bkg
PredictSurvival.com
Predicting Impending Death (less than 72 hours) in Advanced Cancer Patients
This data has only been validated in advanced cancer patients admitted to Acute Palliative Care Units.
These two models may help predict the probability of death over 72 hours. The first model requires a pretest probability of death within 72 hours. Based on the presence of clinical signs associated with impending death3 it will modify that probability using likelihood ratios. The second model was derived from the same data, does not require a pretest probability but does require a Palliative Prognostic Scale, see the reference for details on how it was derived.2

If you are unsure about entering a pretest probability, it may be helpful that the average advanced cancer patient admitted to an Acute Palliative Care Unit in the cited studies had a pretest probabilty of 38% of dying within 72 hours.2,3,4

Estimated Probability of Death in the Next 72 Hours: %
Palliative Perfomance Scale (PPS): Reference to look up the PPS if needed


For each of the following signs, check the box if present and leave blank if not. If unfamiliar scroll below for some explanations.

Present? Present? Present?



Probability of Death in the Next 72 hrs:
RPA Model:
Likelihood Ratio Model:

Definitions for signs that may be new:
  • Drooping of Nasolabial Fold: These are the 'smile lines', that run from the nose to the corner of the mouth. Think of it as looking for relaxed facial muscle tone. See page 10 of [2] for an example. Googling will mostly bring up its relation to cosmetic surgery.
  • Mandibular Breathing: See this brief video for a good overview.
  • Vocal Cord Grunting: "A sound produced predominantly on expiration, related to vibrations of the vocal cords."3
  • Decreased Response to Visual Stimuli: No eye tracking movement.
  • Decreased Response to Verbal Stimuli: No verbal communication (noises don't count), unable to follow commands.
  • Unable to close eyelids: Unable to close when asked, eyes open without observed blinking.
Sources: