Comparing means in a 3-arm study using ML analysis
Posted: Tue Aug 16, 2011 1:03 pm
I hope that the MLwiN user forum could help us in selecting the most appropriate approach to model the findings.
We are conducting a 3-armed clustered randomised trial in Chronic Kidney Disease. The 3 arms are (1) Usual Practice, (2) Guideline and Prompt, and (3) Audit based education. A key outcome under investigation is the reduction of blood pressure (BP) which is a continuous variable. The unit of analysis is at the changes in BP at 2 points in time and at individual patient level, but the randomisation is at the level of GP practices.
The model hopefully is a simple one: to examine the differences of BP between the 3 arms but taking into account the hierarchy of GP practices, as each GP practice has its own ethos, culture and infra-structure for management of chronic diseases, broadly shared deprivation scores and access to local renal specialists in secondary care for its patients. I am not sure how to model this. It has been more than 15 years since I attended a workshop on Multi-level analysis. The simplest might be ANOVA model but I don't know how I could account for the 'GP practice' hierarchy.
Any comments would be gratefully received.
Best wishes
Tom Chan
We are conducting a 3-armed clustered randomised trial in Chronic Kidney Disease. The 3 arms are (1) Usual Practice, (2) Guideline and Prompt, and (3) Audit based education. A key outcome under investigation is the reduction of blood pressure (BP) which is a continuous variable. The unit of analysis is at the changes in BP at 2 points in time and at individual patient level, but the randomisation is at the level of GP practices.
The model hopefully is a simple one: to examine the differences of BP between the 3 arms but taking into account the hierarchy of GP practices, as each GP practice has its own ethos, culture and infra-structure for management of chronic diseases, broadly shared deprivation scores and access to local renal specialists in secondary care for its patients. I am not sure how to model this. It has been more than 15 years since I attended a workshop on Multi-level analysis. The simplest might be ANOVA model but I don't know how I could account for the 'GP practice' hierarchy.
Any comments would be gratefully received.
Best wishes
Tom Chan