Construction and validation of an algorithm to separate focal and generalised epilepsy using clinical variables: A comparison of machine learning approaches.

PURPOSE: Epilepsy type, whether focal or generalised, is important in deciding anti-seizure medication (ASM). In resource-limited settings, investigations are usually not available, so a clinical separation is required. We used a naïve Bayes approach to devise an algorithm to do this, and compared its accuracy with algorithms devised by five other machine learning methods. METHODS: We used data on 28 clinical variables from 503 patients attending an epilepsy clinic in India with defined epilepsy type, as determined by an epileptologist with access to clinical, imaging, and EEG data. We adopted a machine learning approach to select the most relevant variables based on mutual information, to train the model on part of the data, and then to evaluate it on the remaining data (testing set). We used a naïve Bayes approach and compared the results in the testing set with those obtained by several other machine learning algorithms by measuring sensitivity, specificity, accuracy, area under the curve, and Cohen's kappa. RESULTS: The six machine learning methods produced broadly similar results. The best naïve Bayes algorithm contained eleven variables, and its accuracy was 92.2% in determining epilepsy type (sensitivity 92.0%, specificity 92.7%). An algorithm incorporating the best eight of these variables was only slightly less accurate - 91.0% (sensitivity 89.6%, and specificity 95.1%) - and easier for clinicians to use. CONCLUSION: A clinical algorithm with eight variables is effective and accurate at separating focal from generalised epilepsy. It should be useful in resource-limited settings, by epilepsy-inexperienced doctors, to help determine epilepsy type and therefore optimal ASMs for individual patients, without the need for EEG or neuroimaging.

An epilepsy type algorithm developed in India is accurate in Sudan: A prospective validation study.

PURPOSE: The effects of epilepsy are worse in lower- and middle-income countries (LMICs) where most people with epilepsy live, and where most are untreated. Correct treatment depends on determining whether focal or generalised epilepsy is present. EEG and MRI are usually not available to help so an entirely clinical method is required. We applied an eight-variable algorithm, which had been derived from 503 patients from India using naïve-Bayesian methods, to an adult Sudanese cohort with epilepsy. METHODS: There were 150 consecutive adult patients with known epilepsy type as defined by two neurologists who had access to clinical information, EEG and neuroimaging ("the gold standard"). We used seven of the eight variables, together with their likelihood ratios, to calculate the probability of focal as opposed to generalised epilepsy in each patient and compared that to the "gold standard". Sensitivity, specificity, accuracy, and Cohen's kappa statistic were calculated. RESULTS: Mean age was 28 years (range 17-49) and 53% were female. The accuracy of an algorithm comprising seven of the eight variables was 92%, with sensitivity of 99% and specificity of 72% for focal epilepsy. Cohen's kappa was 0.773, indicating substantial agreement. Ninety-four percent of patients had probability scores either less than 0.1 (generalised) or greater than 0.9 (focal). CONCLUSION: The results confirm the high accuracy of this algorithm in determining epilepsy type in Sudan. They suggest that, in a clinical condition like epilepsy, where a history is crucial, results in one continent can be applied to another. This is especially important as untreated epilepsy and the epilepsy treatment gap are so widespread. The algorithm can be applied to patients giving an individual probability score which can help determine the appropriate anti-seizure medication. It should give epilepsy-inexperienced doctors confidence in managing patients with epilepsy.

A Neurology Advanced Referral Management System (NARMS) Reduces Face-to-Face Consultations By Over Sixty Percent.

Background: The COVID-19 pandemic has made neurology clinic waiting times longer. To prevent a build-up of patients waiting, we introduced a neurology advanced referral management system (NARMS) to deal with new referrals from GPs, using advice, investigations, or the telephone, as alternatives to face-to-face (FF) assessment. Methods: For six months, electronic referrals from GPs were triaged to the above categories. We recorded the numbers in each category, patient satisfaction, inter-consultant triage variation, re-referrals, and calculated CO2 emissions. Results: There were 573 referrals. Triage destinations were advice 33%, investigations 27%, telephone 17%, and FF 33%. Of patients referred for MRI, 95% were happy not to be seen if their investigation was normal. Less-experienced consultants triaged 20% and 30% respectively, to advice or investigations, compared with 40% by a triage-experienced neurologist. Four percent were re-referred. Numbers on the waiting list did not increase. CO2 emissions were reduced by 50%. Discussion: Two thirds of neurological referrals from GPs did not need to be seen FF and 50% were dealt with without the neurologist meeting the patient. Carbon emission was halved. This system should be employed more, with FF examination reserved for those patients who need a neurological examination for diagnosis and management.

A children's epilepsy diagnosis aid: Development and early validation using a Bayesian approach.

