Self-reported physical and mental health status and quality of life in adolescents: a latent variable mediation model

We designed a study to: (a) validate adolescents' differentiation of their self-reported physical and mental health status and (b) examine the associations of these variables with global QOL and several relevant life domains, including adolescents' satisfaction with their family, friends, living environment, school, and self. With respect to the first objective, we hypothesized that, relative to one another, self-reported physical health status would be more strongly associated with physical activity, and self-reported mental health status with depressive symptoms. Drawing on the previously mentioned conceptual models and empirical research on health and QOL, we further sought to obtain information about (a) the relative importance of self-reported physical and mental health status with respect to adolescents' global QOL and several life domains and (b) the extent to which the relationships among self-reported physical and mental health statusand global QOL are mediated by the life domains (see Figure 1). Global QOL is viewed here as a unidimensional construct that pertains to individuals' satisfaction with, or appreciation of, their lives overall[18, 30‐32]. The life domains represent adolescents' satisfaction with various conditions in life that have the potential to contribute to their global QOL [33].

The data were obtained via the British Columbia Youth Survey on Smoking and Health 2 (BCYSOSHII), a cross-sectional health survey that was conducted in 2004 to obtain information about tobacco dependence, drug and health-related behavior, and quality of life in adolescents in grades 7 to 12 in schools in the province of British Columbia (BC), Canada. The methods of this survey have been described in detail in several published studies [e.g., [34‐39]]. The survey avoided two regional districts within the province that were known to have very low smoking prevalence rates so as to be cost-efficient in assembling a sample of adolescents that used tobacco (the primary purpose of the principal study). Nineteen of the 60 school districts in BC were sampled to achieve maximal geographic coverage of regional districts (remote and sparsely populated areas were not surveyed). Fourteen of the school district administrators provided permission for their schools to participate. This resulted in a sample of 89 eligible schools, ofwhich 49 (42 secondary schools, 5 alternative schools, and 2 middle schools) agreed to participate. Passive parental consent was obtained by providing parents with letters that informed them of the survey. Ethical approval was granted by the Behavioural Research Ethics Board of the University of British Columbia.

The survey questionnaire was administered by research assistants during class-time hours in pen and paper format (79.6%) or through a computer-based format (20.4%). The format was primarily determined by the availability of computers in the various schools. Less than 1% of the students refused to participate and the response rate within schools was 84%, on average (non-response was mostly due to student absenteeism) [34, 36]. The resulting sample consisted of 8,225 adolescents (smokers and non-smokers).

Self-reported physical and mental health status were measured using two questions, "How would you rate your physical health?" and "How would you rate your emotional or mental health?" with the following response options, which were taken from the general health status question of the SF-36 [25] and which are widely used in the national population health surveys of many countries: "excellent" (coded as "5"), "very good," "good," "fair," or "poor" (coded as "1"). The validity of measuring adolescents' self-reported general health status in this manner is supported by observed associations with various other health status indicators, including physical activity, nutrition, health-risk behavior, and physical disability [40‐43]. Study findings have consistently revealed that a relative increase in adolescents' self-reported general health status is associated with less health-risk behavior and fewer days of limited activity [41, 43].

To validate adolescents' differentiation of their physical and mental health status, we examined the relative importance of these variables with respect to depressive symptoms and the frequency of physical activities. Depressive symptoms were measured using 12 items from the Center of Epidemiologic Studies Depression Scale (CES-D) [44]. The adolescents were asked: "How often have you felt or behaved in the following manner in the past week (7 days)?" (e.g., "hopeful about the future," "happy," "lonely," "sad"). The CES-D provides four response options ranging from "rarely or none of the time (less than one day)" (coded as "0") to "most or all of the time (5-7 days)" (coded as "3"). The total score, with a possible range of 0 (no depressive symptoms) to 36, was used in the analysis. The estimated reliability of the 12 items is .87 in this sample (based on the ordinal Cronbach alpha reliability estimate [45]). Physical activity was measured using the following question adapted from several large surveys (e.g., The USA Youth Risk Behavior Survey [46] and The Ontario Drug Use Survey [47]): "On how many of the last 7 days did you exercise or participate in sports activities for at least 20 minutes that made you sweat and breathe hard? If none, enter '0' days. Please include activities such as basketball, jogging, swimming, cross-country skiing, hockey, or dance, that you participated in either at school or outside of school."

