Table 4.

Gender-specific annual mean expenditures within the case group adjusted for age

	Service component	Male dementia patients (n = 2,328)			Female dementia patients (n = 6,819)			p-values
		Users (as a %)	Cost per user (in €)	Cost per patient (in €)	Users (as a %)	Cost per user (in €)	Cost per patient (in €)	Probability of service use	Cost per user
Approach 1 one-step model	Costs of formal care*	100.0	12,648	12,648	100.0	12,209	12,209	/	0.17
	Health insurance expenditures*	100.0	6,465	6,465	100.0	5,558	5,558	/	0.002
	Medication	99.4	1,394	1,394	99.5	1,277	1,277	/	0.12
	including anti-dementia drugs	18.8	782	127	14.0	770	115	0.12	0.70
	General practitioner	100.0	650	650	99.5	638	638	/	0.18
Approach 2 two-part- model	Medical specialist	88.2	675	595	79.1	447	363	<0.0001	<0.0001
	Hospital treatment	46.3	5,959	2,790	41.0	5,308	2,157	0.0005	0.01
	Non-physician services	28.3	839	225	26.5	677	190	0.25	0.03
	Medical aids	56.9	508	301	68.8	534	349	<0.0001	0.05
	Home health care	14.3	1,761	266	17.9	2,461	392	<0.0001	<0.0001
	Rehabilitation	5.2	3,307	170	4.6	3,459	164	0.81	0.21
	Long-term care services*	59.3	9,027	5,935	68.3	10,267	6,467	0.065	<0.0001

Per user data derive from one-step and two-step Generalized Linear Models with gender as the coefficient of interest.

Patient level data are means based on recycled predictions with gender as the coefficient of interest.

Significance at the patient level is estimated based on p1 and p2 in the two-part models and based on p2 in the one-step models.

p1 derives from the logistic model (approach 2, step 1) and p2 derives from the gamma model (approach 2, step 2 and approach 1).

* Results of model estimation; the addition of mean costs per category yields slightly different figures.