We recently developed a computational tool (EczemaPred) for personalised prediction of AD severity dynamics using PO-SCORAD. We aim to investigate whether EczemaPred can be applied to predict POEM score recorded weekly. We used data from 247 AD patients aged 11 to 65 years in a dose-finding study with up to 6 measurements of POEM and daily symptom scores for POEM over 16 weeks. We investigated the influence of the recency bias in reporting POEM by computing the partial cross-correlation between the observed POEM score and daily symptom scores recorded on the previous seven days using linear mixed-effects models. We also modelled the evolution of individual daily symptom scores using EczemaPred (Bayesian state-space) models and generated POEM predictions. We found a systematic difference between the observed POEM scores and those calculated from the daily symptom scores. The symptom scores on the day of POEM recording had the strongest correlation with the observed POEM score, among those recorded on the previous seven days. This alludes to a recency bias where patients may be more strongly influenced by their current symptoms while they are supposed to report their scores over the past week. EczemaPred models predict next week’s POEM scores with an accuracy of 80% when trained on 15 weeks of daily symptom scores. Prediction of POEM is challenging as it compresses daily presence of symptoms into a weekly average and is subject to recency bias. Recency bias can be mitigated by recording daily symptoms scores for POEM.

ESDR108 - Recency bias in weekly POEM recording and its effects on POEM prediction

The scientific meeting of the European Society for Dermatological Research (ESDR) is held yearly, usually in September. With an international attendance of more than 1000 skin scientists from around the globe, the four-day event includes short oral presentations from peer-reviewed abstracts, keynote lectures from international experts, sponsored symposia, satellite programs, educational events as well as an important focus on posters, networking and building collaborations amongst scientists. For more information about this year’s meeting visit our website.