Delirium Reduction via Scripted Family Voice Recordings in Critically Ill Patients Receiving Mechanical Ventilation

CL Munro. Am J Crit Care. 2025; Nov 1;34(6):429-437. doi: 10.4037/ajcc2025486

Clinical Question

• In mechanically ventilated adult ICU patients, does a structured voice reorientation intervention recorded by family members, compared to usual care, reduce delirium?

Background

• Delirium affects a significant proportion of patients receiving mechanical ventilation. It is an independent predictor of ICU mortality, and is associated with poor long-term cognitive outcomes, longer ICU and hospital stays, and higher healthcare costs
• Pharmacologic interventions have not been shown to prevent delirium and may be associated with worse outcomes
• Non-pharmacologic interventions (e.g. continuous reorientation, family engagement) may help but are supported by low-quality evidence
• Direct family involvement in ICU can be limited by other social demands and visiting policies, especially during the COVID-19 pandemic
• Existing evidence:
  • A small pilot study by the same group, Munro et al. (2017), suggested a family voice-recorded intervention may be a simple, low-cost and easy-to-implement effective intervention to reduce delirium

Design

• Prospective, 2-arm, blinded randomised controlled trial
  • Institutional Review Board (IRB) approvals obtained prior to initiation of trial. Written informed consent in English or Spanish obtained from all subjects’ legally authorized representatives (LAR)
  • Registered at ClinicalTrials.gov (ID NCT03128671) and study protocol published 2021
• Randomisation and allocation concealment:
  • 1:1 allocation to intervention (Family Automated Voice Recording, FAVoR) or usual care using permuted blocks of variable size
  • Randomisation in sealed opaque envelopes
  • Research personnel completing CAM-ICU assessments blinded to assignments, and both groups had wireless speaker present next to bed
• Data collection:
  • Delirium assessment: Blinded outcome assessors conducted CAM-ICU assessments twice a day for a total of 7 days (intervention recordings lasted 5 days) or until ICU discharge for primary outcome measure
  • CAM-ICU is a well-validated delirium assessment tool in ICU
  • Medical record review: Baseline demographics and clinical characteristics; type of ICU (e.g. surgical, cardiovascular etc.) duration of mechanical ventilation, duration of ICU stay and hospital stay
• Sample size and power:
  • Assumptions: attrition rate of 40% to account for early extubation and death
  • Power calculations: Pre-planned comparison of Chi-square test data comparing delirium-free days in two groups with effect size of W = 0.35 (moderate to large effect) with 95% power
  • Sample size calculations: Using PASS software and their pilot study data, sample size of 127 required. Planned recruitment sample size 178 to include 40% attrition rate: 89 in each group
• Statistical analysis:
  • Primary outcome: Delirium-free days compared with Chi-square test between two groups as unadjusted analysis
  • Multivariate modelling analysis with Poisson regression analysis to control for covariates: intubation process, ICU length of stay, total dosage of intervention (number of recordings listened to)
  • Effect size calculator: Online calculator used to calculate standardised mean difference of Poisson rates of delirium-free days
  • Missing data: intention-to-treat principle; multiple imputation analysis using R package MICE (Multiple Imputation by Chained Equations, R Foundation for Statistical Computing) to assess sensitivity of multivariate analysis to missing data

Setting

• 9 ICUS at 2 large hospitals in USA
• April 2018 – November 2020 (enrolment temporarily suspended March 2020-June 2020 due to COVID-19 pandemic)

Population

• Inclusion criteria:
  • Adults ≥18 years; within 48 hours of initial intubation and ICU admission
  • Patient/legally authorised representative able to provide informed consent in English or Spanish
  • Family member able to speak English or Spanish and willing to record scripted audio messages
• Exclusion criteria:
  • Dementia
  • Anticipated imminent patient death
  • Medical contraindication to intervention (psychiatric history of auditory hallucinations; profound deafness)
  • Inability to speak English or Spanish
• Participant flow:
  • Screened/enrolled: 4852 assessed/182 enrolled; 4670 excluded (4427 did not meet inclusion criteria; 111 declined to participate; 132 other reasons)
  • Randomised: 178 randomised – 89 to each group (4 withdrew prior to randomisation)
  • Analysed: 178 for univariate analysis; 165 for multivariate analysis (13 excluded by statistical software due to missing values on covariates)
• Baseline characteristics were broadly similar between groups (FAVoR intervention vs control)
  • Demographics:
    • Age: 58 vs 61
    • Male sex: 65% vs 55%
    • White: 74% vs 82%
    • Hispanic/Latino Ethnicity: 57% vs 63%
    • Black/African American: 25% vs 17%
  • Intubation and clinical severity:
    • Elective intubation: 30% vs 40%
    • Urgent/Emergency intubation: 70% vs 60%
    • APACHE III score, mean 66.5 vs 66.2
  • Type of intensive care unit:
    • Medical ICU: 29% vs 28%
    • Surgical ICU: 24% vs 25%
    • Trauma ICU: 17% vs 12%
    • Neurologic ICU: 10% vs 9%
    • Cardiovascular ICU: 20% vs 26%

