Research Letter
Neurology
Music Listening and Multiple Dimensions of Well-Being in Acute Stroke: A Randomized Clinical Trial
Giacomo Giacalone, Silvia De Boni, Sara Pezzotta, et al
JAMA Netw Open 2026;9;(7):e2623732. doi:10.1001/jamanetworkopen.2026.23732
Introduction
Music listening, by engaging widespread neural networks,1 may be associated with improved mood and cognition.2-6 Data in acute stroke remain limited. We hypothesized that during acute stroke hospitalization early listening of structured and personalized vocal music could improve psychological well-being and conducted a randomized clinical trial to assess its feasibility, acceptability, and effects on patient-reported outcomes (PROMs) at discharge.
Methods
This single-center prospective randomized clinical trial with blinded outcome assessment enrolled patients aged 18 to 85 years with acute ischemic or hemorrhagic stroke within 24 to 96 hours of symptom onset, a National Institutes of Health Stroke Scale (NIHSS) score of less than 10, had preserved comprehension and communication, and were without major cognitive or psychiatric disorders. Participants were randomized 1:1 to music listening or standard care. The intervention consisted of 1-hour weekday sessions of personalized listening of Italian vocal music delivered via tablet and headphones until discharge. Primary outcomes at discharge were feasibility, measured as enrollment rate (>90% of eligible patients), dropout (<10%), listening time (>70% of days-in-study); acceptability, evaluated with a 5-item Likert questionnaire (eTable in Supplement 1); Hospital Anxiety and Depression Scale scores (HADS-a, HADS-d); EQ-visual analog scale (EuroQol-VAS). Secondary PROMs at discharge were Insomnia Severity Index (ISI) and EuroQoL 5-dimensions 3-level (EQ-5D-3L). Sample size estimation was based on the 3 coprimary PROMs: assuming a 2-sided α-level of 0.015 and 80% power, 150 participants were planned. Missingness analyses supported the plausibility of the missing-at-random assumption. Multiple imputation was used for missing data. Between-group differences at discharge were assessed using analysis of covariance adjusted for baseline values, followed by multivariable linear regression for covariate adjustment according to the intention-to-treat approach, including sensitivity analyses. All statistical tests were 2-sided, with statistical significance set at P < .015 to account for the 3 coprimary outcomes. Analyses were conducted using SPSS version 26 (IBM). All participants provided written informed consent. This randomized clinical trial followed the CONSORT reporting guideline and was approved by the San Raffaele Hospital ethics committee (NCT06743412). Data were analyzed from June 2022 to November 2024.
Results
Of 165 eligible patients, 150 (mean [SD] age, 67.9 [11.4] years; 55 females [36.7%] and 95 males [63.3%]) were randomized (75 per group) and included for analyses (Figure). Clinical-demographic characteristics were comparable (Table). Feasibility results were: enrollment rate 91%, dropout 6.7%, median (IQR) cumulative listening time 6 (4-7) hours, intervention delivered on median (IQR) 100% (80%-100%) of days-in-study. More than 80% rated the intervention highly acceptable (4 to 5 on a 5-point Likert scale) for each questionnaire item (eTable in Supplement 1). Perceived quality of care was comparable between groups (P = .37). At discharge, the music group had lower HADS-d (adjusted mean difference [aMD], −2.68; 95% CI, −3.90 to −1.47) and HADS-a (aMD, −1.86; 95% CI, −2.93 to −0.79) scores than controls (P < .001). EQ-VAS was comparable between music and control groups (aMD, −0.19; 95% CI, −6.03 to 5.64; P = .94). Music listening was associated with better sleep quality (ISI) (aMD, −2.83; 95% CI, −4.43 to −1.24; P < .001) and lower EQ-5D-3L-pain/discomfort scores (aMD, −0.35; 95% CI, −0.60 to −0.09; P = .008). In multivariable analyses, music listening (β = −2.6; 95% CI, −3.8 to −1.4; P < .001) and male sex (β = −1.7; 95% CI, −2.9 to −0.5; P = .005) were independently associated with lower HADS-d scores at discharge, whereas higher baseline HADS-d scores were associated with higher HADS-d scores at discharge (β = 0.5; 95% CI, 0.4 to 0.6; P < .001). Music listening was independently associated with lower HADS-a scores at discharge (β = −1.9; 95% CI, −2.9 to −0.8; P = .001), whereas higher baseline HADS-a values were associated with higher HADS-a scores at discharge (β = 0.6; 95% CI, 0.5 to 0.7; P < .001). Music listening was independently associated with lower ISI scores at discharge (β = −2.8; 95% CI, −4.3 to −1.2; P < .001), whereas higher baseline ISI (β = 0.4; 95% CI, 0.3 to 0.5; P < .001) and baseline HADS-d values (β = 0.3; 95% CI, 0.2 to 0.5; P < .001) were associated with higher ISI scores at discharge. In sensitivity analyses, intervention remained associated with lower HADS-d and ISI scores at discharge (P < .001). No adverse events related to the intervention were observed.
