Factors Associated with Seizure Attacks in Epileptic Patients at Felege Hiwot Referral Hospital: Application of Linear Mixed Model

Awoke Seyoum Tegegne^1*, Denekew Bitew Belay¹ and Yirga Mekonnen^2
*Correspondence: Awoke Seyoum Tegegne, Department of Statistics, Bahir Dar University, Bahir Dar, Ethiopia, Email:

Keywords

Average seizure attacks • Linear mixed model • Seizure freedom • Epileptic patient

Abbreviations

AEDs: Anti-Epileptic Drugs; AIC: Akaki Information Criteria; BIC: Bayesian Information Criteria; LRT: Likelihood Ratio Test

Introduction

Epilepsy is one of the most chronic non-communicable disease of the brain which is the conditions of: At least two unprovoked seizures occurring more than 24 h apart; one unprovoked seizure and diagnosis of an epilepsy syndrome [1]. It is known that due to the characterization of recurrent seizures, it needs long term treatment periods [2].

Globally, epilepsy affected 70 million people, with 50.4 per 100,000 people newly diagnosed per year with prevalence of 5 to 8 cases per 1000 depending on the age group of epileptic. Among these, around 85% of people with epilepsy are from developing countries and around 100 million people experienced at least one epileptic seizure in their lives [3-5]. In addition to the above global effect of an epilepsy in sub-Saharan Africa countries like Ethiopia also affects approximately 10 million people annually and it has high prevalence in the study area [6].

Epileptic patients have poor health outcomes and serious physical, psychological, social and economic effects. Although epilepsy is a treatable condition with cheap medication, the treatment gap in developing countries remains alarmingly high [7]. Treatment gap varies from 10% in developed countries to 75% in low-income countries [8].

Approximately 80–90% of epileptic patients have difficulties in accessing appropriate treatment in developing countries which causes exacerbation of social isolation, unemployment, dependent behaviour, psychological issues and reduced quality of life [9]. Maintaining seizure-free state in the absence of treatment related effects is the crucial goal of treating for epileptic patients and using cost-effective approach to reduce the frequency by avoiding drug interaction and side effects [10,11].

In Ethiopia, epilepsy is a major public health burden, which affects about 29.5 per 1000 population [12]. A study conducted in northern Gondar in Ethiopia indicates that epileptic people face different social, psychological and physical problems as a result of their illness which may include stigma and discrimination [13]. Another studies in Ethiopia with regard to the cause and forms of treatment of epilepsy, shows that there is a widespread belief that cause of epilepsy as demon possessions or ancestors’ spirits and this hinders the patients not to use modern antiepileptic drugs [14]. Traditional treatments are commonly used in Ethiopia and only 1.6% had been treated with modern antiepileptic drugs in rural community and 13% in urban community of the country [15].

One of the previous studies indicates that about 58% of the patients who developed generalized tonic–clonic seizure with higher frequency, 23.2% of them died and epileptic patients who were not seizure free increases the accident of disability, death and drug side effects [16]. Most of the studies conducted previously in the study area focused the problems related with this non-communicable disease and didn’t investigate for factors affecting the severity of the disease. Knowing the predictors of the disease helps to health practitioners and families of patients in conducting treatments [17]. Different studies done on the epileptic patients are cross sectional and used binary logistic regression but these methods cannot handle these complications, while there are known difficulties to use these methods, verifying their assumptions, and interpreting the results correctly. And ignoring different sources of correlation has severe consequences which cause s higher false positive rates, invalid confidence intervals and underestimate the standard errors [18]. Therefore, the current study aimed to investigate factors affecting average seizure attacks in epileptic patients considering linear mixed model application in the case of Felege-Hiwot Referral Hospital, Bahir Dar, Ethiopia.

Materials and Methods

Study area and study design

A retrospective cohort study design was conducted for the current investigation on epileptic patients. The study was conducted with relevant information, recorded by health staff during treatment, from the medical charts of epileptic patients. The study was conducted at Felege-Hiwot specialized and Teaching Hospital, North-west Ethiopia. This hospital serves as a specialized, teaching and referral hospital for the people who came from different district hospitals in the surrounding areas of North-central Ethiopia. It is the only governmental hospital in the study area, delivering treatment for epileptic patients. The hospital also provides a training service for health practitioners/students for apprenticeship from various health institutions, including Bahir Dar University.

Source of population and data collection procedures

Epileptic patients whose age is at least 18 years old, started treatment at the hospital indicated above were used as source of population for this study. The study exclusively used secondary sources. For the current study, the data were collected by trained health staff with in the hospital from the medical chart of epileptic patients at the hospital whose follow-ups were from November 2014 up to 31 December 2019. Therefore, a data extraction check-list was designed by the authors in consultation of health staff with in the hospital.

Sample size and sampling procedures

In the study area, among the total of 1652 newly registered epileptic patients reported by neurologic clinic at the hospital in the study period, only 131 patients who satisfied the inclusion criteria were included in the study.

Variables under study

The response variable for this study was the seizure attacks in epileptic patients. This variable was recorded every six months for epileptic patients. Hence, the data were collected at equal interval of period taken every six months.

The potential predictor variables for current investigation were classified as socio-demographic and clinical variables. Sociodemographic variables were: Sex, age at base line, religion, residence area, marital status, and occupation and education level. The clinical variables were: Age at seizure onset, age difference (age at diagnosis–age of seizure onset), drug amount used at base line, drug type, family disease history, Intellectual disability, history of febrile convulsion, co-morbidity, cure belief, drug addiction, traumatic brain (head injury) and medication source. The categories of predictor variables are indicated in Table 1.

Statistical data analysis

The current investigation used longitudinal data to see disease progression and changes of outcome(s) over time. The changes or variation were conducted at individual or group patterns over time. The advantage of longitudinal data over cross-sectional studies is that, it helps to see both group and individual effects over time and have more statistical power [19,20]. Exploratory data analysis was conducted to detect the individual trends through the study period. Univariable and multivariable data analysis were also conducted to assess the significant predictors in current investigation. Data analysis was conducted using R software. The study was conducted for 60 months in such a way that the data was recorded at every six months. Exploratory data analysis was also conducted in order to assess various associations and patterns exhibited in the data. Additionally, the individual profile plots, mean structure plots and variance plots were obtained in order to obtain the pattern at individual and group levels. The plotting of the expected seizure attacks over time, graph of different subgroups/categories were included the investigation to illustrate the relationship between the expected seizure attacks in epileptic patients.

Results

The categorical socio-demographic variables under study are indicated in Table 1. Table 1 shows that about 61.8% of the patients were male, 70.2% of them had family disease history, 74% of patients developed intellectually disable, 20.6% of them had history of febrile convulsion, and about 69.5%believed to cure when they took AEDs (anti-epileptic drugs). Among the study subjects, (34.4%) of them took traditional drugs to cure from the disease, 20.6% of the patients got head injuries, 57.3% of patients got free medication source, 54.5% took PHB (Phenobaribitone) and about the quarter of patients (26%) took a combination of drugs and 13% and 6.9% took PHT (Phenytoin) and others type of anti-epileptic drugs respectively.

Table 1. Baseline characteristics of epileptic patients.

Among the patients, about 74% of them took monotherapy antiepileptic drugs at the beginning of follow ups, 39.7% were high adherent patients whereas 30.5% them were low adherent patients.

The covariates under current investigation are also summarized in Table 2. From Table 2 the average age of epileptic patients included in this study was 30.5 with standard deviation of 12.5 years. The average time of getting AEDs treatment after the onset of seizure was 12.93 years. At the beginning of taking AEDs, they took an average of 100.04 mg drugs. The average age seizure onset of epileptic patients was 17.6 years with standard deviation 12.55 years.

Table 2. Description of continuous variables.

The current investigation indicates that, as visiting time increased the average seizure attacks decreased. For instance, the average seizure attacks at the baseline was 7.51 with standard deviation 8.91 and at the tenth visit, it was reduced to 0.4 with standard deviation of 0.41.

The repeated measurement for each epileptic patient over time indicates that there was a visible pattern for all patients. The individual profile plot also shows that there was larger variability at the beginning/commencement of AEDs and this variability decreased over time or becomes minimal at the end. Hence, their variation about average seizure attack was larger at initial time and comes closer and closure to each other at the end. The variation of the variable of study was high at initial time and very small at the end of study period. The average seizure attack levels for some patients were going down, and others were going up over the time points. This implies that not all the patients responded the same to AEDs treatment and this magnifies the relevance of random component of a mixed effects model to address the random effects part in addition to fixed effects.

The covariance structure for current investigation was also assessed. To identify the appropriate covariance structure, the commonly used covariance structures namely Independence (IND), Compound Symmetry (CS), first order autoregressive AR (1), Toeplitz (Toep) and Unstructured (UN) were considered. Among the potential covariate structures indicated above, first order autoregressive (AR (1)) covariance structure was selected due to its smallest AIC and BIC compared to the remaining covariance structures. Similar to covariance structures, an appropriate random effect model was also selected by using likelihood ratio test and this indicates that model with random intercept and random slope was a better fit model. This model allows the intercept and coefficient to vary randomly among individuals.

A univariable analysis was performed in order to see the effect of each covariate on the study variable. The significant variables that should be included to multivariable data analysis was selected using stepwise variable selection criteria in linear mixed effect model analysis. In such selection procedures, a 20% level of significance was taken in to consideration. Among the potential predictor variables, those were significant at univariate data analysis and included in the multivariable analysis are indicated in Table 3. The multivariate data analysis was conducted using the selected random intercept and random slope model with unstructured covariance structure.

Table 3 revealed that for a unit increased in visiting time, the expected seizure attack of epileptic patients was decreased by 0.088 keeping all other variables constant. Even though cubic time has small regression coefficient, it improves the overall model effects and significance effect on the expected seizure attacks.

Comparing the study variables between intellectual dis abilities with non-mental dis abilities, the expected seizure attacks for intellectual disabilities was increased by 0.081 (P-value=0.005) as compared to non-mental disability, keeping other variables constant. As amount of drug used at commencement of the AEDs treatment increased by hundred mg, the expected seizure attacks for epileptic patients was increased by 0.05 (p-value<0,001), keeping the other variable constant.

The time when a patient comes for diagnosis had also significant effect for expected seizure attacks on epileptic patients. Hence, as patient’s diagnosis for the disease being late by one unit, the expected seizure attack was increased by 0.0033 (P-value=0.0042), keeping the other variables constant. The expected seizure attacks for epileptic patients who got anti-epileptic drugs with payments was greater by 0.062 (p-value=0.04) as compared to those patients who got the drug freely, keeping the other things constant. On the other hand, for epileptic patients who believed to be cured when they started AEDs, the expected seizure attacks were decreased by 0.054 (p-value=0.03) as compared to patients whose belief was not to cure from the disease, keeping all other variables constant.

Table 3. Multivariable linear mixed model results.

For patients who had a febrile convulsion history, the expected seizure attacks were increased by 0.076 (p-value=0.03) as compared to patients without febrile convulsion history, keeping all other conditions constant. Similarly, for epileptic patients, who had a family disease history, the expected seizure attacks were increased by 0.054 (p-value=0.04) as compared to patients without family disease history, keeping other thing constant. The expected seizure attacks, for patients who had other diseases (co-morbidity), was increased by 0.076 (p-value=0,004) as compared to patients who had no comorbidity, keep the other things constant.

