Summary of findings 1. Summary of findings .

Communication intervention compared with usual care for end of life care
Patient or population: people approaching the end of life (within 12 months), their family members and/or carers Settings: any (residential care, hospital (inpatient and outpatient units) and community‐based clinics, palliative care services) Intervention: interventions to improve communication about EoL and/or EoL care Comparison: usual care
Outcomes	Intervention effects	Number of participants (studies)	Certainty of the evidence (GRADE)
Patient, family and/or carer outcomes
Knowledge and understanding   Variable scales: information (amount and type) needs/preferences; discordance between patient and physician survival and curability estimates   Timing: immediately to 1 month post‐consultation	Overall, interventions to improve communication may have little or no effect on measures of knowledge of illness and prognosis, or information needs and preferences   In 1 study (303 participants), discordant estimates of 2‐year survival between patients and doctors (intervention 59% versus usual care 62%) and curability (intervention 39% versus usual care 44%) were similar between groups (Epstein 2017). Another (79 participants) reported similar proportions of patients had their preferences for amount of information met or exceeded across intervention and usual care groups, but that type of information was met or exceeded more often in the intervention group (93% versus 80%) (Walczak 2017). 1 final study (170 participants) reported no differences in patients’ unmet information needs overall (Clayton 2007)	552 (3 studies)c	⊕⊕⊝⊝ lowa,b
Evaluation of the communication: different constructs (perceptions of communication quality; patient‐centredness of communication; involvement preferences; doctor‐patient relationship measures)   Timing: immediately post‐consultation to 18 weeks post‐consultation	Across constructs (patient‐centredness, involvement preferences, doctor‐patient relationship, satisfaction with consultation), there may be minimal or no effects of interventions to improve communication about EoL and EoL care (Agar 2017; Bernacki 2019; Clayton 2007; Epstein 2017; Walczak 2017), and uncertaintye about effects on quality of communication (Au 2012)	6 studiesf	⊕⊕⊝⊝ lowb,d
Discussions of EoL/EoL care: discussion timing and length   EMR review post‐death   Timing: at time of family conference (intervention) in ICU; post‐death	The intervention may lead to longer and earlier discussions of EoL and EoL care, compared with usual care, but each result is based on a single study   1 study (108 participants) reported comparative data: median family conference intervention duration was 30 minutes (IQR 19 to 45 minutes) versus usual care (median 20 minutes, IQR 15 to 30 minutes) (Lautrette 2007)   1 study (376 participants) reported that the first documented Serious Illness Conversation happened earlier among intervention group participants (median 143 days prior to death (IQR 71 to 325) than usual care (71 days, IQR 33 to 166) (Bernacki 2019)	484 (2 studies)	⊕⊕⊝⊝ lowb,d
Discussions of EoL/EoL care: discussion occurrence   EMR review post‐death; coding of consultations; self‐reported occurrence   Timing: immediately, 1 or 2 weeks post‐consultation; after death	Overall, we are uncertain about the effects of interventions to improve discussions about EoL care   2 studies indicated that the intervention increased the occurrence of EoL discussions, compared with usual care (RR 1.96, 95% CI 1.61 to 2.39; 2 trials, 537 participants); the others indicated little or no effect of the intervention on mean total numbers of patient questions in consultations (MD 1.58, 95% CI ‐1.82 to 4.98; 2 trials, 249 participants)	786 (4 studies)	⊕⊝⊝⊝ very lowb,g,h
Adverse (unintended) outcomes	There is insufficient evidence to determine whether adverse (unintended) outcomes are associated with communication interventions. Patient and/or carer anxiety was reported (3 studies), but was judged as confounded, and no other unintended consequences, or worsening of desired outcomes, were reported	‐	‐
CI: confidence interval; EMR: electronic medical record; EoL: end of life; ICU: intensive care unit; IQR: interquartile range; MD: mean difference; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of the effect.

^aDowngraded (‐1) due to inconsistency (different outcome measures and concepts assessed across studies).
^bDowngraded (‐1) for indirectness (all participants were older patients with advanced cancer or chronic obstructive pulmonary disease (COPD) or both, and results may not apply to other populations nearing EoL).
^c1 further study (Lautrette 2007, 108 participants) conducted in an ICU setting did not report useable data.
^dDowngraded (‐1) due to imprecision (results are from a single study and/or a small number of participants).
^eQuality of communication also downgraded (‐1) due to methodological limitations (sequence generation rated at unclear risk of bias).
^fMost of these outcomes under this broad construct were assessed by only 1 study; doctor‐patient relationship was reported by 3 studies (238 participants); and numbers of participants were consistently small across all outcome measures.
^gDowngraded (‐1) for inconsistency (2 of 4 studies indicated that the intervention had no effect, with residual variation despite similar populations and interventions).
^hDowngraded (‐1) for methodological limitations (the largest study rated as at unclear risk of bias on sequence generation).