An abridged version of Huebner's Multidimensional Students' Life Satisfaction Scale (MSLSS) [48] was used to measure adolescents' satisfaction with five life domains, including their family (4 items), school (4 items), living environment (2 items), friends (4 items) and self (4 items) [37]. The original MSLSS consists of 40 items, of which 10 are negatively worded. The psychometric analyses reported by Sawatzky et al. [37] revealed that the adolescents may not have interpreted and responded to all items in the same way. There were inconsistencies in the responses to the negatively worded items and several other items. An abridged 18-item version was developed by identifying those items that were found to be most invariant (all positively worded). Confirmatory factor analyses (CFA) provided support for its construct validity when allowing for a few theoretically defensible modifications [37]. The same measurement structure was used to represent the five life domains as latent factors in the study reported herein. The ordinal Cronbach alpha reliability estimates [45] of the abridged subscales with four items were ≥ .80 in this sample. A six-point ordinal response format (with response options ranging from "strongly disagree" (coded as "1") to "strongly agree" (coded as "6")) was used [49].

Global QOL was measured with two items. The adolescents were asked to appraise their QOL using a picture of an eight-rung ladder (Cantril's self-anchoring ladder [50], referred to here as the QOL-ladder) (see Figure 2). The bottom run was coded as "1" and the top as "8". The adolescents also were asked to rate their agreement with the statement, "I am satisfied with my quality of life" with four response options ranging from "strongly disagree" (coded as "1") to "strongly agree" (coded as "4"). General questions of this nature, including Cantril's self-anchoring ladder, have been widely used in surveys for the measurement of various concepts such as global QOL [51‐53]. A latent factor explaining the variance in both of these variables was used to represent global QOL.

The adolescents were asked to indicate their age and sex, and to answer several questions about their ethnic identity and living arrangements. Ethnic identity was determined by asking, "How would you describe yourself?" The 12 response options were adapted from Statistics Canada's [54] classification of "visible minorities" (e.g., "white/Caucasian," Aboriginal/First Nation, Chinese, South East Asian). The adolescents selected one or more responses, which were subsequently grouped as: "white/Caucasian," Asian (including Chinese, Japanese, Korean, South East Asian, and Filipino), Aboriginal/First Nation, and "other." With respect to their living arrangements, the adolescents were asked, "Which parent or parents do you currently live with most of the time?" with eight response options (i.e., mother, father, step-mother, step-father, guardian(s), foster parent(s), grandparent(s), and other please specify).

Structural equation modeling was used to examine the hypothesized relationships by fitting a latent variable mediation model to the sample data (see Figure 1). The variances of the latent factors were specified to equal one to avoid indeterminancy and to set the metric of the latent factors [55]. Polychoric correlations were used to avoid obtaining biased parameter estimates due to the ordinal distributions of the observed variables [56‐58]. The MPlus 5.2 software [59] was used to estimate the model parameters by specifying a probit link function and using a mean and variance adjusted weighted-least squares estimation method (WLSMV) suitable for ordinal data [60]. Model fit was evaluated with several global fit indices, and by comparing the differences between the implied and the observed polychoric correlation matrices. Adequate model fit was defined by a root mean square error of approximation (RMSEA) of < .06 [61] and a comparative fit index (CFI) of ≥ .95 [62]. In addition, the pattern and magnitudes of the residual correlations were examined to locate any specific areas of misfit [63, 64]. The percentage of residual correlations with absolute values greater than .10 is provided as a summary of this direct comparison.

The relative importance of the explanatory variables was determined by the Pratt index (d) [65], which quantifies each variable's contribution to the explained variance (irrespective of the magnitude of the R-squared), measured as a percentage. The extent to which the two relationships between global QOL and physical and mental health status were mediated by the life domains was evaluated as the division of the indirect-effects (mediated by the life domains) and the total effect (the sum of the direct- and indirect-effects for the associations between global QOL and physical and mental health status), expressed as a percentage [66]. The standard error of the indirect effects was calculated using the Delta method, which is similar to the approach used in the Sobel test [67].

Of the 8,225 adolescents, 920 did not provide responses to any of the MSLSS questions. The analysis was limited to those who responded to the global QOL items, the items measuring mental or physical health status, and at least one of the MSLSS items (N = 6,932). Multiple imputation (MI) [68] was used to impute any remaining missing responses (2.5% imputed data). The results were compared with those obtained using MI for the subsample of adolescents who provided a value for at least one of the analysis variables (N = 8,174; 13.9% imputed data). The SAS 9.2 software package [69] was used to create 10 imputed datasets for the MI analyses, following the guidelines offered by Allison [70] and Enders [71], to assess convergence and to incorporate auxiliary variables (i.e., demographic variables (sex, ethnicity, school grade), symptoms of depression, and two variables pertaining to the adolescents' experiences at school).

