The prevalence of headache disorders in children and adolescents in Ethiopia: a schools-based study

Over multiple iterations between 2000 and 2019, the Global Burden of Disease (GBD) study has demonstrated that headache disorders are a major cause of global ill health, to the extent of being the second highest cause of disability [1‐6]. The data for headache contributing to GBD estimates nevertheless remain deficient: there are still geographical gaps. Further, these data so far have very largely been collected from adults (aged 18–65 years).

A key objective of the Global Campaign against Headache, under the direction of Lifting The Burden (LTB), is to quantify headache-attributed burden worldwide [7‐10]. In adults, the principal contributors to this are migraine, tension-type headache (TTH), medication-overuse headache (MOH) and the group of other disorders characterized by headache occurring on ≥15 days/month. In children (aged 6–11 years) and adolescents (12–17 years), the same are important, but recent studies have shown, in addition, a prevalent headache disorder characterised by mild pain of short duration (< 1 h) [11, 12]. Termed “undifferentiated headache” (UdH), this is not diagnosable as migraine or TTH but believed to be expressions of these by the immature brain. In these age groups, LTB is conducting a global schools-based programme using a standardised protocol [13] in multiple countries in each of the world’s six major regions, thereby cluster-sampling the world.

This national study in Ethiopia, part of the global programme, is the first to be reported from sub-Saharan Africa (SSA) and the first in a low-income country. Half of Ethiopia’s population are aged under 18 years [14]. While no published data exist for child and adolescent headache in this country, a population-based national survey on adult headache disorders showed that they were common: 1-year prevalence estimates were 17.7% for migraine (higher than the global mean [6]), 20.6% for TTH, 0.7% for probable MOH (pMOH) and 2.5% for other headache on ≥15 days/month [15].

This study estimated the prevalence and attributed burden of headache disorders in children and adolescents in Ethiopia. The purposes were to add to knowledge of the global burden of headache, and to inform local health and educational policies. This paper reports prevalence; a later paper will report on burden.

The study, a cross-sectional survey, followed the generic protocol for the global programme [13]. It was conducted in schools selected to be nationally representative. Enquiry was by self-completed questionnaire administered under supervision to pupils within their school classes.

Class lists indicated a total population of 2403, but 54 pupils (male 31 [2.6%], female 23 [1.9%]) were absent on the days of survey. All 2349 who were present agreed to participate, but five (0.2%) were excluded (non-participation proportion = 0.2%) because of incomplete responses. Therefore, the surveyed sample were N = 2344 (male 1157 [49.4%], female 1187 [50.6%]). Children (n = 1011 [43.1%]) were under-sampled in comparison with adolescents (n = 1333 [56.9%]). Overall mean age was 12.0 ± 2.2 years (range: 6–17; median 12).

School variables (Table 1) were clearly influenced by, but did not entirely reflect, Ethiopia’s urban/rural divide (21:79 [21]). A third (36%) of all participating pupils attended urban schools, while the remainder were at schools in semi-rural or rural locations. The majority (59%) were at schools rated lower-middle or low income, as defined above. Teachers rated school-area income similarly: 55% of pupils were from lower-middle- or low-income areas. Most pupils (62%) were at schools close to their homes (fewer than 25% travelling for > 1 h/day).

This was the first study to be reported from SSA within the schools-based enquiry into child and adolescent headache, conducted as a major fact-finding component of the Global Campaign against Headache [7, 8]. The key finding was that headache was prevalent among these age groups, affecting almost three quarters. Migraine was the most often reported type (over one third), followed by TTH (one fifth); UdH, while common at 12.7%, was less so than these specific headache disorders. Headache on ≥15 days/month was reported by 1.3%, very little of this being pMOH (0.2%).

Associations with gender were much as expected but, rather surprisingly, increased prevalence of migraine was not seen among adolescents. In this context we need to say more about UdH. This recently characterised disorder is understood to represent expressions of migraine or TTH by the immature brain [11]. Accordingly, its prevalence as a proportion of all reported headache is expected to decline with increasing age, as seen in Turkey [11] and Austria [12]. In Ethiopia the picture was more complex, but what was clear was that the prevalence of UdH fell sharply after age 14 years as other headache disorders increased. Presumably this change reflected rapid brain development after puberty, but age-related decline in UdH was expected earlier. UdH has been defined as short-duration mild headache, with duration somewhat arbitrarily fixed at < 1 h [11], while children’s sense of time is unreliable. Although we are not presenting headache duration here, reported duration for 64.9% of all headache, including 84.4% of cases diagnosed as probable migraine, was < 2 h.

In relation to this, we must question our finding of 38.6% migraine prevalence, which we acknowledge to be implausible. LTB’s study of adult headache disorders in Ethiopia estimated 1-year prevalence of migraine at 17.7% (with rather more TTH, at 20.6%) [15]. We do not believe that migraine is more than twice as common in these younger age-groups. In children in particular, headache diagnosis is notoriously difficult, dependent as it is on comprehension of language on the one hand (formulation of ICHD criteria into questions that are clear and unambiguous to children is a challenge) and subjective evaluations on the other (over a quarter [26.5%] of cases with reported headache duration of < 1 h were diagnosed as probable migraine, the distinction from UdH resting solely on reported headache intensity). Questions relating to associated symptoms – likely to drive diagnosis – are especially difficult to structure other than as leading questions, with yes/no response options. Although these questions had been successfully tested elsewhere [11‐13], children in particular may be susceptible to suggestion. We saw this, probably, in the almost 80% affirmative response to phonophobia (Table 3), which deprived this symptom of diagnostic utility (although the much lower photophobia response [37.9%] diminished its impact, since both must be present for either to contribute to migraine diagnosis [17]). The responses to questions on nausea (44.8% affirmative) and vomiting (28.0% affirmative), although lower, were probably more influential, since these symptoms are specific to migraine. If it were true that over one quarter of all participants with headache usually vomited during headache episodes, and nearly half were nauseated, then the proportion with migraine was undoubtedly high. But we doubt that it was true.