Figure. Participant Flow Diagram

Table. Clinical and Demographic Characteristics of Patients in the Control and Music Groups

| Variablesa | Patients, No. (%) | |
|---|---|---|
| Control (n = 75) | Music (n = 75) | |
| Age, median (IQR), y | 71 (64-75) | 70 (60-76) |
| Sex | ||
| Female | 29 (39) | 26 (35) |
| Male | 46 (62) | 49 (65) |
| Education, median (IQR), y | 11 (8-13) | 13 (8-15) |
| Diagnosis | ||
| Ischemic stroke | 69 (92) | 66 (88) |
| Cerebral hemorrhage | 6 (8) | 9 (12) |
| Thrombectomy | 11 (15) | 7 (9) |
| Thrombolysis | 12 (16) | 17 (23) |
| Dominant hemisphere affected | 32 (43) | 31 (41) |
| Premorbid mRS, median (IQR) | 0 (0-0) | 0 (0-0) |
| Pre-morbid psychotropic medication | 11 (15) | 12 (16) |
| Arrival scale, median (IQR) | ||
| NIHSS | 4 (2-6) | 3 (2-6) |
| Enrollment scale, median (IQR) | ||
| NIHSS | 2 (1-4) | 2 (1-4) |
| BMRQ | 62 (54-73) | 67 (57-75) |
| HADS-db,c | 8 (4-11) | 9 (4-13) |
| HADS-ab | 6 (3-9) | 7 (4-10) |
| EQ-VASb | 60 (50-80) | 60 (50-80) |
| EQ-5D-3L | ||
| Mobilityb | 2 (1-3) | 2 (1-4) |
| Self-careb,c | 2 (1-4) | 2 (1-4) |
| Usual activitiesb | 2 (1-4) | 2 (1-4) |
| Pain or discomfortb | 1 (1-2) | 1 (1-3) |
| Anxiety or depressionb | 2 (1-3) | 2 (1-3) |
| ISIb,d | 5 (1-10) | 8 (2-13) |
| Onset-to-randomization time, median (IQR), h | 54 (40-76) | 52 (45-71) |
| Days of hospitalization, median (IQR), d | 8 (6-11) | 8 (5-11) |
| Days in study during hospitalization, median (IQR), d | 7 (4-8) | 7 (4-9) |
| Physiotherapy or speech therapy during hospitalization | 34 (45) | 39 (52) |
| Discharge type | ||
| Home | 55 (73) | 56 (75) |
| Rehabilitation clinic | 20 (27) | 18 (24) |
| Death | 0 (0) | 1 (1.3) |
| Discharge scale | ||
| NIHSS, median (IQR) | 1 (0-3) | 1 (0-3) |
| mRS, median (IQR) | 1 (0-3) | 1 (0-3) |
Discussion
Early initiation of structured and personalized vocal music listening during acute stroke hospitalization is feasible, acceptable, and associated with improved mood and self-perceived sleep quality. Limitations include the single-center design and the prevalence of patients with low NIHSS scores, which restrict the generalizability. The absence of an active sham control raises the possibility of nonspecific effects that could be mitigated in light of the similar perceived quality of care between groups. As a low-cost and minimally demanding intervention, music listening could be integrated into acute stroke management, even in resource-limited settings.