Sleep Problems and Psychological Symptoms in Adolescents
Sleep Problems and Psychological Symptoms in Adolescents
Objective To examine the contribution of adolescents' sleep problems and tiredness to psychological symptoms after accounting for shared risk and psychological co-morbidity.
Methods Secondary analyses of cross-sectional data on 12–16-year-old (N = 980) adolescents without chronic illness, functional limitation, or developmental delay. Adolescents rated sleep problems, tiredness, and psychological symptoms. Parents provided information about risk factors, adolescent tiredness, and psychological symptoms.
Results Prior to accounting for psychological co-morbidity, most sleep variables were significant correlates of adolescent-, but not parent-rated, psychological symptoms. After accounting for psychological co-morbidity: nightmares were associated with adolescent-rated anxiety/depression; sleeping more than others was associated with adolescent-rated aggression; trouble sleeping was associated with adolescent-rated attention problems, anxiety/depression, and withdrawal; and adolescent-rated tiredness was associated with adolescent-rated aggression and withdrawal.
Conclusions Studies examining sleep and psychopathology should control for psychological co-morbidity.
Sleep problems and indicators of inadequate sleep (e.g., tiredness) among adolescents have been associated with poorer psychological outcomes, including anxiety, depression, and poorer perceived health (e.g., Moore et al., 2009), cognitive difficulties (e.g., attention problems; Fernandez-Mendoza et al., 2009), and behaviors (e.g., aggression; Johnson & Breslau, 2001) indicative of psychopathology. Gregory and O'Connor (2002), for example, found that higher sleep problem scores, as measured by the sleep problem items on the Child Behavior Checklist (CBCL; Achenbach, 1991a) were associated with higher scores on a range of CBCL symptom scales, including anxiety/depression, attention problems, aggression, and withdrawal. Similarly, Johnson, Chilcoat, and Breslau (2000) found that the single CBCL item "trouble sleeping" predicted higher levels of psychological symptoms, again across a range of CBCL symptom scales. The causality and mechanisms underlying the association between sleep and psychological symptoms, however, are not well understood.
Large-scale, population-based studies can enhance our understanding of sleep and psychopathology by examining associations among multiple sleep variables and psychological symptoms as they occur in the population, while controlling for potential confounds in these associations. In the study at hand, we examine associations among sleep (sleep problems, tiredness) and psychological symptoms (aggression, attention problems, withdrawal, and anxiety/depression) in healthy Ontario adolescents, while controlling for two potential confounds: risk factors for poor sleep and psychopathology ("shared risk factors") and psychological co-morbidity. The possibility that shared risk factors and psychological co-morbidity confound observed associations between sleep and psychological symptoms is described below.
"Shared risk factors" (e.g., stressful life events, poor family functioning, marital disharmony, maternal negative affect) are established risk factors for psychopathology that are also believed to interfere with sleep, possibly by heightening arousal, impairing emotional regulation, or diminishing one's sense of safety (Reid, Huntley, & Lewin, 2009). Failure to account for these variables may result in spurious associations between sleep problems and psychological symptoms in adolescents.
Psychological co-morbidity presents a similar "third variable" problem (see Ivaneko & Johnson, 2008). That is, sleep problems associated with one type of psychopathology (e.g., anxiety/depression) may appear to be associated with another type of psychopathology (e.g., attention problems) when: (1) the two types of psychopathology co-occur (i.e., they are co-morbid) and (2) the co-morbid psychopathology is not accounted for in analyses. For example, it may be that adolescents with symptoms of anxiety/depression endorse having trouble sleeping because they lie awake worrying or ruminating; a significant proportion of these adolescents may also have attention problems. If associations between sleep problems and attention problems were examined in these adolescents, without controlling for symptoms of anxiety/depression, it would appear that trouble sleeping and attention problems were directly related when they were only indirectly related through their shared association with anxiety/depression.
