Risk factors associated with the epilepsy treatment gap in Kilifi, Kenya: a cross-sectional study

1More than 62 million people with epilepsy live in countries with low and middle incomes and most of them do not receive appropriate treatment, an issue known as the epilepsy treatment gap. The epilepsy treatment gap is defined as the number of people with active epilepsy who have not accessed biomedical services (ie, epilepsy not diagnosed) or who are not on treatment or are on inadequate treatment, expressed as a percentage of the total number with active epilepsy. In a systematic review of studies done in countries with low and middle incomes, the overall estimate of the epilepsy treatment gap based on self-reporting was 56%, but the 95% CI was wide, ranging from 31% to 100%, with a higher proportion in rural areas than in urban areas. A more recent review reported an epilepsy treatment gap of more than 75% in low-income countries compared with less than 10% in high-income countries.

2The epilepsy treatment gap could be caused by poor access to diagnosis and treatment of epilepsy or non-adherence to antiepileptic drugs (AEDs). Adherence to a drug regimen is defined as the extent to which patients take drugs as prescribed by their health-care providers. Unintentional non-adherence might be due to forgetfulness or inability to follow treatment instructions because of poor understanding or impairment (eg, poor eyesight), whereas intentional non-adherence arises when the patient rejects either the doctor's diagnosis or recommended treatment. Between 25% and 75% of people with epilepsy do not follow prescribed drug regimens, leading to uncontrolled seizures and reduced quality of life.

3Health-care seeking behaviour and adherence are complex issues affected by many factors. They can be explored with Andersen's behavioural model, which considers that health service use and adherence are a function of four categories: health-care system factors (eg, distance to health facilities); predisposing factors (eg, health beliefs); enabling factors (eg, transportation); and need, which refers to severity of illness and whether people judge their illness to be of sufficient magnitude to consult health services.

4We determined the treatment gap, investigated the reliability of self-reporting of adherence, and examined factors associated with failure to seek biomedical treatment and non-adherence in people with epilepsy in a rural area of Kenya.

5We undertook a cross-sectional survey and risk-factor analysis of the epilepsy treatment gap. The study was done in the Kilifi health and demographic surveillance system (KHDSS) with 233 881 residents in 2008. The residents are mainly Mijikenda, a Bantu group of nine tribes with Giriama (45%) dominating. The average per head income is about 700 Kenyan Shillings (KES; US$8) per month and about 55% of the population is regarded as poor. Most people (80%) depend on subsistence farming. Literacy is low (45%). KHDSS is served by one district hospital (Kilifi District Hospital, KDH), one health centre, and 12 dispensaries. KDH serves as a primary care centre and first-level referral facility for the district and stocks four AEDs: phenobarbital, phenytoin, carbamazepine, and sodium valproate. The health centres and dispensaries stock only phenobarbital.

6People with epilepsy were identified in a three-stage cross-sectional survey that was done to establish the prevalence of active convulsive epilepsy as previously described. In the first stage, household heads were asked if anybody in the household had convulsions. People who were reported to have seizures were interviewed with an epilepsy-specific questionnaire and those who were classed as positive were invited to be assessed by a clinician. Active convulsive epilepsy was defined as two or more unprovoked tonic or clonic seizures in a lifetime, of which at least one seizure was within the 12 months preceding the survey, since this criterion is used for prescription of AEDs in Kenya, is used in other parts of Africa,, and is recommended by the International League Against Epilepsy for calculation of the treatment gap,, and there is poor documentation of previous seizures and recall beyond a year in this region. The study was reviewed and approved by the Kenya Medical Research Institute and national scientific and ethical review committees and the Swiss Tropical and Public Health Research Committee. All participants or their carers provided written informed consent.

7Questionnaires based on those used in previous studies of epilepsy in this area,,,, and elsewhere, were developed in English and translated into local languages. In addition to the outcomes of not seeking treatment and non-adherence to AEDs, we obtained data for several other items: demographic details of the person with epilepsy (age and sex), social characteristics (religion, education, occupation, and marital status) of either the person with epilepsy or their carer if the carer made decisions about seeking treatment and taking AEDs if prescribed; socioeconomic data for the households based on assets; accessibility factors (distance from home to health facility, and payment for AEDs); severity of epilepsy (duration of epilepsy, seizure frequency, type of seizures, and injury during a seizure); stigma; and beliefs and attitudes about epilepsy. People with learning difficulties were defined as those who were not orientated for person, place, or the period of day (morning or afternoon) and who had difficulty following instructions—eg, finger nose test—without a neurological deficit as assessed by the clinician. For analysis of non-adherence, details of prescribed regimen were obtained (number of prescribed AEDs, frequency of taking AEDs per day, side-effects of AEDs, and number of years a person with epilepsy has taken AEDs). The field staff administered questionnaires to all people with a confirmed diagnosis of active convulsive epilepsy after written informed consent. If the person with epilepsy was a child or cognitively impaired adult, a caregiver was interviewed. For the risk factor analysis, people who sought biomedical treatment were compared with those who had not sought treatment, and those who adhered (assessed by presence of AEDs in blood) were compared with those who did not have AEDs detected in their blood.

