Epilepsy prevalence in rural Zambia: a door-to-door survey
Gretchen L. Birbeck, No. 138 Service Road, A 217, East Lansing, MI 48824-1313, USA. Tel.: +1 517 353 8122 (extn 141); Fax: +1 517 432 9414; E-mail: firstname.lastname@example.org (corresponding author).
Ellie M. N. Kalichi, Chikankata Health Services, Private Bag S2, Mazabuka, Zambia.
Objectives To identify people with epilepsy (PWE) in our Zambian catchment area of 55 000 people.
Methods A nine-item, previously validated screening instrument for detecting epilepsy in developing countries was forward-and-back translated into Chitonga. Early piloting indicated poor specificity among children, so three questions were added. Local census data were used to estimate the population at risk. Community health workers conducted screening interviews with household heads. All positive screens were referred for physician assessment. A blinded neurologist assessed a randomly selected subset (100 positives, 50 negatives) to determine screening instrument characteristics.
Results We identified 799 people with possible epilepsy (unadjusted prevalence 14.5/1000). The adapted instrument exhibited 86% specificity (adjusted prevalence 12.5/1000). False positives occurred primarily among children who had experienced multiple malaria-associated seizures. Age-specific rates were highest for children aged 5–15 years (26.2/1000) and for people over 65 years (15.9/1000). Males were disproportionately represented (55.8%vs. 44.2%, P < 0.05), although this trend reversed after childbearing age.
Conclusion Even using a relatively conservative definition, we identified almost 700 PWE. Use of the recommended epidemiological definitions would likely have yielded higher prevalence rates. The age-specific prevalence did not follow patterns described where neurocysticercosis is the commonest cause of epilepsy. Trends in age- and gender-specific prevalence may offer a clue to the aetiology of epilepsy in this region.
Before establishing an Epilepsy Care Team to provide hospital- and community-based services, we conducted a door-to-door survey to identify the population of people with epilepsy (PWE) and developed a population-based registry for our rural catchment area of 55 000. The World Bank classifies Zambia as one of the poorest countries in the world. Most health indicators have worsened over the past 20 years with rural people suffering poorer health than urban populatons (CSO-Zambia 2002). In such an environment, resources for both needs assessments and health-care provision are very limited. Previous hospital-based investigations indicated that epilepsy is a major cause of neurological morbidity and mortality in this region (Birbeck 2001). To facilitate population-based screening, a previously validated World Health Organization screening instrument was used. This questionnaire, with some adaptation and variation in case definition, has been used to assess epilepsy prevalence rates in over 20 developing countries.
The Chikankata catchment area, encompassing approximately 7800 km2 and 55 000 people, comprised our sampling frame. Denominator data (i.e. details regarding the population at risk) were available through the 1999 local insurance scheme census. This insurance scheme provides people in the catchment area with subsidized care relative to out-of-catchment patients. Chikankata Health Services offers the only option for healthcare within the region (excluding traditional healers). All community leaders within the catchment area chose to participate in the insurance scheme. Each village provided household level data including name, age and community (for adult females) or name, age and dependents’ names and demographic details (for adult males). After census registration, all enrollees received an identification card providing access to free and/or subsidized care.
The nine-item screening instrument used was originally developed from a longer (20-item) instrument validated in Ecuador (Placencia et al. 1992a). In Ecuador, the questionnaire demonstrated 92.9% specificity, 79.3% sensitivity, a positive predictive value of 18.3% and a negative predictive value of 99.6%. We translated it forward-and-back and reconciled discrepancies before piloting it in general medicine and epilepsy clinics. Comprehension of the questionnaire was satisfactory, but high rates of false-positive screens were evident, primarily among children with seizures during recurrent malaria-related fevers. Given the few physicians in this region during the time of the survey (3.5 full-time physicians for the entire population of 55 000), we required an instrument with a high positive predictive value. Therefore, three questions were added to eliminate febrile and malaria-associated seizures (see Appendix).
