A Three-Item History-Based Model to Estimate Fatigue Severity: A Large Cross-Sectional Community Study across Five Neurological Conditions

BackgroundFatigue is prevalent and debilitating across neurological disorders yet remains under recognised in routine care. We characterised and compared fatigue across five long term neurological conditions using a common assessment, and developed a brief, history-based model to estimate fatigue severity. MethodsAdults with epilepsy, Huntingtons disease (HD), multiple sclerosis (MS), Parkinsons disease (PD), or motor neurone disease (MND) completed an online cross-sectional survey (n=652). The outcome was self-rated fatigue severity (1-10 numerical rating scale). The development sample comprised complete cases (n=481); multiple imputation sensitivity analyses (m=20) were undertaken. Model performance was assessed with adjusted R2, root mean squared error (RMSE), mean absolute error (MAE), and calibration (slope/intercept) with bootstrap resampling and 10-fold cross-validation (CV). ResultsClinically significant fatigue (>=6/10) affected 51.1% of participants. Highest rates were in MND and MS, and lowest in PD. Fatigue episodes occurred at least twice weekly in 70.4% of participants (epilepsy 60.1%, HD 58.3%, MS 79.0%, PD 69.2%, MND 82.3%). Fatigue impacted social life and concentration "often" or more frequently in 38.5% and 34.5% of participants, respectively. Overall, 69.8% had not discussed fatigue with a healthcare professional. In multivariable models adjusted for demographics and diagnosis, three patient-reported items independently predicted severity: episode frequency, frequency of impact on social life, and frequency of impact on concentration. Full model performance: CV adjusted R2=0.45; RMSE=1.59; MAE=1.27; slope=0.92 (0.84-1.01), intercept=+0.39 (-0.06 to +0.91). A three-item history-based model performed similarly (CV adjusted R2=0.47; RMSE=1.57; MAE=1.25; slope=0.96 (0.87-1.05); intercept=+0.21 (-0.30 to +0.73). Diagnosis was not independently associated with severity after adjustment; episode frequency was the strongest predictor. ConclusionsFatigue is frequent and functionally disruptive across neurological conditions, yet seldom discussed in clinic. We derived a three-item history-based model that explained nearly half of the variance in self-rated severity, showing stable internal performance with good calibration. Because episode frequency and frequency of impact are rarely captured by existing fatigue scales, eliciting them as simple history items offers a pragmatic, transdiagnostic triage approach for clinical and digital pathways. External validation across neurological and other long-term conditions is now required.