Design and implementation of electronic health record common data elements for pediatric epilepsy: Foundations for a learning health care system

Abstract

Objective: Common data elements (CDEs) are standardized questions and answer choices that allow aggregation, analysis, and comparison of observations from multiple sources. Clinical CDEs are foundational for learning health care systems, a data-driven approach to health care focused on continuous improvement of outcomes. We aimed to create clinical CDEs for pediatric epilepsy.

Methods: A multiple stakeholder group (clinicians, researchers, parents, caregivers, advocates, and electronic health record [EHR] vendors) developed clinical CDEs for routine care of children with epilepsy. Initial drafts drew from clinical epilepsy note templates, CDEs created for clinical research, items in existing registries, consensus documents and guidelines, quality metrics, and outcomes needed for demonstration projects. The CDEs were refined through discussion and field testing. We describe the development process, rationale for CDE selection, findings from piloting, and the CDEs themselves. We also describe early implementation, including experience with EHR systems and compatibility with the International League Against Epilepsy classification of seizure types.

Results: Common data elements were drafted in August 2017 and finalized in January 2020. Prioritized outcomes included seizure control, seizure freedom, American Academy of Neurology quality measures, presence of common comorbidities, and quality of life. The CDEs were piloted at 224 visits at 10 centers. The final CDEs included 36 questions in nine sections (number of questions): diagnosis (1), seizure frequency (9), quality of life (2), epilepsy history (6), etiology (8), comorbidities (2), treatment (2), process measures (5), and longitudinal history notes (1). Seizures are categorized as generalized tonic-clonic (regardless of onset), motor, nonmotor, and epileptic spasms. Focality is collected as epilepsy type rather than seizure type. Seizure frequency is measured in nine levels (all used during piloting). The CDEs were implemented in three vendor systems. Early clinical adoption included 1294 encounters at one center.

Significance: We created, piloted, refined, finalized, and implemented a novel set of clinical CDEs for pediatric epilepsy.

Figure 1.

Consensus Process to create and disseminate Pediatric Epilepsy Learning Health System registry questions for use at the point of care. NINDS, National Institute of Neurological Disorders and Stroke; CDE, Common Data Elements; NISC, National Infantile Spasms Consortium; ELES, Early Life Epilepsy Study; NSR, Neonatal Seizure Registry; pSERG, Pediatric Status Epilepticus Group; EHR, Electronic Health Record

Figure 2.

Responses to questions during the pilot (blue; 10 centers) and implementation (orange; one center) about the most recent seizure (top row) and the overall frequency of seizures (bottom row) for four seizure types (columns). For infantile spasms, treatment response is all-or-none, thus frequency of epileptic spasms is not collected. Of note, three labels at the implementation site were slightly different than in the pilot: (1) “Too many to count” was labelled “Many per day”; (2)“Multiple per day” was labelled “Several per day”; and (3) “Frequency not well defined” was not included as a response.

Figure 3.

Relationship between Pediatric Epilepsy Learning Healthcare System (PELHS) seizure outcomes and the International League Against Epilepsy (ILAE) seizure classification system (Fisher et al, Epilepsia 2017). Seizure types are in general mapped to three concepts: generalized tonic clonic seizures, regardless of onset (black lines, grey box), motor seizures (blue), and non-motor seizures (orange). For children under 3 years old, epileptic spasms are tracked separately (green). Aware vs impaired awareness is not explicitly captured. Unclassified seizures are categorized based on presence / absence of movement.

Figure 4.

Examples of the PELHS question about the last generalized tonic clonic seizure, as implemented in three electronic health record systems at four centers: (A, B) Epic Systems Corporation (Verona, WI), (C) Cerner Corporation (North Kansas City, MO), (D) athenahealth Inc (Watertown, MA). The question and answers are conceptually identical; however, there are variations in interface (buttons vs radio boxes), wording, capitalization, and the order of answer choices. In one implementation (B), the question appears only if the patient is known to have tonic-clonic seizures, and so there is no option “Never/Does not have this seizure type”.