Intervention

• Family Automated Voice Recording (FAVoR)
  • Set of 10 recorded messages; each 2 minutes long, played hourly during usual daytime waking hours by speaker next to patient
  • Recorded by family member using standardised script describing ICU environment; day-night cues; presence of healthcare staff/family members to re-orientate patient
  • Played once an hour; up to 8 doses per day
  • Maximum duration 5 days or until ICU discharge if before

Control

• Usual ICU care
  • Wireless speaker placed similarly at bedside to maintain blinding

Outcome

• Primary outcome of delirium-free days:
  • Higher percentage in FAVoR intervention group: 34.8% vs 33.7%
  • Unadjusted analysis: initial unadjusted Chi-square analysis result not statistically significant (P = 0.07)
  • Adjusted analysis: Poisson regression model using ICU length of stay as offset and adjusting for total intervention dosage (number of recordings heard) showed significant difference (P <0.001)
  • Comparison of adjusted Poisson regression models: FAVoR group higher rate of delirium free days (β = 1.094)
  • Effect size: Standardised mean difference 0.354 (95% CI 0.182–0.608)
  • Potential dose relationship: Analysis showed significant positive correlation (r = 0.62, P < 0.001) between dose of intervention and number of delirium-free days
  • Exclusions from multivariate analysis: 13 patients excluded by statistical software due to missing values on covariates
  • Impact of missing data: pre-specified multiple imputation analysis used and P values remained <0.001 suggesting missing data did not bias the outcome of raw data analysis

Authors’ Conclusions

• The FAVoR intervention is a non-pharmacological, low-resource strategy to re-orient patients receiving mechanical ventilation in ICU and was effective in reducing delirium

Strengths

• Easily reproduceable low-risk, low-resource intervention that may also benefit patients’ families experiences
• Clear patient-centred outcome of delirium-free days
• Good representative sample frame for diverse ICU population with RCT across multiple different types of unit (Medical, Surgical, CV etc.)
• Sound methodology with published protocol and clear randomisation
• Met pre-planned adequate power calculations and recruitment numbers
• Reliable delirium assessment method will well-validated CAM-ICU instrument
• Intention-to-treat analysis
• Excellent adherence to intervention arm
• Analysis suggests a dose-response relationship
• Multiple imputation analysis suggested minimal bias from missing data
• Included both English and Spanish native speakers with appropriate message

Weaknesses

• Statistical significance emerged only after adjusted analysis with Poisson regression modelling. Initial unadjusted analysis was not statistically significant with only a modest benefit
  • Results therefore dependent on model assumptions
  • However, this is an appropriate model to use
• Single country with only two US hospitals may limit external generalisability
• Small chance of misclassification of delirium-free days as fluctuating condition only assessed twice a day (although unlikely to differ between groups)
• Unclear if effect due to content of message or familiar voice
• Although baseline characteristics were similar, there were differences in e.g. trauma ICU percentage (17% vs 12%)

The Bottom Line

• The FAVoR trial demonstrates a reduction in delirium for ventilated patients receiving scripted family-recorded voice message to re-orientate them
• This intervention may be a practical, relatively low-cost addition to delirium prevention strategies in ICU
• There was only a 1.1% improvement in delirium free days, a low number of patients recruited to the trial, and although small the differences in baseline characteristics between the intervention & control groups may be significant given the small difference in outcome. I would therefore want to see further data before pushing for widespread adoption of this technique

External Links

• Article: Delirium Reduction via Scripted Family Voice Recordings in Critically Ill Patients Receiving Mechanical Ventilation
• Study protocol: Family Automated Voice Reorientation (FAVoR) Intervention for Mechanically Ventilated Patients in the Intensive Care Unit: Study Protocol for a Randomized Controlled Trial 

Metadata

Summary author: Mark Giza
Summary date: 18th December 2025
Peer-review editor: David Slessor

Picture by: Image by Studio_Iris from Pixabay