Neuroprognostication: a conceptual framework

Abstract

Neuroprognostication, or the prediction of recovery from disorders of consciousness caused by severe brain injury, is as critical as it is complex. With profound implications for mortality and quality of life, neuroprognostication draws upon an intricate set of biomedical, probabilistic, psychosocial and ethical factors. However, the clinical approach to neuroprognostication is often unsystematic, and consequently, variable among clinicians and prone to error. Here, we offer a stepwise conceptual framework for reasoning through neuroprognostic determinations - including an evaluation of neurological function, estimation of a recovery trajectory, definition of goals of care and consideration of patient values - culminating in a clinically actionable formula for weighing the risks and benefits of life-sustaining treatment. Although the complexity of neuroprognostication might never be fully reducible to arithmetic, this systematic approach provides structure and guidance to supplement clinical judgement and direct future investigation.

Fig. 1 |. Schematization of neuroprognostication: steps 1 and 2.

When prognosticating after acute brain injury, step 1 is to identify the patient’s current level of neurological function, which corresponds schematically to the depth of the curve’s downward deflection and the starting point for recovery. Step 2 is to synthesize the prognostic data to identify the patient’s likely recovery trajectory. Because trajectories cannot be anticipated precisely, the estimation of recovery resembles a ‘probability cloud’ of possible outcomes, where darker regions within the cloud represent outcomes of higher probability, and lighter regions represent outcomes of lower probability.

Fig. 2 |. Sources of information about a patient’s goals of care.

Decisions about the objectives of medical intervention (that is, a patient’s goals of care) should prioritize the patient’s interests, but in the case of severe brain injury, the patient’s current interests are typically unknown. Thus, clinicians must rely on proxies to approximate those interests: the interests that the patient had expressed in the past (for example, advance directives), the patient’s family or surrogates, and population data on the quality of life of other patients with similar brain injury. The advantages of each proxy are highlighted in blue, and the disadvantages in red.

Fig. 3 |. Schematization of neuroprognostication: steps 3 and 4.

Step 3 of neuroprognostication is to identify the patient’s goals of care, both in terms of the minimum level of neurological function that would make life worth living for the patient (represented as the position of the orange line on the y axis) and the deadline to attain that level of function (represented by the duration of the orange line on the x axis). Step 4 is to partition the range of possible outcomes into good outcomes (that is, those in which the patient reaches the goals of care by the deadline (outlined in blue)) and bad outcomes (that is, those in which the patient does not reach the goals of care, or does so after the deadline (outlined in red)).

Fig. 4 |. Schematization of neuroprognostication: steps 5 and 6.

The probability of a good outcome (P_GO) (part a) is defined by the range of good outcomes (the weighted area, or amount of black, outlined in blue) divided by the range of all outcomes. The probability of a bad outcome (P_BO) (part b) is defined by the range of bad outcomes (the weighted area, or amount of black, outlined in red) divided by the range of all outcomes. Step 5 of neuroprognostication is to identify value modifiers, both for good outcomes (part c) and bad outcomes (part d). Step 6 (part e) is to synthesize the considerations above to guide whether life-sustaining treatment should be continued or withdrawn.