As the time for treatment of seizure onset increased by one year from their age, the expected seizure attacks of epileptic patients reduced by 0.003 (P-value=0.02), keeping all other variables constant. Adherence had significant effect for the amount of expected seizure attacks. Hence, the expected seizure attacks for high adherent epileptic patients was decreased by 0.11 (p-value<0.001) as compared to low adherent patients, keeping the other variables constant.

From the random part of model, the correlation between random intercept and random slope (-0.962755) shows that epileptic patients who had higher initial seizure attacks value had small slopes, which suggests that epileptic patients who are initially high seizure attacks reduced at a greater rate for the consecutive visits.

Discussion

The current investigation identified important predictor variables for the variable of study. The baseline seizure attack is one of the significant variables in current investigation. Epileptic patients with higher seizure attacks before diagnosis, leads a risk of its control, but for high adherent patients in consecutive visiting times, the magnitude of seizure attacks decreased through time. Hence, seizure attacks are treatable. This is agreed with one of the previous studies. The longitudinal study, conducted in current investigation, revealed that time has significant contribution for improvement of the variable of interest. The magnitude of seizure attack decrease as visiting time/ follow ups increased.

Traumatic brain injured epileptic patients are worsen prognosis of seizure attacks compared to no traumatic brain injury patients without considering their sex difference. The result obtained in current investigation agreed with one of the previous studies.

The amount of drug use at the commencement of their treatment had significant effect for the variable of interest. Patients with high initial drug use have higher seizure attack as compared to low drug users. Hence, patients should start treatments with lower among of drugs. This result is similar with one of the previous studies. Epileptic patients with family disease history are significantly affected the study variables, hence patients with family disease history have severe seizure attacks as compared to without family diseases history.

Patients who believed that the disease is curable have low severity of seizure attacks as compared to patients who do not believe that the disease is curable. This result agreed with results obtained from one of the previous studies. However, the current result contradicted with the results obtained in one of the other study, which states that no different for the two groups. The potential reason for this difference may be difference of study subjects taken in different areas or different cultures, methods of study, length of the study and sample size. This needs further investigation. Patients who had co-morbidity are experience more seizure attacks compared to non-co-morbid patients this result is in line with the studies conducted previously.

Epileptic patients who got free access of drug for treatment have better improvement with regard to seizure attacks. Hence, income plays significant role for easy improvement of the disease. Related with access of drugs, higher adherent epileptic patients highly reduced the expected seizure attacks compared to medium and lower adherent patients. Hence, non-adherent epileptic patients develop severe seizure attacks and low adherent patients took longer time to cure from the disease.

The difference of age at seizure onset and age of diagnosis which is duration of seizure and mental disability are associated with severity of seizure attacks. Hence, epileptic patients who come lately for treatment have severe seizure attacks. The result obtained in current study is supported by another study. Mental disable epileptic patients have severe seizure attacks as compared to the others. This result is supported by another study conducted previously.

Age has an effect on the expected seizure attack, which indicates that as age of seizure onset increased, the severity of its attack decreased which is supported by another study. Amount of drugs used at the commencement of AEDs have an association with the reduction of seizure-attacks. This is consistent to another study conducted previously but contradicted with another study. The potential reason for this difference may be difference of study area, length of study and sample size differences. Patients who have history of febrile convulsion seizure have severe seizure attacks. This result is supported by another study but it is contradicted with another previously conducted study.

Conclusion

Seizure attack in epileptic patients was severe at the commencement of AEDs and the severity decreased through time because of the treatment and adherence of AEDs. Hence, the disease is treatable and manageable for high adherent epileptic patients considering proper medication described by the health staff. The adherence level in current investigation revealed that there were higher seizure attacks for epileptic patients at commencement of their treatment and rapid reduction of the disease for epileptic patients with high adherence level as compared to low and medium adherence level of epileptic patients.

From the linear mixed model; increasing of follow-up time, the increase of age at seizure onset, being high adherence and believing to cure epileptic patients at the start of taking antiepileptic drugs significantly reduced the expected seizure attacks. And higher difference of age at diagnosis and age at seizure onset or duration of seizure till diagnosis, intellectual disability/retardation, having a history febrile convulsion, patients who didn’t get drug access freely and patients who had family disease history associated with increase of seizure attacks.

