Table 2. Summary based on data from 74 doctors of robust Poisson conditional likelihood analyses comparing T2 (3 month) to T3 (12 months post-T2 assessment).

Behaviour	RR (95% CI)	P-value
Fewer leading questions	0.89 (0.65, 1.24)	0.5	Maintenance of skill shown to be higher than that of controls at 3 months in RCT
More focused questions	0.97 (0.81, 1.15)	0.69	Maintenance of skill shown to be higher than that of controls at 3 months in RCT
More focused open questions	0.96 (0.83, 1.10)	0.56	Maintenance of skill shown to be higher than that of controls at 3 months in RCT
Appropriate response to patient led cues	a1.02 (0.67, 1.54)	0.93	Maintenance of skill shown to be higher than that of controls at 3 months in RCT
Expressions of empathy	0.46 (0.33, 0.65)	<0.001	Attrition of skill shown to have been higher than that of controls at 3 months in RCT
More summarising of information	1.38 (1.02, 1.87)	0.038	Improvement in skill not previously seen to be higher than that of controls at 3 months
Fewer interruptions of patient	0.19 (0.10, 0.35)	<0.001	Improvement in skill not previously seen to be higher than that of controls at 3 months
Checking understanding	1.21 (0.88, 1.67)	0.24	No change, that is, no difference in behaviour at any time point compared with controls

^aA patient cue has to be given for the doctor to make a response, therefore the statistic reported here is OR (odds ratio) from conditional binomial analysis T2 – T3.