Factors Predicting Outcome in Group Treatment of Alcohol Use Disorders (AUDs)

HYPOTHESES, AIMS AND OBJECTIVES

The general objective of the study is to increase the knowledge of addiction treatment, and how therapy can be made more effective, especially in the case of AUD. In particular, the project will study how patient characteristics interact with treatment and influence therapy outcome. The main research questions are:

1. How do patient factors moderate treatment outcome in group therapy with patients with AUD? The following variables will be investigated: Severity of alcohol dependence, symptom disorders, personality disorders, cognitive deficits and demographic variables.
2. Do positive treatment outcomes last over time? The participants will be followed up after one and three years post-treatment to see if recovery persists.
3. Is Montreal Cognitive Assessment (MoCa) feasible as a brief screening instrument for mild cognitive impairments in patients with AUD? At face value, this instrument contains rather crude tasks, which seems to make it more sensitive to large cognitive deficits than small. By comparing MoCa results to more extended neuropsychological testing we will assess the correlations, sensitivity and specificity of impairment assessed with MoCA.

The project has two main aims. The first primary aim is predictors of successful treatment completion operationalized as percentage of participation in therapeutic activities. The second primary aim is predictors for effect of group therapy one year after treatment termination. Primary outcome variable is alcohol and substance use reduction, measured with AUDIT and DUDIT. Secondary outcomes are symptom level measured with SCL-90 and quality of life measured with WHOQOL-bref. In addition, register data concerning use of health services after finishing treatment, and participation in working life, will be collected three years after treatment completion.

PROJECT METHODOLOGY

The present study is designed as a quasi-experimental, multi-centre study on treatment in ordinary clinical practice. The study will include at least 120 patients (approximately 40 participants pr. fall and spring-inclusion term). Four data collection sites are included in the study:

1. Patients enrolled in treatment at the Department of Addiction treatment, Vestfold Hospital Trust
2. Patients enrolled in residential treatment at the Borgestad Blue Cross Clinic
3. Patients enrolled in residential treatment at the Blue Cross Clinic, Treatment center Eina
4. Patients enrolled in residential treatment at A-senteret, The Church City Mission

Participants in the study will be patients group treatment for a primary diagnosis of AUD. Treatment is administered in a time-limited format at all sites. The therapy format may vary to some degree, but is similar in overall structure and content. In this quasi-experimental study research is carried out in an ordinary clinical situation, and study attempts to investigate the complexity of clinical practice. As a result, the researchers do not have control over all of the variables, but the results will be more ecologically valid with a higher degree of generalizibility.

PROJECT ARRANGEMENTS, METHOD SELECTION AND ANALYSES

The study will use well-researched tools and methods, used in both research and ordinary clincial practice. The following information will be entered in the registry:

Patient pre-therapy background: Earlier treatment episodes and diagnoses, demographic variables (age, sex, education, marital status, economical situation) will be retrieved from participants' journal.

Screening pre-treatment:

1. The Mini International Neuropsychiatric Interview (MINI). Short structured diagnostic interview for Diagnostic and statistical manual of mental disorders IV (DSM-IV)-diagnoses.
2. Severity indices of personality problems (SIPP-118). Self-report questionnaire focusing on core components of mal-adaptive personality functioning. 118 questions scored on a 4 point
3. Montreal Cognitive assessment (MoCa). A 10 minute screening tool to assist clinicians in detecting mild cognitive impairments. Yields a maximum score of 30. Cut-off under 26, indicating cognitive impairment.
4. Wechsler Adult Intelligence Scale (WAIS-IV). Measures core aspects of intelligence (estimate of General Ability Index). The following subtests will be administered: Similarities, Information, Visual puzzles, Block design and Digit span.
5. Delis-Kaplan Executive Function System (D-KEFS). Measures executive functioning. (Color Word Interference Test and tasks 2-4 from Trail Making Test).

Screening pre- and post-treatment, and after 1 and 3 years:

1. Alcohol disorders identification test (AUDIT). 10 item self-report questionnaire for identifying harmful drinking and dependence. Responses are scored on a scale from 0 to 4.
2. Drug disorders identification test (DUDIT). 11 item self-report questionnaire for drug problems. Responses are scored on a scale from 0 to 4.
3. Symptom check list 90-R (SCL-90-R). Self-report inventory, 90 questions. Scored on a 4 point scale. Measures severity of psychological distress. Will be administered pre-treatment, post-treatment and after 1 and 3 years.
4. World health organization quality of life scale (WHOQOL-BREF). Measures quality of life on four dimensions: Sensory abilities, autonomy in the past, present and future activities and social participation. 26 items scored on a 1 to 5 Likert scale.

Screening post-treatment:

1. Treatment satisfaction: 10 questions about treatment satisfaction.

Register data:

In addition, after 3 years, register data from patient journals and from NAV will be collected concerning working status, economical benefits and if the patient have undergone more treatment.

ANALYSES

Both intention to treat and per protocol analyses will be carried out on all participants and those who complete the treatment and data collection, until the last point of assessment. Mixed methods analyses may be used to include patients with incomplete protocols.

The outcome measures will be subjected to regression analyses to test:

1. How much of the outcome can be explained by substance use before treatment (AUDIT, DUDIT)?
2. How much are explained by psychological distress (SCL-90-R: GSI)
3. How much are explained by cognitive deficits?
4. How much are explained by demographic variables? (Age, sex, education level)

The study will seek to use hierarchical regression analyses transcending the single predictor domains by entering those measures from each domain that correlates highest with outcome and lowest with predictors from other domains. For the first analyses there are only one outcome measure (participation in group treatment). For the second set of analyses (function after one year) there are multiple outcome measures (drinking/substance abuse, symptom level, function and quality of life). Thus, we will seek to construct a gross overall composite outcome measure based on normative data from the tests used.

For non-continuous data (axis 1 diagnosis, gender and marital status) analyses of variance will be computed with group-characteristics as independent variable and outcome as dependent. It is reasonable to assume that patients with combinations of substance abuse profiles, symptom level, cognitive impairments and demographic factors may constitute specific groups (clusters) with different prognosis. Cluster analysis will be applied to unravel distinct combinations of factors associated with good or poor prognosis. As an experiment, one substance use measure, one symptom distress measure, one cognitive functioning measure and one or two demographic markers will be included in the analysis. Cluster analysis is merely an explorative method designed to uncover post-hoc empirical groups with combinations of features, but may discern individual patterns that will not be evident in analyses of group means.

STATISTICAL POWER

It is important that the planned number of participants is sufficiently large to answer the research questions. It is a quasi-experimental design in the sense that division into subgroups are determined by attributes of the participants not under complete control by the researchers. Regarding statistical power, the the following examples are computed: If it is expected that 35 % (n=42) of the participants have an axis 1 diagnosis of major depression and that this group will have somewhat over half a standard deviation (.55) higher score on AUDIT/DUDIT one year after completed treatment, this difference will be statistically significant on the five percent alpha-level if there are 41 and 76 persons in the two groups (depression vs. absence of depression), which will then be within the planned sample size. Continuous data will be analyzed with correlations and hierarchical regressions. Small effect sizes (Cohen's d: .30) will be statistically significant on the five percent level with 85 participants.