Table 3.

Results of univariate and multivariate Cox proportional hazard models, and sensitive analyses

Model		HR (95% CI)	p value
Univariate model		1.20 (1.17, 1.23)	< 0.001
Multivariate model a		1.19 (1.16, 1.22)	< 0.001
Sensitive analyses	model 1b	1.20 (1.16, 1.24)	< 0.001
	model 2c	1.17 (1.14, 1.20)	< 0.001
	model 3d	1.18 (1.15, 1.21)	< 0.001

Notes: HR: hazard ratio; CI: confident interval

a: Model adjusted for gender, age group, marital status, education, insurance, outpatient services, inpatient services, residence, family size, family economic level, and socioeconomic development level

b: Based on the multivariate model, in this model, the definition of CHE (household-medical OOP expenditure exceeding 45% household total expenditure net of food expenditure) was replaced as household-medical OOP expenditure exceeding 25% of total household expenditure

c: Based on the multivariate model, in this model, we changed the original selected main respondents in households with 2 eligible respondents to the 5270 excluded participants

d: Based on the multivariate model, in this model, we used the nighttime light level to indicate socioeconomic development rather than gross regional products