The evolution of depression: Exploring the mismatch hypothesis

The evolution of depression and sleep: depressive

symptoms in university students can be predicted by

sleep quality

Abstract

Depression is a common mood disorder and is becoming increasingly prevalent in students. Furthermore, the lifetime prevalence of depression in university students is around 30.6%. There is a negative correlation between depression and sleep quality and it is even thought that insomnia might be an independent risk factor for developing depression. The co-occurrence between depression and sleep quality could be explained by the mismatch hypothesis. The mismatch hypothesis states that individuals in the western world suffer more from neuropsychiatric disorders, like depression because the environment in which they live is completely different from the one in which our species evolved. The current study contains two parts. The first part examines if sleep quality and mismatches can predict depressive symptoms in students. This is measured through questionnaires. Therefore, the USDI is used to examine depressive symptoms in students, the PSQI to determine the sleep quality and the mismatch questionnaire which is newly developed to measure mismatches. In the second part, interviews are used to study the longitudinal effects of depression and sleep quality and to investigate the effects of smartphone use before bedtime. It is found that sleep quality and mismatches have a significant predictive value on depressive symptoms. Furthermore, it is not clear whether poor sleep quality is a cause or an effect of depression and there are no effects found on smartphone use. The results show that the number of mismatches and sleep quality are good predictors for developing depressive symptoms. The interviews and the survey should be optimized to study the longitudinal effects of sleep on depression and to determine the effects of smartphone use on sleep quality.

1. Introduction

Depression is a common mood disorder that affects individuals of all ages worldwide. The World Health Organization (WHO, 2015) declared depression as the world’s leading cause of disability since generally 350 million individuals worldwide are affected by the disorder. Depression is considered a multi-problematic disorder because it leads to impairments in occupational, social and interpersonal functioning (Baglioni et al., 2011). The average duration of a depressive episode is estimated at eight months and 20% of people develop a chronic depression (Van Randenborgh et al., 2012). Furthermore, 30% of people with depression experience a recurrence, meaning that the chance of relapse is very high in depression (Richards, 2011). Depression has a wide variety of symptoms such as having a depressed mood for most of the day, reduction of interests in (almost) all activities, low energy, poor appetite, insomnia, feelings of worthlessness, lack of concentration, feelings of hopelessness and thoughts of suicide (Association American Psychiatric, 2013). The lifetime prevalence of depression, the number of people who once in their life suffered from depression or still have depression, has a wide variability in the world and even among groups (Ibrahim, Kelly, Adams, & Glazebrook, 2013). Furthermore, the lifetime prevalence in university students in the United States is estimated at 30.6%, where it is 9% in the general population (Ibrahim et al., 2013).

It has been found that there is a correlation between depression and sleep disruption. 50% to 90% of the people with depression suffer from poor sleep quality (Riemann, Berger, & Voderholzer, 2001; Tsuno, Besset, & Ritchie, 2005) and researchers suggest that increased sleep disruption is one of the most important predictors for remitting a new depressive episode (Tsuno et al., 2005). Most individuals with depression experience early morning awakenings, frequent nocturnal awakenings and difficulties falling asleep (Riemann et al., 2001). Several longitudinal studies investigated the link between insomnia and depression and found that they are comorbid (Riemann & Voderholzer, 2003; Tsuno et al., 2005). Additionally, Baglioni et al. (2011) found that non-depressed individuals with insomnia have a twofold risk for developing depression, compared with individuals who have no sleep disruptions. Insomnia predicts subsequent depression in individuals of all ages and it might even be an independent risk factor for depression (Baglioni et al., 2011).

The co-occurrence of depression and sleep disruptions and the wide variability in lifetime prevalence of depression could be explained with the mismatch hypothesis. The mismatch hypothesis is based on an evolutionary perspective that states that individuals in the western world suffer more from depressive symptoms because the environment in which they live is completely different from the one in which our species evolved (Hahn-Holbrook & Haselton, 2014). A mismatch arises when past selection produces behavior that has no enhanced fitness in the present (Marlowe, 2010). It is

thought that our species evolved as hunter-gatherers (Stearns, 2012) and therefore mismatches arise from poor adjustment between psychological and physiological requirements between the hunter-gatherers and the western modernization. This can occur because the environment where individuals of the western world live in nowadays evolves faster to due cultural and technological changes than human genes and biology evolves (Stearns, 2012). The inability to adapt along with these changes can be associated with neuropsychiatric disorders like depression (Hahn-Holbrook & Haselton, 2014).

Sleep habits and the circadian rhythm are considered to be one of the key differences between modern civilization and the hunter-gatherers who possibly causes a mismatch (Hahn-Holbrook & Haselton, 2014). There are some small communities still living similarly to our ancestors, for instance the Hadza, and it is thought that their circadian rhythm is more similar to these hunter-gatherers. By measuring the circadian rhythm of the Hadza, it was found that they have shorter sleep durations, though they spent more time in bed compared to people who live in the western world (Italians) (Samson, Crittenden, Mabulla, Mabulla, & Nunn, 2017). The sleep onset of the Hadza was around ten in de evening and their sleep offset around seven in the morning, with a total sleep duration of six hours. The shorter total sleep duration compared to the longer total time in bed could be explained by nighttime wake-bouts. To compensate for the nighttime wake-bouts the Hadza frequently nap during daytime (Samson et al., 2017). In contrast, Italians had a sleep onset around midnight and sleep offset around eight in the morning with a total sleep duration of seven to eight hours (Samson et al., 2017). The later sleep onset of the Italians could be an effect of more widespread use of artificial light, which results in a “social jetlag” because of the misalignment between the biological clock and the social clock (Nunn, Samson, & Krystal, 2016; Touitou, Reinberg, & Touitou, 2017; Yetish et al., 2016). Overall, the Hadza still has a shorter sleep duration compared with the Italians. This could be explained by the body mass index (BMI) since individuals with a higher BMI exhibit longer sleep durations. The average BMI of the western population is higher than those of the Hadza (Samson et al., 2017). The use of artificial light and social jetlag could affect circadian rhythms. Electric light exposure contributes to later sleep onset, but not to later sleep offset (de la Iglesia et al., 2017). This difference becomes larger during winter, which suggests that later sleep onsets are an effect of electric light due to exposure to brighter light intensities during winter evenings (de la Iglesia et al., 2017). Piosczyk et al. (2014) investigated the effects of sleep-wake behavior in individuals who lived in Stone Age conditions for two months where they had no access to electricity or any modern conveniences. Living in these conditions resulted in earlier sleep onset, but not a later sleep offset (Piosczyk et al., 2014).