The veracity of our participants’ responses is questionable again in the over-reporting of HY, at least by those with episodic headache. Headache frequencies reported over the preceding 7 and 28 days were perfectly feasible for all headache types, but, other than in those with headache on ≥15 days/month, HY claims were not: approximately double the predictions based on the former and four times the predictions based on the latter. HY reporting should be free from recall error: it is from this, essentially, that it derives its value as a measure [16]. One interpretation, therefore, is that HY reporting was correct, while accuracy diminished as memory faded over 7 then 28 days. But we doubt this also: it is implausible that more than one quarter of these young participants had headache on the day in question (which was, presumably, no different from any other day). Were they dissembling? We doubt this too, but, if the explanation lies elsewhere, we do not know what it is. No problems were encountered with the same question in adults [22].

In future surveys of children and adolescents, some questions may need to be reformulated, even though they were not problematic in Turkey or Austria [11‐13]. However, as this series of studies is continued elsewhere in the world, it may transpire that epidemiological diagnosis of headache type in these age groups is an inherently inexact process. This would not be surprising, if our belief is correct that headache is evolving from an undifferentiated form. Neither would it be a revelation: many past surveys have failed to diagnose large proportions of reported headaches, often without comment (for discussion of this, see [11]).

While, therefore, some uncertainty surrounded headache type, prevalence of headache overall (1-year: 72.8%) was the same as was found in children and adolescents in Turkey (73.7%) with the same questionnaire [11]. The Austrian study of adolescents (10–18 years), also using the same questionnaire, reported a 1-year headache prevalence of 75.7% [12], not different from our 77.6% in adolescents (12–17 years). These values appear robust, and leave no doubt that, at least in terms of prevalence, headache is no less a problem in these age groups than in adults. It is interesting to compare our data with the adult data from Ethiopia: 1-year prevalence of headache was 44.9%, with a female preponderance of 4:3, of migraine 17.7%, with a female preponderance of 3:2, and of TTH 20.6%, with marginal female preponderance [15]. The difference between these is in migraine only, although there is the diagnostic caveat. These two datasets indicate that not all childhood headache survives into adulthood, and what may appear as migraine in children may not persist as migraine in adults. This is somewhat contrary to received wisdom, raising a query for future studies.

The strengths of this study lie in its adequate sample size, the very high participation proportion and the tested and validated methodology [11‐13], although the last proved imperfect in Ethiopia in one important respect. We have already acknowledged this limitation, which has no easy solution. Whether parents might give a better account of their children’s symptoms than the children themselves is very questionable: clinical experience rather suggests the opposite. In any case, on a general level, parental questioning requires a quite different protocol; in this highly rural country characterised by low income and poor education, it would have been impractical. The questionnaire could not be validated in the languages used in Ethiopia for the same reason as in the adult study [15]: with very few headache experts in the country, validation would have required at least one to take time out from the health service in order to re-interview a subset of participants in the field, and it could not be done. A third limitation was that we surveyed in only two of more than 80 active languages in Ethiopia [23]. Respecting the pupils’ preference, we used the local language(s) of each school: Amharic in Addis Ababa, Hawassa and Bahir Dar; Oromiffa and Amharic (the latter preferred by most adolescents) in Asella. In fact these are the most widely spoken languages in Ethiopia – Oromiffa by 33.8% and Amharic (the national working language) by 29.3% of the total population according to 2007 census data [23]. Nevertheless, our survey was, on this account, arguably representative of only a majority (63%) of the population. The additional resources required to add just one language could not be justified when the next most common (Somali) is used by only 6.2% [23]. Finally, schools-based sampling is valid when education is compulsory and uptake is high [16]. We acknowledge a further, possibly important, limitation in this regard. Primary education in Ethiopia is, officially, from 7 to 14 years of age (grades 1–8), and secondary education from 15 to 18 years (grades 9–12). According to Ministry of Education data for 2018/19, school intake among those aged 7 years exceeded 90% nationally (96.5% among males and 88.2% among females), with a higher apparent intake rate of 136.6% reflecting enrolment in grade 1 of many younger or older children [24]. But high drop-out rates from grade 2 onwards greatly reduce transition to secondary education (nationally, 32.0% in 2018/19: 48.5% for grades 9–10 and only 14.8% for grades 11–12 [25]). In a low-income country, economic factors weigh heavily: for many, even basic educational expenses may be unaffordable, while parents may depend on their children’s labour at home or for income generation [26]. Children lost to our study were probably, largely, from the poorest homes, with Table 2 suggesting some bias might result, but no other sampling method is likely to obviate this.

Headache is very common in children and adolescents in Ethiopia – apparently more so than in adults – a finding with major public-health implications for a country in which half the population are aged under 18 years [14]. With no similar study yet from elsewhere in the whole of SSA, or from another low-income country, this finding also contributes to knowledge and understanding of child and adolescent headache globally.