Return to work after medical rehabilitation in Germany: influence of individual factors and regional labour market based on administrative data

Journal for Labour Market Research

Table 4 Fractional logit regression models for RTW in psychosomatics

Predictors	M1		M2 (M1 + personal and rehabilitation biography)		M3 (M2 + employment biography)		M4 (M3 + interaction)
Predictors	b {AME}	s.e	b {AME}	s.e	b {AME}	s.e	b {AME}	s.e
Unemployment rate (UR)	− 0.019 ^***	0.004	− 0.025 ^***	0.005	− 0.002	0.004	− 0.011 ^*	0.005
	{ − 2.2}		{ − 2.2}		{ 0.0}
Prerehabilitation employment status [not employed, reference: employed]					− 1.229 ^*** { 209.9}	0.018	− 1.341 ^*** { 209.9}	0.042
UR * [not employed]							0.018 ^**	0.006
{AME, subgroup „employed “}							{ − 0.5}
{AME, subgroup „not employed “}							{ 0.6}
Random effects
τ_{00, labour market region}	0.01		0.01		< 0.01		< 0.01
τ_{00, rehabilitation departments}	0.13		0.12		0.04		0.04
Pseudo-R²	0.029		0.206		0.386		0.386
AIC	157,162		138,530		117,661		117,655

Method is cross-classified fractional logit regression with n _{labour market region} = 257, n _{rehabilitation departments} = 202, n _patients = 117,386, * p < 0.05, ** p < 0.01, *** p < 0.001; estimators for intercept, for personal / rehabilitation biography predictors and employment biography predictors in Additional file 1
M model, b coefficients, s.e. standard error, AME average marginal effects (in days), τ₀₀ variance component of labour market region or rehabilitation department, R² square of the correlation between the model’s predicted values and the actual values, AIC Akaike-criterion