Table 1.

Setting, study characteristics and key findings of included studies

Study	Study design	Country	Setting	Sample	Intervention and professional involved	Primary outcome measure	ACP outcomes	Health-care outcomes	Key findings
Caplan et al. [17]	Controlled before and after study	Australia	Nursing homes (NHs) (n = 32) and hospitals (n = 3)	Nursing home residents (NHRs), their families, staff and general practitioners	Education for residents, families, staff and GPs	Not specified	Number of ACP referrals	Number of Emergency Department presentations and hospital admissions	ACP. In the first year, 63 NHRs referrals were received; 45 NHRs agreed to proceed with ACP. Five were judged to have capacity: discussions and decisions documented in their notes
				Potential nursing home participants n = 1,344			Number of ACP discussions	Number of emergency ambulance calls from all the nursing homes (NHs)	Of the 40 without capacity: 3 ‘persons responsible’ completed a plan of treatment, 20 NHRs had discussions and preferences were recorded in notes but no document was signed, 10 NHRs completed discussions but nothing formally documented, 6 persons responsible declined discussion of ACP, 1 NHR had previously completed ACP
							Number of discussions that proceeded to a written form by either the resident or a proxy	Number of emergency and elective admissions to hospital and bed-days occupied by residents in hospital	Mortality. No significant change in mortality until the final third year when mortality rose in the control nursing homes (30.4 versus 41.6 deaths per 100 nursing home beds; P = 0.0425)
								Number of deaths registered to the address of each NH	Health care use. Annual rate of admission of residents per nursing home bed initially higher in the intervention hospitals, but lower by the final year (0.865 versus 1.254; RR = 0.89; 95% CI: 0.85–0.93; P < 0.0001). Risk of a resident being in an intervention hospital bed for a day compared with control group fell by a quarter from being initially similar (RR = 1.01; 95% CI: 0.98–1.04; P = 0.442) to being lower (RR = 0.74; 95% CI 0.72–0.77; P ≤ 0.0001)
Hanson et al. [19]	Controlled before and after study	USA	Nursing homes (n = 9)	Nursing home residents (n = 458)	Recruitment and training (in palliative care practices) of palliative care leadership in each facility	Not specified	Number of documented ACP discussions	Percentage of residents receiving hospice or palliative services, pain assessment, pain treatment among residents in pain	ACP. Documented ACP discussions were rare at baseline. Intervention NHRs had a significant increase in documented discussions from 4 to 17% (P < 0.001)
	Two nursing homes were chosen at random for training						Number of orders on life-sustaining treatments, living wills and health-care powers of attorney		DNR orders increased slightly in intervention nursing homes (58–65%, P = 0.04), as did use of DNR flags. Living wills (31% versus 30%) and health-care powers of attorney (27% versus 33%) were not significantly changed
									Patient health. Significant increase in the intervention group in pain assessment from 18 to 60% (P ≤ 0.001); no change in orders for pain medication
									Health-care use. Significant increase in hospice enrolment in the intervention group from 4.0% at baseline to 6.8% post intervention (P ≤ 0.01)
Molloy et al. [18]	Randomised control trial	Canada	Nursing homes (n = 6)	Nursing home residents (n = 1,133)	Two days educational workshop for nurses in nursing homes who then trained other staff and counselled patients. Used Let Me Decide. Nurses carried out the discussions; Doctors reviewed and signed the forms	Residents’ and families’ satisfaction with health care and health-care utilisation over 18 months	Rates of completion of advance directives	Residents’ and families’ satisfaction with health care and	Satisfaction with healthcare. No significant differences between intervention and control homes
	Six nursing homes were randomly allocated to the two trial arms							Health care utilisation over 18 months	ACP . Overall, 49% of participants completed ADs. At the end of the study, 70% of NHRs in the intervention group and 57% in the control group completed ADs. In the control homes 71% of ADs were DNR orders and in the intervention homes 89% of directives were Let Me Decide
									Mortality. Similar in both intervention and control groups (24% versus 28%; P = 0.20)
									Health-care use. Mean hospitalisation rates per patient lower in the intervention group compared with the control (0.27 versus 0.48; P = 0.001)
									Average total cost per patient (health-care resource use) significantly less in the intervention group compared with the control group (CAD 3,490 versus CAD 5,239; P = 0.01); hospital costs per patient also less in the intervention group (CAD 1,772 versus CAD 3,869; P = 0.003).
Morrison  et al. [20]	Non-randomised controlled trial	USA	Nursing home	Nursing home residents (n = 139)	Education, using interactive methods, in ACP for intervention group social workers. Social workers (usual care) received a talk on state law and medical decision-making	Not specified	Nursing home chart documentation of advance directives and DNR orders	Concordance of treatments received with documented preferences	ACP. Intervention residents were more likely to have their preferences documented in nursing home chart than control group residents. Significant increases in documentation of patent preferences in the intervention group compared with control: cardio pulmonary resuscitation (40% versus 20%; P = 0.005); artificial nutrition and hydration (47 versus 9%; P ≤ 0.01); intravenous antibiotics (44 versus 9%; P = 0.01) and hospitalisation (49 versus 16%; P ≤ 0.01)
	Four social workers were randomly allocated to two arms						Documentation of patient preferences for care		Health-care use. Control residents were significantly more likely than intervention residents to receive treatments discordant with their prior stated wishes (P = 0.04)

*Int, intervention; **Con, control; ***CI, confidence interval; RR, relative risk.