In some examples, determining a heart failure status using a medical device comprising one or more sensors includes determining a first value of a heart beat variability metric of a patient while an activity state of a patient satisfies an inactivity criterion based on a signal received from the one or more sensors, and determining, within a predetermined period of time after further determining that the activity state of the patient no longer satisfies the inactivity criterion, a second value of the heart beat variability metric while the activity state of the patient no longer satisfies the inactivity criterion based on the signal. A difference between the first value of the heart beat variability metric and the second value of the heart beat variability metric may be determined and the heart failure status of the patient may be determined based on the difference.