8Data were obtained from households for assets and indicators of household characteristics such as livestock and source of energy for cooking and lighting. Principal component analysis was done to construct a homestead wealth assets index from the range of assets and household characteristics. Households were classified into socioeconomic status quintiles on the basis of this index.

9Perceived stigma and epilepsy beliefs and attitudes were measured with Likert scales. The Kilifi stigma scale for epilepsy had 15 questions and measured only perceived stigma. The Kilifi epilepsy beliefs and attitude scale had 34 questions, divided into five subscales: causes of epilepsy, biomedical treatment of epilepsy, cultural treatment of epilepsy, risk and safety concerns, and negative stereotypes about epilepsy.

10To calculate distance to health facilities, homesteads and health facilities were mapped with a global positioning system and the Euclidean distance was calculated with the coordinates as a reflection of the time taken to travel the distance. The health facility used was defined as the one where people with epilepsy reported seeking treatment most frequently. Treatment-seeking behaviour was established by asking whether the person had ever sought treatment for epilepsy from a health facility before the 2008 epidemiological survey. People answering “yes” were asked to identify the health facility from which they sought treatment.

11Questions were asked about the AEDs that patients were currently taking (actual tablets were shown on a board to aid recognition). Reported adherence was assessed with the Morisky medication adherence scale, a four-item scale with high reliability and validity that has been used to measure adherence in epilepsy.,

12To measure adherence, blood samples (2 mL) were collected. Phenobarbital was measured in all samples, with phenytoin and carbamazepine measured in those from participants taking these AEDs, with a TDx FLx analyser (Abbott Laboratories, Abbott Park, IL, USA). An individual was defined as adherent if AEDs were detected in their blood, with detectable limits of 1·1 μg/mL for phenobarbital, 1·0 μg/mL for phenytoin, and 0·5 μg/mL for carbamazepine. Since the half-lives of phenytoin and carbamazepine (the AEDs with the shortest elimination half-lives used in this area) are greater than 5 h, and it takes at least five half-lives to clear the drug after one dose, the absence of AED in the blood suggests that the drug has not been taken within at least the 24 h before the blood sample was taken. The optimum therapeutic ranges were defined as 10–40 μg/mL for phenobarbital, 10–20 μg/mL for phenytoin, and 4–12 μg/mL for carbamazepine.

13Data were double-entered and verified with a MySQL V5 database. Statistical analyses were done with Stata versions 11 and 12. We investigated treatment seeking for all study participants interviewed, whereas non-adherence was investigated only in those who reported taking AEDs and gave a blood sample. The epilepsy treatment gap was defined as the number of people with active epilepsy who had not accessed biomedical services or who were not on treatment or were on inadequate treatment, expressed as a percentage of the total number with active epilepsy. The sensitivity, specificity, and confidence intervals were calculated in Stata. Concordance was assessed between self-reported adherence and AED detection in blood with the κ statistic. In exploratory analyses, we examined the distributions of the dependent variables before selecting the appropriate test statistic. We used Pearsons' χ2 test to measure associations between categorical sociodemographic characteristics and failure to seek treatment or non-adherence.

14Univariate associations for two outcomes (failure to seek treatment or non-adherence to AED) were investigated with logistic regression, adjusted for age. We included variables with p≤0·20 in a multivariable model using a backward elimination process. Variables with p≤0·05 in multivariable analysis were retained in the final model. Attributable fractions were calculated with the final multivariable logistic regression model by a user-written command “punaf” in STATA version 12.,

15The sponsor of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in this study and had final responsibility for the decision to submit for publication.

16The epidemiological survey included 232 176 (99%) of 233 881 people living in the KHDSS region during 2008. 699 people with epilepsy were identified, of whom 673 (96%) were interviewed (figure). 392 (58%) participants were younger than 18 years. Many adults (133 [47%]) had no formal education.

