Table 2.

Results of generalized linear models estimating likelihood of aggregate joint trajectory group membership.

	Aggregate Group 1 Medicaid stops; Cash pay minimal n = 325 (17%) RR	Aggregate Group 2 Medicaid drops; Cash pay minimal n = 948 (49%) RR	Aggregate Group 3 Medicaid stops; Cash pay increases n = 264 (14%) RR	Aggregate Group 4 Medicaid drops; Cash pay increases n = 399 (21%) RR
Age
18–30 (reference)
31–40	0.68**	1.38**	0.89	0.91
41–50	0.48**	1.58**	0.56**	1.06
51–64	0.55**	1.63**	0.57**	0.90
Female	0.84	1.07	0.92	1.04
Nonwhite	1.26*	0.91	0.97	1.05
Metropolitan residencea	1.12	0.91	1.28	0.97
Baseline clinical diagnoses
Substance use disorder	1.29*	0.87**	1.12	1.00
Chronic pain	0.63**	1.07	1.43	1.31
Anxiety disorder	0.77*	1.01	0.90	1.29**
Depression	0.67**	1.02	1.22	1.14
Other mental illness	1.01	0.99	1.27	0.87
Charlson comorbidity indexb
0 (reference)
1	0.95	1.02	0.94	1.05
2 or more	0.72*	1.05	0.80	1.21
LIP eligibility
Opioid Rx fills	0.91	0.94	0.91	1.32
Benzodiazepine Rx fills	0.35	0.92	0.91	1.76**
Pharmacy shopping	1.25	0.85**	1.32	1.16
No criteria metc	1.68*	0.78	1.09	0.93

^**Note. p < .01,

^*p < .05; RR: Relative risk; Rx: Prescription. MxCx group notations indicate a joint trajectory group, which pairs a Medicaid-covered opioid and benzodiazepine fill trajectory group (M1 -M4) with a cash pay opioid and benzodiazepine fill trajectory group (C1 -C4) depicted in Figure 1. Relative risks were estimated in each model using modified Poisson generalized linear models estimating the association of patient characteristics with membership in an aggregate joint trajectory group, compared to all other aggregate joint trajectory groups.

^aResidence in a county with a Rural-Urban Continuum Code designation of 1 −3.

^bCharlson scores were calculated using the Quan modification.¹⁴

^cSubject did not meet any of the three LIP eligibility criteria based on claims data review. We assumed these subjects were referred for LIP enrollment by a healthcare provider.