The hormone melatonin can explain the finding of De la Iglesia et al. (2017) and Piosczyk et al. (2014). Melatonin plays a crucial role in homeostatic mechanisms by signaling whether it is light or dark (Mundey, Benloucif, Harsanyi, Dubocovich, & Zee, 2005). Moreover, it is thought that melatonin

directly affects human sleep (de la Iglesia et al., 2017) since the secretion of melatonin is associated with increased sleep propensity, the preparation to transit from wakefulness to sleep, or the capability to stay asleep when already sleeping (Cardinali, Brusco, Lloret, & Furio, 2001). Melatonin secretion is light-dependent and it is increased during the night when it is dark and decreased during the day when it is light (Nose et al., 2017). Artificial light in the late evenings, like smartphone use, results in less melatonin and later sleep onset. This may be because a suppression or delay of melatonin secretion occurs (Cajochen et al., 2011; Nose et al., 2017). Interestingly, it was also found that late-night light exposure increased cognitive performance, which was associated with sustained working memory and declarative memory and attention. This finding is useful for late-night works, but it is not for falling asleep (Cajochen et al., 2011) (for more information about melatonin see appendix A).

Thus, there is a correlation between sleep disruption and depression and it is found that sleep disruptions could be an effect of mismatches. Furthermore, it is thought that smartphone use can suppress melatonin and therefore could cause poor sleep quality. However, it is still unidentified if sleep quality and mismatches can predict depressive symptoms. It is also still unknown if depressive symptoms are a cause or an effect of poor sleep quality. Therefore, the current study will investigate if sleep quality and mismatches can predict depressive symptoms in students and if poor sleep quality is a cause or an effect of depression. Furthermore, the current study will investigate if smartphone and computer use before bedtime could lead to poorer sleep quality. It is hypothesized that students who do not get enough sleep and have more mismatches will have more depressive symptoms than students who get enough sleep and experience fewer mismatches. Furthermore, it is hypothesized that poor sleep quality is a cause of depression and that participants who use their smartphone before bedtime will have poorer sleep quality. There is specially chosen for students since they are a high-risk population for developing depressive symptoms (Ibrahim et al., 2013). Furthermore, it is thought that students experience many mismatches because they are going to live on their own in big cities for college.

This research question was studied in two parts, wherein the first part questionnaires were used to investigate the predictive value of sleep quality and the mismatch hypothesis on depression. In the second part, participants with a low and high score at the questionnaires were interviewed to determine the longitudinal timetable of depression and the effects of artificial light before bedtime. Furthermore, the interviews are used as a qualitative investigation to improve the mismatch questionnaire. In the second part, participants got to wear an activity tracker as well to measure sleep quality and total sleep duration. The common way to classify depression is through questionnaires such as the Beck Depression Inventory, the Hamilton Depression Inventory, and the Zung Depression scale, but these are argued to measure dominant symptoms of depression, while

depression at university students has been known to be a consequence of situational stressors (Romaniuk & Khawaja, 2013). Since this study is specifically focused on students, the University Student Depression Inventory (USDI) was used because the USDI has been found to measure students unique depressive symptomatology (Habibi, Khawaja, Moradi, Dehghani, & Fadaei, 2014). Sleep quality was measured with the Pittsburgh Sleep Quality Index (PSQI). The PSQI is primarily intended to measure sleep quality and not to provide accurate clinical diagnoses and is easy in use for the participants (Buysse, Reynolds, Monk, Berman, & Kupfer, 1989). The number of mismatches was measured with the mismatch questionnaire, which was specially developed for this research.

Based on the aforementioned studies, it is expected that individuals with poor sleep quality will have more depressive symptoms and individuals with a higher mismatch score will have more depressive symptoms. It is even proposed that sleep quality and mismatches could predict depressive symptoms at students. Furthermore, it is predicted that poor sleep quality is a cause of depression since it is thought that insomnia might be an independent risk factor for developing depression. Additionally, it is expected that phone and computer use before bedtime will lead to poor sleep quality and sleepiness.

2. Methodology I

2.1 Participants

Prior power analysis for multiple linear regression with two predictor variables showed that at least 68 participants had to be included, preferably, with an equal sex ratio. This would mean 34 females and 34 males. All participants were MBO, HBO, bachelor WO or master students and as inclusion criteria, the participant had to be between sixteen and thirty years old, student and Dutch as their native language. All participants filled out an informed consent before participating.

2.2 Materials

2.2.1 USDI

The questionnaire to measure depressive symptoms at students is the USDI (see appendix B). The USDI contains thirty items on a 5-point Likert scale from experiencing an item not at all (1) until all the time (5) over the past two weeks. Furthermore, the USDI is subdivided into three sub-scales: academic motivation, cognitive and emotional and lethargy. These subscales produce a total scale score between the 30 and a 150 for every participant, where a higher score indicates a higher level of depressive symptoms. The total mean score from the English version is 71.1 with a standard deviation

of 19.3 (Khawaja & Bryden, 2006). There is a strong internal consistency found for the Dutch USDI (Cronbach's  = 0.95)(Heerink, 2019).

2.2.2 Mismatch questionnaire

The mismatch questionnaire measures mismatches in people in modern societies. The mismatch questionnaire contains 78 items subdivided into fourteen subscales: sleep rhythm, lack of exercise, processed food, lack of sun exposure, lack of social contacts, materialism, performance-orientating and perfectionism, lack of time, lack of freedom, lack of happiness, fear of missing out, worrying, deviant youth and unhealthy habits (see appendix C). These subscales are based on common mismatches in people in modern societies (Marlowe, 2010). Individuals nowadays have different sleep rhythms compared with hunter-gatherers, which is partly due to artificial light. People in modern societies experience less exercise and lack of sun exposure since they do not hunt and because most of the jobs are inside. Nowadays most of the food is processed so it has a longer expiration date. People in modern societies experience fewer social contacts since there is less food trade and individuals do not go foraging. Moreover, individuals currently experience less time and they care more about material goods. They all want to be the best and are afraid to miss out on something. Children have a different and less independent youth compared with the most hunter-gatherers and individuals in modern societies experience less freedom since it is harder to just move away in difficult situations. At least, people in modern societies probably worry a lot more, since they are more aware of the problems all over the world (Marlowe, 2010). The mismatch questionnaire is newly developed.

2.2.3 PSQI

The PSQI measures sleep quality over the last month. The PSQI contains nineteen items and is subdivided into seven subscales: subjective sleep quality, sleep onset latency, sleep duration, sleep efficiency, sleep disturbances, use of sleep medication and daytime dysfunction (Backhaus, Junghanns, Broocks, Riemann, & Hohagen, 2002) (see appendix D). A PSQI score smaller or equal to five indicates for a good sleeper and a score above five indicates for a poor sleeper (Buysse et al., 1989). The English version of the PSQI is a valid measurement with a high internal consistency (Cronbach's  = 0.83) and a high test-retest reliability ( = 0.87)(Buysse et al., 1989).

2.3 Procedure

Ethical approval was obtained from the University of Amsterdam. The study was part of a larger investigation into the evolution of depression: Exploring the mismatch hypothesis. All participants

filled in the PSQI, USDI and mismatch questionnaire which were advertised by social media for nineteen days. The survey was taken in October, which is important to note since depression is seasonally dependent. The survey was created in Qualtrics and beside the PSQI, USDI and mismatch questionnaires also a food, high sensitivity, and quick big five questionnaire were implicated for other studies (see Figure 1). Before participating all participants filled in informed consent. At the end of the survey participants were asked to fill in their email if they were willing to participate in further investigation.

