Using EHRs to advance epilepsy care

Seizure frequency

Understanding seizure frequency is essential to clinical epilepsy care. This information may be collected through questionnaires or acquired from patient seizure detection devices. Several families use online seizure diaries (epilepsy.com; seizuretracker.com; texting4control.com; WebEase.org; patientslikeme.com), smart watches, and other devices to record their seizure data. The Responsive Neurostimulation System (a cranially implanted neurostimulator device used for intractable partial epilepsy) can remotely transfer seizure incidence data from a patient to a physician for potential therapy management.¹¹ Although many medical centers collect seizure frequency data generated by patients, it has been challenging to develop a system to automatically download the data into the EHR due to technical constraints. Potential barriers to patients or families completing the data prior to the visit include high completion rates include data complexity or primary language other than English. Nevertheless, completion of information prior to a visit is common practice in other disciplines (e.g., mental health, nutrition) and can be overcome with well-designed systems.¹²

Demographic and historical data

Some practices have patients provide information on portable electronic devices prior to being seen by the clinician. The information is then populated into the patient's health record and is available for the clinician to review during the visit and may therefore save time during the encounter.

Developmental and behavioral screens

Developmental and behavioral screens could be incorporated in the information collected prior to the visit while in the waiting room. A simple checklist completed by the parent may facilitate developmental screening in an epilepsy clinic with high sensitivity and specificity.¹³ With developmental and behavioral information available prior to a visit a physician may be able to identify and address comorbidities such as mood and anxiety disorders, attention deficit hyperactivity disorder, social problems or autistic spectrum disorders.

Rationale for multicenter collaboration

Pooling patient data across centers results in larger patient populations and allows for improved understanding of the heterogeneity of disease. This is particularly important in pediatric epilepsy, considering that many of the conditions are rare. There are several models for how to leverage EHRs to support large research efforts while protecting patient privacy and abiding by HIPAA regulations.

Practice Based Research Networks

A Practice Based Research Network is a group of practicing physicians that collect data together to answer questions related to care delivery. An active neurology Practice Based Research Network has successfully shared EHR tools across multiple centers to pool data for 11 common neurologic indications.^22,23 This initiative includes over 15 institutions whose pooled data are used for subgroup based adaptive assignment of treatments and captures outcomes data at the point of care using the EHR; for example, the treatments among 3 anti-epileptic drugs (AEDs) (lamotrigine, levetiracetam and valproic acid) with outcome measures of survival and 50% reduction of seizure frequency is being compared.

Prospective Registries

By pooling data from multiple centers and enrolling patients prospectively cohorts of patients with relatively rare conditions can be developed to allow for research that would not otherwise be possible at a single center. Infantile Spasms is an example of a rare condition where there is variability in initial treatment. Through the National Infantile Spasms Consortium database developed in 2012, the Pediatric Epilepsy Research Consortium determined that adrenocorticotropic hormone was a more effective treatment than other standard therapies.³¹ Other examples of epilepsy specific prospective registries are the Pediatric Status Epilepticus Research Group's (pSERG) cohort that has been used to investigate variability in treatment of children with refractory status epilepticus³² and the Early Life Epilepsy Study which found that levetiracetam is superior to phenobarbital for new onset nonsyndromic epilepsy in infants aged 1 month to 1 year.³³

The use of multicenter collaboration and research to improve clinical care has been successfully demonstrated in other diseases by various groups. The Children's Oncology Group is an organization devoted to childhood and adolescent cancer research; they have a long history of clinical trials and collaborative research that has led to an overall significant increase in survival rates for childhood cancer.³⁴ Another collaborative group, the Vermont Oxford Network has used their Neonatal Encephalopathy Registry to investigate pertinent topics such as the antecedents of neonatal encephalopathy.³⁵

EHR Vendor Initiatives

Another initiative for collaboration across centers is the development of a pediatric epilepsy registry by the Cerner Corporation in collaboration with the Children's Hospital of Orange County and the Children's National Medical Center to close gaps in care and measure evidence based practices. The registry allows for assessing the adherence to best practice measures for epilepsy including the provision of annual emergency seizure action plans, reproductive health education for all females over 12 years of age and on AEDs, and surgical referrals for patients with intractable epilepsy (Z. Danner, personal communication, January 31 2017).

