Clinical Application of Listening to Music to Prevent Delirium in the Intensive Care Unit

BACKGROUND AND RATIONALE Delirium is a prevalent form of acute brain dysfunction in the Intensive Care Unit (ICU), affecting a significant proportion of older critically ill adults, including those receiving invasive mechanical ventilation. It is independently associated with increased mortality, long-term cognitive impairment, and higher healthcare costs. Current pharmacological strategies for delirium prevention have limited efficacy; consequently, clinical guidelines recommend multicomponent non-pharmacological interventions. This study aims to evaluate the clinical effectiveness of a structured music listening intervention as a preventive strategy in older ICU patients.

STUDY DESIGN AND SETTING This study is a single-center, prospective, randomized, controlled trial with three parallel groups conducted in the mixed medical-surgical Intensive Care Unit (27 beds) at the Hospital Británico de Buenos Aires, Argentina.

RECRUITMENT AND RANDOMIZATION Patients admitted to the ICU are screened daily. To ensure inclusion of high-acuity patients (including those receiving invasive mechanical ventilation), screening continues until the patient becomes evaluable. The recruitment window opens during the first 24 hours after the patient first achieves an adequate level of consciousness (RASS -2 to +2) and is confirmed CAM-ICU negative. Randomization must occur within 24 hours of the patient first becoming evaluable and no later than 120 hours (5 days) from ICU admission. Patients who are CAM-ICU positive are not eligible for randomization. Patients who do not become evaluable within 120 hours of ICU admission are not eligible. Allocation is managed via a secure, web-based automated randomization module (REDCap) hosted on the Hospital Británico server. To ensure allocation concealment, the sequence is generated by an independent statistician using permuted blocks of random sizes and loaded into the system. The assignment is revealed to the investigator only after the participant's baseline data has been entered into the system.

INTERVENTION PROTOCOL The protocol consists of music sessions scheduled twice daily in two predefined time windows (Morning: 08:00-12:00; Afternoon: 14:00-20:00) to allow controlled delivery without interfering with ICU care. Each session lasts at least 30 minutes for up to 7 days after randomization, and the first session is started ideally within 2 hours (maximum 4 hours) after randomization unless clinically contraindicated or delayed by procedures. All protocolized sessions are delivered exclusively using study-provided MP3 players and circumaural headphones to ensure technical standardization. Outside the protocolized sessions, participants may listen to additional music using either study equipment or personal devices, with source and minutes recorded separately. Research staff perform a visual check 15 minutes into each session to support protocol fidelity. Sessions may end before 30 minutes if delirium develops (e.g., agitation) or at the participant's request. Study staff ensure that the volume is comfortable and not excessively loud (approximately equivalent to normal conversation, 60-70 dB). Headphones and MP3 players undergo cleaning/disinfection per institutional Infection Control standards between patients. Minutes of total music exposure are recorded daily. In the Relaxing Music arm, any listening that is not consistent with the assigned relaxing-music intervention is recorded as a protocol deviation ("contamination"), with minutes documented. Personalized Music (MP): Playlists are curated based on the participant's prior music preferences (songs provided for loading onto the study MP3 player) and aim to evoke positive emotions. Relaxing Music (MR): Standardized playlists feature slow-tempo (60-80 bpm) instrumental tracks without lyrics. Intervention delivery and daily outcome assessment cease at the earliest of day 7, documented ICU transfer order, death, or withdrawal of consent.

CONTROL GROUP AND STANDARD OF CARE The Control arm receives the institutional standard of care for delirium prevention (ABCDEF Bundle). Importantly, this standard of care is provided to all study participants regardless of group assignment. To preserve the integrity of the comparison, Control participants and families are asked to avoid using music devices during the fixed 7-day study window; however, music is not actively prohibited if the patient requests it or uses personal equipment. Any such exposure is recorded as a protocol deviation ("contamination"), including duration (minutes) and type of exposure. Television or spoken-word radio (e.g., news, podcasts) is allowed as part of usual care.

