Page 228 - DECISION-MAKING IN SEVERE TRAUMATIC BRAIN INJURY PATIENT OUTCOME, HOSPITAL COSTS, AND RESEARCH PRACTICE

Chapter 12
to individual patient preferences, needs, and values. 32 Choosing an acute treatment strategy that is proportional and leads to best possible patient outcome is however difficult. This is mainly caused by uncertainties on future patient outcome, especially regarding outcome prediction and outcome valuation.
Patient outcome prediction
Because providing healthcare is about patient outcome in the future, it is necessary to
use a prediction of that outcome for acute treatment decisions. Knowing what specific
outcome will be achieved after a specific treatment is likely to improve decision- making. 30,31,33-35
Unfortunately, physicians appear to be unable to make accurate outcome predictions (Table 1). 22,33,36,37 Validated prognostic models, such as IMPACT and CRASH 38,39, have been developed to assist physicians with TBI outcome prediction, but they have not been widely implemented in clinical care. 40-44 Although IMPACT and CRASH models display good discriminative ability in validation studies 40,41, they are, like experienced physicians, considered to be too inaccurate on individual level predictions. Heterogeneity between individual patients with variable injuries, pathophysiology, and treatments makes prognostication difficult and uncertain. Another limitation of available prognostic models is that they only include robust short-term outcome measures like mortality and functional outcome. Although robustness is a good epidemiological attribute of clinical studies it misses personal human properties like long-term physical, cognitive, emotional and behavioural outcome, or satisfaction with life. 33,38-45 This is problematic, because these long-term consequences of s-TBI are highly relevant to include in outcome assessment. 46
Table 1. Difficulties in outcome prediction in TBI patients (chapter 6) 47
# Difficulties in random order.
1 The heterogeneous nature of s-TBI and concurring comorbidities and their unknown effect on outcome.
2 Unclear/incomplete clinical information, including the patient’s neurological state and level of consciousness.
3 Largely unknown pathophysiological mechanisms of brain injury and inherent degree of brain plasticity.
4 Prediction models do not include long-term (health-related) quality of life, although long-term changes have been reported and patients/proxies are known to value this outcome.
5 Prediction models are based on large retrospective data sets that do not necessarily reflect current or future treatment strategies.
High prognostic accuracy is indispensable when a prediction is used to substantiate individual acute treatment decisions. Relatively small mathemathical inaccuracies
226