As a result, patients who experience seizure episodes should visit health facilities immediately. And those patients who had treated with payments should be supported by the health institutions/governments in order to decrease the progression of the number of seizure attacks. Additionally, health professionals should underline the newly diagnosed epileptic patients who were coming to lately after the start of seizure onset, those patients who had family history of seizure and febrile seizure should be diagnosed regularly to decrease the progression of seizure attacks. Due attention should be given for Epileptic patients who had family disease history, febrile history and low adherent patients. The result obtained in current investigation helps for policy makers and health staff for amendment of policy issues and proper intervention in reducing the severity of seizure attacks in epileptic patients.

This study was not without limitation. The data were taken at one treatment site and including more sites may have additional information. The authors recommend further investigation for future studies by including additional treatment sites.

Ethical Approval and Consent of Participate

Ethical approval certificate had been obtained from Bahir Dar University Ethical approval committee, Bahir Dar University, Ethiopia with reference number: RCS/1412/2012. In data collection, there was no written or verbal consent from participants because of use of secondary data.

Consent for Publication

This manuscript has not been published elsewhere and is not under consideration by any other journal. Authors agreed this manuscript to be submitted in this journal for publication.

Availability Data and Privacy Regulations

The data used in current investigation is available with in the corresponding authors. The data accessed in current investigation complied with relevant data protection and privacy regulations, and that this study was conducted in accordance with the Declaration of Helsinki.

Competing Interests

There is no financial and non-financial competing interest between an author and institutions.

Funding

Not applicable.

Author Contributions

YM contributed to proposal development, supervision during data collection, data analysis and drafting, DB contributed data analysis, writing the manuscript and editing. AS Tegegne contributed in writing, correctly editing and finalizing the manuscript.

Authors have agreed to submit the manuscript to this journal to publish as original research. Authors also agreed on their order and agreed to be accountable for all aspects of the work.

Disclosure

The author declares that there is no conflict of interest regarding the publication of this manuscript.

References

1. Fisher, Robert S, Carlos Acevedo, Alexis Arzimanoglou, and Alicia Bogacz, et al. "ILAE official report: a practical clinical definition of epilepsy." Epilepsia 55 (2014): 475-482.

   [Crossref] [Google Scholar] [PubMed]

2. Narayanan, Jaishree, Robert Frech, Shaun Walters, and Vimal Patel, et al. "Low dose verapamil as an adjunct therapy for medically refractory epilepsy–an open label pilot study." Epilepsy Research 126 (2016): 197-200.

   [Crossref] [Google Scholar] [PubMed]

3. Dessie, Getenet, Henok Mulugeta, Cheru Tesema Leshargie, and Fasil Wagnew, et al. "Depression among epileptic patients and its association with drug therapy in sub-Saharan Africa: A systematic review and meta-analysis." PloS One 14 (2019): e0202613.

   [Crossref] [Google Scholar] [PubMed]

4. Katchanov, Juri, and Gretchen L. Birbeck. "Epilepsy care guidelines for low-and middle-income countries: From WHO mental health GAP to national programs." BMC Med 10 (2012): 1-5.

   [Crossref] [Google Scholar] [PubMed]

5. Wahab, Abdul. "Difficulties in treatment and management of epilepsy and challenges in new drug development." Pharmaceuticals 3 (2010): 2090-2110.

   [Crossref] [Google Scholar] [PubMed]

6. Garnett, William R. "Antiepileptic drug treatment: outcomes and adherence." Pharmacotherapy 20 (2000): 191S-199S.

   [Crossref] [Google Scholar] [PubMed]

7. Kerr M, M Scheepers, M Arvio, and J Beavis, et al. "Consensus guidelines into the management of epilepsy in adults with an intellectual disability." J Intellect Disabil Res 53 (2009): 687-694.

