Table 4 Sensitivity of meta-analyzed childhood predictors to unmeasured confounding

From: A cross-national analysis of childhood predictors of daily smoking in adulthood

  

Total sample (N = 202,898)

Smoker sample (N = 38,290)

  

Binary

Continuous

Variable

Category

E-value

E-value limit

E-value

E-value limit

Relationship with mother

(Ref: very bad/somewhat bad)

    

Very good/somewhat good

1.31

1.12

1.31

1.00

Relationship with father

(Ref: very bad/somewhat bad)

    

Very good/somewhat good

1.32

1.11

1.34

1.00

Parental marital status

(Ref: parents married)

    

Divorced

1.84

1.68

1.29

1.00

Single, never married

1.59

1.31

1.05

1.00

One or both parents had died

1.42

1.20

1.04

1.00

Subjective financial status of family growing up

(Ref: got by)

    

Lived comfortably

1.07

1.00

1.22

1.00

Found it difficult

1.13

1.00

1.13

1.00

Found it very difficult

1.31

1.00

1.38

1.00

Abuse

(Ref: no)

    

Yes

1.84

1.66

1.39

1.05

Outsider growing up

(Ref: no)

    

Yes

1.70

1.53

1.19

1.00

Self-rated health growing up

(Ref: good)

    

Excellent

1.22

1.00

1.13

1.00

Very good

1.11

1.00

1.14

1.00

Fair

1.20

1.00

1.28

1.00

Poor

3.26

1.00

1.25

1.00

Immigration status

(Ref: born in this country)

    

Born in another country

1.45

1.00

1.28

1.00

Age 12 religious service attendance

(Ref: never)

    

At least 1/week

1.24

1.00

1.55

1.30

1–3/month

1.29

1.00

1.73

1.50

<1/month

1.05

1.00

1.47

1.24

Age; year of birth

(Ref: age 18–24; 1998–2005)

    

age 25–34; 1988–1998

2.13

1.85

1.89

1.69

age 35–44; 1978–1988

2.46

2.05

2.62

2.22

age 45–54; 1968–1978

2.39

1.95

2.85

2.37

age 55–64; 1958–1968

2.44

1.93

3.19

2.56

age 65–74; 1948–1958

1.54

1.00

3.04

2.38

age 75–84; 1938–1948

8.46

1.00

2.14

1.38

85 or older; 1938 or earliera

55.48

4.83

1.91

1.00

Gender

(Ref: male)

    

Female

5.77

3.26

2.06

1.55

Othera

18.20

1.00

3.10

2.09

  1. aGroup is very small (<0.1% of the observed sample) within several countries leading to high uncertainty in this estimate—be cautious about interpreting this estimate; E-value is the minimum strength of the association an unmeasured confounder must have with both the outcome and the predictor, above and beyond all measured covariates, for an unmeasured confounder to explain away an association47.
Back to article page