –0.02 (–0.14 to 0.09) | 0.21 (0.11 to 0.32) |
| Model 2a: direct effects | ||
| Direct effect of trauma and covariates | ||
| Trauma (c’ path) | 0.11 (–0.01 to 0.23) | 0.20 (0.08 to 0.32) |
| Age | 0.03 (–0.07 to 0.13) | 0.33 (0.23 to 0.43) |
| Gender | –0.08 (–0.18 to 0.02) | 0.25 (0.15 to 0.35) |
| Direct effect of potential mediator | ||
| Perceived needs (b path) | 0.60 (0.48 to 0.72) | 0.26 (0.14 to 0.38) |
| Model 2b: indirect effect | ||
| Perceived needs | 0.31 (0.24 to 0.39) | 0.12 (0.06 to 0.19) |
a. Standard multivariate regression with the predictor (trauma exposure) and the covariates (age, gender).
The indirect effects in Jordan and Nepal are consistent with previous research. In particular, they support Miller & Rasmussen's mediation model, Reference Miller and Rasmussen16 in that the addition of unmet needs to the model increases the overall explanatory power and it weakens the direct association between trauma exposure and distress. Reference Miller and Rasmussen12 It suggests that the model, which advocates an increased focus on ongoing stressors arising from current difficulties and challenges that life poses compared with a focus largely on past traumatic exposure, may extend beyond conflict and post-conflict settings to include wider humanitarian settings. These data suggest that for populations that have undergone large-scale traumatic experiences, it is important to look beyond the direct effect model in guiding intervention strategies. As the present study involved samples that had undergone the potentially traumatic experiences several years prior, a plausible explanation for these findings is that the potentially traumatic stressors were distal, whereas perceived needs were proximal. The results also tie in with previous studies that have emphasised the importance of post-displacement stressors compared with past traumatic experiences in predicting mental health outcome, both in post-emergency settings themselves, Reference Betancourt, Brennan, Rubin-Smith, Fitzmaurice and Gilman34,Reference Miller, Omidian, Rasmussen, Yaqubi and Daudzai35 as well as among refugees and asylum seekers in other countries, including Western ones. Reference Laban, Gernaat, Komproe, Schreuders and de Jong36–Reference Porter and Haslam38 Finally, the results support the relevance of measuring perceived unmet needs, using the newly developed HESPER Scale, Reference Semrau, van Ommeren, Blagescu, Griekspoor, Howard and Jordans15 also in light of the prevailing broad mental health and psychosocial support framework. 2
Although the results of this study provide further support for the importance of current stressors in explaining the association between past traumatic exposure and mental health outcomes in humanitarian settings, further research across different settings is required to shed more light on this association. Conceptually, reverse causation in the tested model is unlikely, but could nonetheless be examined through longitudinal study. Regarding statistical analyses, an alternative mediation analysis method for handling multiple mediators is structural equation modelling. However, given the small sample size, the bootstrapping technique employed here made multiple mediator regression analysis the preferred analytical approach. Future studies need to include analyses with additional covariants (for example time since displacement) included in the model.
The finding that older age and female gender independently predicted distress in one of the two settings fits in with the inconsistent results previous research has produced. These variables, in particular female gender, have been found to be predictive of mental health outcome in some studies, Reference Foa, Stein and McFarlane4,Reference Porter and Haslam38,Reference Brewin, Andrews and Valentine39 but not in others. Reference Steel, Chey, Silove, Marnane, Bryant and van Ommeren5,Reference Brewin, Andrews and Valentine39 An explanation for this may be that both older age and female gender as risk factors for poor mental health outcome may be context-dependent, for example according to culture, or type of mental health outcome. Reference Kohrt, Speckman, Kunz, Baldwin, Upadhaya and Acharya33
Implications
This study has several implications for policy and practice. If poor mental health (such as distress) may in part be mediated by post-emergency current stressors, as the findings of this study suggest, it is highly plausible – and testable – that positive changes to the post-emergency environment may potentially attenuate the impact of past traumatic experiences. Inter-sectoral humanitarian action targeting perceived unmet needs may therefore have a substantial impact in reducing (or in absence potentially worsening) mental health complaints experienced by emergency-affected populations. Reference Porter and Haslam38 As has been advocated recently by several authors, 2,Reference Mollica, Lopes Cardozo, Osofsky, Raphael, Ager and Salama3,Reference van Ommeren, Saxena and Saraceno10,Reference Miller and Rasmussen12–Reference Jordans, Tol, Komproe, Susanty, Vallipuram and Ntamatumba14 rather than relying solely on resource-intensive trauma-focused interventions to mitigate the impact of traumatic exposure, the results indicate that interventions that broadly affect the recovery environment may hold promise. Multidisciplinary interventions that aim to reduce current stressful social and material conditions caused or worsened by humanitarian emergencies may potentially buffer against the negative impact of traumatic experiences on an individual's mental health. Clearly, trauma-focused care is indicated for those with serious post-traumatic stress complaints; however, it does confirm the need for a more integrated approach to mitigate the impact of potential traumatic exposure in humanitarian settings. In line with a recent global research priority setting exercise for mental health and psychosocial support in humanitarian settings, Reference Tol, Patel, Tomlinson, Baingana, Galappatti and Panter-Brick40 future research should increasingly evaluate the distinctive pathways by which emergencies have an impact on mental health. The better science unravels the mechanisms of that association (i.e. influence of contextual or protective factors), the easier it becomes to match scarce resources with adequate intervention.
Funding
Data collection in Jordan was funded by the Jordanian Nursing Council, WHO Jordan, and the University of London Central Research Fund. Data collection in Nepal was funded by WHO Geneva and United Nation High Commissioner for Refugees Nepal; WHO Nepal providing further support. M.S. is funded by a PhD studentship grant of the Medical Research Council (UK). G.T. is funded by a National Institute for Health Research (NIHR) Applied Programme grant awarded to the South London and Maudsley NHS Foundation Trust and is affiliated with the NIHR Specialist Mental Health Biomedical Research Centre at the Institute of Psychiatry, King's College London and the South London and Maudsley NHS Foundation Trust.
Acknowledgements
The steering group of the HESPER project includes M.v.O. and Dr Andre Griekspoor at WHO Geneva, and G.T., Professor Louise Howard, Dr Heidi Lempp, Dr Morven Leese and M.S. at the Institute of Psychiatry, King's College London. M.J. would like to thank Dr Brandon Kohrt and Dr Ivan Komproe for their important contributions. Data collection in Jordan was organised by Dr Anita Marini of WHO Jordan, and was implemented by Accurate Opinion; United Nation High Commissioner for Refugees provided advice on sampling. Data collection in Nepal was organised by HealthNet TPO/TPO Nepal.
eLetters
No eLetters have been published for this article.