An active epilepsy case was defined as anyone on treatment for epilepsy (or) anyone with a history of recurrent seizures with the most recent occurring in the past 12 months if the seizure was not provoked by an acute fever in a child (<7 years old), and the seizure did not occur solely during severe malaria (malaria requiring admission) in adults.
Consecutive point-prevalence rates were determined for the period September 15th 2000–January 28th 2001.
This work was approved by the Chikankata Health Services Review Board, which includes community representatives. Twenty-three community health workers, seven clinical officers and three nurses were trained to use the screening instrument. After training, research staff observed trainees using the instrument to assure adequate understanding and appropriate delivery. The catchment area was divided into four zones already commonly used for community-based planning and therefore detailed maps were available. Subdivisions of these zones were assigned to each screener based upon geographical access from their own home. This required each staff member to interview approximately 340 households in approximately 5 months. Screeners walked or used bicycles (sometimes rented for the purpose) to access the population. Rarely, motorcycle rental was required to reach more distant regions. Verbal consent was obtained from respondents prior to questioning. Data collected for positive screens, including age, gender, location, and treatment histories, were entered on standardized forms. The zones’ clinical officers delivered these forms to the research team when they travelled to the hospital for their monthly pay trip.
Data management and analysis
Data from paper forms were entered into a Microsoft Access database already containing census data. This information was stored on CD-ROM and imported into Epi Info 2002 for analysis of age- and gender-specific prevalence rates.
A clinic-based validation of a subset of a random sample of 150 screened individuals (100 positives, 50 negatives) was assessed by a neurologist (blinded to screening results) who conducted a complete neurologic history, physical examination, and review of prior medical records before determining if the patient met the case definition for active epilepsy. The adapted questionnaire exhibited 86% specificity. All those who screened negative were judged not to have epilepsy by the neurologist. False-positive screens occurred primarily among children who had experienced more than one provoked seizure during recurrent malaria infections. The positive predictive value of the instrument was 92%.
During the survey, 799 people initially screened positive for epilepsy (unadjusted prevalence 14.5/1000) with an adjusted prevalence of 12.5/1000 based upon our validation of the adapted instrument. Details are provided in Table 1. Six people screened positive who were not permanent residents of the region, had no insurance scheme identification card, and had not been previously captured in the census (denominator) data. Therefore, these individuals were excluded from the prevalent cases. Age-specific prevalence rates were highest among children 5–15 years old with a second smaller peak in the over 65-year olds. In this population-based registry, males were disproportionately represented (55.8%vs. 44.2%, P < 0.05) overall. However, the male predominance among PWE appears to be age-related, with female prevalence rates rising after age 45.
Table 1. Age, prevalence and gender distribution of people with epilepsy (PWE)
|0–4||18.5 (15.7–21.5)|| 6.9||57.1|
|5–10||25.6 (22.4–28.9)||26.2 (among 5–15 years)||56.0|
|16–20||10.1 (8.1–12.6)||15.1 (among 16–35 years)||54.8|
|36–45||4.0 (2.8–5.8)|| 7.9||55.2|
|45–55||4.3 (3.0–6.1)|| 7.3 (among 45–65 years)||41.9|
The primary purpose of this door-to-door survey was to determine the number of people in our catchment area requiring epilepsy care. Almost 800 PWE were initially identified, although prior to the survey the hospital registration information indicated only 32 PWE receiving chronic care for the disorder. Our approach did not provide epidemiological information in the preferred format (Shorvon 1996), but active epilepsy prevalence in the past year likely represents the burden of disease requiring treatment in this region and allowed us to determine the need for healthcare – an approach more suitable for our underlying purpose (i.e. planning for health services).
We believe the estimate provided likely underestimates the full epidemiological burden of epilepsy in the region. The instrument used is best able to detect generalized tonic clonic seizures, so individuals with absence or myoclonic epilepsy who do not experience generalized seizures were probably under recognized. Furthermore, epilepsy-associated stigma may have resulted in a reluctance to admit to the disorder resulting in further underestimation of the number of prevalent cases.