Few studies of sleep and psychopathology have controlled for psychological co-morbidity; those that have, have focused primarily on only one or two types of psychopathology as an outcome. Corkum, Moldofsky, Hogg-Johnson, Humphries, and Tannock (1999) and Mick, Biederman, Jetton, and Faraone (2000) have found that the presence of co-morbid anxiety and oppositional disorders, respectively, may account for observed associations between sleep and attention/hyperactivity. Hyperactivity, conversely, does not appear to account for associations between period limb movements during sleep and conduct problems (Chervin, Dillon, Archbold, & Ruzicka, 2003), although there is limited research on this association. In our own work with children (aged 4–11 years; Coulombe, Reid, Boyle, & Racine, 2009), we found that controlling for psychological co-morbidity, and considering sleep problems as individual predictors of psychopathology (vs. total sleep problem scores; c.f. Gregory & O'Connor, 2002), allowed specific associations among sleep problems (e.g., trouble sleeping, nightmares, sleeping less than other children, and sleeping more than other children), tiredness, and psychological symptoms (aggression, attention problems, and anxiety/depression) to emerge. For example, after controlling for aggression and attention problems, trouble sleeping, nightmares, and tiredness were significantly associated with parent-rated symptoms of anxiety/depression (Coulombe et al., 2009). It is unclear whether similar associations will be found in adolescents (aged 12–16 years) using adolescent-report, as well as parent-report, data.
Hypotheses Based on the results of the existing literature and of our own study conducted with children (Coulombe et al., 2009), it was predicted that, after accounting for shared risk factors and psychological co-morbidity, (1) sleep problems and tiredness would be significantly associated with symptoms of anxiety/depression, withdrawal, and aggression and (2) that tiredness alone would be significantly associated with symptoms of attention problems. As results of studies using cross-informant outcomes have been inconsistent (c.f., Blunden & Chervin, 2008; Johnson et al., 2000), these hypotheses were examined separately for both adolescent-rated and parent-rated symptoms of psychopathology. At least one study suggests that adolescents may be better reporters of their internal states than their parents (Angold, Weissman, John, & Merikangas, 1987).
Secondary data analyses of the Ontario Child Health Study (OCHS; Boyle et al., 1987) were conducted. The primary purpose of the OCHS was to estimate the prevalence of emotional and behavioral disorders in a representative sample of 4–16-year-olds living from the province of Ontario, Canada; (see Boyle et al., 1987 for a description of the methodology for the OCHS). Demographic information, child physical health, parent-rated questionnaires for risk factors (e.g., family functioning), and parent-rated psychopathology checklists were collected from the female head of the household; adolescent self-report data were also collected. Parent nonresponse (missing checklists and questionnaires) was low (less than 1.5% of the total sample, aged 4–16 years); adolescent non-response was also low (2.6 %) (see Boyle et al., 1987).
The current study was approved by the Research Ethics Board at McMaster University.
Participants Adolescents with chronic medical conditions, significant physical limitations, mental retardation, developmental delay, and for whom this information was missing (n = 322) were excluded from analyses because sleep problems and physical health problems tend to co-exist (Mindell & Owens, 2003) and because sleep problems of these adolescents likely differ from those who do not have these health conditions (Stores, 1999). The final sample size for this study was 980 adolescents (49% male; age M = 13.88 years, SD = 1.41 years) from 771 households.
Measures and Key Variables Shared Risk Factors Risk factors with consistent and robust associations with psychopathology that may also affect sleep (e.g., single parent status, stressful life events, maternal depression or negative affect, poverty, poor family functioning) were identified from the pediatric sleep (e.g., Warren, Howe, Simmens, & Dahl, 2006) and developmental psychopathology literatures (e.g., Mash & Dozois, 2003; described in greater detail in Coulombe et al., 2009). The following variables were included: adolescent sex, adolescent age, family income (in increasing increments of $5000), maternal education (in years), maternal negative affect (using the Bradburn Affective Balance Scale; Bradburn, 1969), stressful life events (count of stressful events over the previous year that families experienced; see Boyle et al., 1987), marital disharmony (a composite of 4 items measuring overall relationship, frequency of quarrelling, enjoyable activities, and caring), and family functioning (measured using the General Functioning Scale of the McMaster Family Assessment Device; Byles, Byrne, Boyle, & Offord, 1988). All variables other than family income and adolescent sex were continuous, with higher scores indicating increasing age, more maternal education, greater maternal negative affect, more stressful life events, greater marital disharmony, and poorer family functioning.
Sleep Problems, Tiredness, and Psychological Symptoms. Sleep problems (adolescent-rated only), tiredness, and psychological symptoms scores (aggression, attention problems, anxiety/depression, and withdrawal) were derived from Child Behavior Checklist (CBCL; Achenbach, 1991a) and Youth Self Report Form (YSR; Achenbach, 1991b). Parents and adolescents rated how each item described the target child "now or within the past 6 months" using a 3-point scale (0 = "never or not true", 1 = "sometimes or somewhat true", 2 = "often or very true").
