Musical activity in a subsample of the German National Cohort study

Musical activities (MA) such as singing, playing instruments, and listening to music may be associated with health benefits. However, evidence from epidemiological studies is still limited. This study aims at describing the relation between MA and both sociodemographic and health-related factors in a cross-sectional approach. A total of 6717 adults (50.3% women, 49.7% men, median age: 51 years (IQR 43–60) were recruited from the study center Berlin-Mitte of the German National Cohort (NAKO), a population-based prospective study. This study is based on a sample randomly selected from the population registry of Berlin, Germany, aged 20 to 69 years. 53% of the participants had been musically active at least once in their life (56.1% women, 43.9% men). Playing keyboard instruments (30%) and singing (21%) were the most frequent MA. Participants listened to music in median 90 min per day (IQR 30.0–150.0). Musically active individuals were more likely to have a higher education, higher alcohol consumption, were less likely to be physically active, and had a lower BMI compared to musically inactive individuals. This large population-based study offers a comprehensive description of demographic, health, and lifestyle characteristics associated with MA. Our findings may aid in assessing long-term health consequences of MA.


Study population
The present study used data from the German National Cohort (NAKO), a population-based, prospective study that enrolled more than 200,000 participants in 18 study centers across Germany with the aim of identifying causes of common chronic diseases and detecting their preclinical stages 26 .Participants were between 20 and 69 years of age at the timepoint they were included in the study 27 .The study design of the NAKO has been described in detail elsewhere 26,28 .NAKO was approved by the ethical review committees of all participating study centers including the Charité-Universitätsmedizin Berlin and all procedures followed the guidelines as represented in the Declaration of Helsinki in its current form.Written informed consent was obtained from all participants.As stated above, the MusA questionnaire was applied in the study center NAKO Berlin-Mitte.The baseline examination in the NAKO Berlin-Mitte took place from 2014 to 2019 (n = 11,048).The MusA questionnaire was only used from 2016 onwards.879 participants completed the MusA questionnaire during their baseline examination (2016-2019).5913 individuals completed the MusA questionnaire during their first follow-up examination (2019-2022).Since the present study focused on amateur musicians only, professional musicians (n = 75) were excluded from the present analyses.The total study population comprised 6717 participants (women = 3336, men = 3381).

Assessment of musical activity
In the MusA questionnaire, participants indicated how many hours/minutes a day they listened to music in the last 12 months and which music genres they preferred.The music genres were predefined: Rock/Pop, Classical music, German Schlager, Jazz, Oldies/Evergreens, Dance/Hiphop/Rap, Hard Rock/Heavy metal, Techno/House, Folk music/Brass music, Opera/Operetta/Singing, Musicals or other.Further, MusA assessed the number of years of music making and the respective weekly practice hours in five individual life stages (≤ 10 years, 11-20 years, 21-30 years, 31-50 years, > 50 years).Participants counted as non-musically active if they had not played an instrument or sung at any stage of their life.In contrast, participants counted as musically active if they had played an instrument or sung at a single or at multiple stages of their life.We calculated the total music-hours for each life stage (years of music making multiplied by annual active music hours) based on the information about the years of music making and the weekly active music hours (conversion into annual active music hours by factor 52.14).Further, we calculated the cumulative lifetime music-hours as the sum of total music-hours of all life stages.Lifetime active music years are the summation of the years of music making of all life stages.Practice

