Implementations described herein disclose a method of predicting oxygen level desaturation including generating an input sequence of oxygen levels based on an input signal sequence, the input signals indicative of a physiological condition of a patient, generating an input feature sequence based on at least one of the input signal sequence and the input sequence of oxygen levels, generating, using a neural network, a predicted value sequence of the oxygen levels based on the input feature sequence, comparing the predicted value sequence of the oxygen levels with the input sequence of oxygen levels for a predetermined temporal window to generate a predicted sequence confidence value, and generating, in response to determining that the predicted sequence confidence value is above a threshold confidence value, an oxygen level desaturation prediction based on the predicted value sequence.
