Table 3.

MSMs of the number of ER visits

	M1 Poisson			M2 Negative Binomial			M3 Zero-Inflated Poisson
	RR	[95% CI]		RR	[95% CI]		RR	[95% CI]
ᾱ = (0, 0, 1)	2.107	0.923	4.809	2.099	0.924	4.767	2.218	0.986	4.990
ᾱ = (0, 1, 0)	2.173**	1.236	3.820	2.194**	1.241	3.879	1.670	0.854	3.265
ᾱ = (0, 1, 1)	1.116	0.965	1.292	1.115	0.963	1.291	1.134	0.966	1.332
ᾱ = (1, 0, 0)	1.146**	1.037	1.267	1.151**	1.040	1.272	0.984	0.873	1.109
ᾱ = (1, 0, 1)	1.206**	1.064	1.368	1.209**	1.066	1.372	1.025	0.882	1.191
ᾱ = (1, 1, 0)	1.265**	1.172	1.364	1.266**	1.174	1.366	1.078	0.987	1.177
ᾱ = (1, 1, 1)	0.984	0.956	1.013	0.984	0.956	1.013	0.983	0.948	1.019
ᾱ = (0, 0, 1)							1.090	0.967	1.229
ᾱ = (0, 1, 0)							0.603	0.220	1.658
ᾱ = (0, 1, 1)							1.028	0.829	1.276
ᾱ = (1, 0, 0)							0.757**	0.639	0.896
ᾱ = (1, 0, 1)							0.741**	0.595	0.924
ᾱ = (1, 1, 0)							0.748**	0.660	0.846
ᾱ = (1, 1, 1)							0.997	0.958	1.038
No. Obs.	199,682			199,682			199,682
QIC	3.518			2.953			3.107
BIC	−1.952e+06			−1.85E+06			−1.82E+06
log-likelihood	−351214			−294827			−310192
p	<0.001			<0.001			0.0563
Expected no. of zeros+	108,800			139,686			139,167

All regressions adjust for propensity score and marginal structural weights. For the ZIP, the top panel indicates the Poisson model for the counts, while the bottom panel indicates the parameters for the any/no outcome component of the model using a logistic function.

⁺Actual number of zeros is 138,901.

^*p<.01,

^**p<.001 based on Wald tests