Objective: To determine the incidence of clinically important subdural hematoma (SDH), hydrocephalus not associated with a recent intracranial bleed, and intracranial tumor that is not obviously metastatic, and to test the sensitivity of a promising decision rule for computerized tomography (CT) in dementia.
Design: Population-based, retrospective, sequential case series.
Setting: Staff model health maintenance organization (HMO).
Patients: Patients aged 65 years and older with one of the three lesions, diagnosed over a 4.5-year period, identified mainly through computerized databases of hospital discharge diagnoses and a registry of malignant tumors.
Measurements: Clinical data were based on chart review. The decision rule, based on one that had been previously proposed and tested, stated that CT would be required if any one of 11 clinical criteria were met by a patient with cognitive impairment. Rule sensitivity was evaluated using clinical information recorded before CT.
Main results: One hundred forty-five clinically important lesions were identified among 137,100 person-years at risk. Average annual incidence per 100,000 was 46.7 (95% CI 36.0, 59.6) for SDH, 5.8 (95% CI 2.5, 11.5) for hydrocephalus not associated with a recent intracranial bleed, and 53.2 (95% CI 41.7, 66.9) for intracranial tumors that were not obviously metastatic. Using the 65 to 74- year age strata as a reference, the relative risk for SDH was 4.8 (95% CI 2.7, 8.5) in 75 to 84-year-olds and 13.1 (95% CI 7.7, 22.5) in the 85 and older strata. Among 59 patients who presented with cognitive impairment, without altered sensorium or physical evidence of trauma, decision rule sensitivity was 93.2% (95% CI 83.5%, 98.1%). Sensitivity was 90.7% (95% CI 77.9%, 97.4%) in the subgroup that presented to an ambulatory care clinic rather than to an emergency department.
Conclusions: These three lesions, which are the most common surgical lesions that may present as dementia, are rare. Most cases have presentations that easily distinguish them from typical Alzheimer's disease. This case series indicates that it may be feasible to develop a decision rule for the selective use of CT in dementia. Disease spectrum will influence measures of decision rule performance such as sensitivity and specificity.