17Of the 673 people with epilepsy who were interviewed, 499 (74%) reported seeking treatment for epilepsy from a health facility, of whom 395 (79%) used KDH as their primary source of AEDs (appendix p 1). 502 (75%) gave blood samples. People who did not give blood samples were significantly younger, were more likely to have traditional religious (animist) beliefs, were more likely to believe that epilepsy responds to biomedical treatment, were more likely to be married (if adults), less likely to pay for AEDs, had epilepsy for a shorter period, had less frequent seizures, and were less likely to injure themselves during the seizures compared with people who provided samples (appendix pp 2–3). Of the 502 people in whom blood concentrations of AEDs were measured, 393 reported seeking treatment, of whom 117 (40%) were on polytherapy. Phenobarbital was detected in the blood of 151 people with epilepsy, phenytoin in 50, and carbamazepine in 48. Four patients reported taking sodium valproate (which was not measured in blood), but two of these patients were taking one of the other AEDs, and the other two had not collected their drug from KDH (the only source of sodium valproate) within the past 3 months and were classified as non-adherent. The epilepsy treatment gap based on AEDs detected in blood samples was 62·4% (95% CI 58·1–66·6).

18We studied 19 variables as potential predictors of people not seeking biomedical treatment (table 1). Of these, 16 had univariate p values of 0·20 or less and were included in the multivariable logistic regression model. From the multivariable analysis, people with epilepsy were less likely to seek treatment if they had traditional animistic religious beliefs, had negative attitudes about biomedical treatment, lived 20 or more km from where they sought health care, reported paying for AEDs, had epilepsy for longer than 10 years, had learning difficulties, and had focal seizures (table 2). Attributable fraction analysis indicated that treatment seeking could improve by 11% (95% CI 1–21) if people with epilepsy thought that biomedical treatment would help, 40% (21–54) if AEDs were free, and 12% (3–21) if the health facilities where people obtained AEDs were less than 20 km from their homes.,

19Of the 499 people who reported seeking biomedical treatment, only 385 (77%) were prescribed AEDs and the person with epilepsy or their carer were able to answer the questions on the Morisky scale. On the basis of Morisky scores, 179 (46%) people with epilepsy were adherent (score of 0) and 206 (54%) were non-adherent (score >0; table 3). Of the 502 who had AEDs measured in blood, 189 (38%) had AEDs detected of whom 72 reported taking AEDs. In the 313 who did not have any AED detected, 60 (19%) reported taking AEDs. 48 (25%) of the 189 people with AEDs detected in their blood had concentrations in the detectable but not optimum range and 141 (75 %) were in the optimum range. Overall, the sensitivity and specificity of reported adherence compared with AEDs detected in the blood were 38·1% (95% CI 31·1–45·4) and 80·8% (76·0–85·0), respectively. The concordance between self-reported adherence and detection in blood for the three prescribed AEDs occurred in 72 (55%) of 132 people with epilepsy. There was little agreement between adherence as measured by blood samples and the Morisky scores.

20We studied 25 variables as potential predictors of non-adherence (table 4). 15 were retained in the multivariable logistic regression model. People with epilepsy were less likely to adhere if they had negative attitudes such as feeling that they were rejected or had taken AEDs for more than 5 years (table 5). Attributable fraction analysis indicated that adherence could be improved by 20% (95% CI 7–30) if attitudes could be modified.

21In Kenyans with active convulsive epilepsy who had similar characteristics to people with epilepsy in other parts of Africa,, we found that the epilepsy treatment gap was 62% as measured by detection of AEDs in blood samples. This finding represents a reduction from 74% measured during a survey in 2003 and could be attributed to setting up an epilepsy clinic with a continuous supply of AEDs and sensitisation of the community. However, the sensitivity and specificity of self-reporting remained poor. The factors associated with reduced treatment seeking are different to those associated with reduced adherence to AEDs. Modifiable risk factors for failure to seek treatment are negative attitudes about biomedical treatment, payment for AEDs, and distance from health facilities providing AEDs. Adherence to AEDs could be improved by reduction of negative attitudes to epilepsy (panel). Systematic review We searched Medline (1966, to May, 2012), ISI Web of Science (1966, to May, 2012), and Cochrane reviews with the search term “epilepsy treatment gap”. Searches were restricted to human studies. We found 281 references, most of which we reviewed in a previous paper and a recent review. We reviewed the outstanding 31 references. We found that the epilepsy treatment gap was attributed to cultural beliefs, inadequate skilled manpower, cost of treatment, and unavailability of antiepileptic drugs. Interpretation This study showed that self-reporting of adherence to antiepileptic drugs (AEDs) might not be reliable in rural Africa, and that risk factors for the failure to seek biomedical treatment for epilepsy and the poor adherence to prescribed treatment are different. Treatment seeking could be improved by modification of attitudes to epilepsy and provision of free AEDs in health facilities near to the residence of the person with epilepsy. Adherence could be improved by changes in attitudes towards epilepsy.