The age distribution for PWE in Zambia reveals a younger peak than has been reported in other developing countries (Kaiser et al. 1996). This is surprising, especially in light of our additional questions aimed at a more rigorous screen to avoid inclusion of children with febrile or malaria-provoked seizures. Such an age-distribution for epilepsy is not consistent with neurocysticercosis (NCC) as a major underlying aetiological factor, as NCC-associated seizures result in epilepsy primarily after age 25 (Garcia et al. 1993). Only 7.9% of households in Zambia's southern province keep pigs (Thom Jayne, Food Security Research Project, Personal communication) compared with up to 80% in Latin America where per capita pork production is 2–10 times higher than in Zambia (United Nations 2003). The age-specific epilepsy prevalence identified suggests a unique underlying cause for the high-rates of epilepsy seen in this catchment region.
Among the PWE identified in this population-based survey, males predominated overall. Similar studies in Saudi Arabia (Al Rajeh et al. 2001) and Kenya (Kaamugisha & Feksi 1988) have also reported male predominance. However, females appear to have higher epilepsy prevalence in Guatemala (Garcia-Noval et al. 2001) and Bolivia (Nicoletti et al. 1999), and no gender differences were identified in Tunisia (Attia-Romdhane et al. 1993), Nigeria (Osuntokun et al. 1987), or Ecuador (Placencia et al. 1992b). Whether these gender differences represent different aetiologies and gender-dependent risk factors within these various regions or are simply an artefact of case ascertainment or competing mortality risks remains unclear. The reversal of the male to female trend in Zambia occurring after childbearing age is intriguing and may offer a clue as to the underlying aetiology of epilepsy in the region.
This study provides epilepsy prevalence estimates for rural Zambia – a region with no previously available population-based epilepsy data. Our findings are similar to previous prevalence reports within Africa which, although exhibiting a wide range of values, have been estimated at approximately 15 of 1000 (Jallon 2002). If our estimate is applied to all of Zambia, then over 150 000 people with active epilepsy presently reside in Zambia with most cases occurring among children and youths. Ideally, subsequent surveys might have allowed us to examine time-trends and incidence rates. However, we are now experiencing significant immigration into the catchment area by families who are seeking services for a family member (or members) with epilepsy and this may limit the external validity of future population-based studies.
Funding for this work was provided by the Rockefeller Brothers Fund through the Charles E. Culpeper Medical Scholars Programme (GB). Any opinions expressed represent those of the authors and not necessarily Michigan State University, Chikankata Health Services, or the Rockefeller Brothers Fund.
Appendix 1 Screening questionnaire
Original questions (any positive response makes screen positive)
1. In the past year, have you or anyone in this household had attacks of shaking of the arms and legs which could not be controlled?
2. In the past year, have you or anyone in this household had attacks in which you fall and become pale?
3. In the past year, have you or anyone in this household lost consciousness?
4. In the past year, have you or anyone in this household had attacks in which you fall with loss of consciousness?
5. In the past year, have you or anyone in this household had attacks in which you fall and bite your tongue?
6. In the past year, have you or anyone in this household had attacks in which you fall and lose control of your bladder?
7. In the past year, have you or anyone in this household had attacks of shaking or trembling in one arm or leg or in the face?
8. In the past year, have you or anyone in this household had attacks in which you lose contact with the surroundings and experience abnormal smells or sensations?
9. In the past year, have you or anyone in this household had a diagnosis of epilepsy or epileptic fits?
Additional questions asked if above screen is positive (any positive response to these three questions negates above and makes screen negative)
10. Has such an attack occurred only once ever?
11. Did this/these attacks occur in a child (7 years or younger) ‘only’ during an illness with fever?
12. Did this/these attacks occur ‘only’ during an acute infection with malaria that resulted in hospital admission?