Table 3.

Cox regressions for mortality and 1-year landmark analysis among Chinese herbal medicine and Western medicine cohorts.

Clustera	Unadjusted				Adjustedb
	HRc,d	99% CI	P value	aHRe,f		99% CI	P value
Cluster 1 (n=5272)	0.39	0.35-0.43	<.001	0.40		0.36-0.44	<.001
Cluster 2 (n=2275)	0.37	0.32-0.43	<.001	0.37		0.32-0.44	<.001
Cluster 3 (n=2139)	0.38	0.33-0.45	<.001	0.39		0.34-0.46	<.001
Cluster 4 (n=905)	0.23	0.15-0.36	<.001	0.24		0.16-0.35	<.001
Cluster 5 (n=1144)	0.33	0.26-0.43	<.001	0.34		0.27-0.44	<.001
Cluster 6 (n=665)	0.38	0.28-0.50	<.001	0.38		0.29-0.51	<.001
Cluster 7 (n=173)	0.43	0.25-0.74	<.001	0.48		0.30-0.76	<.001
Cluster 8 (n=375)	0.22	0.10-0.49	<.001	0.23		0.11-0.47	<.001
Cluster 9 (n=243)	0.32	0.19-0.54	<.001	0.32		0.20-0.50	<.001
Cluster 10 (n=286)	0.33	0.20-0.53	<.001	0.32		0.21-0.48	<.001
Cluster 11 (n=196)	0.45	0.27-0.73	<.001	0.45		0.29-0.69	<.001

^aEach cluster contained patients who took different groups of Chinese herbal medicines.

^bGender, age, geolocation, insured level, comorbidities, and medications were used as covariates in the adjusted regression models.

^cHR: hazard ratio.

^dThe hazard ratio of each cluster was estimated after inverse probability treatment weighting in contrast to the Western medicine cohort.

^eaHR: adjusted hazard ratio.

^fThe adjusted hazard ratio was calculated by a Cox regression model considering patient gender, age, comorbidities, medications, insured level, and geolocation. Inverse probability treatment weighting was estimated from the same covariates to relieve the accessible confounding bias between Chinese herbal medicine users and nonusers.