22Systematic review

23We searched Medline (1966, to May, 2012), ISI Web of Science (1966, to May, 2012), and Cochrane reviews with the search term “epilepsy treatment gap”. Searches were restricted to human studies. We found 281 references, most of which we reviewed in a previous paper and a recent review. We reviewed the outstanding 31 references. We found that the epilepsy treatment gap was attributed to cultural beliefs, inadequate skilled manpower, cost of treatment, and unavailability of antiepileptic drugs.

24Interpretation

25This study showed that self-reporting of adherence to antiepileptic drugs (AEDs) might not be reliable in rural Africa, and that risk factors for the failure to seek biomedical treatment for epilepsy and the poor adherence to prescribed treatment are different. Treatment seeking could be improved by modification of attitudes to epilepsy and provision of free AEDs in health facilities near to the residence of the person with epilepsy. Adherence could be improved by changes in attitudes towards epilepsy.

26People with epilepsy who held traditional animistic religious beliefs were less likely to seek biomedical treatment than were Christians and Muslims, possibly because of misconceptions and superstitions associated with epilepsy in Kilifi. People with epilepsy with negative beliefs and attitudes about biomedical treatment of epilepsy were less likely to seek treatment from health facilities, suggesting that modification of patients' beliefs and attitudes is an important step in improvement of treatment seeking.,

27Other factors associated with failure to seek treatment were distance to health facilities and payment for AEDs. The geographical proximity of health services to people's homes is an important factor that affects use of health services, particularly in resource-poor settings where density of biomedical health facilities is often low. The 1997 Kenyan health policy strategic framework states that all households should have access to health services within a 5 km range. In KHDSS, there is a high density of biomedical health facilities within reach by people with epilepsy, but only KDH has an epilepsy clinic with a continuous supply of AEDs and AEDs other than phenobarbital. The erratic supplies of AEDs in the peripheral clinics (JAC, unpublished) might have reduced the use of these peripheral facilities. People with epilepsy who reported paying for AEDs were less likely to seek treatment than were people who did not pay. The cost of AEDs in peripheral facilities was 10 KES (US$0·13) for a 1-month supply of phenobarbital in 2008, whereas at KDH it was 40 KES ($0·51). This cost of AEDs, although small, is significant when combined with indirect costs of seeking care—eg, transport costs.

28Of the epilepsy-specific factors, people with epilepsy who had seizures for more than 10 years might be less likely to seek treatment compared with people with disease of shorter duration because they had learnt to cope with their disorder. People with focal seizures were less likely to seek treatment than were those with generalised seizures, although focal seizures are associated with symptomatic causes and are more likely to be associated with behavioural problems. Most of the seizures were focal becoming generalised, and being aware at the onset of seizure might have made people less inclined to seek treatment because of the stigma. People with epilepsy and learning difficulties were less likely to seek treatment, perhaps because they were not brought to biomedical facilities by their families or because of their difficulties in understanding the importance of treatment.

29Patient self-reporting using reliable and valid questionnaires is thought to be an efficient and cost-effective method to assess adherence, although some patients might underestimate or overestimate their adherence. We found that the sensitivity and specificity of self-reporting compared with detection of AEDs in blood samples was low, suggesting that self-reporting is not a reliable measure of adherence in people with epilepsy in this community and probably in many countries with low and middle incomes. This finding questions the reliability of data in systematic reviews of the epilepsy treatment gap., Furthermore, we did not find that the Morisky scale was useful for measuring adherence in this population. Although the scale has been used in other studies of epilepsy,, it has not been compared with AED detection in blood. There was no significant difference in adherence based on blood samples among the different AEDs (data not shown), despite the different half-lives (phenobarbital, 48–168 h; phenytoin, 7–42 h; carbamazepine, 5–26 h).

30Non-adherence was associated with established epilepsy, which is in accord with previous reports that people with epilepsy who had taken AEDs for a longer duration were less likely to adhere. Non-adherence was also associated with negative beliefs and attitudes about epilepsy as reported previously.

31Our study had several limitations. Non-adherence was reported at only one timepoint as opposed to monitoring over time, which might capture variations such as differences in timing of administration, missing doses of AEDs, or taking AEDs just before clinical assessment. We did not study people with non-convulsive epilepsy since the survey was designed to capture only those with active convulsive epilepsy. Recall and reporting bias of self-reported answers might have occurred, and there were significant differences between people who gave blood samples for measurement of adherence and those who did not. This study was done in an area where we have been studying epilepsy since 2003, although the treatment gap is still 62%. We did not control for quality of care, which might play a part in determining health service use.

32In this rural area of Kenya, almost two-thirds of people with epilepsy were not adequately treated and the sensitivity and specificity of self-reported adherence were poor. Modification of beliefs and attitudes and provision of a free and constant supply of AEDs to nearby health facilities could substantially reduce the treatment gap.