Psychological Symptoms and Psychological Co-morbidities. Parent- and adolescent-rated aggression, attention problems, anxiety/depression, and withdrawal scores were calculated for each adolescent. No sleep problem or tiredness item was included in any of the psychological symptom scales. So as not to artificially inflate associations among scales, items that appeared in more than one scale (e.g., "nervous or tense" appears as both an anxiety/depression and attention item) were not included in any scale score (Hinden, Compas, Howell, & Achenbach, 1997). Two items were removed from the adolescent-rated scales and three items were removed from the parent-rated scales; no more than two items were removed from any one scale. Prior to imputing missing data, internal consistencies for the parent-rated scales in our sample were: aggression (α = .87), attention problems (α = .77), anxiety/depression (α = .83), and withdrawal (α = .73). Internal consistencies for the adolescent-rated scales in our sample were: aggression (α = .81), attention problems (α = .71), anxiety/depression (α = .84), and withdrawal (α = .57). It should be noted that, through the remainder of this paper, a scale is referred to as a type of "psychological symptom" when the scale is used as an outcome variable; the term "psychological co-morbidity" is used when a scale is being used as a control variable.
Sleep Problem and Overtired Scores. Four sleep problem items were rated by adolescents: "I have trouble sleeping" (trouble sleeping), "I have nightmares" (nightmares), "I sleep less than most kids" (sleeping less), and "I sleep more than other kids during the day and/or night" (sleeping more). One tiredness item was rated by adolescents ("I feel overtired") and by parents ("overtired without good reason"). Each of these items (rated as 0 = "never or not true", 1 = "sometimes or somewhat true", 2 = "often or very true") was treated as a separate predictor of psychological symptoms (see Coulombe et al., 2009).
Data Analyses Analyses were conducted using Statistical Package for the Social Sciences (SPSS) version 16.0 and, for the regressions, STATA (Statistics/Data Analysis, 2006) version 10.1.
Missing Data. Less than 5% of the values were missing for any parent-rated variables (i.e., risk factors, parent-rated psychological symptoms, parent-rated "overtired"), with the exception of marital disharmony, which was systematically missing for all single parents (10.4% of households, 10.6% of adolescents). Less than 5% of the values were missing for any adolescent-rated variables (i.e., adolescent-rated sleep and tiredness items, adolescent-rated psychological symptoms). Few differences among data imputation methods have been found when less than 10% of data is missing, either randomly or systematically (Roth, 1994). In our sample, no differences in the patterns of results were found when adolescents with missing data were excluded from analyses or when missing values were replaced with the mean for that variable. Thus, with the exception of single parents missing marital disharmony scores (who were given scores indicative of "no marital disharmony"), all missing data was imputed using mean substitution (George & Mallery, 2002); interested readers are referred to the companion article for more detailed information on how missing data was handled (Coulombe et al., 2009).
Data Transformation. Psychological symptom/psychological co-morbidity scores, particularly for parent-rated data, were non-normally distributed. Square root transformations produced skewness and kurtosis values closest to 0 for parent-rated data (Tabachnik & Fidell, 2007). Transformation did not improve the skewness and kurtosis of adolescent-rated data.
Multivariate Analyses. Linear regressions examined the contribution of risk factors, psychological co-morbidity, sleep problems, and tiredness to the prediction of symptoms of adolescent-rated and parent-rated aggression, anxiety/depression, attention problems, and withdrawal. Some households had more than one adolescent participating in the study. As data from two siblings are not independent, all regressions controlled for the clustering of siblings within households. All variables were standardized prior to entry, allowing the regression coefficients to be interpreted as beta weights (Bring, 1994).
Predictors were entered in three blocks (Models 1–3). Shared risk factors and demographic variables were entered into the first block (Model 1). Sleep problem and tiredness scores were entered into the second block (Model 2), permitting a comparison of our findings with those of previous studies that have not controlled for co-morbidity. Parent-rated tiredness was included as a cross-informant predictor of psychological symptoms. Psychological co-morbidity was entered into the third block (Model 3), testing the hypothesis that sleep problems and tiredness would be significantly associated with psychological symptoms when shared risk factors and co-morbidity were considered. To better understand the effects of controlling for psychological co-morbidity, changes to the regression coefficients from Model 2 to Model 3 for each sleep variable were examined. Regressions were completed separately for adolescent-rated and parent-rated psychological symptoms.