Lifestyle factors and other variables
During baseline assessment, all study participants underwent an extensive examination program: Age, sex, and nationality were assessed via self-report.Various information about socio-economic characteristics were assessed by a standardized, computer-assisted face-to-face interview i.e., household size, living situation, education, employment status, family net income, and migration background.Education level was determined using the International Standard Classification of Education 29 , further categorized as 'low' , 'medium' or 'high' education.The equivalized income was calculated by dividing the family net income by the number of household members (converted into equalized adults).Smoking status, alcohol consumption and physical activity were assessed via touchscreen-based self-administered questionnaires.In addition, the Alcohol Use Disorders Identification Test short version (AUDIT-C) was applied to assess hazardous drinking.Risky alcohol consumption was defined as AUDIT-C Score > 4 for men, and > 3 for women 30 .Physical activity was measured using the Global Physical Activity Questionnaire 31 .Based on this information Metabolic Equivalents (MET)-min/week were calculated.MET minutes per week indicate how much energy is expended on various activities throughout the week and are a surrogate for activity intensity.Trained study staff (according to standard operating procedures) measured weight and height when participants were wearing underwear without shoes using a calibrated integrated measurement station (SECA model 764, Seca®, Hamburg, Germany).BMI was calculated by dividing body weight (kg) by height in meters squared (m 2 ).In addition, BMI was also categorized according to common BMI categories, as follows: underweight (BMI < 18.5 kg/m 2 ), normal weight (18.5 ≤ BMI < 25.0 kg/m 2 ), overweight (25.0 ≤ BMI < 30.0 kg/m 2 ), obesity (≥ 30.0 kg/m 2 ).Waist circumference was measured at the midpoint between the lower rib and the iliac crest.The subjective life satisfaction and subjective health status were assessed by questionnaires 32,33 .Likert scales were used to have the respondents rate the extent to which they agreed with their personal life satisfaction, ranging from completely unsatisfied (0) to completely satisfied (10), and health status, ranging from excellent (1) to bad (5) 32,33 .

Statistics
Normal distribution was tested for all continuous variables.As all variables were skewed, they were reported as median and interquartile range (IQR) in all tables.Categorical variables were reported as n (percentages).The present study characterized the sample of musically active participants (amateur musicians), as well as music listeners at baseline in terms of socio-demographic data, physical activity, BMI and other lifestyle factors such as smoking and alcohol consumption.To investigate differences in basic characteristics of participants who were musically less vs. more active, a median split procedure was carried out for the cumulative lifetime music-hours (median: 1042.8 h).A Mann-Whitney U test was used to compare the continuous variables between different groups, and a chi-square test was used for categorical variables.Possible associations between music listening minutes (minutes/day) and other variables were investigated by using categorization into tertiles of music listening minutes per day (T1 (Min-Max): 0-50 min/day, T2: 60-120 min/day, T3: 125-1440 min/day).Orthogonal polynomial contrasts were used in analysis of variance (ANOVA) or multivariable adjusted analysis of covariance (ANCOVA) to test for linear trends across tertiles of music listening minutes per day.ANCOVA was only performed for cumulative lifetime music-hours, lifetime active music years and practice density, adjusted for sex, age, education, smoking, physical activity, BMI, employment status, and monthly equivalized income.As all variables were skewed, variables were log-transformed prior to ANCOVA and afterwards back-transformed.As the geometric means useful summaries skewed data, cumulative lifetime music-hours, lifetime active music years and practice density were expressed as geometric means and 95%-confidence intervals (95% CI).As extended analyses, potential predefined determinants (i.e.sex, age, education, smoking, physical activity, BMI, employment status, monthly equivalized income) of cumulative lifetime music-hours were investigated using regression analyses, examined in Model 1 (unadjusted) and further mutually adjusted in Model 2 using data of all participants (n = 6717).The exposure variable (cumulative lifetime music-hours) and all continuous determinants (i.e.age, physical activity, BMI, monthly equivalized income) were log-transformed prior to the analyses.All statistical analyses were performed using SAS software, version 9.4 (SAS Institute, Cary, NC, USA).P-values were not corrected for multiple testing.

