The temperature of a patient is a significant predictor of death in heart failure patients. Temperature provides a window into the physiology of the patient's underlying condition and may be used as an early marker for CHF exacerbations. The patient's temperature is taken to form a time series of temperature values. In accordance with some embodiments, the time series of temperature values is converted to the frequency domain by, for example, a discrete Fourier Transform. The frequency domain representation then is analyzed for a marker indicative of the worsening condition of the patient. In accordance with other embodiments, the patient's time series of temperature values is analyzed for a marker using, for example, Cosinor analysis. In yet other embodiments, both the time and frequency domain temperature data is analyzed for markers of the patient's worsening medical condition.