Path analysis of cognitive behavioural models in substance misuse. What is the relationship between concepts involved?

Introduction: The aims were (i) to assess if data support the propositions about the assumed causal relationships deduced from Cognitive Behaviour Therapy models; (ii) to explore the potential inclusion of negative expectancies in those; and (iii) to assess if data supports one model better than others.

Method: Patients dependent on alcohol, opioids and stimulants were recruited. Mediational Path analysis was performed. Change of dependence at 3 months was the outcome.

Results: Ninety-six alcohol, ninety-four opioid and seventy-seven stimulant participants were recruited. Multi group analysis indicated that the three substance groups could be collapsed and analysed together. The multiple regression model, based on Marlatt’s relapse prevention model, was the best to fit the data. Urges had a strong negative effect and negative expectancies a weaker positive effect on outcome. An exploratory model incorporating causal relationships between the variables was tested, with urges and negative expectancies as the final mediators. This model though was not better than the multiple regression one.

Conclusion: Comparison across models suggested that the regression model based on the original Relapse prevention was the best, with urges and negative expectancies having the highest effect in opposite directions.