Reviewed and revised 26 August 2015
Appropriate outcome measures is vitally important in ICU research
Types of outcomes
- ICU mortality
- hospital mortality
- 90 day mortality
- 1 year functional outcome
- simple, single metric
- concrete endpoint which is already available in hospital databases
- death is an important endpoint
- aggregation of a large number of diagnoses with a small number in each increases power to detect variation
- variation over time may reflect institutional and organizational events or characteristics- budget cuts, bed pressure etc. and be able to detect true quality deficiencies
- May be useful when combined as part of an overall quality program
- Definition of ICU is very hospital specific which can influence mortality (e.g. non-ICU step-down areas in some hospitals)
- As a consequence can be ‘gamed’ e.g. transfers out to die in the ward or other units
- Poor correlation between mortality and quality of care in some diagnoses – alternatives available e.g. diagnosis specific risk models e.g. EuroSCORE for CABG, APACHE SMR, trauma scores. Can mask problems in low volume diagnostic groups
- Difficult to draw hospital comparisons and or allow league table construction
- False conclusions can be drawn unless robust statistical methods used
Compared to ICU mortality, avoids many problems of censoring at ICU discharge. Hospital mortality can often be 50% higher than ICU mortality, and is a reasonable surrogate (90%) for 90 days mortality.
- gets over differences in definition of ICU and ICU discharge thresholds.
- Still a simple and robust endpoint which is easy to obtain from existing hospital databases.
- can confound intensive care outcomes with deficiencies in ward or other post ICU care.
- Does not address in any way functional outcomes [so discharge from hospital to a nursing home in a vegetative state is counted as a positive outcome]
90 DAY MORTALITY
Simple robust endpoint which addresses the issue of ongoing mortality after hospital discharge (though this difference is about 10% relative in recent large trials).
- simple robust endpoint
- data may be available by linkage with external registries (e.g. Births, Deaths and Marriages)
- still an arbitrary time point [while 28 days is clearly inadequate, 90 days may still be insufficient to accurately measure the attributable mortality from an episode of critical illness]
- Problems with loss to follow up after ICU discharge.
- Ethical implications of contacting patients after discharge (especially for research studies).
ONE YEAR FUNCTIONAL OUTCOME
- a ‘POEM’ (patient oriented endpoint that matters)
- Takes into account disability and true long-term consequences of critical illness.
- no ideal scoring tool available – existing tools all have problems; some measure particular functional domains well
- problems with face validity
- All functional outcome measures are time consuming to apply
- Problems with loss to follow up
- Time consuming, labour intensive, costly face to face vs. phone vs. mail follow up
- Depending on disease may reflect more the natural history of the disease rather than the ICU care per se