Table 3.

As Per protocol-analysis

	1. Outcome (45 day RR.)				2. Outcome (365 day RR.)
	Resp	Non-Resp	Total	RR	Resp	Non-Resp	Total	RR
IG	43	49a	92	46.7 %	45	47	92	51.1 %
CG	32a	93	125	25.6 %	36	89	125	28.8 %
Total	75	142	217b		81	136	217
Test	p	OR [95 % CI]:		ΔRR	p	OR [95 % CI]:		ΔRR
Mid-p	0.001	2.535 [1.432, 4.540]		21.1 %	0.003	2.354 [1.343, 4.167]		22.3 %
Fisher’s	0.001				0.003
χ 2	0.001				0.002

The table shows the 2x2-contingency tables w.r.t. 1. and 2. outcomes, with total figures for respondents (Resp) and non-respondents (Non-Resp) in the intervention- and control groups as defined in the APP-analysis. Response rates (RR) for the two study arms are also provided. Below the contingency tables we report the OR with 95 % CI’s, the absolute difference in response rates (ΔRR), and the p-values from three standard 2-sided tests for effect provided by the oddsratio-function from the epitools-package for R (CI’s are computed w.r.t. the mid-p value)

^aThe IG non-responders includes the 5 respondents who were successfully contacted, but refused participation in the questionnaire study; the CG responders includes 2 responders who were randomized to be pre-contacted, who were not successfully reached, but who nevertheless returned their questionnaires

^bOriginally 235 patients were included; 10 patients (7 from the IG and 3 from the CG) died before it was time to send them a questionnaire; 6 (3 in the IG and 3 in the CG) were discovered to be demented; 1 IG patient was discovered to have been mis-diagnosed with stroke, and one CG patient was discovered to have been mis-assessed as eligible. These 18 were excluded prior to analysis