BLINDING AND ASSESSMENT Delirium assessment is performed daily by a blinded investigator unaware of group assignment. The assessment integrates direct observation using CAM-ICU and a structured chart review of nursing notes from the previous 24 hours using a predefined delirium chart-review algorithm (Chart Review method). The routine blinded assessment is performed daily at 12:00 (±2 hours). Any positive assessment (in-person CAM-ICU or Chart Review) classifies that calendar day as a delirium day (non-free day). To reduce misclassification due to sedatives/analgesics, CAM-ICU assessment is deferred until at least 2 hours after a sedative/analgesic bolus or discontinuation of a continuous infusion.

PRIMARY ENDPOINT (OPERATIONAL DEFINITION) The primary endpoint is Days Alive and Free of Delirium and Coma (DCFDs; range 0-7) in a fixed 7-day window starting on the calendar day of randomization (day 1). Coma days are defined as RASS -3, -4 or -5 and are considered non-free days (CAM-ICU is not performed on coma days). If a participant receives a medical discharge order (transfer order) from the ICU before day 7, remaining days are considered free days (regardless of physical discharge delays); if death occurs before day 7, remaining days are considered non-free days.

KEY SECONDARY ENDPOINT DEFINITIONS While full secondary outcomes are listed in the specific data fields, the following operational definition is specified to distinguish it from the primary endpoint logic: ICU Length of Stay (ICU LOS): Defined as the duration from ICU admission until physical discharge from the unit (actual transfer out of the ICU). Unlike the primary endpoint (DCFDs), which uses the documented medical transfer order to define ICU discharge for endpoint adjudication (regardless of physical discharge delays), ICU LOS captures the total duration of ICU bed occupancy, including administrative delays, to reflect resource utilization.

SAMPLE SIZE ASSESSMENT A total of 330 participants is planned (1:1:1 allocation; 110 per arm). Sample size was defined to ensure adequate power for the primary confirmatory objective under a hierarchical gatekeeping strategy, assuming a minimally clinically relevant difference of 1 day in the primary endpoint (DCFDs) with a common standard deviation of 2.5 days and a two-sided alpha of 0.05, using an asymptotic normal approximation as a conservative reference. The pre-specified primary confirmatory comparison is Combined Music (MP+MR; n=220) versus Control (n=110), providing >90% power under these assumptions. The theoretical reference size for a 1:1 comparison is approximately n≈99 per group for 80% power (relevant for MP vs MR to detect a 1-day difference). The final planned size (110 per arm) includes a 10% increase over the theoretical reference to account for missing data within the 7-day evaluation window and preserve ITT power. Because DCFDs is discrete and bounded (0-7) and the confirmatory test is non-parametric (Mann-Whitney U), the normal-approximation sizing is considered conservative; where appropriate, a simulation-based check under plausible DCFDs distributions may be performed as an internal robustness assessment.

STATISTICAL ANALYSIS PLAN AND GATEKEEPING The primary analysis follows the Intention-to-Treat (ITT) principle. The primary endpoint (DCFDs, 0-7) will be compared using a two-sided Mann-Whitney U test (Wilcoxon rank-sum). Results will be reported as medians [IQR]. Effect size will be reported using the Hodges-Lehmann shift estimate with 95% confidence intervals, and a probability-of-superiority measure (e.g., Cliff's delta or an equivalent estimator) will be provided to facilitate clinical interpretation. To control the global type I error rate and address the clinical questions in a pre-specified priority order, a hierarchical gatekeeping strategy is employed (two-sided alpha=0.05): Step 1 (Confirmatory): Combined Music (MP + MR) versus Control for DCFDs. Step 2 (Key Secondary, under gatekeeping): Performed only if Step 1 is statistically significant (p<0.05); MP versus MR for DCFDs, using the same two-sided Mann-Whitney U test (alpha=0.05). Sensitivity and consistency analyses are pre-specified. Because Mann-Whitney is unadjusted, a sensitivity analysis using robust regression (or quantile regression) will adjust for pre-specified baseline covariates (age, APACHE II, and baseline invasive mechanical ventilation) to confirm robustness to chance baseline imbalances; however, the unadjusted Mann-Whitney test remains the primary basis for confirmatory conclusions. Independently of the gatekeeping outcome, individual comparisons (MP vs Control and MR vs Control) will be reported descriptively with point estimates and 95% confidence intervals as consistency analyses. Secondary outcomes will be interpreted as exploratory, except for two key supportive endpoints (incidence of delirium and days with delirium), for which Holm-Bonferroni adjustment will be applied within that predefined family (family-wise alpha=0.05); supportive endpoints do not determine confirmatory conclusions.

