Table 3.

Two-part model for hospitalization costs of individuals using the public sector. ISA-Capital 2003–2015, São Paulo city, Brazil.

Variables	Categories	Logit			GLM			Marginal Effects
		β	SE	Sig.	β	SE	Sig.	dy/dx	SE	Sig.
Sex	(1 = female)	−0.049	0.151		−0.556	0.175	**	−26.25	11.30	*
Age	(years)	0.001	0.004		0.016	0.004	***	0.79	0.28	**
Skin color/ethnicity	(1 = white)	−0.015	0.158		−0.263	0.193		−12.08	11.16
Marital status	(1 = accompanied)	0.220	0.167		0.157	0.161		16.14	10.13
Education	(years)	0.012	0.018		−0.015	0.024		−0.12	1.31
Occupational status	(1 = employed)	−0.373	0.162	*	0.440	0.182	*	3.36	10.57
Income per capita	($PPP)	−0.328	0.069	***	−0.070	0.115		−16.91	5.63	**
Leisure-time PA	(1 = yes)	−0.475	0.220	*	−0.252	0.253		−31.03	14.68	*
Time spent in sitting position (ref. = lower tertile)	(1 = middle tertile)	−0.080	0.186		0.343	0.285		11.25	15.17
	(1 = upper tertile)	0.054	0.176		0.180	0.176		9.74	10.39
Obesity	(1 = yes)	−0.281	0.313		1.004	0.535		31.78	27.69
Tobacco use	(1 = yes)	0.492	0.161	**	0.304	0.221		33.98	12.95	**
Health insurance	(1 = yes)	−0.891	0.232	***	0.026	0.261		−36.45	14.40	*
Hypertension	(1 = yes)	−0.041	0.235		1.670	0.463	***	70.83	26.56	**
Diabetes	(1 = yes)	1.219	0.407	**	1.574	0.645	*	119.81	36.94	**
CVD		1.136	0.331	**	0.285	0.317		60.30	19.90	**
Year (ref. = 2003)	(1 = 2008)	0.243	0.221		−0.013	0.245		7.14	10.26
	(1 = 2015)	0.611	0.204	**	0.258	0.231		38.09	14.16	**
Constant		−1.503	0.449		6.724	0.612
N		7104			298			7104

GLM = generalized linear model (gamma family with log-link); β = regression coefficient; SE = robust standard error; dy/dx = average marginal effects. Statistical significance: *** = p < 0.001; ** = p < 0.01; * = p < 0.05.

PA = physical activity; CVD = cardiovascular diseases.