   [Crossref] [Google Scholar] [PubMed]

8. Almu, Shitaye, Zerihun Tadesse, Paul Cooper, and Richard Hackett. "The prevalence of epilepsy in the Zay Society, Ethiopia-an area of high prevalence." Seizure 15 (2006): 211-213.

   [Crossref] [Google Scholar] [PubMed]

9. Berhanu, Shibru, Shitaye Alemu, Jilalu Asmera, and Martin Prevett. "Primary care treatment of epilepsy in rural Ethiopia." Ethiop J Health Dev 16 (2002): 235-240.

   [Crossref] [Google Scholar]

10. Berhanu, Shibru, and Martin Prevett. "Treatment of epilepsy in rural Ethiopia: 2 year follow-up." Ethiop J Health Dev 18 (2004).

    [Google Scholar]

11. Worku, Dawit Kibru, Yared Mamushet Yifru, Douglas G. Postels, and Fikre Enquselassie Gashe. "Prevalence of depression in Parkinson’s disease patients in Ethiopia." J Clin Mov Disord 1 (2014): 1-12.

    [Crossref] [Google Scholar] [PubMed]

12. Amare, Amanuel, Guta Zenebe, Julie Hammack, and Gail Davey. "Status epilepticus: clinical presentation, cause, outcome, and predictors of death in 119 Ethiopian patients." Epilepsia 49 (2008): 600-607.

    [Crossref] [Google Scholar] [PubMed]

13. Nasir, Beshir Bedru, Yared Mamushet Yifru, Ephrem Engidawork, and Meron Awraris Gebrewold, et al. "Antiepileptic drug treatment outcomes and seizure-related injuries among adult patients with epilepsy in a tertiary care hospital in Ethiopia." Patient Relat Outcome Meas 11 (2020): 119-127.

    [Crossref] [Google Scholar] [PubMed]

14. Gibbons, Robert D, Donald Hedeker, and Stephen DuToit. "Advances in analysis of longitudinal data." Annu Rev Clin Psychol 6 (2010): 79-107.

    [Crossref] [Google Scholar] [PubMed]

15. Tan, Xianming, Mariya P. Shiyko, Runze Li, Yuelin Li, and Lisa Dierker. "A time-varying effect model for intensive longitudinal data." Psychol Methods 17 (2012): 61.
16. Zeger, Scott L, and Kung�Yee Liang. "An overview of methods for the analysis of longitudinal data." Stat Med 11 (1992): 1825-1839.

    [Crossref] [Google Scholar] [PubMed]

17. Hitiris, Nikolas, Rajiv Mohanraj, John Norrie, and Graeme J. Sills, et al. "Predictors of pharmacoresistant epilepsy." Epilepsy Res 75 (2007): 192-196.

    [Crossref] [Google Scholar] [PubMed]

18. Mekonnen, Fenta Haile, Workie Demeke Lakew, Zike Dereje Tesfaye, and Prafulla Kumar Swain. "Statistical models for longitudinal zero-inflated count data: application to seizure attacks." Afr Health Sci 19 (2019): 2555-2564.

    [Crossref] [Google Scholar] [PubMed]

19. Waja, Tsegereda, Jemal Ebrahim, Zegeye Yohannis, and Asres Bedaso. "Prevalence of alcohol use disorders and associated factors among people with epilepsy attending Amanuel Mental Specialized Hospital, Addis Ababa, Ethiopia." Neuropsychiatr Dis Treat 12 (2016): 2989-2996.

    [Crossref] [Google Scholar] [PubMed]

20. Zewudie, Ameha, Yitagesu Mamo, Desalegn Feyissa, and Mohammed Yimam, et al. "Epilepsy Treatment Outcome and its Predictors among Ambulatory Patients with Epilepsy at Mizan�Tepi University Teaching Hospital, Southwest Ethiopia." Neurol Res Int 2020 (2020): 8109858.

    [Crossref] [Google Scholar] [PubMed]