INTRODUCTION: The diagnosis of epilepsy in children is difficult and misdiagnosis rates can be as much as 36%. Diagnosis in all countries is essentially clinical, based on asking a series of questions and interpreting the answers. Doctors experienced enough to do this are either scarce or absent in very many parts of the world so there is a need to develop a diagnostic aid to help less-experienced doctors or non-physician health workers (NPHWs) do this. We used a Bayesian approach to determine the most useful questions to ask based on their likelihood ratios (LR), and incorporated these into a Children's Epilepsy Diagnosis Aid (CEDA). METHODS: Ninety-six consecutive new referrals with possible epilepsy aged under 10 years attending a pediatric neurology clinic in Khartoum were included. Initially, their caregivers were asked 65 yes/no questions by a medical officer, then seen by pediatric neurologist and the diagnosis of epilepsy (E), not epilepsy (N), or uncertain (U) was made. The LR was calculated and then we selected the variables with the highest and lowest LRs which are the most informative at differentiating epilepsy from non-epilepsy. An algorithm, (CEDA), based on the most informative questions was constructed and tested on a new sample of 47 consecutive patients with a first attendance of possible epilepsy. We calculated the sensitivity and specificity for CEDA in the diagnosis of epilepsy. RESULTS: Sixty-nine (79%) had epilepsy and 18 (21%) non-epilepsy giving pre-test odds of having epilepsy of 3.83. Eleven variables with the most informative LRs formed the diagnostic aid (CEDA). The pre-test odds and algorithm were used to determine the probability of epilepsy diagnosis in a subsequent sample of 47 patients. There were 36 patients with epilepsy and 11 with nonepileptic conditions. The sensitivity of CEDA was 100% with specificity of 97% and misdiagnosis 8.3%. CONCLUSION: Children's Epilepsy Diagnosis Aid has the potential to improve pediatric epilepsy diagnosis and therefore management and is particularly likely to be useful in the many situations where access to epilepsy specialists is limited. The algorithm can be presented as a smartphone application or used as a spreadsheet on a computer.

Usefulness of a smartphone application for the diagnosis of epilepsy: Validation study in high-income and rural low-income countries.

INTRODUCTION: In low- and middle-income countries (LMIC), the diagnosis of epilepsy should be made by Non-Physician Health Workers (NPHW) who are widely available in these settings. Recently a smartphone app (Epilepsy Diagnosis Aid) has been developed and validated to be used by NPHW, in order to confirm the diagnosis of epilepsy. The aim of our study was to perform a validation of the app in two different contexts: a hospital-based setting of a high-income country (HIC) and a population-based setting of the rural communities of a LMIC. MATERIAL AND METHODS: For the hospital-based setting, the app was administered to a sample of patients with epilepsy (PWE) and to a sample of subjects affected by syncope attending the epilepsy center of the University of Catania. For the population-based setting, performed in the rural communities of the Gran Chaco region in Bolivia,the app was administered by NPHW to a sample of PWE previously identified. Sensitivity and specificity were calculated for the diagnosis of epilepsy. RESULTS: In the hospital-setting, the app was administered to 100 PWE and 20 syncopes. A probability score > 80 showed a sensitivity of 76% (95%CI 66.4-84) and a specificity of 100% (95%CI 83.2-100) for the diagnosis of epilepsy; higher values were found for active epilepsy with tonic-clonic seizures. In the rural-setting, the app was administered to 38 PWE, giving a sensitivity of 92.1% (95%CI 78.6-98.3). CONCLUSION: The app for epilepsy could represent a valuable instrument, which can be easily employed by trained NPHW to diagnose epilepsy in primary health-care settings of LMIC.

The development of a smartphone application to help manage epilepsy in resource-limited settings.

PURPOSE: Epilepsy is treatable but in low- and middle-income countries (LMICs) it goes untreated with dire consequences for people with it and their families. There are not enough available doctors to treat it so it has been suggested that non-physician health workers (NPHWs) take a role in diagnosis and management. Tools will be essential to help them. A smartphone application (app) for episode diagnosis has proved safe and effective; this paper describes an app for epilepsy management. METHODS: Questions were devised which captured temporal characteristics of episodes, diagnosis of episodes, seizure types, and epilepsy type, together with information on previous investigations, treatment, drug reactions, and current treatment. For untreated patients a management plan was suggested. The app generated a summary which could be sent to a remote specialist for advice. The finished app was evaluated in 23 people presenting with possible epilepsy by four doctors in training and one NPHW; its summary was compared to face-to-face evaluation by a neurologist. RESULTS: The app was correct in 22 of 23 (96 %) patients for episode diagnosis, 2 of 2 for symptomatic seizures,18/20 (90 %) for epilepsy type and 9/10 (90 %) for treatment suggestion in untreated patients. The app took less than 15 min to complete. CONCLUSION: The initial results suggest that this management app is a worthwhile tool to help inexpert doctors or NPHWs manage suspected epilepsy. Its accuracy is well within reported inter-observer agreement. In its present form it requires input from a remote epilepsy specialist. This combination is a potential solution to managing epilepsy in LMICs.