Results
Table 1 shows the basic characteristics of the total study population (n = 6717) with a median of 51 years of age (IQR 43.0-60.0).In the last 12 months, 23.5% were musically active.A total of 53% (n = 3512) of the study population were musically active across one or more stages of life, whereas 3109 individuals were never musically active (N = 6621, missing information n = 96).Musically active and inactive groups were found similar with respect to their age ranges.Table 1 also shows that MA was more frequent in females as compared to males.Indeed 70.9% of singers and a similar majority of woodwind players were female, who also dominated keyboard playing with 58.6%.By contrast, plucked, brass and percussion instruments were more male-dominated, leaving strings as the only category with a balanced sex distribution (see Supplemental Table 1, for details).With 98.6% indicating Caucasian descent, 15.9% also indicated a migration background.The latter proportion is lower than the total proportion in the entire city of Berlin, which amounts to 24.3% 34 .The socioeconomic variables revealed an advantage of the MA group with respect to both education and family income (Table 1).Accordingly, no differences have been seen in sex-stratified analyses (Supplemental Table 2).Individuals in the MA group were more likely to work part time (applies mainly to men, Supplemental www.nature.com/scientificreports/than three people (Table 1).Overall, with 69.8% full employment and an unemployment rate of 3.1% in our total study sample, their unemployment rate was lower than that of the entire city of Berlin, where the current statistics indicates an unemployment rate of more than 9% 35 .Concerning lifestyle variables, smoking was similar across groups, but MA individuals were in median less physically active, showed lower BMI and waist circumference values, but consumed more alcohol than their musically inactive peers (Table 1).Moreover, even after adjusting for age and sex MA individuals showed a lower BMI compared to the musically inactive group (data not shown).Further, basic characteristics according to musical activity stratified sex were presented in Supplemental Table 2.

Musical activity
Considering MA across one or more stages of life we observed that 60.4% (n = 2121) of all musically active individuals (n = 3512) commenced this activity in their early childhood (≤ 10 years of age). Figure 1    stages (median 2 to 3 h/week).To combine the information of extent and weekly active music hours the present study focused on the cumulative lifetime music-hours (n = 3121, missing information n = 391).In median, the cumulative lifetime music-hours in musically active participants accounted for 1042.8 h (IQR 417.1-2659.1).The musically active study population (i.e.instrumentalists and singers) had a median of 7.0 lifetime active music years (IQR 4.0-14.0),corresponding to a median practice density of 156.4 h per active music year (IQR 104.3-231.2).In the total population (N = 6717) the main determinants in individuals with high cumulative lifetime music-hours in the unadjusted model 1 (Supplemental Table 4) were female sex, lower age, lower physical activity, low BMI, high monthly equivalized income, high education and employment.Only slight variations occurred in the mutually adjustment model (Supplemental Table 5).

Musical activity and lifestyle
Table 2 presents a comparison between musically less active vs. more active participants by means of a median split of cumulative lifetime music-hours (< 1042.8 h vs. ≥ 1042.8 h).The former group accumulated only 3 years of playing music/singing in median in their life course and had about half of the practice density compared to the more active group, i.e. hours of practice per active year of singing or playing an instrument (Table 2).We observed no differences between the musically less active and the more active group according to sex, age, or any other lifestyle factor, as well as household size living situation (all p ≥ 0.05, Table 2).However, the musically more active group had a higher education status and was more likely to work part-time (< 35 h/week).In addition, the musically more active group tended to have a better subjective health status and subjective life satisfaction (Table 2).
At the individual peak of MA among all musically active participants, keyboard instruments represented the most frequently played instrument group (30%, n = 1020), followed by singing (21%, n = 731) and plucked instruments (19%, n = 639) (Fig. 2).As shown in Supplemental Table 2, there were some peculiarities in the characteristics of participants depending on the main instrument.

Music listening
In addition to active music making, music listening was assessed by the MusA questionnaire.In the present study 6533 participants reported that they listened to music in median for 90 min per day (IQR 30.0-150.0).An ANOVA/ANCOVA across tertiles of music listening minutes per day (T1 (Min-Max): 0-50 min/day, T2: 60-120 min/day, T3: 125-1440 min/day) showed trends of linear associations in basic characteristics across the tertiles (Table 3).
We observed that the amount of music listening was associated with higher age, BMI, waist circumference, as well as practice density across the tertiles (Table 3).In contrast, monthly equivalized income and education were inversely related to music listening.Individuals with low music listening time were more frequently non-smokers as compared to the intense listeners (Table 3).In addition, there was a positive association between smaller household size (i.e.single person household, household with 2 persons) with time spent listening to music.However, households with larger numbers of individuals showed lower affinity to music listening.Moreover, participants who were employed tended to spend less time with listening to music (inverse association across tertiles), but contrasting information was observed in unemployed or retired participants (Table 3).
With regard to the most popular music genres, it was found that Rock/Pop (43.8%),Classical music (10.3%) and German Schlager (9.6%) were preferred most often (Fig. 3).As shown in Supplemental Table 6 the preferences for music genres were equally distributed across sex and age with few exceptions.Musicals were more often preferred by women, and Hard rock/Heavy metal, Folk music/Brass music and Techno/House were more often favored by men.Younger people tended to listen to Techno/House, while older participants preferred Opera/ Operetta/Singing and German Schlager.

