The Norwegian Brøset-Violence-Checklist (BVC) is one of the few instruments that is suitable for short-term prediction of violence of psychiatric inpatients by nursing staff in routine care. The instrument assesses the presence or absence of six behaviours or states frequently observed before a violent incident. We conducted a study to elucidate whether the predictive properties of the BVC are retained in other psychiatric settings than the original north-Norwegian validation dataset. During their admission period, 219 consecutive patients admitted to six acute psychiatric wards were assessed as to the risk for attack using a German version of the BVC (BVC-G). Data on preventive measures were concurrently collected. Aggressive incidents were registered using an instrument equivalent to the Staff Observation of Aggression Scale (SOAS-R). Fourteen attacks towards staff were observed with incident severity ranging from 5 to 18 of a possible 22 points. BVC-G sensitivity was 64.3%, the specificity 93.9%, the positive predictive value 11.1%, and the area under the receiver operating characteristic curve 0.88. In some false positive cases intense preventive measures had been implemented. The predictive accuracy of the BVC-G proved consistent with the Norwegian original.