Figure 1. Schematic overview of the questionnaires used in the survey. In the survey, six questionnaires (PSQI, food questionnaire, mismatch questionnaire, High Sensitive Person questionnaire, Quick big five and USDI) were used. The questionnaires were taken in this order and also the questions were in a set order. The survey took 25 minutes in total with informed consent in the beginning and at the end a question if you would be willing to leave your email address for further investigation.

2.4 Statistical analysis

The dependent variables were the scores on the USDI questionnaire and the independent variables were sleep quality and the number of mismatches. The survey results were available in Qualtrics and exported to check for missing data. After deleting participants with missing data, the mean age and standard deviation were calculated. The number of males and females and the percentage of all study levels were calculated as well to investigate if the data is normally distributed.

The total scores of the USDI, PSQI and mismatches questionnaire were calculated, as well as the scores of every subscale. Furthermore, the mean and standard deviation for the USDI, PSQI and mismatch questionnaire were calculated. The Cronbach alpha for all three questionnaires and subscales were calculated to measure the internal consistency.

Firstly, the data was analysed to check the assumptions of normality, linearity, homoscedasticity, and multicollinearity. Thereafter the data was checked for sex differences. Afterwards, correlations between the USDI and PSQI and between the USDI and mismatch questionnaire were calculated. Additionally, correlations between the USDI and all mismatch subscales were calculated. If the correlations between the USDI and PSQI and between the USDI and mismatch questionnaire were significant, with a significance level of 0.05, a multiple linear regression was executed. The USDI score was the predicted Y-variable and the PSQI and mismatch scores the predictor X-variables. With a summary, the predictive values of sleep quality and mismatches on depressive symptoms were measured. A significance level of 0.05 was applied.

A principal component analysis (PCA) was executed to investigate underlying factors. This was only done for the mismatch questionnaire since the questionnaire is self-developed and never used before. Before executing the PCA the Kaiser-Olkin (KMO) and Bartlett’s test were executed, to test if a PCA is convenient. The PCA is convenient, if the KMO is above 0.50 or if Bartlett’s test had a p-value below 0.05. The correlation matrix table was used to determine the rotation. If there was a low correlation between the questions, varimax rotation was be used. The oblique rotation was used if there is a high correlation between the questions. The questionnaire contains fourteen categories, so it was expected for the PCA to contains fourteen components as well.

3. Results I

3.1 Participants

148 participants filled out the questionnaires (36 males, 107 females, 5 unknown). 43 participants (32 females, 6 males, 5 unknown) were excluded for further analysis because they did not complete all questionnaires or there was data missing. This means that 71% was female and 29% was male. 18% of the participants were doing their master, 52% their WO bachelor, 25% HBO and 4% MBO. The main age of the participants was 22.0 with a standard deviation of 2.4.

3.2 Behavioral data

The minimum obtained score on the PSQI was one and the maximum obtained score was thirteen with a mean score of 6.5 and a standard deviation of 2.5 (see Figure 2). The minimum obtained score

on the mismatch questionnaire was nineteen and the maximum obtained score was 46 with a mean score of 30.7 and a standard deviation of 6.1. The minimum obtained score on the USDI was 33 with a maximum obtained score of 111. The mean score on the USDI was 64.3 with a standard deviation of 17.7 (see Figure 2). One question was removed from the mismatch questionnaire since the question was only relevant for females, and this resulted in missing data for the males. The internal consistency of the USDI was .92 and .66 of the PSQI. The internal consistency of the entire mismatch questionnaire was .67 with different internal consistencies at every subscale (see Table 1).

The assumptions of normality and the homogeneity of variance were calculated before investigating if there were sex differences. The data of the PSQI for both males and females were normally distributed, W = .968, p = .428; W = .969, p = .075, and homogenous, F(1,103)= .318, p=.574. The data from the USDI for both males and females were normally distributed, W = .946, p= . 101; W=.972, p= .113, and homogenous, F(1,103)=1.477, p= .227, as well. The data from the mismatch questionnaire contained homogeneity, F(1, 103)= .892, p= .347, and the males were normally distributed, W= .982, p = .833. The females were not normally distributed, W = .956, p= . 013. There were no sex differences found in the USDI, t = .520, p = .605, PSQI, t = 1.085, p = .282, and mismatch questionnaire, W = 1287.5, p = .494.

Figure 2. The average score and standard deviation of the three questionnaires. On the x-axis, the three different questionnaires are shown and on the y-axis the main score. The main score of 105 participants was 6.5 on the PSQI with a standard deviation of 2.5. The main score of the 105 participants was 30.7 on the mismatch questionnaire with a standard deviation of 6.1 and 64.3 on the USDI with a standard deviation of 17.7.

Table 1

The internal consistency (Cronbach’s alpha) for the mismatch questionnaire with all subscales. The Cronbach’s alpha is calculated for the entire questionnaire and each of the fifteen subscales. Only the subscale fear of missing out had a high internal consistency with a Cronbach alpha > 0.70.

3.3 Statistical tests

The data of the USDI was not normally distributed by the Shapiro-Wilk test, W = .972, p = .026. From the histogram, it was found that the data were positively skewed. The data from the PSQI and mismatch questionnaire calculated with the Shapiro-Wilk test were normally distributed, W = .978, p=.082; W= .982, p = .185, and the data also seemed normally distributed by the histograms. The overall data calculated with a normal Q-Q plot was normally distributed as well. The data did not contain homoscedasticity, BP = 4.644, p = .098; Chi-square = 3.829, p = .050, and was linear (see Figure 5 and 6). Furthermore, there was multicollinearity found between the PSQI and mismatch scores, cor=.283, p = .004.

Cronbach’s alpha

Cronbach’s alpha Entire questionnaire .67 Performance-orientating

and perfectionism

.66

Sleep rhythm .07 Lack of time .70

Lack of exercise .48 Lack of freedom .51

Processed food .28 Lack of happiness .38

Lack of sun exposure .43 Fear of missing out .80

Lack of social contacts .22 Worrying .44

Materialism .46 Deviant youth .41

3.4 Correlations

Correlations were investigated between different questionnaires and subscales. There was a significant correlation found between the USDI and the PSQI, p < .001, rho = .38, and between the USDI and the mismatch questionnaire, p < .001, rho = .50 (see Table 2). Moreover, there were significant correlations found between the USDI and mismatch subscales lack of freedom, worrying, lack of time, processed food, sleep rhythm and performance- orientating and perfectionism, p< .001, rho = .46; p=.006, rho = .264; p= .005, rho = .272; p = .005, rho = .270; p= .016, rho = .235; p = .023, rho = .222 (see Table 2). There were no significant correlations found between the USDI and the lack of exercise, lack of sun exposure, lack of social contacts, materialism, unhealthy habits, lack of happiness, fear of missing out, deviant youth and remaining questions subscales of the mismatch questionnaire (see Table 2). Furthermore, there was a significant correlation between the PSQI and the sleep rhythm subscale of the mismatch questionnaire, p= .041, rho= .200 (see Table 2).

Figure 5. Scatterplot of the PSQI scores compared with the USDI scores. On the x-axis, the PSQI scores are displayed and on the y-axis the scores on the USDI. The dots are the data points of every participant and the line is the trendline.