Discussion
The study presents a comprehensive description of demographic, health-and lifestyle factors associated with active and receptive musical activity in a cohort of over 6000 participants.The findings also take into account the instruments being played, the engagement in singing, and an assessment of preferred music genres when listening to music.The present work focuses on amateur musicians only, professional musicians were excluded from the analyses.These findings are important, because the level of detailed information with regard to different kinds of MA can rarely be found in previous studies.However, this work is aligned with recent developments of inventories to address both active and receptive MA in recent cohort studies on healthy ageing 36 .
In the current sample, about every second person (53%) of the population had been musically active at least once in their life.This proportion is similar to findings from a Danish cohort study with respect to the participating amateur musicians 25 .MA is most common in the first decade of life, but decreases after the age of 20, presumably due to other life tasks such as continued education, starting a profession, or raising a family.It is likely that MA in the early years is at least partially driven by socioeconomic and educational factors, which may disadvantage families with low income and/or education in this regard.However, the importance of MA in childhood and adolescence with respect to cognitive and social-emotional development has been implicated in previous work using data from panel studies 37,38 .These lines of research suggest a greater role of parental education for the amount of MA in children, and a more mixed picture regarding family income.Interestingly, a statistical survey of amateur musicians in Germany noticed that most of those who are musically active at the age of 30 usually remain so into old age 2 .Indeed, also the present study reveals that 75% of the participants who are musically active during the life stage of 31-50 years are still active beyond their 50th birthday.The lifestyle characteristics of musically active individuals showed a trend of drinking more alcohol, being less physically active, and having a lower BMI in comparison to musically inactive individuals.This pattern resembles previous observations concerning social groups that are differentiated by socioeconomic status and education 39,40  www.nature.com/scientificreports/Previous studies demonstrated that playing music instruments may result in physiological activation and energy expenditure [41][42][43] .Whether or not this activation has an influence on the BMI, has not yet been investigated.In addition, the potential influences of MA on lifestyle, leading to lower BMI, for example, should be a subject of future investigations.Those may also address the question, whether musical activity can help developing social skills and networks with more active lifestyles, especially when it comes to playing music in families 38,44 .
The detailed comparison between groups with lower and higher MA in the current study shows that the latter group accumulated a four-fold greater number of years of musical engagement over the life course and that they had a double practice density.
However, no differences have been observed between the two groups in the majority of lifestyle factors, with the exception of education.Participants with a higher musical engagement had a higher level of education.This underlines the proposed connection between education and MA, a finding that is mirrored in a recent survey on amateur music in Germany.Therein, 25% of teenagers with a higher socio-economic status played music, but only 12% from the lower and 14% from the middle class 2 .In addition, MA of younger individuals seems to depend on the educational status and income of their parents 45 .
The current study allows singers and instrumentalists to be considered independently.The main instruments in the musically active population were keyboards and plucked instruments, a finding that partially mirrors the previously cited report, which also entails a higher prevalence for males playing electronic and brass instruments and females being more active with woodwind instruments 2 .We also note a sex imbalance favoring female singers, which could be expected from previous work on amateur singing in Germany 46 .Interestingly, we further observed a lower education in the group of singers and plucked instruments than in those playing keyboard instruments.However, these results need replication in other population-based samples.Our data also show a trend towards a slightly lower socioeconomic status of the singers compared to the MA group as a whole.
Music listening was found to be associated with greater physical activity, but showed an inverse relationship with socioeconomic indicators.The observation that frequent listening was found more often in individuals with   lower socioeconomic could mirror their reduced access to MA 46 .Alternatively, music listening could function as a preferred emotion regulation strategy in this group, whereas a more selective music listening might take place in so-called high brow listeners, i.e. listeners with higher levels of education and socioeconomic status 47 .Finally, enhanced levels of music listening which were found associated with greater physical activity could also reflect its wide use in sports activities like jogging 48 .The question of whether and to what extent music listening is a factor to promote more active lifestyles is of interest due to positive effects on mood, physical performance, perceived exertion and oxygen consumption 49 .
The preferences for music genres in the present study were almost equally distributed across sex and age, with women showing more interest in musicals, whereas men were more interested into Hard Rock and Heavy metal.The development of musical taste is influenced by cultural and social environment as well as age, gender, income, personality style and musical experience, which could partially explain the observed gender patterns 50,51 .However, a more detailed analysis will be required to assess genre preferences in different age groups.Knowledge about such music preferences and, for example, familiarity, may be helpful in the application of music listening as strategy in stress regulation 52 .
A strength of our study is the large sample size and the availability of high-quality data.At the present time the use of the MusA questionnaire provided the most comprehensive data on active and receptive MA.Through its use in the NAKO study center Berlin-Mitte, these data were linked with comprehensive information on demographic, health, and lifestyle characteristics.Nevertheless, the cross-sectional design of the present study does not allow a causal inference.Furthermore, we used data from one study center only.Therein, recruitment takes place across a range of different communities across the city of Berlin.However, the results might be not generalizable to other populations.In addition, the MusA survey was applied prior to the COVID-19 pandemic shows that the highest percentage of MA was seen between 11 and 20 years of age (75.0%, n = 2633).MA is present in all adult age groups (21-30 years: n = 1084; 31-50 years: n = 1005; > 50 years: n = 492).In total, 476 individuals reported continued MA for their entire life (participants age ≤ 20 years: n = 1; ≤ 30 years: n = 107; ≤ 50: n = 230; > 50 years: n = 138).Supplemental Table3presents the extent (years of music making) and active music hours (hours/week) of musical activities in individual life stages.Active music hours per week differed only slightly across individual life