Abstract and Introduction
Abstract
Objective To examine the contribution of adolescents' sleep problems and tiredness to psychological symptoms after accounting for shared risk and psychological co-morbidity.
Methods Secondary analyses of cross-sectional data on 12–16-year-old (N = 980) adolescents without chronic illness, functional limitation, or developmental delay. Adolescents rated sleep problems, tiredness, and psychological symptoms. Parents provided information about risk factors, adolescent tiredness, and psychological symptoms.
Results Prior to accounting for psychological co-morbidity, most sleep variables were significant correlates of adolescent-, but not parent-rated, psychological symptoms. After accounting for psychological co-morbidity: nightmares were associated with adolescent-rated anxiety/depression; sleeping more than others was associated with adolescent-rated aggression; trouble sleeping was associated with adolescent-rated attention problems, anxiety/depression, and withdrawal; and adolescent-rated tiredness was associated with adolescent-rated aggression and withdrawal.
Conclusions Studies examining sleep and psychopathology should control for psychological co-morbidity.
Introduction
Sleep problems and indicators of inadequate sleep (e.g., tiredness) among adolescents have been associated with poorer psychological outcomes, including anxiety, depression, and poorer perceived health (e.g., Moore et al., 2009), cognitive difficulties (e.g., attention problems; Fernandez-Mendoza et al., 2009), and behaviors (e.g., aggression; Johnson & Breslau, 2001) indicative of psychopathology. Gregory and O'Connor (2002), for example, found that higher sleep problem scores, as measured by the sleep problem items on the Child Behavior Checklist (CBCL; Achenbach, 1991a) were associated with higher scores on a range of CBCL symptom scales, including anxiety/depression, attention problems, aggression, and withdrawal. Similarly, Johnson, Chilcoat, and Breslau (2000) found that the single CBCL item "trouble sleeping" predicted higher levels of psychological symptoms, again across a range of CBCL symptom scales. The causality and mechanisms underlying the association between sleep and psychological symptoms, however, are not well understood.
Large-scale, population-based studies can enhance our understanding of sleep and psychopathology by examining associations among multiple sleep variables and psychological symptoms as they occur in the population, while controlling for potential confounds in these associations. In the study at hand, we examine associations among sleep (sleep problems, tiredness) and psychological symptoms (aggression, attention problems, withdrawal, and anxiety/depression) in healthy Ontario adolescents, while controlling for two potential confounds: risk factors for poor sleep and psychopathology ("shared risk factors") and psychological co-morbidity. The possibility that shared risk factors and psychological co-morbidity confound observed associations between sleep and psychological symptoms is described below.
Shared Risk Factors for Psychopathology and Sleep Problems
"Shared risk factors" (e.g., stressful life events, poor family functioning, marital disharmony, maternal negative affect) are established risk factors for psychopathology that are also believed to interfere with sleep, possibly by heightening arousal, impairing emotional regulation, or diminishing one's sense of safety (Reid, Huntley, & Lewin, 2009). Failure to account for these variables may result in spurious associations between sleep problems and psychological symptoms in adolescents.
Psychological co-morbidity presents a similar "third variable" problem (see Ivaneko & Johnson, 2008). That is, sleep problems associated with one type of psychopathology (e.g., anxiety/depression) may appear to be associated with another type of psychopathology (e.g., attention problems) when: (1) the two types of psychopathology co-occur (i.e., they are co-morbid) and (2) the co-morbid psychopathology is not accounted for in analyses. For example, it may be that adolescents with symptoms of anxiety/depression endorse having trouble sleeping because they lie awake worrying or ruminating; a significant proportion of these adolescents may also have attention problems. If associations between sleep problems and attention problems were examined in these adolescents, without controlling for symptoms of anxiety/depression, it would appear that trouble sleeping and attention problems were directly related when they were only indirectly related through their shared association with anxiety/depression.
