Introduction Small primary intracerebral haemorrhages (PICHs) cause mild stroke symptoms and resolve rapidly on CT. Delays in imaging in stroke incidence studies may therefore have inadvertently led to an underestimate of the frequency of small PICHs. Objective To determine whether the rate and timing of CT in community-based stroke incidence studies was adequate to determine accurately the proportion of strokes due to PICH. Methods A systematic review of community-based stroke incidence studies that included details on pathological type of stroke, excluding subarachnoid haemorrhage. We extracted information on the proportions of patients scanned, the timing of scans, characteristics of patients that were less likely to be scanned, and the proportion of ischaemic, or haemorrhagic, or unknown strokes. Results In the 25 studies identified, scanning methods were poorly documented. When mentioned, the median proportion of patients scanned was 63% (95% confidence intervals (CI) 60 to 85%) and mostly performed outside the time for reliable distinction of PICH from ischaemic stroke (median 18.5 days, 95% CI 7 to 30 days). Patients particularly likely to miss scanning were older, those not admitted to hospital or who died early after stroke. Conclusion The scanning strategy documentation, the proportion of patients scanned and the timing of scanning in stroke incidence studies has been suboptimal. The frequency of a primary intracerebral haemorrhage, and its distribution in different age groups of patients or severities of stroke, has been underestimated. Future incidence studies should adopt more rigorous scanning policies and describe these policies more precisely.