BACKGROUND. Comparison of the predictive value of four 'diagnostic tests' for the outcome of shunting in patients with normal-pressure hydrocephalus (NPH). METHODS. Ninety-five NPH patients who received shunts were followed for 1 year. Gait disturbance and dementia were quantified by an NPH scale and handicap by a modified Rankin scale. Primary outcome measures were differences between the preoperative and last scores on both the NPH scale and the modified Rankin scale. Clinical and computed tomographic (CT) findings typical of NPH, absence of cerebrovascular disease, and a resistance to outflow of cerebrospinal fluid (CSF) ≥ 18 mmHg/ml/minute were designated as a positive test outcome; clinical and CT findings compatible with NPH, presence of cerebrovascular disease, and an outflow resistance < 18 mmHg/ml/minute as a negative test outcome. RESULTS. For each of the four tests the percentage of patients classified as improved was significantly greater for those with positive than with negative test results. Measurement of CSF outflow resistance was the only significant prognostic factor for the improvement ratio in NPH scale and CT in the modified Rankin scale according to multivariate logistic regression analysis. The accurate predictive value of the combination of typical clinical and CT findings was 0.65, that of the positive test results of outflow resistance, clinical and CT findings was 0.74. CONCLUSION. The best strategy is to shunt NPH patients if their outflow resistance is ≥ 18 mmHg/ml/minute or, when the outflow resistance is lower, if their clinical as well as their CT findings are typical of NPH. Copyright (C) 2000 Elsevier Science Inc.