The sample consisted of an approximately equal proportion of boys and girls in grades 7 through 12 (see Table 1). The average age was 15.2 years (SD = 1.5, n = 8,054) with 7,964 adolescents being between 12 and 18 years. Although most of the adolescents who identified their ethnicity (n = 7,882) self-identified as "white/Caucasian" (72.6%), the sample also included Aboriginal adolescents (16.5%), Asian adolescents (Chinese, Japanese, Korean, Filipino, or South-East Asian) (5.8%), and adolescents belonging to one or more other groups (5.1%). A sizeable percentage (17.3% of 7,994 adolescents) indicated regularly speaking a language other than English, and 6.9% of 8,058 reported being born in a country other than Canada.

Most of the adolescents agreed or strongly agreed to being satisfied with their QOL (82.3% of 7,606 adolescents) (see Table 1). The mode of the QOL-ladder responses was at level 6 of 8 rungs (36.7%), with 11.9% of the adolescents reporting the best possible life, and 14.0% providing a rating at or below the middle of the scale (≤ 4) (n = 7,675).

The joint- and marginal-distributions of self-reported physical and mental health status are provided in Table 2. The corresponding conditional distributions provide support for adolescents' ability to differentiate these variables. For example, 9.5% of the adolescents who rated their physical health as good or better rated their mental health as fair or poor, and 5.3% of the adolescents who rated their mental health as good or better rated their physical health as fair or poor. The polychoric correlation was .55, indicating a shared variance of only 30% among these two (underlying) variables. With respect to the differentiation of mental and physical health status (see Table 3), we found that 94% (d, Pratt Index) of the explained variance in depressive symptoms (R² = 35.5%) could be attributed to mental health status (the remaining 6% was attributed to physical health status). Conversely, relative to self-reported physical health status, self-reported mental health status accounted for only 18% of the explained variance in physical activity (R² = 7.7%) (see Table 3).

The hypothesized model with the life domains operating as mediators of the relationships between self-reported physical and mental health status and global QOL resulted in acceptable overall fit (WLSMV χ² ranging from 2,083.22 to 2,010.02 for the 10 MI datasets (N = 6,932), RMSEA = .049, CFI = .951, residual correlations ranging from -.07 to .07) (see Figure 1). Satisfaction with family, friends, school, living-environment, and self, and self-reported physical and mental health status explained 76.1% of the variance in global QOL. Although self-reported physical and mental health status were bivariately significantly correlated with global QOL (r = .49 and .70, respectively), their associations were substantially smaller, albeit statistically significant, in the multivariate model (see Table 4). The life domains also were bivariately significantly correlated with global QOL. However, relatively small and statistically non-significant regression coefficients were obtained for satisfaction with friends and satisfaction with school in the multivariate model (see Table 4). These variables accounted for less than 2% (d, Pratt Index) of the explained variance relative to the other variables in the model (see Table 4). Global QOL was mostly explained by satisfaction with self (d = 42%), self-reported mental health status (d = 30%), and satisfaction with family (d = 20%). Self-reported physical health status accounted for only 3% of the explained variance.

Self-reported physical and mental health status were significantly correlated with each of the life domains (r _{physical health} ranging from .22 to .45; r _{mental health} ranging from .27 to .54), and they predominantly explained satisfaction with self (R² = 33.0%), and, to a lesser extent, satisfaction with family (R² = 16.9%), friends (R² = 11.3%), and living environment (R² = 14.2%) (see Table 5). Only 7.9% of the variance in satisfaction with school was explained by self-reported physical and mental health status. Relative to self-reported physical health status, most of the variance in each of the life satisfaction dimensions could be attributed to the adolescents' self-reported mental health status (d ranging from 68% to 87% for each of the life domains) (see Table 5).

The parameters for the relationships between physical and mental health status, the life domains, and global QOL were used to determine the magnitude of the total and the indirect relationships between physical and mental health status and global QOL as mediated by each of the life domains (see Table 6). The standardized total effect on global QOL was larger for self-reported mental health status (β = .61), while adjusting for self-reported physical health status, than for self-reported physical health status (β = .17), while adjusting for self-reported mental health status. These relationships were partially mediated by the life domains (67.8% total mediation for physical health and 45.4% total mediation for mental health status). The relationships between the two health status variables and global QOL were primarily mediated by satisfaction with self (54.0% mediation for self-reported physical health and 29.1% mediation for self-reported mental health) and, to a lesser extent, by satisfaction with family (10.8% mediation for self-reported physical health and 13.7% mediation for self-reported mental health).