Decreasing Delirium Through Music

A Randomized Pilot Trial

Sikandar H. Khan, DO, MS; Chenjia Xu, PhD; Russell Purpura, MD; Sana Durrani, MBBS; Heidi Lindroth, PhD, RN; Sophia Wang, MD; Sujuan Gao, PhD; Annie Heiderscheit, PhD, MT-BC, LMFT; Linda Chlan, PhD, RN; Malaz Boustani, MD, MPH; Babar A. Khan, MD, MS


Am J Crit Care. 2020;29(2):e31-e38. 

In This Article


We conducted a 3-arm, single-blind, randomized controlled clinical trial; details of the protocol have been published elsewhere.[20] The Indiana University institutional review board reviewed and approved the study. In brief, we included English-speaking adult patients (≥ 18 years old) admitted to the ICU and receiving mechanical ventilation for at least 24 hours but not more than 48 hours. We excluded patients who had been receiving mechanical ventilation for longer than 48 hours because delirium develops early during the ICU stay, and our intervention may have preventive and therapeutic effects. Patients were excluded if they had neurologic injury, chronic neurologic disease, or uncorrected hearing or vision impairments; were intoxicated by or in withdrawal from alcohol or drugs; were in a coma after cardiac arrest, pregnant or nursing, or incarcerated; or the primary team did not consider the patient appropriate for the study (eg, patient soon enrolling in comfort care). We obtained consent from the patient or, if the patient was unable to provide consent, from their legally authorized representative (LAR). If initial consent was obtained from the LAR, the patient was approached for reconsent once they were able to communicate.[20] To randomize patients, we used permuted block randomization with various block sizes and computer-generated random numbers. Patients were assigned to 1 of 3 arms: (1) personalized music (PM) playlists incorporating patients' preferences based on information obtained from their LAR; (2) nonpersonalized relaxing, STM (60–80 beats per minute) consisting of piano, guitar, and classical music and Native American flute sounds (eg, Lifescapes: Relaxing Piano, by John Story, and Watermark, by Enya) preselected by a board-certified music therapist; and (3) audiobooks for attention control (AC).[20] Patients in the AC group were randomly assigned 1 of the following audiobooks: Treasure Island, by Robert Louis Stevenson; Harry Potter and the Chamber of Secrets, by J.K. Rowling; or Dr. Seuss's Oh the Places You'll Go!. These books were chosen for their readability, their broad appeal, the quality of the audiobook narration, and the high audiobook ratings (as reviewed on commercial websites).

All patients received two 1-hour sessions each day (between 9 and 11 AM, and between 2 and 4 PM) for up to 7 days. The music or audiobook was delivered through noise-canceling headphones attached to Apple iPod Shuffle mp3 players. These sessions continued until the patient was transferred out of the ICU, was discharged, or died. In-hospital follow-up to measure delirium, pain, anxiety, and clinical and mobility outcomes continued until discharge or day 28, whichever occurred first.

We evaluated 4 primary outcomes of the pilot study: (1) recruitment rate (enrollment of 5 patients per month; 60 patients in 12 months), (2) adherence to the prescribed intervention (80% of sessions delivered), (3) acceptability of the intervention (patient survey), and (4) feasibility (retention of 80% of participants). Secondary outcomes were the estimated effect of music on the number of delirium/coma-free days, delirium severity, anxiety, pain, physiological stress, and mobility. (See the published protocol for details.[20])

Data Collection

We collected demographic data, baseline cognitive and functional statuses, clinical data (including medications), and blinded outcome assessments, as described in the published protocol.[20] We obtained music preferences from all patients or their LARs at enrollment using a Music Assessment Tool.[12] Research assistants assessed patients' delirium and delirium severity twice daily (after each intervention) using the Confusion Assessment Method for the ICU (CAM-ICU) and the CAM-ICU-7, respectively.[2,21–23] They assessed patients' anxiety once daily (after the morning intervention) using a self-report visual analog scale (0, no distress; 4, very severe distress).[24,25] Finally, the research assistants assessed patients' pain twice daily (after each intervention) using the Critical Care Pain Observation Tool.[26] To measure adherence, research staff, who were blinded to the type of intervention, recorded the duration of each music/audiobook session, including start and stop times, and reasons for any interruptions. Vital signs (heart rate, blood pressure, respiratory rate) were also recorded before and after each session. Staff obtained patients' mobility milestones from inpatient therapy notes. Patients were randomly surveyed by telephone after hospital discharge to assess the acceptability of the audio selections, the fit and comfort of the headphones, and whether they would enroll in the study again.

We conducted a 3-arm randomized controlled trial to test feasibility and acceptability of music in intensive care patients.

Statistical Analysis

We performed an intention-to-treat analysis. We compared baseline characteristics using the Fisher exact test and the Wilcoxon rank sum test. We considered adherence to the intervention as the percentage of sessions delivered, adjusted for the number of days the patient was eligible. We present delirium outcomes as delirium/coma-free days because delirium and coma fluctuate over hours or days, because delirium is difficult to assess when a patient is in a coma, and because death and discharge affect delirium/coma outcomes. We defined delirium/coma-free days as the number of days a patient was alive and free from coma or delirium; we compared delirium/coma-free days among the 3 groups using the Wilcoxon rank sum test. Delirium/coma-free days provide an estimate of the duration of normal brain function (free from coma and delirium), and hence they function as a surrogate of delirium duration not confounded by coma or death. Previous high-impact studies have used delirium/coma-free days as an outcome, and the variable accounts for confounding by death or discharge. For patients discharged from the hospital before day 7, the remaining days until day 7 were counted as delirium/coma-free.

To provide a conservative estimate of the intervention's effects on delirium, and to be consistent with methods applied in prior studies, for patients who died or withdrew before day 7, we counted their subsequent delirium/coma-free days as 0; this managed the conflicting effects of the intervention on delirium and survival.[7,9,22] Similarly, to provide a conservative estimate of music's or an audiobook's effects on delirium severity, we imputed values missing from the CAM-ICU-7 on the basis of the patient's worst coma/delirium status. We chose the patient's worst delirium severity score, rather than the group's mean score, because of the small sample size in this feasibility trial. We present delirium severity during the intervention period as the median daily CAM-ICU-7 score for each patient. We report level of consciousness as the median of the mean daily scores on the Richmond Agitation- Sedation Scale for each patient during the intervention period. We define ventilator-free days as the number of days the patient was alive and breathing without mechanical ventilation. We present medication exposure as the percentage of patients receiving at least 1 dose of medication during the intervention period. A mean daily dose was calculated from the total amount of a drug administered in 24 hours. We converted benzodiazepine doses to lorazepam equivalents and opioid doses to morphine intravenous equivalents.

Patients received music or an audio book through noise-canceling headphones and mp3 players for 1 hour twice daily for up to 7 days.

We analyzed changes in heart rate and blood pressure as mean differences before and after each intervention session, and we used fixed effects models to analyze changes in anxiety and pain scores over time. We used a Cox proportional hazards model to analyze time to ambulation. We calculated length of stay using dates of admission to and discharge from the ICU, date of death, or date of withdrawal from the study; we compared length of stay among the 3 groups by using the Wilcoxon rank sum test.