Figure 4. Scatterplot of the mismatch scores compared with the USDI scores. On the x-axis, the scores on the mismatch questionnaire are displayed and on the y-axis the scores on the USDI. The dots are the data points of every participant and the line is a trendline.

Table 2

The correlations between the USDI and the mismatch questionnaires and all mismatch subscales, the USDI and PSQI and between the PSQI, and the sleep rhythm subscale of the mismatch questionnaire. In the table, all the correlations are displayed. The correlations were calculated with the non-parametric function since the USDI data and the sleep rhythm data were not normally distributed.

USDI () USDI ()

MM questionnaire .499 *** MM:

Performance-orientating and perfectionism

.222 *

MM: sleep rhythm .235* MM: Lack of time .272**

MM: Lack of exercise - .102 MM: Lack of freedom .463***

MM: Processed food .270** MM: Materialism .085

MM: Lack of sun exposure .121 MM: Fear of missing out

.123

MM: Lack of social contacts

.166 MM: Worrying .264**

MM: Lack of happiness .166 MM: Deviant youth .182

MM: Unhealthy habits .140 USDI () PSQI Rho = 0.375 *** MM: Sleep rhythm () PSQI Rho= 0.200* * .010 < p < .050 ** .001 < p < .010 *** p < .001

3.5 Main analysis

Since there were significant correlations found between the different questionnaires a multiple linear regression was executed to investigate if sleep quality and mismatches can predict the score on the USDI. The intercept (USDI score) was significant as well as the two slopes (mismatch and PSQI score), standard error= 7.959, t = 2.199, p =.0284; standard error= 1.153, t = .256, p <.001; standard error=1.710, t= .633, p = .007 (see Table 3). The mismatch questionnaire and the PSQI explain 27.17% of the variance of the USDI. The multiple linear regression for every subscale of the mismatch questionnaire on the USDI was executed to investigate the predictive value of all subscales on the USDI. A significant difference was found in the intercept (USDI score), p = .002, and in the processed food subscale, p=.036 (see Table 4). The subscales sleep rhythm, lack of exercise, lack of freedom, lack of sun expose, lack of social contacts, materialism, unhealthy habits, performance-orientating and perfectionism, lack of time, lack of happiness, fear of missing out, worrying and deviant youth were not significant (see Table 4). The different mismatches explained 30.25% of the variance of the USDI.

Table 3

Data from the multiple linear regression with the USDI as intercept and the mismatch questionnaire and PSQI as slopes. The standard error, t-value and p-value for the intercept and slopes were calculated with multiple linear regression.

Standard error t-value Pr(>|t|)

Intercept 7.96 2.199 .028 *

Slope Mismatch 1.15 .256 > .001 ***

Slope PSQI 1.71 .633 .007 **

* .010 < p < .050 ** .001 < p < .010 *** p < .001

Table 4

Data from the multiple linear regression with the USDI as intercept and all mismatch subscales as slopes. The standard error, t-value and p-value for the intercept and slopes were calculated with multiple linear regression

*.010 < p < .050 ** .001 < p < .010 *** p < .001

3.6 Factor analysis

The principal component analysis (PCA) determines the number of components in a questionnaire. There was sampling adequacy of .318 found on the KMO and Bartlett’s Test was significant, approximal Chi-Square = 3997.159, df= 2926, p>.001. From the scree plot, it was found that 27 components had an eigenvalue above 1 and 13 components an eigenvalue above 2. There were no high correlations found in the correlation matrix, which indicates that the items were independent. The varimax rotation with 13 components was used for further analysis. There was chosen for an eigenvalue above 2 since there was a clear level off and because the mismatch questionnaire contained 14 categories.

The 77 questions are ordered in 13 components using the rotated component matrix (see appendix E). Table 5 shows how many questions in each category fell within one component. It was found that components one, two, three, five, six and ten could be compared with a category from the mismatch questionnaire. Therefore, component one could be compared with the category lack of time, component two with fear of missing out and component three with deviant youth. Furthermore, component five should be compared with unhealthy habits, component six with

Standard error t-value Pr(>|t|)

Intercept 9.97 2.804 .006 **

Sleep rhythm 1.70 1.919 .058

Lack of exercise 1.26 -1.254 .213

Processed food 1.58 2.128 .036*

Lack of sun exposure 1.60 1.122 .265

Lack of social contacts 1.80 1.018 .311

Materialism 1.33 -.458 .648 Deviant youth 1.79 .926 .357 Performance-orientating and perfectionism 1.40 1.307 .195 Lack of time 1.26 .257 .797 Lack of freedom 1.90 1.808 .074 Lack of happiness 1.75 .926 .359

Fear of missing out 1.00 .339 .735

Worrying 1.33 1.642 .104

performance-orientating and perfectionism and component ten with materialism. Component one was equal with a lack of time because the component contained all questions from this category. Furthermore, there were three more questions included from different categories. Component two contained all questions from the category of fear of missing out together with three other questions. There was one question missing from the category deviant youth in component three and three questions from other categories were included. Furthermore, four questions of the category unhealthy habits were missing in component five and it contained three more questions from different categories. Component six missed one question from the category performance-orientating and perfectionism and there was one question from the category lack of happiness added. Component ten only existed of questions from the category materialism, but two of the five questions were missing.

Components four, seven, eight and twelve contained related questions but they could not be compared with a category from the mismatch questionnaire. The questions in these components were forming new categories, where component four assembles questions from the categories lack of freedom and lack of happiness. Component seven formed a category with the subscales lack of social contacts, worrying and remaining questions and component eight contained questions about healthy food intake and sleep behavior. Furthermore, component twelve was about lack of exercise and lack of sun exposure. At least, components nine, eleven and thirteen did not have an obvious link with each other. These components cannot be linked with a specific category.

Table 5

The thirteen different components which followed from the PCA with an eigenvalue above 2. The components contained questions from the different categories from the mismatch questionnaire. For

every component, it is shown which categories belong to the component and how many of the questions from the category were loaded at the component.

Comp-onent

1 5: Lack of time 1: Worrying 1: Lack of freedom 1: Lack of social contacts 2 5: Fear of missing out

2: Sleep rhythm 1: Remaining questions

3 3: Deviant youth 1: Lack of social contacts 1: Unhealthy habits 1: Lack of happiness 4 3: Lack of freedom 2: Lack of happiness 1: Materialism 5 4: Unhealthy habits 2: Processed food 1: Worrying 6 4: Performance-orientating and perfectionism 1: Lack of happiness 7 1: Lack of social contacts 1: Worrying 2: Remaining questions 8 2: Processed food

2: Sleep rhythm 1: Lack of social contacts 1: Lack of sun exposure 9 1: Lack of freedom 2: Unhealthy habits 1: Lack of sun exposure 1: Lack of social contacts 1: Performance-orientating and perfectionism 10 3: Materialism

11 1: Deviant youth 1: Lack of happiness 2: Unhealthy habits 1: Processed food 1: Lack of exercise 1: Sleep rhythm 12 3: Lack of sun exposure 4: Lack of exercise

13 1: Deviant youth 2: Worrying 1: Materialism

4. Conclusion I

From the data, it can be concluded that there is a strong correlation between the USDI and the PSQI and between the USDI and the mismatch questionnaire. There was multicollinearity found between

the PSQI and the mismatch questionnaire. Furthermore, significant correlations were found between the USDI and the mismatch subscales lack of freedom, sleep rhythm, processed food, performance-orientating and perfectionism, lack of time and worrying. The multiple linear regression presented that sleep quality and mismatches explained 27.17% of the variance in developing depressive symptoms in a student population.