Few studies of sleep and psychopathology have controlled for psychological co-morbidity; those that have, have focused primarily on only one or two types of psychopathology as an outcome. Corkum, Moldofsky, Hogg-Johnson, Humphries, and Tannock (1999) and Mick, Biederman, Jetton, and Faraone (2000) have found that the presence of co-morbid anxiety and oppositional disorders, respectively, may account for observed associations between sleep and attention/hyperactivity. Hyperactivity, conversely, does not appear to account for associations between period limb movements during sleep and conduct problems (Chervin, Dillon, Archbold, & Ruzicka, 2003), although there is limited research on this association. In our own work with children (aged 4–11 years; Coulombe, Reid, Boyle, & Racine, 2009), we found that controlling for psychological co-morbidity, and considering sleep problems as individual predictors of psychopathology (vs. total sleep problem scores; c.f. Gregory & O'Connor, 2002), allowed specific associations among sleep problems (e.g., trouble sleeping, nightmares, sleeping less than other children, and sleeping more than other children), tiredness, and psychological symptoms (aggression, attention problems, and anxiety/depression) to emerge. For example, after controlling for aggression and attention problems, trouble sleeping, nightmares, and tiredness were significantly associated with parent-rated symptoms of anxiety/depression (Coulombe et al., 2009). It is unclear whether similar associations will be found in adolescents (aged 12–16 years) using adolescent-report, as well as parent-report, data.
Hypotheses Based on the results of the existing literature and of our own study conducted with children (Coulombe et al., 2009), it was predicted that, after accounting for shared risk factors and psychological co-morbidity, (1) sleep problems and tiredness would be significantly associated with symptoms of anxiety/depression, withdrawal, and aggression and (2) that tiredness alone would be significantly associated with symptoms of attention problems. As results of studies using cross-informant outcomes have been inconsistent (c.f., Blunden & Chervin, 2008; Johnson et al., 2000), these hypotheses were examined separately for both adolescent-rated and parent-rated symptoms of psychopathology. At least one study suggests that adolescents may be better reporters of their internal states than their parents (Angold, Weissman, John, & Merikangas, 1987).
Materials and Methods
Secondary data analyses of the Ontario Child Health Study (OCHS; Boyle et al., 1987) were conducted. The primary purpose of the OCHS was to estimate the prevalence of emotional and behavioral disorders in a representative sample of 4–16-year-olds living from the province of Ontario, Canada; (see Boyle et al., 1987 for a description of the methodology for the OCHS). Demographic information, child physical health, parent-rated questionnaires for risk factors (e.g., family functioning), and parent-rated psychopathology checklists were collected from the female head of the household; adolescent self-report data were also collected. Parent nonresponse (missing checklists and questionnaires) was low (less than 1.5% of the total sample, aged 4–16 years); adolescent non-response was also low (2.6 %) (see Boyle et al., 1987).
Study-specific Methodology
The current study was approved by the Research Ethics Board at McMaster University.
Participants Adolescents with chronic medical conditions, significant physical limitations, mental retardation, developmental delay, and for whom this information was missing (n = 322) were excluded from analyses because sleep problems and physical health problems tend to co-exist (Mindell & Owens, 2003) and because sleep problems of these adolescents likely differ from those who do not have these health conditions (Stores, 1999). The final sample size for this study was 980 adolescents (49% male; age M = 13.88 years, SD = 1.41 years) from 771 households.
Measures and Key Variables Shared Risk Factors Risk factors with consistent and robust associations with psychopathology that may also affect sleep (e.g., single parent status, stressful life events, maternal depression or negative affect, poverty, poor family functioning) were identified from the pediatric sleep (e.g., Warren, Howe, Simmens, & Dahl, 2006) and developmental psychopathology literatures (e.g., Mash & Dozois, 2003; described in greater detail in Coulombe et al., 2009). The following variables were included: adolescent sex, adolescent age, family income (in increasing increments of $5000), maternal education (in years), maternal negative affect (using the Bradburn Affective Balance Scale; Bradburn, 1969), stressful life events (count of stressful events over the previous year that families experienced; see Boyle et al., 1987), marital disharmony (a composite of 4 items measuring overall relationship, frequency of quarrelling, enjoyable activities, and caring), and family functioning (measured using the General Functioning Scale of the McMaster Family Assessment Device; Byles, Byrne, Boyle, & Offord, 1988). All variables other than family income and adolescent sex were continuous, with higher scores indicating increasing age, more maternal education, greater maternal negative affect, more stressful life events, greater marital disharmony, and poorer family functioning.
Sleep Problems, Tiredness, and Psychological Symptoms. Sleep problems (adolescent-rated only), tiredness, and psychological symptoms scores (aggression, attention problems, anxiety/depression, and withdrawal) were derived from Child Behavior Checklist (CBCL; Achenbach, 1991a) and Youth Self Report Form (YSR; Achenbach, 1991b). Parents and adolescents rated how each item described the target child "now or within the past 6 months" using a 3-point scale (0 = "never or not true", 1 = "sometimes or somewhat true", 2 = "often or very true").