Figure 1 .
Figure 1.Percentage of musical activity across different life stages (musically active study population n = 3512).Results are given for musically active persons (n = 3512) only, musically inactive persons are not considered (n = 3109).Blue boxes display participants who were musically active in the respective life stages.Red boxes display participants who were not musically active in the respective life stages.Blue-or red-fading streams indicate changes of musical activity between different life spans of participants.A stream from blue to red indicates a change from a musically active to a musically inactive life stage.A stream from red to blue indicates a change from a musically inactive to musically active life stage.Grey-fading streams display an age restriction of study participants which were younger than the considered life stage.

Table 2 .
Differences between musically less active and more active participants (assessed by cumulative lifetime music-hours median split) (n = 3121, missing information n = 391).Data expressed as median (interquartile range (IQR)) or percentage (n), Group comparisons were calculated by chi-square tests or Mann-Whitney U tests. a Risky alcohol consumption was defined as AUDIT-C Score > 4 for men, and > 3 for women.

Table 3 .
Basic characteristics across tertiles of music listening minutes (minutes/day) (n = 6533, missing information n = 184).Data expressed as median (IQR) or percentage (n), p-values were calculated by chisquare tests or ANOVA/ANCOVA testing for linear trend.a Risky alcohol consumption was defined as AUDIT-C Score > 4 for men, and > 3 for women.b Data expressed as geometric mean (95%-CI), ANCOVA adjusted for sex, age, education, smoking, physical activity, BMI, employment status, monthly equivalized income.c Considering only musically active participants.

Figure 3 .
Figure 3. Percentage rates of music genres a most often indicated as favorites (n = 6576, missing information n = 141).a List of music genres predefined in the MusA Questionnaire.

Table 1 .
Basic characteristics of the total study population and according to musical activity.Data expressed as median (IQR) or percentage (n).a Participants count as non-musically active if they have not played an instrument or sung at any stage of life.b Participants count as musically active if they have played an instrument or sung at a single or multiple stages of life.c Missing information on musical activity (n = 96).d Risky alcohol consumption was defined as AUDIT-C Score > 4 for men, and > 3 for women. .