Telemedicine in epilepsy: How can we improve care, teaching, and awareness?

Telemedicine (TM) is the use of telecommunications' technologies to provide medical information and services. Telehealth (TH) permits broader and psychosocial support for patients and their families. We aimed to highlight the importance of the use of TH for all aspects of epilepsy, either for the scientific aspects (e.g., research, education, care, management, etc.) or for the social matters (e.g., education, sensitization, association support, etc.). There is a deep gap in knowledge and use of TH in the developing and developed countries. Epilepsy is a condition responsible for 1% of the global burden of disease. More than 50 million people have epilepsy, and barriers to care include shortage of human resources, medical facilities, and resources. Eighty (80) percent of people with epilepsy (PWE) live in low- and middle-income countries. Telehealth has the potential of addressing limited resources and improving access to PWE across the globe.

Managing Epilepsy by Telemedicine in Resource-Poor Settings.

Epilepsy is a common and treatable disease; in rich countries the expectation is that two-thirds of people will have their seizure episodes controlled on medication. In low- and middle-income countries (LMICs) however most people are not on treatment either because no doctors live near them or the logistics of affordable drug supply is absent. People with epilepsy then are prone to the bad effects of this disease-death, disfigurement from accidents and burns, and social problems due to the stigma with which the disease is associated. So this represents a failure of conventional face-to-face medicine. Might a telemedicine approach do better? The World Health Organization has suggested that non-physician health workers are empowered to diagnose and manage epilepsy; to do this they will need considerable medical support, which might be provided by telemedicine through the telephone, smartphone applications or a combination. This paper sets out what telemedicine does at present for people with epilepsy in LMICs and suggests how it might be developed in the future.

Comparing long-term outcomes of epilepsy patients from a single-visit outreach clinic with a conventional epilepsy clinic: A cross-sectional observational study from India.

PURPOSE: To compare long-term treatment outcomes in epilepsy patients from a single-visit outreach clinic on the Lifeline Express (LLE) with a conventional hospital (AIIMS) based epilepsy clinic in India. METHODS: Using a cross-sectional observational study design, consecutive epilepsy patients from fifteen LLE clinics conducted from 2009 to 2014 were compared to epilepsy patients registered in the same duration at the AIIMS epilepsy clinic. The primary outcome was to determine if patients were still taking AEDs. To determine current AED status, patients from the LLE clinic were contacted telephonically. For the AIIMS patients, hospital records were reviewed and phone calls made to those patients who had not followed-up for more than a year. RESULTS: In the 5 years under review, 1923 and 1257 patients had consulted at the LLE and AIIMS clinics respectively. Long-term outcomes were available for analysis in 688 AIIMS and 531 LLE clinic patients. Of the AIIMS patients, 581(87%) were continuing AEDs, 49(7%) had discontinued AEDs after being seizure-free for at least 5 years, 39(6%) had discontinued AEDs without medical advice and 19(2.8%) were dead. Outcomes in 531 LLE patients revealed that 351(72%) continued to be on AEDs, 34(7%) had discontinued AEDs on advice, 106 (22%) had discontinued AEDs without any medical advice and 40 (7.5%) were dead. The treatment gap in the LLE patients was reduced from 49% at first contact to 22% at follow-up 2-8 years later. CONCLUSIONS: Even single-visit epilepsy clinics may be an effective option for reducing treatment gap in limited-resource regions of the world.

Epilepsy field workers, a smartphone application and telephone telemedicine: Safe and effective epilepsy care in rural Nepal.

PURPOSE: Most people with epilepsy live in low- or middle-income countries (LMICs) where there are relatively few doctors. Over 50% of people with epilepsy in these countries are untreated so other models of care are needed. In this report we evaluate a novel model of care. METHODS: We trained four residents of Myagdi, a rural district in Nepal as epilepsy field workers (EFWs). They provided epilepsy awareness to their communities. When they identified someone with possible epilepsy they used a smartphone application (app) to determine the probability score for an episode being epileptic and contacted an epilepsy specialist by phone. If the specialist thought treatment was indicated this was arranged by the EFW. We recorded mortality, change of diagnosis at face-to-face consultation and drug-related events as measures of safety. Seizure frequency and general wellbeing were also recorded, and a questionnaire was devised to measure satisfaction. RESULTS: 112 patients with app scores suggesting epileptic seizures were identified and managed in 18 months, of whom 15 had provoked seizures. Forty-three percent of epilepsy patients were untreated. At follow-up one had died of a cause other than epilepsy. Diagnostic agreement at face-to-face assessment was 93%. Overall 5% had side-effects of medication. Seizures were stopped in 33% and reduced in 57%. Ninety-six percent of patients preferred this service to travelling to other doctors. CONCLUSION: This novel service met all criteria of safety and was effective in reducing frequency of seizures. Patients preferred it to conventional services. It should be transferable to other LMICs.

