Regional variation in healthcare spending and mortality among senior high-cost healthcare users in Ontario, Canada: a retrospective matched cohort study

Table 3 Regression results: total public healthcare expenditures (adjusted, log scale)

Total costs
Variables	HCU#	Non-HCU&
	Coefficient (SE)	Coefficient (SE)
Variance in probability of incurring costs, σ2u1		0.065 (0.026)*
Variance in costs once incurred, σ2u2	0.0009 (0.0004)*	0.001 (0.001)*
Covariance between σ2u1 and σ2u2		0.006 (0.003)
Probability (costs≠0)
Intercept		4.49 (0.16)*
Age		−0.067 (0.002)*
Sex, M		− 0.205 (0.03)*
ADG		1.018 (0.009)*
Low income status		−0.129 (0.035)*
Costs > 0
Intercept	9.74 (0.02)*	5.946 (0.025)*
Age	0.008 (0.0002)*	0.016 (0.001)*
Sex, M	0.064 (0.003)*	0.044 (0.005)*
ADG	−0.011 (0.0004)*	0.081 (0.001)*
Low income status	0.018 (0.004)*	0.134 (0.006)*
log_theta	0.788 (0.003)*	0.473 (0.004)*
R² (part 1), %		87.0%
R² (part 2), %	1.6%	19.7%
LRT (Chi2 dist, p < 0.05)	2333.0	88,500.57

& - Estimated through GLMM; # - Estimated through a mixed effects two-part model; * - estimates were statistically significant at p < .05; ADGs- Aggregated Diagnosis Groups; HCU- high-cost user; LHIN – Local Health Integrated Network; Log-theta- the logarithm of the shape parameter of gamma distribution; LRT- likelihood ratio test; R²- coefficients of determination; SE- Standard Error

ISSN: 1471-2318