Common use of Predictor variables Clause in Contracts

Predictor variables. The self-report questionnaires covered anxiety, physical health, personality pathology, negative life events, post-traumatic stress, social support, self-efficacy, and coping style. The level of anxiety was measured with the subscale of the Dutch version of the Hospital Anxiety and Depression Scale (HADS-A; ▇▇▇▇▇▇▇ & ▇▇▇▇▇▇, 1983). A cut- off of 8 is recommended to distinguish between high and low anxiety levels. The α in this sample was .80. As an indication of physical health the presence of chronic medical conditions was assessed at pre-treatment and at the 14-month FU; this was done with a checklist of nine chronic medical conditions covering cardiovasculair diseases, pulmonary conditions, brain damage, diabetes, rheumatism, arthrosis, dysplasia (Central Bureau of Statistics, 1989). Furthermore, the scales for pain and physical functioning of the Medical Outcome Study Short Form General Health Survey (MOS-SF-20; ▇▇▇▇▇▇, 1992; ▇▇▇▇▇▇▇ & ▇▇▇▇, 1988) were used as indications of physical health. Personality pathology was assessed with the Questionnaire of Personality Traits - VKP (in Dutch: Vragenlijst voor Kenmerken van de Persoonlijkheid), an inventory with items based on the DSM-IV and ICD-10 definitions and criteria of personality disorders (Duijsens, Eurelings-▇▇▇▇▇▇▇▇, & ▇▇▇▇▇▇▇▇, 1996; Duijsens, Haringsma, & Eurelings-▇▇▇▇▇▇▇▇, 1999). At pretrreament the DSM-IV section which consisted of 149 items (including the passive-aggressive and the depressed personality disorders) was administered. The VKP yields a diagnosis and a dimensional score for each specific personality disorder (PD). The latter can be summed into a dimensional score for each cluster and into a total sumscore (PD-NOS). The cluster scores and the sum score were used as predictor variables. The experience of negative life events at pretrreament was measured with a checklist based on the Negative Life Events Questionnaire used by ▇▇▇▇▇▇ and ▇▇ ▇▇▇▇▇ (2001). It covers different developmental periods, such as childhood, adulthood, and events in the past year. A sumscore was calculated for the whole life span. Current posttraumatic stress was assessed with the Dutch version of the Impact of Event Scale (IES; ▇▇▇▇, ▇▇▇▇▇▇, & ▇▇▇▇▇▇▇, 1986; ▇▇▇▇▇▇▇▇, ▇▇▇▇▇▇, & ▇▇▇▇▇▇▇, 1979). It has 15 items; in this sample the α was .94. Social support was assessed with the abbreviated version of the Social Support List-Interaction (SSL112-I), which is intended for use with elderly adults (▇▇▇▇▇▇ & ▇▇▇ ▇▇▇▇, 1995). The sum scale in this sample had an α of .86. Self-efficacy was measured with the Dutch version of the General Self-Efficacy Scale (GSES; ▇▇▇▇▇▇▇▇▇▇, 1997, 1998), a 10-item questionnaire. In our sample α was .89. The habitual coping style, one of the targets in the course was measured with the Utrechtse Coping List (UCL; Schreurs, Willige, & ▇▇▇▇▇▇▇▇▇, 1993). It has 47 items and measures seven coping strategies: active-problem-solving (α =.79), palliative-responses ((α = .71), avoidance-strategies (α =.74), seeking-social- support ((α = .79), depressive-reaction-pattern (α = .74), expression-of-emotions (particular anger) (α = .55), and comforting-cognitions (α = .60). New negative life events were checked at every assessment; these were summed to get an estimate of adverse events experienced since the conclusion of the course. The 14-month FU assessment also contained a checklist for newly developed medical conditions. Stress-buffering effects of positive life events and improved physical health that may protect against depression were similarly checked. Preliminary analyses included checks for normality and the computation of descriptive statistics. All variables except those considering personality pathology (cluster A, cluster B, Cluster C, and PD-NOS) appeared to be distributed acceptably close to normal. Distributions of personality pathology variables were improved by applying square root transformations, which were used in the analyses. Only variables that showed significant (p < .05) effects will be reported. Random coefficient regression models (RCRMs) were used to examine the contribution of the various predictor variables to the immediate and maintenance effect. Rrepeated measures were considered to be nested within individuals, nested within CWD-groups. Because this research focuses on: (1) the immediate effect; and (2) the maintenance effect, it was decided to study the two corresponding trajectories in two different linear models, instead of fitting a less adequate non-linear trend over four time points. The model for the immediate effect covered the first three measurements (pre-, post- and two-month FU). The maintenance effect was modelled using the post- treatment, two-month FU and 14-month FU measurements. Hence, data on two time points – post-treatment and 2-month FU measurements – were used twice. In the model for the first trajectory, intercept and slope can referred to as average pretrreament score and average improvement rate, respectively. In the second trajectory they can be referred to as average post-treatment score and average change rate, respectively. Both models contained variance components estimating the amount of variation of individual (linear) trends around these average lines. Predictor variables for both models were selected in a three-step approach. The first step was testing each predictor variable separately by adding it to the model with Time as the only predictor (referred to as the baseline model). Time was measured in weeks; pre-, post-, two-month FU and 14-month FU had the values of 1, 10, 20 and 72 respectively. The variables showing a significant weight (p < .05) were retained for the final model. The final model was simplified using likelihood-ratio tests (χ2 derived from deviance values) and tests for separate fixed effects. Finally the most appropriate model was selected. Fixed effects were tested using one-tailed t-tests. Variance components were tested using likelihood-ratio tests as well. Potential predictors for the immediate effect were socio-demographic, mental health, and physical health variables, the sum of adverse life events, and coping variables; all variables were assessed at pre-treatment. Stable characteristics unlikely to have changed during the time of the course, for instance socio-demographic variables and the variables pertaining to mental health history and to coping resources, were selected as predictors for the maintenance effect. The effects of unpredictable events that might have influenced the level of depression at 14-month FU, such as new chronic illness, new stressful life events, improved physical health and positive life events were also analyzed. The possible contribution of CWD-group differences to the variance of the response variable was examined by estimating the intra-class correlation. Hierarchical logistic regression models were used to predict diagnostic status at the 14-month FU. Two subgroups were formed to study the different prevention goals. First, the participants who were at risk for developing a MDD (indicated prevention)