Discussion
This study is the first nationwide population-based survey in Nepal to estimate the prevalence and burden of MDBs. The main purpose of this paper is to report how it was carried out in a country with rather unusual physiographic challenges, multiple languages and wide socioculture diversity. Because of the widespread illiteracy and lack of dependable communication services and poor infrastructure, a door-to-door survey with stratified, multistage cluster random sampling and face-to-face interviewing was the only feasible method. We carefully selected and trained interviewers for data collection. We used the well-established HARDSHIP questionnaire [6] with some cultural adaptations. It has a modular design and separate question sets to capture demographic characteristics, screen for caseness, diagnose headache type according to ICHD-3 beta criteria [16] and assess various quantifiable elements of burden. It has already been used successfully in population-based surveys of headache in China [20], India [21], Pakistan [22] and Russia [23] and proven to be valid and acceptable in similar sociocultural settings [21,22]. The Nepali-translated version appeared to be culturally acceptable and inoffensive. To measure anxiety and depression levels, HADS was incorporated into the HARDSHIP questionnaire as an additional module, both screening for caseness and assessing symptom severity. HADS is one of the practical and short rating scales that can capture both disorders, with good screening properties in the general population [24]. EPQ-N was also incorporated into the HARDSHIP questionnaire to explore the association between headache and neuroticism. It is a short and practical questionnaire with proven reliability and validity to measure the psychometric properties of neuroticism in an adult population in different socio-culture settings [25-27].
Nepal is home to many different ethnic groups; there are more than 100 indigenous languages [28,29]. The translation of all these instruments into multiple languages would be not only enormously resource-consuming but also of questionable value, because every translation would need validating and it was doubtful that the means to do this existed. Nepali is the lingua franca among the different ethno-linguistic groups [28,29], and the official language of Nepal, used in national surveys [15,30,14]. We believe the pragmatic decision to conduct the survey in Nepali was sensible, and vindicated by the participation rate of 99.6%.
This participation rate was exceptionally high, and a guarantee of freedom from participation bias [13]. Yet there was apparent under-representation of males aged 18–34 years, according to national census data from 2011. We believe this reflects different survey methods: we would have excluded family members temporarily resident outside the country as migrant workers (mostly young males), but the census might not [30,15]. Employing interviewers familiar with the local culture and sentiments, and who spoke the local language, and enlisting the help of FCHVs who were respected in the community, undoubtedly encouraged participation in the survey (Risal et al. submission of manuscript in Journal Headache and Pain), which exceeded that of studies recently conducted in other countries in Asia [20-22]. So, probably, did the strategy of data collection in early mornings, late evenings, weekends and public holidays. The very high participation rate in this study might also be attributed in part to the collaboration with Dhulikhel Hospital, well-known for its outreach health services throughout the country.
Quality assurance plays a vital role in estimating the prevalence of diseases and their burdens, especially when decisions are to follow on allocation of scarce health resources. Unfortunately, studies such as this are vulnerable to fraud, as has recently been shown [31]. Many quality-control measures were built into the methods of this study, beginning with careful selection and training of the interviewers. We prepared a working manual for the field. We planned and took preventive and detective measures throughout the study period: we were in daily mobile telephone communication with each team leader; we made surprise visits to sites during data collection; we checked data and households. This was labour-intensive work, adding cost to the study, but experience had shown it was important [31].
If our inability to validate the survey instruments was a limitation of the study, it was forced upon us. In the case of the headache diagnostic questions, there were no headache experts in Nepal to be the gold standard. This is a situation encountered elsewhere and, if it is seen as an insurmountable barrier, there is no way forward. While HARDSHIP reliably estimates the prevalence of headache, and the attribution of burden to headache disorders, which in public-health terms are of first importance, the relative contributions of migraine and TTH are also of considerable interest because somewhat different health-care provisions are needed for their management. The diagnostic questions of HARDSHIP are based on ICHD-3 beta [16], and have worked with good specificity and reasonable sensitivity (though less so for TTH) in several cultures and languages [6]. The diagnosis of MOH, or more correctly probable MOH, rests on the association in individuals of headache on ≥15 days/month and medication overuse, since there cannot be evidence of causation. The coexistence of these in itself signals ill health regardless of the precise diagnosis [32]. Similar considerations apply to HADS and EPQ-N. These are instruments that have been used and validated widely, and were used with the expectation, but not proof, that they would perform no less well in Nepal than elsewhere.