The data showed that the mismatch questionnaire contained 27 components with an eigenvalue above 1 and 13 components with an eigenvalue above 2. The categories lack of time and fear of missing out had a strong coherency since all questions fell within one component. The categories deviant youth, performance-orientating and perfectionism and materialism constitute an average consistency because most of the questions fell within the same component. Furthermore, the subscales lack of freedom and lack of happiness coincide in a component and could form the new category lack of freedom and happiness. The subscales lack of exercise and lack of sun exposure coincide together as well and can constitute the category lack of outside sports. Questions from the categories lack of social contacts, worrying and remaining questions assembles the category favorable living situations and the subscales processed food and sleep rhythm composed the new category healthy lifestyle. Moreover, the subscales unhealthy habits and processed food constituted the new category of unhealthy food intake.

2.

Methodology II

2.1 Participants

Six participants were contacted to participate in an interview (4 female, 2 male) of whom four agreed to participate (2 female, 2 male). Two of the participants (1 male, 1 female) were also asked to wear a Xiaomi Mi Band 3 - Activity Tracker to measure sleep. As inclusion criteria, the participant had to be between sixteen and thirty years old, student and Dutch as the native language.

2.2 Materials

2.2.1 Interviews

The interviews were semi-structured with mostly open-ended questions to encourage the participants to give detailed and in-depth answers. The questions were developed using the book Qualitative Interviewing: The art of hearing data (Rubin & Rubin, 1995). During the interview, there were three main themes: sleep quality, depressive symptoms, and mismatches. Besides the main questions, there were also probes and follow-up questions to contain a good flow for an in-depth and detailed interview. The sleep and depressive symptoms questions were developed to investigate the

longitudinal schedule and correlation between sleep quality and depressive symptoms. Furthermore, sleep questions were developed to investigate the influence of telephone use before sleep. The mismatch questions were developed to create more insight into the different subscales.

2.2.2 Activity tracker

Participants were asked to wear a Xiaomi Mi Band 3 – Activity Tracker to measure sleep. The Xiaomi Mi Band 3 – Activity Tracker is a light weighted activity tracker that should be worn around the wrist. The tracker contained an accelerometer, an optical heart rate sensor and the display contained a light-emitting diode (Puri et al., 2017). There were different displays as time and date, step counts, exercises, heart rate, timer, weather, and extra options. The Xiaomi Mi Band 3 – Activity Tracker was linked with the application Mi Fit where gender, age, height, and weight were programmed. In the application the calories, step count, heart rate and sleep quality were visible. Sleep quality was measured by the optical heart rate sensor and the application measured the amount of REM sleep, non-REM sleep, and nighttime awakenings.

2.3 Procedure

Participants with a high and low score on the USDI, PSQI and mismatch questionnaire and who left their email behind were invited for further analysis. Participants who wanted to participate were interviewed. The interviews took place at places where the participants felt comfortable, like their own house. The interviews were conducted by one interviewer and each interview lasted for approximately thirty-five minutes (see Figure 8). All interviews were recorded with the permission of the participants and they all remained anonymous. During the interview, the timeline of the depressive symptoms and sleep quality were determined as well as the effects of smartphone use before bedtime. Furthermore, the interviews were used to create insight into the mismatch questionnaire. To achieve this, sleep quality, depressive symptoms, and mismatches were used as themes. All the interviews were individually adjusted using the scores obtained from the foregoing survey.

Furthermore, two participants were asked to wear the Xiaomi Mi Band 3 – Activity Tracker on their wrist for fourteen days to determine their sleep quality. To generate the data as accurate as possible participants were asked to wear the tracker all day and night (see Figure 8). The sleep quality and the answers in the interview together with the score on the USDI were compared with each other.

Figure 8. Schematic overview of the interviews and activity tracker. Four participants were interviewed for approximately 35 minutes. Afterward, two of the four participants were asked to wear the Xiaomi Mi Band 3 – activity tracker for 14 days to measure sleep quality. The activity tracker was linked with the application Mi Fit. The activity tracker measured the total sleep duration, duration of light and deep sleep, night awakenings, time falling asleep and time waking for every day and night during the 14 days.

2.4 Analysis

The interviews were transcribed on the same day or one day afterward. The interviews were transcribed using the standard verbatim, which means that the interview involved a detailed transcription with small editing. This method ensures that the transcript was highly accurate, but not overloaded with unnecessary details and summaries. Afterward, a summary of all interviews was made so the main themes could be compared. From the summary of the themes, the longitudinal effects between sleep quality and depressive symptoms were determined as well as the effects of artificial light before bedtime. Furthermore, the mismatch questions were used to determine which mismatch categories apply to people with a high and low score on the USDI. They were also used to investigate if any categories were missing. There was no statistical analysis conducted since only four participants were interviewed.

The sleep quality and total sleep duration were analyzed by the application of Mi Fit. The application was connected with the Xiaomi Mi Band 3 – Activity Tracker and measured the total sleep duration, the hours light, and deep sleep and the nighttime wake bouts. The data of fourteen days were collected and compared with each other since one person had no sleep problems according to the PSQI and had a low score on the USDI. The other person did experience sleep problems according to the PSQI and had a high score on the USDI. Besides comparing the data with each other the data was also compared with the results on the PSQI and with the given answers in the interview.

3.

Results II

3.1 Participants

Of the 6 participants, 4 participants agreed with an interview (2 male, 2 female) and two participants (1 male, 1 female) agreed to wear a Xiaomi Mi Band 3 - Activity Tracker to measure sleep quality. There was an equal gender distribution for both the interviews and the measurement of sleep quality. 50% of the participants were doing their WO bachelor, 25% HBO and 25% MBO. The main age of the participants was 23.3 with a standard deviation of 1.0.

3.2 Behavioral data

The participants were chosen on their scores on the USDI, PSQI and mismatch questionnaire. Table 5 shows their age, gender, study level and total scores on the USDI, PSQI and mismatch questionnaire. Besides the total score on the mismatch questionnaire also the sub-scores were added to compare the answers given in the interview with the answers on the questionnaires. The duration of the interviews was between 29:13 and 36:00 minutes with a mean duration of 35:44 minutes and a standard deviation of 0.27 minutes. The interviews were transcribed by standard verbatim (see appendix VI).

Table 5

Behavioral data of the four participants who were interviewed. In the table, every participant their gender, age, study level, and the total scores on the three questionnaires are shown. Besides the total scores, the scores of the mismatch subscales are shown as well. The total score of all subscales except unhealthy habits and remaining questions is 5. The total score for unhealthy habits is 9 and the total score of remaining questions is 3.

