Prediction of impulse control disorders in Parkinson's disease through a longitudinal machine learning study

Impulse control disorders (ICD) in Parkinson's disease (PD) patients mainly occur as adverse effects of dopamine replacement therapy. Despite several known risk factors, ICD development cannot yet be accurately predicted at PD diagnosis. We aimed to investigate the predictability of incident ICD by baseline measures of demographic, clinical, dopamine transporter single photon emission computed tomography and single nucleotide polymorphisms data of medication-free PD patients, obtained from the Parkinson's Progression Markers Initiative (PPMI; n = 311) and Amsterdam University Medical Center (UMC; n = 72) longitudinal datasets. We trained machine learning models to predict incident ICD at any follow-up assessment. The highest predictive performance (AUC = 0.66) was achieved by clinical features only. We observed significantly higher performance (AUC = 0.74) when classifying patients who developed ICD within four years from diagnosis compared with those tested negative for seven or more years. Overall, prediction accuracy for later ICD development at the time of PD diagnosis is limited, but increases for shorter time-to-event predictions.