Clinical diagnosis of multiple system atrophy is challenging and many patients with Lewy body disease (i.e. Parkinson's disease or dementia with Lewy bodies) or progressive supranuclear palsy are misdiagnosed as having multiple system atrophy in life. The clinical records of 203 patients with a clinical diagnosis of multiple system atrophy were reviewed to identify diagnostic pitfalls. We also examined 12 features supporting a diagnosis of multiple system atrophy (red flag features: orofacial dystonia, disproportionate antecollis, camptocormia and/or Pisa syndrome, contractures of hands or feet, inspiratory sighs, severe dysphonia, severe dysarthria, snoring, cold hands and feet, pathological laughter and crying, jerky myoclonic postural/action tremor and polyminimyoclonus) and seven disability milestones (frequent falls, use of urinary catheters, wheelchair dependent, unintelligible speech, cognitive impairment, severe dysphagia, residential care). Of 203 cases, 160 (78.8%) were correctly diagnosed in life and had pathologically confirmed multiple system atrophy. The remaining 21.2% (43/203) had alternative pathological diagnoses including Lewy body disease (12.8%; n = 26), progressive supranuclear palsy (6.4%; n = 13), cerebrovascular diseases (1%; n = 2), amyotrophic lateral sclerosis (0.5%; n = 1) and cerebellar degeneration (0.5%; n = 1). More patients with multiple system atrophy developed ataxia, stridor, dysphagia and falls than patients with Lewy body disease; resting tremor, pill-rolling tremor and hallucinations were more frequent in Lewy body disease. Although patients with multiple system atrophy and progressive supranuclear palsy shared several symptoms and signs, ataxia and stridor were more common in multiple system atrophy. Multiple logistic regression analysis revealed increased likelihood of multiple system atrophy versus Lewy body disease and progressive supranuclear palsy if a patient developed orthostatic hypotension or urinary incontinence with the requirement for urinary catheters [multiple system atrophy versus Lewy body disease: odds ratio (OR): 2.0, 95% confidence interval (CI): 1.1-3.7, P = 0.021; multiple system atrophy versus progressive supranuclear palsy: OR: 11.2, 95% CI: 3.2-39.2, P < 0.01]. Furthermore, autonomic dysfunction within the first 3 years from onset can differentiate multiple system atrophy from progressive supranuclear palsy (multiple system atrophy versus progressive supranuclear palsy: OR: 3.4, 95% CI: 1.2-9.7, P = 0.023). Multiple system atrophy patients with predominant parkinsonian signs had a higher number of red flag features than patients with Lewy body disease (OR: 8.8, 95% CI: 3.2-24.2, P < 0.01) and progressive supranuclear palsy (OR: 4.8, 95% CI: 1.7-13.6, P < 0.01). The number of red flag features in multiple system atrophy with predominant cerebellar signs was also higher than in Lewy body disease (OR: 7.0, 95% CI: 2.5-19.5, P < 0.01) and progressive supranuclear palsy (OR: 3.1, 95% CI: 1.1-8.9, P = 0.032). Patients with multiple system atrophy had shorter latency to reach use of urinary catheter and longer latency to residential care than progressive supranuclear palsy patients, whereas patients with Lewy body disease took longer to reach multiple milestones than patients with multiple system atrophy. The present study has highlighted features which should improve the ante-mortem diagnostic accuracy of multiple system atrophy.