Prognosis in functional and recognised pathophysiological neurological disorders - a shared basis

Oliver Shipston-Sharman; Stoyan Popkirov; Christian H Hansen; Jon Stone; Alan Carson

doi:10.1016/j.jpsychores.2021.110681

Prognosis in functional and recognised pathophysiological neurological disorders - a shared basis

J Psychosom Res. 2021 Nov 20:152:110681. doi: 10.1016/j.jpsychores.2021.110681. Online ahead of print.

Authors

Oliver Shipston-Sharman¹, Stoyan Popkirov², Christian H Hansen³, Jon Stone⁴, Alan Carson⁴

Affiliations

¹ Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, United Kingdom. Electronic address: oliver.shipston-sharman@nhs.net.
² Department of Neurology, University Hospital Knappschaftskrankenhaus Bochum, Ruhr University Bochum, Bochum, Germany.
³ Faculty of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London, United Kingdom.
⁴ Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, United Kingdom.

PMID: 34872006
DOI: 10.1016/j.jpsychores.2021.110681

Abstract

Objective: To compare self-reported outcomes, clinical trajectory and utility of baseline questionnaire responses in predicting prognosis in functional and recognised pathophysiological neurological disorders.

Methods: Baseline data on 2581 patients included health-related quality of life, psychological and physical symptoms, illness perceptions, consultation satisfaction and demographics. The prospective cohort included neurology outpatients classified with a functional (reporting symptoms 'not at all' or 'somewhat explained' by 'organic disease'; n = 716) or recognised pathophysiological disorder ('largely' or 'completely explained'; n = 1865). Logistic regression and deep neural network models were used to predict self-reported global clinical improvement (CGI) at 12-months.

Results: Patients with functional and recognised pathophysiological disorders reported near identical outcomes at 12-months with 67% and 66% respectively reporting unchanged or worse CGI. In multivariable modelling 'negative expectation of recovery' and 'disagreement with psychological attribution' predicted same or worse outcome in both groups. Receipt of disability-related state benefit predicted same or worse CGI outcome in the functional disorder group only (OR = 2.28 (95%-CI: 1.36-3.84) in a group-stratified model) and was not related to a measure of economic deprivation. Deep neural network models trained on all 92 baseline features predicted poor outcome with area under the receiver-operator curve of 0.67 in both groups.

Conclusions: Those with functional and recognised pathophysiological neurological disorder share similar outcomes, clinical trajectories, and poor prognostic markers in multivariable models. Prediction of outcome at a patient level was not possible using the baseline data in this study.

Keywords: Deep neural networks; Functional neurological disorders; Logistic regression; Machine learning; Neurological symptoms; Neurology; Neurology outpatients; Neuropsychiatry; Prognosis; Socioeconomic analysis.