Diagnosis of epileptic seizures by community health workers using a mobile app: A comparison with physicians and a neurologist.

PURPOSE: The World Health Organisation (WHO) strategy for non-physician health workers (NPHWs) to diagnose and manage people with untreated epilepsy depends on them having access to suitable tools. We have devised and validated an app on a tablet computer to diagnose epileptic episodes and now examine how its use by NPHWs compares with diagnosis by local physicians and a neurologist. METHODS: Fifteen NPHWs at Jan Swasthya Sahyog (JSS) a hospital with community outreach in Chhattisgarh, India were trained in the use of an epilepsy diagnosis app on a tablet computer. They were asked to determine the app scores on patients in their communities with possible epilepsy and then refer them first to their local JSS doctors and then to a visiting neurologist. With the neurologist's opinion as the "gold standard", the misdiagnosis rate from the NPHWs was compared with that of the local physicians. RESULTS: There were 96 patients evaluated completely. The NPHWs misdiagnosed eight and the physicians seven. There were more uncertain diagnoses by the NPHWs. In the 22 patients who presented for the first time during the study, the NPHWs misdiagnosed three and the physicians five. CONCLUSIONS: NPHWs using an app achieved similar misdiagnosis rates to local physicians. Both these rates were well within the range of misdiagnosis in the published literature. These results suggest that task-shifting epilepsy diagnosis and management from physicians to NPHWs, who are enabled with appropriate technology, can be an effective and safe way of reducing the epilepsy treatment gap.

Telephonic review for outpatients with epilepsy-A prospective randomized, parallel group study.

PURPOSE: Our objective was to assess how telephonic review of outpatients with stable epilepsy compared with conventional face-to-face clinic management. METHODS: We constructed a randomized parallel group study of suitable patients attending our Epilepsy Clinic and compared telephonic review with conventional clinic visit based management. Primary outcomes were the percentage of patients with breakthrough seizures and total number of breakthrough seizures. We also compared cost, patient satisfaction and numbers defaulting. RESULTS: A total of 465 patients were randomized and 429 were included in the final analysis. There was no significant difference in breakthrough seizures between the two groups. Mean time spent in the consultation was 10min in the telephone group (FT) and 22h in the face-to-face group (FC) and cost was INR 865 more expensive on an average in the FC group. Satisfaction was over 90% in the FT group. Significantly more people in the FC group were lost to follow-up. CONCLUSION: This study provides Class I evidence that the number of stable epilepsy patients who have breakthrough seizures and the total number of breakthrough seizures remain the same irrespective of whether patients are reviewed telephonically or face-to-face in the clinic. Clinicians managing epilepsy patients should consider using telephonic review for selected patients. Telephonic reviews have the potential of effectively reducing the secondary treatment gap in millions of patients who do not have easy access to doctors.

Effective clinical classification of chronic epilepsy into focal and generalized: A cross sectional study.

PURPOSE: Investigations such as EEG and brain imaging are often difficult to obtain in primary care settings of resource-limited regions impacting millions of epilepsy patients. We wanted to test the hypothesis that classification of chronic epilepsy into focal and generalized based on clinical history and examination alone would be comparable to making such a classification with additional inputs from EEG and brain imaging. METHODS: Two investigators independently classified consecutive chronic epilepsy patients into focal, generalized and unclassified epilepsy. Investigator 1 made this determination using clinical history and examination alone whereas Investigator II additionally used EEG and brain imaging too. We calculated inter observer agreement between the two investigators and also looked at the predictors of focal and generalized epilepsy. RESULTS: Five hundred and twelve patients were recruited. Inter observer agreement between the two investigators in making the focal versus generalized classification was 96.8%, kappa 0.91 (p<0.0001). When EEG and neuroimaging findings were added to clinical information, there was a change in classification in 3.2% patients. Several predictors of focal and generalized epilepsy were identified. CONCLUSIONS: Classification of chronic epilepsy into focal and generalized can be done reliably in most patients using clinical information alone. Investigating chronic epilepsy patients with EEG and brain imaging may not be necessary in every patient. The results of our study are especially significant for epilepsy patients living in resource-limited regions where such investigations may not always be available.