Skip to main content

Table 4 Estimates of the health effect of unemployment on health for different variable setups for the logistic regression, G-computation and inverse propensity weighting (IPW) estimators for self-reported long-term unemployment

From: Methodological perspectives on the study of the health effects of unemployment – reviewing the mode of unemployment, the statistical analysis method and the role of confounding factors

Model

Unemployment during

follow-upa

Risk difference

G-computation

IPW standard

IPW augmented

IPW

doubly robust

Model 1 (full)b

No censor

0.134*

0.129*

0.128*

0.125*

Censor

0.111*

0.114*

0.108

0.124*

Model 2:

model 1—education level

No censor

0.126*

0.130*

0.130*

0.129*

Censor

0.103*

0.117

0.115

0.126*

Model 3:

model 1 – marital status

No censor

0.138*

0.136*

0.136*

0.133*

Censor

0.111*

0.116

0.109

0.131*

Model 4:

model 1—previous health

No censor

0.148*

0.145*

0.142*

0.143*

Censor

0.117*

0.119*

0.113

0.125*

Model 5:

model 1—occupation

No censor

0.136*

0.127*

0.128*

0.125*

Censor

0.111*

0.116*

0.110*

0.120*

Model 6:

model 1—sex

No censor

0.133*

0.135*

0.136*

0.129*

Censor

0.109*

0.121*

0.122

0.122*

Model 7:

model 1 – AVAT

No censor

0.134*

0.128*

0.127*

0.122*

Censor

0.112*

0.105*

0.099*

0.107*

Model 8:

model 1 – AVSI

No censor

0.132*

0.132*

0.129*

0.127*

Censor

0.109*

0.110*

0.103

0.118*

Model 9:

model 1—cash margin

No censor

0.133*

0.120*

0.118*

0.117*

Censor

0.110*

0.104*

0.096

0.115*

Model 10:

model 1 – smoking

No censor

0.137*

0.126*

0.125*

0.121*

Censor

0.115*

0.115*

0.108*

0.120*

Model 11:

model 1—alcohol intake

No censor

0.129*

0.125*

0.128*

0.125*

Censor

0.107*

0.111*

0.111*

0.124*

Model 12:

model 1—obesity

No censor

0.135*

0.132*

0.131*

0.130*

Censor

0.115*

0.118

0.115

0.125*

Model 13: significant terms in full modelc

No censor

0.129*

0.123*

0.123*

0.120*

Censor

0.113*

0.103*

0.103*

0.101*

Model 14: model 13—education level

No censor

0.121*

0.127*

0.125*

0.126*

Censor

0.105*

0.107*

0.107*

0.107*

Model 15: model 13 – marital status

No censor

0.136*

0.133*

0.133*

0.130*

Censor

0.116*

0.107*

0.106*

0.105*

Model 16: model 13 – previous health

No censor

0.159*

0.154*

0.152*

0.153*

Censor

0.132*

0.121*

0.121*

0.120*

Model 17: model 13—occupation

No censor

0.136*

0.123*

0.123*

0.122*

Censor

0.116*

0.105*

0.104*

0.102*

Model 18: model 13 + sex

No censor

0.130*

0.118*

0.117*

0.116*

Censor

0.114*

0.094

0.093

0.096

Model 19: model 13 + AVAT

No censor

0.129*

0.123*

0.122*

0.121*

Censor

0.112*

0.107*

0.107*

0.107*

Model 20: model 13 + AVSI

No censor

0.131*

0.122*

0.122*

0.118*

Censor

0.115*

0.105*

0.105*

0.102*

Model 21: model 13 + cash margin

No censor

0.131*

0.132*

0.132*

0.130*

Censor

0.115*

0.115*

0.114*

0.111*

Model 22: model 13 + smoking

No censor

0.126*

0.126*

0.128*

0.126*

Censor

0.109*

0.106*

0.107*

0.106*

Model 23: model 13 + alcohol intake

No censor

0.132*

0.127*

0.126*

0.122*

Censor

0.115*

0.105*

0.105*

0.102*

Model 24: model 13 + obesity

No censor

0.129*

0.122*

0.122*

0.118*

Censor

0.110*

0.097

0.097

0.096

  1. Note: “- “ means that this variable was excluded in analyses compared to either model 1 or 13 and “ + ” means that the variable was added to the significant variables in the analyses
  2. * p < 0.05 based on the 2.5% and 97.5 Bootstrap percentiles
  3. a Participants were censored (n = 620) or not censored (n = 805) for unemployment during the follow-up period between 1995 and 2007. b Full model controlling for alcohol intake, Availability of Attachment (AVAT), Availability of Social Integration (AVSI), body mass index, cash margin, education level, gender, marital status, previous health status, occupation and smoking.
  4. c The significant variables in model 13–24 were education level, marital status, previous health status and occupation. IPW = inverse probability weighting estimator. Estimates represent the health effect on unemployed compared to employed individuals where estimate above 0 is poorer health for unemployed
Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us