Participant 1 Participant 2 Participant 3 Participant 4

Gender Male Female Male Female

Age 24 23 22 24

Study level MBO WO bachelor WO bachelor HBO

Score USDI 103 85 36 71

Score PSQI 10 9 2 8

Score MM 31 36 27 46

Mismatch score per category

Participant 1 Participant 2 Participant 3 Participant 4

Deviant youth 3 0 0 2

processed food 1 1 0 3

Fear of missing out 0 3 1 4

Lack of exercise 3 2 2 5

Lack of social contacts 2 4 2 1

Lack of freedom 3 2 1 2

Lack of happiness 4 1 0 2

Lack of sun exposure 3 2 2 2

Materialism 2 1 2 3 Performance-orientating and perfectionism 3 5 4 5 Sleep rhythm 4 2 1 3 Lack of time 0 5 5 5 Worrying 0 2 1 4 Unhealthy habits 2 3 3 2

Besides the interviews, participants 2 and 3 also wore the Xiaomi Mi Band 3 - Activity Tracker for two weeks. Participant 2 had a mean total sleep duration of 7.4 hours with a standard deviation of 1.5 (see Table 6). The average time of falling asleep was 00:40 with a standard deviation of 0.2 and the average time of waking up was 09:33 with a standard deviation of 0.1 (see Table 6). The participant had a mean of 0.3 hours of deep sleep and 6.9 hours of light. The participant had 9 nighttime wake-bouts during the fourteen nights with a mean of 10.8 minutes awake (see Table 6). Three days of data had to be excluded for participant 3 because data of these days were missing. Participant 3 had a mean total sleep duration of 7.2 hours with a standard deviation of 1.2 (see Table

6). The average time of falling asleep was 00:18 and the average time of wake up was 07:44 (see Table 6). The mean duration of the deep sleep was 2.0 hours and the mean duration of the light sleep was 5.0 hours (see Table 6). Participant 3 was only once awake during the night.

Table 6

The mean and standard deviations of the sleep quality measured with the Xiaomi Mi Band 3 – Activity Tracker. The mean (M) and standard deviation (SD) are shown for REM sleep, non-REM sleep, night time wake bouts, total sleep duration, time falling asleep and time waking up.

Participant 2 Participant 3

REM sleep M= 0.3; SD= 0.3 (hours) M= 2.0; SD= 0.6 (hours) Non-REM sleep M= 6.9; SD= 1.5 (hours) M= 5.0; SD= 1.3 (hours) Total sleep duration M=7.4; SD= 1.5 (hours) M= 7.2; SD= 1.2 (hours) Night time wake bouts M= 10.8; SD=12.6 (minutes) M= 0.2; SD=7.8 (minutes) Time falling asleep M= 00:40; SD= 0.2 M= 00:18; SD= 0.1 Time waking up M= 09:33; SD= 0.1 M= 07:44; SD= 0.0

3.3 Main analysis

The main goals of the interviews were determining if smartphone and laptop use before bedtime resulted in poorer sleep quality and investigating the longitudinal effects of sleep quality and depressive symptoms. Furthermore, the mismatch questions were used to investigate if all mismatches have an effect on developing depressive symptoms and if the newly developed questionnaires missed categories. The summaries of the transcripts were used to determine the main goals.

3.3.1 Sleep

All participants used their phones before they went to bed and two of them used night modus. All the participants rarely read a book before bedtime and none of them did know if there was a difference in sleep quality when they did not use a smartphone or laptop before bedtime. Participant 3 indicated that reading a book helped with a better sleep quality since his head was calmer after reading.

Participants 1, 2 and 4 had a poor sleep quality and they all had problems with getting up in the morning. Sometimes they had trouble falling asleep. Normally they fell asleep within 15 to 30 minutes, which is longer compared with the 5 minutes of participant 3. The problems with falling asleep were correlated with depressive symptoms and stress. When the depressive symptoms and

stress levels were higher they had more trouble falling asleep. All participants tried to sleep 7 till 8 hours every night and participants 2 and 3 took naps if they failed to receive 7 till 8 hours of sleep. Furthermore, participants 1 and 2 suffered from depressive symptoms, but participant 4 did not. There was a different underlying cause for the trouble falling asleep. For participant 4 the underlying cause was stress-related and at participants 1 and 2 the underlying problem was related to their depression. There was also a difference found in sleep rhythm between participant 3 and the other participants. Participant 3 had no trouble getting back in his old sleep rhythm and the other participants did. Furthermore, participant 3 did never wake up in the night were participants 1 and 4 did most of the nights.

3.3.2 Depressive symptoms

Participant 1 did not know whether there is a relationship between sleep quality and depressive symptoms. The participant struggled with depression for his whole life and therefore did not know differently than feeling the way he is feeling nowadays. His sleep complaints have been flurrying all his life. Participant 2 experienced poorer sleep quality during the depressive symptoms. She thinks that poor sleep quality is a result of depressive symptoms since she is overstimulated during the episodes.

An environment where you could be yourself with supporting parents and friends could say a lot about developing depressive symptoms. Participant 4 had close contact with their parents, she had supportive family members and friends and a living situation where she could be herself. Participant 3 also had supporting friends and he had a good living situation as well. They both asked help when things became too much. On the contrary, participants 1 and 2 did not always ask for help when they needed it. Moreover, the parents of participant 2 did blame her for her eating disorder and depressive symptoms. This led to a negative and non-supporting environment, which created the feeling that she was standing on her own.

3.3.3 Mismatches

Participants 1 and 2 both said that they worry a lot, have unhealthy habits and experiences a lack of exercise. Furthermore, participants 1 had problems with his sleep rhythm and participant 2 with lack of socials contacts, performance-orientating and perfectionism, lack of time, fear of missing out and she had a deviant youth. Participant 3 mentioned that the categories processed food, materialism, performance-orientating and perfectionism, lack of time, unhealthy habits by alcohol and drug use and fear of missing out affected him. Participants 4 indicated to have mismatches in the categories lack of exercise, lack of sun exposure, processed food, performance-orientating and perfectionism, lack of time, worrying, unhealthy habits and lack of freedom. The answers in the interview and the answers on the mismatch questionnaire for participant 1 and 2 were compared with participant 3 and

it was found that participants 1 and 2 scored higher at lack of freedom, lack of happiness, deviant youth, lack of social contacts, lack of happiness and sleep rhythm. Comparing participants 1 and 2 with participant 4 it was found that participants 1 and 2 scored higher at lack of socials contacts and that only participant 1 scored higher at lack of happiness.

Participant 2 said that besides mismatches there are also protective factors as having dreams, having goals, having a supportive family and having a protective living situation. Having a supportive family and having a protective living situation were already included in the mismatch questionnaire, but having dreams and goals were not. Furthermore, participant 3 thought that having performance-orientating and perfectionism also could have a positive effect if you can handle the pressure which comes with a performance orientation. Participants 3 and 4 said that pressure from society may also be a mismatch factor.