Psychological Symptoms and Psychological Co-morbidities. Parent- and adolescent-rated aggression, attention problems, anxiety/depression, and withdrawal scores were calculated for each adolescent. No sleep problem or tiredness item was included in any of the psychological symptom scales. So as not to artificially inflate associations among scales, items that appeared in more than one scale (e.g., "nervous or tense" appears as both an anxiety/depression and attention item) were not included in any scale score (Hinden, Compas, Howell, & Achenbach, 1997). Two items were removed from the adolescent-rated scales and three items were removed from the parent-rated scales; no more than two items were removed from any one scale. Prior to imputing missing data, internal consistencies for the parent-rated scales in our sample were: aggression (α = .87), attention problems (α = .77), anxiety/depression (α = .83), and withdrawal (α = .73). Internal consistencies for the adolescent-rated scales in our sample were: aggression (α = .81), attention problems (α = .71), anxiety/depression (α = .84), and withdrawal (α = .57). It should be noted that, through the remainder of this paper, a scale is referred to as a type of "psychological symptom" when the scale is used as an outcome variable; the term "psychological co-morbidity" is used when a scale is being used as a control variable.
Sleep Problem and Overtired Scores. Four sleep problem items were rated by adolescents: "I have trouble sleeping" (trouble sleeping), "I have nightmares" (nightmares), "I sleep less than most kids" (sleeping less), and "I sleep more than other kids during the day and/or night" (sleeping more). One tiredness item was rated by adolescents ("I feel overtired") and by parents ("overtired without good reason"). Each of these items (rated as 0 = "never or not true", 1 = "sometimes or somewhat true", 2 = "often or very true") was treated as a separate predictor of psychological symptoms (see Coulombe et al., 2009).
Data Analyses Analyses were conducted using Statistical Package for the Social Sciences (SPSS) version 16.0 and, for the regressions, STATA (Statistics/Data Analysis, 2006) version 10.1.
Missing Data. Less than 5% of the values were missing for any parent-rated variables (i.e., risk factors, parent-rated psychological symptoms, parent-rated "overtired"), with the exception of marital disharmony, which was systematically missing for all single parents (10.4% of households, 10.6% of adolescents). Less than 5% of the values were missing for any adolescent-rated variables (i.e., adolescent-rated sleep and tiredness items, adolescent-rated psychological symptoms). Few differences among data imputation methods have been found when less than 10% of data is missing, either randomly or systematically (Roth, 1994). In our sample, no differences in the patterns of results were found when adolescents with missing data were excluded from analyses or when missing values were replaced with the mean for that variable. Thus, with the exception of single parents missing marital disharmony scores (who were given scores indicative of "no marital disharmony"), all missing data was imputed using mean substitution (George & Mallery, 2002); interested readers are referred to the companion article for more detailed information on how missing data was handled (Coulombe et al., 2009).
Data Transformation. Psychological symptom/psychological co-morbidity scores, particularly for parent-rated data, were non-normally distributed. Square root transformations produced skewness and kurtosis values closest to 0 for parent-rated data (Tabachnik & Fidell, 2007). Transformation did not improve the skewness and kurtosis of adolescent-rated data.
Multivariate Analyses. Linear regressions examined the contribution of risk factors, psychological co-morbidity, sleep problems, and tiredness to the prediction of symptoms of adolescent-rated and parent-rated aggression, anxiety/depression, attention problems, and withdrawal. Some households had more than one adolescent participating in the study. As data from two siblings are not independent, all regressions controlled for the clustering of siblings within households. All variables were standardized prior to entry, allowing the regression coefficients to be interpreted as beta weights (Bring, 1994).
Predictors were entered in three blocks (Models 1–3). Shared risk factors and demographic variables were entered into the first block (Model 1). Sleep problem and tiredness scores were entered into the second block (Model 2), permitting a comparison of our findings with those of previous studies that have not controlled for co-morbidity. Parent-rated tiredness was included as a cross-informant predictor of psychological symptoms. Psychological co-morbidity was entered into the third block (Model 3), testing the hypothesis that sleep problems and tiredness would be significantly associated with psychological symptoms when shared risk factors and co-morbidity were considered. To better understand the effects of controlling for psychological co-morbidity, changes to the regression coefficients from Model 2 to Model 3 for each sleep variable were examined. Regressions were completed separately for adolescent-rated and parent-rated psychological symptoms.
