Schneiderman 2003.

Methods	Randomised trial
Participants	Participants Patients in randomised trial: n = 551 (intervention: n = 278; control: n = 273) Patients in survey: n = 2 Family/surrogates in survey: n = 106 Healthcare providers in survey: n = 255 Gender: 47.5% female; control: 45.2% female Setting: adult ICUs from 7 hospitals Country: USA
Interventions	Comparison: ethics consultation vs usual care Experimental intervention Name: proactive ethics consultation Aim: the intervention should have addressed the ethical issues involved in a specific clinical case. Its central purpose was to improve the process and outcomes of patient care by helping to identify, analyse and resolve ethical problems. Process: non‐standardised process. 1. Receiving consultation request. 2. Assessment of the request. 3. Ethical diagnosis. 4. Recommendations for next steps. 5. Documentation of the consultation. 6. Follow‐up until discharge. 7. Evaluation. 8. Record‐keeping for education and reflection. Providers: single ethics consultant or teams. Consultant with medical, doctoral or law degree, social workers and theologians. All had to be formally schooled in ethics and philosophy. Access: value‐laden issues were identified by a nurse. Control intervention Name: usual care Possible components: family meetings or other consultations considered appropriate by the healthcare providers.
Outcomes	Postintervention Mortality All further outcomes were measured for patients who failed to survive only: Length of ICU stay in days (low score was good) Length of hospital stay in days (low score was good) Length of stay on the floor in days (low score was good) (reported in detail in Gilmer 2005) Days receiving ventilation (low score was good) Days receiving artificial nutrition and hydration (low score was good) Number of full code/comfort care orders (high score was good) Number of CPR attempts (low score was good) Treatment costs (low score was good) (reported in detail in Gilmer 2005) Daily hospital costs (low score was good) (reported in detail in Gilmer 2005) Patients' or families/surrogates' satisfaction with ethics consultation using Likert scale from 1 to 5 (high score was good) (reported in detail in Cohn 2007) Healthcare providers satisfaction with ethics consultation using Likert scale from 1 to 5 (high score was good) (reported in detail in Cohn 2007)
Funding	Study was supported by the Agency for Healthcare Research and Quality (Grant no. 1 R01H510251).
Study date	November 2000 to December 2002
Declaration of Interest	No statement given.
Notes	Other reports of this review: Cohn 2007; Gilmer 2005
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐supported block randomisation was used.
Allocation concealment (selection bias)	Unclear risk	No information given.
Similar baseline outcome measurements (selection bias)	Low risk	No baseline outcome measurement were performed.
Similar baseline characteristics (selection bias)	Low risk	Baseline characteristics were similar.
Protection against contamination (performance bias)	High risk	Patients were randomised in ICUs from 7 hospitals. 76 patients in the intervention group received no intervention and 77 in the control group received ethics consultation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All data were analysed with intention‐to‐treat analysis.
Knowledge on allocated intervention (Performance and detection bias)	Low risk	Participants and personnel could not be blinded due to the character of the intervention. Research assistant reviewed medical records and interviewed the patients and his surrogates, who received the intervention.
Selective reporting (reporting bias)	High risk	More outcomes were assessed than reported.
Other bias	High risk	Power might be low for this study. Fewer participants were recruited than a priori power analysis deemed necessary.

CPR: cardiopulmonary resuscitation; DNAR: do not attempt resuscitation; ICU: intensive care unit; n: number of participants.