3.4 Activity tracker

There was no significant effect found between the total sleep durations of participants 2 and 3, W=112, p= .535. The non-parametric test was executed, because the data of participant 3 was not normally distributed, W = .857, p= .028. The data of participant two was normally distributed, W=.959, p= .703, and the data was homogenous, F(1, 26)= .659, p= .424. There was a significant effect found between the amount of deep sleep of the two participants, W = 0, p < .001. For this test the non-parametric t-test was used since the data of participant 2 was not normally distributed, W = . 798, p=.005, and the data did not contain homogeneity, F(1, 26)= 7.043, p= 0.0134. The data of participant 3 was normally distributed, W = .936, p= .367. Furthermore, there was a significant difference found between the amount of light sleep, t = 3.676, p= .001. The parametric test was executed because the data of both participants was normally distributed, W = .929, p= .300; W = . 936, p= .367, and the data contain homogeneity, F(1, 26)= .309, p= .583.

4.

Conclusion II

Depressive symptoms and sleep quality have a relationship with each other, but the longitudinal timeline is still unclear. Participant 1 did not know what the relationship was between depressive symptoms and sleep quality, but participant 2 thought that sleep problems followed out a depression. Furthermore, there is no evidence found that smartphone use before bedtime harmed sleep quality. Interestingly, participants 3 who only used his phone for 15 minutes before bedtime was also the person who had the least problems with falling asleep. Comparing the mismatches, it was found that the participants without depressive symptoms scored lower at a lack of happiness and lack of social contacts.

There was no significant difference found between the total sleep duration of participants 2 and 3, but there was a significant difference found in the amount of deep and light sleep. Furthermore, participant 2 had a lot more nighttime wake-bouts than participant 3 and she fell asleep twenty-two minutes later. Participant 3 gets up 109 minutes earlier than participant 2.

5.

Discussion

From the questionnaires, depressive symptoms in students can be predicted by both sleep quality and mismatches (explained variance of 27.27%). The mismatch subscales performance-orientating and perfectionism, lack of time, lack of freedom, worrying, processed food and sleep rhythm had a positive correlation with depressive symptoms and processed food predicted depressive symptoms the best. According to the interviews, the categories lack of socials contacts and lack of happiness were related to depressive symptoms. Furthermore, the PCA with an eigenvalue above two contained thirteen components and it was found that not all subscales had a strong coherency. The longitudinal timeline of depression and sleep is still not clear, although there are sleep differences found between participants with a high and a low score on the USDI. Moreover, the results do not review if smartphone use before bedtime causes poor sleep quality.

The fact that depressive symptoms can be predicted by sleep quality and mismatches is in line with the hypothesis. Baglioni et al. (2011) already found that there was a co-occurrence between sleep and depression and that individuals with insomnia had a twofold risk to develop depression than individuals without sleep disruptions. Furthermore, it was already thought that the co-occurrence could be explained by the mismatch hypothesis. Moreover, it was expected that all mismatch subscales correlated positively with depressive symptoms, but unfortunately only six subscales did. This could be explained by the newly developed mismatch questionnaire. The questions were developed to measure all kinds of mismatches and not only mismatches who were relevant for developing depressive symptoms. Furthermore, Cronbach’s alpha shows that only the subscales’ lack of time and fear of missing out were homogeneous and that only these subscales measured the same construct. From the PCA it was found that thirteen components had an eigenvalue above two instead of fourteen and that not all subscales had a strong coherency.

Not all components from the PCA did match with the categories from the mismatch questionnaire. The categories lack of time and fear of missing out matched completely with the components and the categories deviant youth, performance-orientating and perfectionism, and materialism matched almost completely with the components. Furthermore, the categories worrying and lack of social contacts were combined as well as lack of exercise and lack of sun exposure. This

could be explained by the fact that the category lack of exercise was not only about sports but about all kinds of exercise and that most people got their exercise outside. The combination of worrying and lack of social contacts could be clarified by the fact that most participants who worry a lot probably worry about their situation with friends and family as well. The category of processed food must be separated into healthy and unhealthy food choices. Moreover, the question of smartphone use before bedtime should be removed from the category sleep rhythm. Lastly, there must be carefully looked at the category lack of happiness because almost none of the questions were found in the same component. This could be explained by the fact that the questions measure a broad, not coherent spectrum.

Overall, not all mismatch subscales affected depression. Therefore, in a further investigation, the mismatch questionnaire could be further developed. The new subscales should be based on the components from the PCA, correlations, linear regression, Cronbach's alphas, and the interviews. From the correlations and linear regression, it would be interesting to preserve the categories: sleep rhythm, processed food, performance-orientating and perfectionism, lack of time, worrying and lack of freedom. From the interviews, the categories lack of happiness and lack of social contacts should be kept as well. It is also found that youth and social pressure could affect developing depressive symptoms. Taken this together with the PCA and Cronbach alpha’s the new mismatch questionnaire should contain the categories lack of time, performance-orientating and perfectionism and worrying. The categories sleep rhythm and a part of processed food will become a healthy lifestyle and the other part of processed food together with some questions from unhealthy habits will become unhealthy consumptions. Furthermore, the categories lack of social contacts and deviant youth should become lack of kinship and friends and the categories lack of freedom and lack of happiness should be combined in lack of happiness and freedom. An example of a newly developed mismatch questionnaire is shown in appendix G.

The study was not able to identify the longitudinal effects of sleep quality on depressive symptoms. Until now, the research field had failed to determine the exact effects of sleep on depression (Baglioni et al., 2011; Riemann & Voderholzer, 2003; Tsuno et al., 2005) and this investigation neither clarified the effects. In this study, only two participants with depressive symptoms were interviewed. Furthermore, the questionnaires were unable to measure the timeline since not enough of these questions were included. To investigate the longitudinal relationship in more detail, more participants should be interviewed and the questionnaires must be extended. Another way to determine the longitudinal effects of sleep quality on depressive symptoms is to follow participants who are diagnosed with insomnia and do not have depressive symptoms. If insomnia is an independent risk factor for developing depression, participants with insomnia will develop depressive symptoms as well (Baglioni et al., 2011). However, it is important to mention that

there are different types of insomnia (Walker, 2017) and up until now it is unclear whether all types or just a part of the types are related to depression.

There was a significant difference found between light and deep sleep durations in participants with and without depressive symptoms. The participant with a good sleep quality according to the PSQI had significantly more deep sleep compared with the participant who had poor sleep quality according to the PSQI. There was no significant difference found between the total sleep duration. An explanation for the differences in deep and light sleep could be that disruptions of rapid eye movement (REM) sleep are specific for people with depression (Wichniak, Wierzbicka, & Jernajczyk, 2013). The human sleep cycle is subdivided into REM sleep and in non–rapid eye movement (NREM) sleep. The NREM sleep contains four stages, where stages 3 and 4 are related to deep sleep and stage 1 and 2 together with REM sleep are related to light sleep (Šušmáková, 2004; Walker, 2017). The four NREM stages and the REM stage are followed by a short awakening. After the awakening, a sleep cycle is completed and the next sleep cycle will follow (Wichniak et al., 2013; Walker, 2017). Why depressed people have altered sleep cycles and whether other factors are involved is still unknown. Therefore, not only the melatonin levels should be investigated in further research, but the amount of REM and NREM sleep must be included as well. To investigate this, firstly, many more people should be included since now only two participants were investigated. Secondly, participants should be measured for multiple days. Lastly, the participants should not only wear an activity tracker. An activity tracker is not capable of measuring the different sleep stages.

