We have observed significant differences in the high frequency content of the spectrum between healthy and pathological voices. Pathologies like larynx cancer, vocal fold lesions, and patients with larynx or vocal fold removal are examples of this. This finding invites to use high sampling frequencies in voice pathology classification systems to benefit from this high frequency information, which has been traditionally ignored. With a GMM classifier fed with MFCCs and a sampling frquency of 48 kHz we are able to improve AUC almost a 5% compared to a system using a sampling frequency of 8 kHz and more than 2% compared to a system using a sampling frequency of 16 kHz.