A complex intervention to reduce hospital admissions for people living with dementia in shared-housing arrangements in Germany: results of the multicenter, cluster-randomized controlled DemWG-study

Table 5 Extended GEE-Poisson model for the primary endpoint with hierarchical random effect

	Number of hospital admissions at t1			Number of hospital admissions at t2
*Predictors (at t0)*	*Estimates*	*95% CI*	p	*Estimates*	*95% CI*	p
(Intercept)	0.05	0.01–0.45	0.007	0.00	0.00–0.02	< 0.001
Number of hospital admissions	1.24	0.94–1.64	0.129	0.98	0.49–2.00	0.965
Intervention group (reference: control group)	0.47	0.22–0.99	0.048	0.70	0.25–1.98	0.498
Care levels 1 & 2 (reference: care level 3)	0.53	0.24–1.15	0.109	1.29	0.55–3.05	0.558
Care levels 4 & 5 (reference: care level 3)	0.70	0.35–1.39	0.313	0.71	0.23–2.19	0.552
Number of general practitioners per participant	0.84	0.06–11.44	0.895	79.56	2.59–2444.25	0.012
Proportion of voluntary social year staff and volunteers per participant	0.33	0.11–1.05	0.060	0.72	0.19–2.71	0.631
MNA®-SF total score	1.03	0.89–1.19	0.701	1.29	0.98–1.69	0.068
MMSE total score	1.00	0.95–1.06	0.942	1.00	0.92–1.09	0.993
Number of permanently prescribed medications	1.23	1.12–1.36	< 0.001	1.24	1.08–1.42	0.002

Care level was included in the model as a categorical variable, with care level 3 as the reference category, No care level, care level 1, and care level 2 were merged into one group because the terms “no care level” and “care level 1” occurred comparatively rarely and would otherwise have led to a convergence problem of the used models; SHA: Shared-housing arrangement, Number of general practitioners for the SHA: Number of general practitioners who are responsible for the different residents of the relevant SHA, MNA®-SF: German version of the Mini Nutritional Assessment – Short Form, normal nutritional status defined as ≥ 12 points and a possible under nutritional status ≤ 11; MMSE: Mini-Mental State Examination, Range: 0–30, lower scores indicate higher cognitive impairment, σ² = variance component of the residual; τ00 nursing service = variance component of the nursing service; τ00 SHA = variance component of the SHA; ICC = intraclass correlation; nservice = number of different nursing services; nSHA = number of different SHA; nobserv = number of observations; Random effects for t1: σ² = 1.92, τ00SHA = 0.29, τ00 nursing service = 0.40, ICC = 0.26, nservice = 52, nSHA = 78, nobserv = 236; Random effects for t2: σ2 = 2.66, τ00 nursing service = 0.73, τ00 SHA = 0.00, nservice = 43, nSHA = 65, nobserv = 168, Marginal R2/Conditional R2 = 0.332/NA

ISSN: 1741-7015