In a recently released issue of Pediatrics, Dr. Eric van Diessen and colleagues (10.1542/peds.2018-0931) from the Netherlands share their work in creating a decision model that facilitates making the diagnosis of epilepsy. The model is highly pragmatic, and uses clinical information and results of the initial interictal electroencephalogram (EEG) study, which are readily available to clinicians. Primary care physicians who elect to obtain an initial EEG study may find this model particularly helpful. The model was developed using retrospective data from a cohort of 451 children evaluated in outpatient neurology clinic for one or more paroxysmal events (2008-2013, “training cohort”). The authors then tested their model on a second cohort of 187 children (2013-2016, “validation cohort”), also seen in the outpatient setting for evaluation of a paroxysmal event. All children were followed for at least one year following assessment, and a complete and detailed explanation of how the model was created is provided. The model itself is a web application available here and I could not resist clicking on it even before finishing the article.
In the real world of practice there are some children who present with a paroxysmal event that has a fairly clear path to diagnosis, for example the well-functioning school age child who begins to have brief episodes of staring or “day dreaming”: a diagnosis of absence seizures is highly likely. On the other end of the spectrum are infants and children, for example, with known underlying neurological disease who present with multiple brief paroxysmal events per day: the chance of recording the EEG correlate of the events is high enough that a video-EEG admission is likely optimal. But the great gray area of patient care between these two ends of the spectrum is the important zone that the work of van Diessen and colleagues fills: the challenge of a child presenting with one or two events that “kind of-maybe-might” be a seizure (or not) is the clinical niche of their prediction tool.
What appeals to me is how practical the approach and tool are. In their “training cohort” the diagnosis was initially uncertain in 45%, but following additional investigations and follow up, ultimately a certain diagnosis of either epilepsy vs. not epilepsy could be made in 94.2% of children. Results were similar for the “validation cohort” with 29% initially inconclusive and ultimately 92% with a clear diagnosis. In other words, it took time and additional studies to get to the diagnosis, which is often indeed what happens. The prediction tool’s questions codify common sense and thoughtful pediatric care, in that a history of neurological impairment, a genetic syndrome or a psychiatric disorder are included, as are 5 “event descriptors” that a clinician would be likely to ask during history taking. The tool gives a “probability of epilepsy,” not a diagnosis, and the authors are careful not to overstate its clinical value or accuracy. However, I can easily see how this brief but carefully developed prediction tool will be very useful for all who practice clinical pediatrics.