MISSING DATA Proactive procedures (daily monitoring, eCRF alerts, and redundancy via chart review) are implemented a priori with the goal of keeping missingness below 5%. Missing data handling for DCFDs is pre-specified and follows the operational definition of the endpoint. Days with RASS ≤ -3 are classified as coma (non-free days) without requiring CAM-ICU. For evaluable days (RASS ≥ -2) with missing CAM-ICU, the primary ITT analysis uses Multiple Imputation by Chained Equations (MICE) with a longitudinal multilevel approach (including a subject-level random effect), generating m ≥ 20 imputed datasets. Predictors include baseline variables (e.g., age, APACHE II, baseline invasive mechanical ventilation), available daily measures (e.g., RASS, sedation dose), and prior-day status. DCFDs is derived within each imputed dataset and combined using Rubin's rules. Sensitivity analyses include complete-case analysis if overall missingness is <5% and a conservative worst-case scenario imputing all missing evaluable days as non-free (delirium positive).

SUBGROUP AND EXPLORATORY ANALYSES A pre-specified subgroup analysis evaluates heterogeneity of treatment effect in participants receiving invasive mechanical ventilation (IMV) at baseline using a Group × Baseline IMV interaction term adjusted by APACHE II. For time-to-event outcomes in ventilated participants (time to successful extubation), death is treated as a competing risk; the primary analysis uses a Fine-Gray competing risks model (subdistribution hazard ratio, sHR), with a cause-specific Cox model as sensitivity analysis. A pre-specified landmark sensitivity analysis at 4 hours post-randomization will be performed, restricting to participants who remain mechanically ventilated at that time; additional landmark thresholds between 2 and 6 hours will be evaluated as sensitivity analyses. Ventilator-free days at 28 days (VFD-28) will be compared using Mann-Whitney U due to its composite distribution. An exploratory dose-response analysis will assess the association between total music minutes (continuous) and DCFDs within the intervention arms using robust regression (or quantile regression), adjusting for baseline severity (APACHE II) and a sedation-depth indicator (e.g., average RASS), recognizing potential confounding by illness severity and reverse causality; results will be interpreted as exploratory.

Inclusion criteria

• Age 65 years or older.
• Negative Confusion Assessment Method for the ICU (CAM-ICU) at the time of randomization.
• Expected ICU length of stay of at least 48 hours based on clinical judgement at admission (e.g., need for mechanical ventilation, vasopressors, invasive monitoring).

Exclusion criteria

• Positive CAM-ICU prior to randomization.
• Acute primary Central Nervous System pathology (e.g., ischemic or hemorrhagic stroke, traumatic brain injury, meningoencephalitis, intracranial hypertension) presenting with altered sensorium (Glasgow Coma Scale at admission < 14) or focal deficit preventing cognitive assessment (e.g., severe aphasia).
• Severe chronic cognitive impairment or advanced dementia, defined by a known history of severe functional dependence prior to admission (requiring permanent assistance for basic activities of daily living such as feeding or grooming)
• Unresolved auditory or visual impairment.
• Suspected or confirmed intoxication with drugs or alcohol at admission.
• Uncontrolled psychiatric disease (at least one acute episode requiring intervention in the last 6 months).
• Richmond Agitation-Sedation Scale (RASS) score < -2 or > +2 at the time of randomization.
• Medical condition precluding the safe use of headphones (e.g., burns, skull fracture, skin lesions).
• Inability to establish effective communication due to a language disorder or unresolved language barrier.
• Inability of the patient or legal representative to provide informed consent.
• Patients with imminent death, end-of-life care, or limitation of therapeutic effort.