Table 3.

Associations between TCM Use and the Endpoint.

Analysis	NRICM101	NRICM102
	Intubation or ICU Admission	Death
No. of events/no. of patients at risk (%)
TCM + Usual Care	0/164 (0.00)	7/126 (5.56)
Usual Care	14/181 (7.73)	42/240 (17.50)
Relative Risk (95% CI)	--&^	40.80% (20.54%-81.12%)^
Propensity score analyses - with matching
TCM + Usual Care (%)	0/151 (0.00)	7/123 (5.69)
Usual Care (%)	14/151 (9.27)	27/123 (21.95)
Relative Risk (95% CI)	--&*	25.93% (11.73%−57.29%)*
Hazard Ratio (95% CI)	--$#	23.17% (10.36–51.82%)#

& Seriously underestimated relative risk (95% CI) = 15.8% (3.6%−68.3%) for unmatched data and 14.3% (3.3%−71.8%) for matched data when we included 2 censored cases as the endpoint.

^ The chi-square test was used for unmatched data (p value = 0.002 for death and p value = 0.006 when we set 2 censored cases as intubation or ICU admission).

* McNemar’s test compared the proportion of intubation or ICU admission (p value = 0.003) and death (p value < 0.001) for matched data. The power of McNemar’s test being larger than 0.852 for NRICM101 and 0.929 for NRICM102 indicates that the significance of both is not due to chance.

$ Seriously underestimated hazard ratio= 13.58% (3.40–54.21%) when we set 2 censored cases as the endpoint by the marginal Cox model.

# Hazard ratio by marginal Cox regression and p value < 0.001 by stratified log-rank test for both NRICM101 and NRICM102.