There is no conclusive answer about smartphone use before bedtime because all participants who were interviewed used their smartphones before bedtime. It was expected that participants who used their phones less would sleep better, but it turned that all participants used their phones before bedtime even the participants with good sleep quality. The reason why smartphone use before bedtime would result in poor sleep quality is because of a lack of melatonin. The secretion of melatonin is correlated with increased sleep propensity and it is thought that melatonin directly affects human sleep (de la Iglesia et al., 2017). The use of artificial light and so smartphone use will reduce the secretion of melatonin and thereby could result in problems falling and staying asleep (Cardinali et al., 2001). In further research, the effects of smartphone use on sleep quality could be investigated by following participants for two weeks, wherein the first week participants should not use their smartphone before they go to bed and in the second week, they should use their smartphone before bedtime. In both weeks the melatonin levels should be measured. Urine is a reliable measurement to measure melatonin levels and it is convenient since participants can do this at home (Benloucif et al., 2008). The measurement of melatonin is an important aspect since participants who used their phones before bedtime had good sleep quality. Besides measuring the melatonin levels, participants should also keep a logbook to determine their sleep behavior.

Investigating the relationship between sleep quality and smartphone use is important because research already found that smartphone use and sleep quality could predict depression and anxiety (Demirci, Akgönül, & Akpinar, 2015).

Until now, depressive symptoms were only investigated in university students. In the Netherlands, there are different study levels. From the survey, there are significant differences found in the USDI scores between the different study levels (MBO, HBO, WO bachelor, and master). Besides the aforementioned further investigations, it would be interesting as well to investigate if all study levels have an equal change in developing depressive symptoms.

Unfortunately, 43 participants had to be excluded for further analysis which resulted in only 105 participants who executed the survey. This is above the recommended participants according to the prior analysis, but it is below the recommended sample size for a PCA (Pallant, 2013). According to Pallant (2013), at least 150 participants should execute the survey before a PCA could be conducted. Another limitation of the present study is the use of questionnaires. Questionnaires are inexpensive, quick, comparable and easy to analyze but there is also a considerable change of dishonest answers because of social desirability (Kazi & Khalid, 2012). Furthermore, questionnaires are sensitive for differences in interpretations and understandings, they do not capture emotional feelings and they are impersonal (Kazi & Khalid, 2012). Besides questionnaires participants were interviewed as well. Interviews encourage honest answers, counteract differences in interpretations and are personal (Rubin & Rubin, 1995). However, only four participants were interviewed of whom all were acquaintances of the interviewer.

The study had strong and weak points. The strengths of the study were that not only questionnaires were used, but that participants were interviewed as well. Furthermore, there was a questionnaire specially developed to measure mismatches as accurately as possible and sleep quality was not only questioned but also measured. A weakness of the study was the number of participants in both parts. There were not enough participants to conduct a PCA and there were only four participants interviewed. Moreover, sleep quality was only measured in two participants. Another weakness was the reliability of the interviews since the participants were acquaintances of the interviewer.

This study was not able to determine the longitudinal timeline of sleep quality and depression and the effects of smartphone use before bedtime are still not clear. On the contrary, this study was capable of determining the effects of sleep quality and mismatches on depressive symptoms in students. As aforementioned the sample size for both parts should be improved just like the mismatch questionnaire and measurements to measure sleep quality. A longitudinal study could be done with more participants and improved questionnaires. Furthermore, more participants with sleep disruptions and depression should be included to determine the longitudinal effects of sleep

quality and depression. Moreover, the effects of smartphone use should not only be questioned in interviews but also measured with test sessions and more participants.

If the predictive values of sleep quality and all different mismatch categories on depressive symptoms were known, depressive symptoms in students could be reduced. With just some adjustments in daily life and creating good sleep quality, students may reduce depressive symptoms which prevents them from developing depression. Knowing more about the longitudinal effects between depression and sleep quality and knowing which mismatches could predict depressive symptoms could help improve the health of students.

References

Arendt, J., Aldhous, M., & Marks, V. (1986). Alleviation of jet lag by melatonin : preliminary results of controlled double blind trial. British Medical Journal, 292(6529), 1170.

Association American Psychiatric. (2013). Diagnostic and statistical manual of mental disorders. 5th ed. Washington: American Psychiatric Association.

Backhaus, J., Junghanns, K., Broocks, A., Riemann, D., & Hohagen, F. (2002). Test-retest reliability and validity of the Pittsburgh Sleep Quality Index in primary insomnia. Journal of Psychosomatic Research, 53(3), 737–740. https://doi.org/10.1016/S0022-3999(02)00330-6

Baglioni, C., Battagliese, G., Feige, B., Spiegelhalder, K., Nissen, C., Voderholzer, U., … Riemann, D. (2011). Insomnia as a predictor of depression: A meta-analytic evaluation of longitudinal epidemiological studies. Journal of Affective Disorders, 135(1–3), 10–19.

https://doi.org/10.1016/j.jad.2011.01.011

Benloucif, S., Burgess, H. J., Klerman, E. B., Lewy, A. J., Middleton, B., Murphy, P. J., … Revell, V. L. (2008). Measuring melatonin in humans. Journal of Clinical Sleep Medicine, 4(1), 66–69. Buysse, D. J., Reynolds, C. F., Monk, T. H., Berman, S. R., & Kupfer, D. J. (1989). The Pittsburg Sleep

Quality Index: A new instrument for psychiatric practice and research. Psychiatry Research, 28, 193–213.

Cajochen, C., Frey, S., Anders, D., Späti, J., Bues, M., Pross, A., … Stefani, O. (2011). Evening exposure to a light-emitting diodes (LED)-backlit computer screen affects circadian physiology and cognitive performance. Journal of Applied Physiology, 110(5), 1432–1438.

https://doi.org/10.1152/japplphysiol.00165.2011

Cardinali, D. P., Brusco, L. I., Lloret, S. P., & Furio, A. M. (2001). Melatonin in sleep disorders and jet-lag. Neuroendocrinology Letters, 23(1), 9–13.

Choy, M., & Salbu, R. L. (2011). Jet lag:Current and potential therapies. Pharmacy and Therapeutics, 36(4), 221–231.

Costello, R. B., Lentino, C. V, Boyd, C. C., Connell, M. L. O., Crawford, C. C., Sprengel, M. L., & Deuster, P. A. (2014). The effectiveness of melatonin for promoting healthy sleep : a rapid evidence assessment of the literature. Nutrition Journal, 13(1), 106.

de la Iglesia, H. O., Fernández-Duque, E., Golombek, D. A., Lanza, N., Duffy, J. F., Czeisler, C. A., & Valeggia, C. R. (2017). Access to electric light is associated with shorter sleep duration in a traditionally hunter-gatherer community. Journal of biological rhythms, 30(4), 342–350. https://doi.org/10.1111/mec.13536.Application