Common Data Elements

The NINDS CDEs provide a useful tool for collaborative research initiatives that utilize standardization of templates and data reporting. An essential need for data aggregation, however, is the development and implementation of CDEs that are clinically meaningful and feasible to be incorporated into clinical practice. A collaborative initiative by Nationwide Children's Hospital and the Children's Hospital of Philadelphia is working to develop a standardized template for EEG reporting within the EHR to standardize reporting and terminology, improve efficiency and billing as well as to facilitate collaborative research. An initiative such as this would benefit from the NINDS CDE module for scalp EEG which includes 105 distinct CDEs in discrete format covering breadth of details that would need to be included in a comprehensive EEG report. The NINDS Epilepsy CDEs include other modules for demographic, diagnostic, and therapeutic data elements that may be adaptable for practice.

Learning Healthcare Systems

In a Learning Healthcare System, clinicians enter data in a structured format at the point of care. These data are then aggregated, analyzed, and fed back to clinicians rapidly in order to improve care, for example through quality improvement projects or comparative effectiveness research. In pediatric epilepsy, there has been active work to measure quality of care for refractory status epilepticus by the pSERG.³⁶ Furthermore, a recent effort to develop a learning health care system is underway.³⁷ Several national efforts to standardize and share clinical data may help to provide the core informatics infrastructure for these initiatives, including public and commercial clinical health information exchange systems³⁸ and research based Clinical Data Research Networks Patient Centered Outcomes Research Institute's PCORnet infrastructure.³⁹

Resources

Several resources are required to create an effective EHR system that can simultaneously serve the clinical and research missions of a practicing pediatric epilepsy group. The modification of an existing EHR to include discretized data with CDS may require an intensive period of consensus building, content development, software EHR modification, and implementation into the clinical workflow.²² The costs associated with developing these systems may be prohibitive but may be overcome through sharing of existing tools between centers and joining already established data collection initiatives. Parent institutions could be incentivized to fund the development of the EHR systems within their Epilepsy Centers by the potential direct benefits including improvements in care navigation, more efficient note writing, timely communications, value based payment, patient safety, quality improvement and others.²³ Additional sources of funding support that could be considered include the Small Business Innovation Research (SBIR) programs within the National Institutes of Health (NIH), the patient centered outcomes research institute (PCORI), or the Centers for Disease Control.

Collaboration across institutions

Institutions that are collaborating or sharing information have to agree on the terms and means of collaboration. Promoting buy in from other physicians and centers may present a challenge to building networks and data collection due to the natural question of “what's in it for our institution?” An overall benefit of standardized data collection and recording is that it may lead to improvements in care and the ability to improve outcomes by enabling quality research. The use of commercial EHR systems, each with their own proprietary software and restrictions, across different institutions pose a challenge to collaboration. EHR toolkits for data sharing may be designed to work with different versions of the EHR (i.e., historical and current) to allow for broader implementation.²³ Data sharing between institutions using different EHR systems may be overcome through the development of platform agnostic data registries, i.e., registries that are easily accessible and usable by all EHR systems by using standard terminology equivalents. Overall, it would be more feasible to develop tools that may be used within existing EHRs across different institutions than to expect institutions to switch to a single EHR for collaboration. However, even with these modifications the data shared between systems may be limited and therefore the goals of collaborations have to be well defined. An additional challenge for existing EHR systems, is the changing concepts of epilepsy classification. The International League Against Epilepsy recently published a revised operational classification of seizure types⁴⁰; centers would need to ensure consistency in patient and cohort classification even with a potential change in description of the diagnoses.