Quality control in the treatment of trauma patients often consists in comparisons of survival rates. The trauma population under study is seldom defined with regard to severity of injury. Therefore crude survival rates are of little help when the quality of care is discussed. Trauma scores attempt to summarize the severity of injury of trauma patients in a single number. They attempt to translate differences in the severity of injury into a common language. This study tested the validity of seven common trauma score systems in the setting of Cologne in 1987. Six hundred and twelve trauma patients treated by physicians at the scene of the accident were prospectively followed up in 32 hospitals. Final diagnosis, treatment, complications, and survival were evaluated. Sensitivity and specificity in predicting survival were calculated for the following systems: Glasgow Coma Scale, Trauma Score, Revised Trauma Score, Injury Severity Score, TRISS, Prehospital Index, Polytraumaschlüssel. The average time from emergency call to arrival of the emergency physician at the scene of the accident was 6.5 min. Four hundred and one patients (65.5%) were male. One hundred and thirty-seven patients (22.4%) suffered from apnoea, 61 (10.0%) had a systolic blood pressure lower than 90 mmHg, 117 (19.1%) had had a cardiac arrest and 174 (28.4%) were unconscious. Four hundred and twenty-three patients (69.1%) left hospital alive. All trauma score systems tested showed sensitivities and specificities greater than 83%. They all proved their ability to classify trauma patients according to severity of injury. The TRISS performed best of all, with sensitivity of 93.1% and specificity of 93.7% at a cut-off point of 0.85.(ABSTRACT TRUNCATED AT 250 WORDS)