Does music improve sleep?
In one study, sleep quality was improved by listening to music at bedtime.
Yet, we know that individual scientific studies are not conclusive. Theories need to be based on a consensus of research. In order to support a claim, you need multiple, carefully designed and controlled studies.
For this, we turn to meta-analyses. As the name implies, this is an analysis of previous analyses. Investigators dug through large databases full of previous research, and found comparable studies with the same focus. Researchers insured that the comparison was valid by having strict criteria for which studies can be included and which must be left out. These criteria included how participants were chosen, how the study was designed, and the statistical methods used to analyse the data.
Meta-analyses take things one step further, not just comparing similar studies; instead, researchers do their own statistical analysis, using each individual study as a single data point. Basically, it is a new study, but instead of individual people as participants, other studies serve as the unit of analysis.
This allows researchers to see if the effects hold up across multiple controlled tests. Particular attention is paid to the effect size (or Cohen’s d) which indicates the percentage of difference between two groups.
In individual studies, focus is put on the statistics benchmark (P-Value) which, if achieved, means that a result is most likely not by chance. However, effect size is different. A result can be more than chance, but still be so small that it is not useful in making real changes in people’s lives.
Effect size (Cohen’s d) lets researchers know the real size of a change between groups, by showing you the percentage of overlap between the two comparison groups.
In multiple studies, music has been shown to improve sleep quality. However, we are not sure how large this effect is. If the music group had only a very small or negligible improvement in sleep, this could still be statistically significant (P-Value) enough to be published in a scientific journal, but would not be useful for people who were actually trying to find effective ways of improving sleep.
To see if the effect is practically or clinically useful, you need to look at the size of the effect (Cohen’s d), or the degree to which the two groups’ distributions overlap. You also need to look at the combined results of multiple studies, after doing careful work to make sure they are comparable.
We already know that studies have shown music improves sleep quality. Below is a summary of a meta-analysis that investigated whether these studies had more than a small effect size when analysed together. This meta-analysis shows that music’s effect on a sleep quality across studiesis large enough to be practically and clinically useful.
De Niet G., Tiemens B., Lendemeijer G. B., Lendemei Jer B. & Hutschemaekers G. (2009). Music-assisted relaxation to improve sleep quality: meta-analysis. Journal of Advanced Nursing 65(7), 1356–1364doi: 10.1111/j.1365–2648.2009.04982.x
Investigators searched through several medical and psychological peer-reviewed databases for randomized and controlled studies that examined the relationship between presleep music exposure and sleep quality. Particularly, music assisted relaxation (MAR), and its effects on adult and elderly peoples’ sleep quality were assessed. Researchers operationalized sleep quality largely based on the Pittsburgh Sleep Quality Index (PQSI) (Buijsse et al. 1989). Studies dealing with participants diagnosed with serious cognitive or neurological disorders were excluded. Most studies included accompanying relaxation instructions. Included studies were divided into two groups, 1) studies using music alone and 2) music in addition to other interventions.
No statistically significant difference between these subgroups was found, implying that additional relaxation instructions were not very effective, although this is a tentative conclusion. Five studies meeting the methodological criteria were analysed collectively. One of these studies included both a music and music video condition, so ultimately six intervention conditions were examined in this meta-analysis. One study used music specifically designed for relaxation.
The setting of the intervention (hospital or home), as well as the age range of the participants, and the amount of time spent on intervention before sleep, all varied across selected studies. All studies used subjective, self-report measures of sleep. Four studies used the PSQI and one study used the Richards-Campbell Sleep Questionnaire (RCSQ) (Richards, 1987). These two measures are not directly comparable, because they are inverted with high scores indicating opposite qualities of sleep depending on the scale. To deal with this, researchers subtracted scores on the RCSQ from the maximum score, as a way of inverting.
Five of the six studies showed a statistically significant effect of pre-sleep music on sleep quality. The sixth study approached statistical significance (P=.06). Assuming the standard cutoff of (P=.05), this study was one percent more likely to have occurred by chance then is allowed by the cutoff. This cutoff is a somewhat arbitrary statistical boundary.
Statistical software was used to determine individual study’s effect size (Cohen’s d). Harmat et al. (2008), the student sleep study from write up #1, was included in this analysis. In this meta-analysis, based on the aggregate result of five methodologically sound studies, music was found to moderately improve sleep.
“Music is one of the most-used self-help strategies to promote sleep” (De Niet, Tiemens, Lendemeijer, Lendemei Jer & Hutschemaekers, 2009)
“Clinical studies have shown that music can influence treatment outcome in a positive and beneficial way. Music holds the promise of counteracting psychological presleep arousal and thus improving the preconditions for sleep” (De Niet, Tiemens, Lendemeijer, Lendemei Jer & Hutschemaekers, 2009)
“the use of music could be beneficial for people with sleep (onset) problems. Even in patients with chronic sleep problems, whose frustration about not being able to fall asleep might be a perpetuating factor, music could potentially be beneficial”
“Music might be a valuable contribution to the range of non-pharmacological nursing interventions to promote sleep.” (De Niet, Tiemens, Lendemeijer, Lendemei Jer & Hutschemaekers, 2009)
“most non-pharmacological interventions require a relatively large investment in training. The systematic application of music interventions does not involve large investments in training or tools. These interventions are ‘relatively inexpensive, readily available, portable, and completely subject controlled’” (Mornhinweg & Voigner 1995, p 252).
“music-assisted relaxation is an effective aid for improving sleep quality in patients with various conditions” (De Niet, Tiemens, Lendemeijer, Lendemei Jer & Hutschemaekers, 2009)
“We found scientific support for the effectiveness of the systematic use of music-assisted relaxation to promote sleep quality” (De Niet, Tiemens, Lendemeijer, Lendemei Jer & Hutschemaekers, 2009)
This study is a meta-analysis, meaning the researchers aggregated multiple studies and ran their own statistical analysis using the results of each constituent study as a single data point. This allows for a general sense of effect sizes over multiple studies. Meta-analyses are useful because they access research trends in a way not otherwise visible. However, this methodology is problematic in that they can either compare things that are not legitimately comparable, or limit their data so much as to make the analysis redundant due to pretest homogenization of the data. Data homogeneity and publication bias for the individual studies were statistically measured and found to be within a reasonable range for this meta-analysis.
During the final round of study selection, researchers disqualified five studies because either the data was not comparable, or the methodological quality of the study was poor (i.e. no control). Selected studies were “assessed using the Delphi list for quality assessment of (Verhagen et al. 1998), which assesses randomization and blinding. Population variation adds validity to this analysis. Additionally, this meta-analytic test may suffer from publication bias, especially given the small number of studies examined and their small sample sizes. Based on the nature of the study, double blind conditions were impossible:
“All included studies suffered from some methodological flaws. The Delphi list score was mainly compromised by the requirement for blinding. In high quality RCTs, a double blind process is used: neither participant nor administer should be aware of whether the participant is in the intervention or control group. However, the nature of the intervention makes blinding of participants virtually impossible; when patients are informed about the goal and procedure of the trial, as good ethical practice demands, it is impossible to hide the condition to which they are allocated. Randomization was blinded in all included studies”
“Because there is evidence that music has the potential to reduce anxiety, it holds the promise for counteracting psychological presleep arousal and thus improving the preconditions for sleep. Moreover, Johnson (2003) has suggested that music can decrease the frustration and dread associated with sleep complaints”
Pzizz Researcher’s note: Anxiety reduction as opposed to general relaxation is an important distinction.
“Many people experience slow rhythm music, without a heavy beat, as relaxing. However, the effect is strongly dependent on personal preferences”