Work and Mental Health

Preview:

DESCRIPTION

- An Analysis of Canadian Community Health Survey. Work and Mental Health. Miao Fang McMaster University May 13, 2005. Outline. Introduction Data, Sampling and Variables Analyses Results Discussion and Conclusion. Background. - PowerPoint PPT Presentation

Citation preview

Work and Mental Health - An Analysis of Canadian Community Health Survey

Miao Fang

McMaster University

May 13, 2005

Outline

Introduction Data, Sampling and Variables Analyses Results Discussion and Conclusion

Background Over 30% of Canadians reported that most days

at work were quite a bit or extremely stressful.

12% of Canadians aged from 15 to 64 suffer from a mental disorder or substance dependence.

The estimated cost of poor mental health in workplace is in billions of dollars.

Work Poor mental health disability and loss of productivity in workplace

Objective

1. Describe the relationship between work and mental health.

2. Describe the relationship between work and mental health care use for people with mental disorders and substance dependences.

3. Further explore covariates and interactions of any relationships.

Data Canadian Community Health Survey

CCHS1.1 Response rate: 84.7% Sample size: 131,535 (Ontario workers:23,110)

CCHS1.2 Response rate: 77.0% Sample size: 36,984 (Ontario workers: 8,008)

Sampling

Complex survey – multistage stratified cluster design

All estimates were weighed to represent the target Ontario workers population

Variables Dependent variables - Mental disorders and substance dependences (MDSD) CCHS1.1 – Depression (0,1) CCHS1.2 – Any mental disorder or substance dependence (AMDSD) (0,1)

Mental disorders Substance dependences Major depressive episode Alcohol dependence Manic episode Illicit drug dependence Panic disorder Social phobia Agoraphobia

Variables- Mental health care utilization CCHS1.1 – Consultation with a m.h professional (0,1) CCHS1.2 – Utilization of any resource (for mental health)

(0,1)

Main predictors (exposures) Work stressors (0-48)

Covariates Age, sex, BMI, race, marital status, education, income,

type of smokers (all are categorical variables).

Analyses

Descriptive Analysis

Bivariate analysis

Logistic regression analysis

Descriptive Analysis

Two-sample t test

01010 : : XXHXXH

2 where~

01

20

212

)2(

0

2

1

2

0101

nn

SSSt

n

s

n

s

xxT pnn

pp

group outcome theofeach within normal app.

)1,0(~

X

Y

Contingency Table Analysis

Pearson Chi-Squared test for independence

0

2

)1)(1(1 1

2

2 Hunder ~

cr

r

i

c

j ij

ijij

E

EOX

OOOE jiij /

Measuring Association The coefficient of contingency

Goodman and Kruskal’s Gamma - test for trend

)/( 22 NXXC

)/()( QPQPG

P - concordant pairs Q – discordant pairsP - concordant pairs Q – discordant pairs

Multiple Logistic Regression Model

),...,,( 21 pxxxx )1,0(~Y

pp

k

ljljl

j

pp

xDxx

kj

e

e

xxx

j

1

122110

th

g

)g(

22110

)g(

levels, has variable If

1

-1 log)(g

x

x

x

xx

x

x

)|1( xx YP

Point Estimation of Coefficient

)(ˆ,ˆObtain

.,1,2, ,0)( , 0)(

:equations likelihood 1

)(1ln)1(ln)(ln)(

)](1[)()(

),,( nsobservatiot independen

11

1

1

1

i

n

iiiij

n

iii

n

iiiii

yi

yi

n

i

ii

pjyxy

p

yylL

l

yn

ii

xx

xxββ

xxβ

x

Variance Estimation

Maximum likelihood estimation

ii

n

iij

j

xL

11

2

2

2

βββ

βIβ

ˆat )Var( evaluatingby obtained are )ˆr(aV

)()(Var

1

1

1

2

iiil

n

iij

lj

xxL

Variance Estimation

Bootstrap method

2

1

ˆˆ1ˆˆ

B

bbBOOT B

V

BOOTVARE_V21.SPS

1) Calculate the point estimate using the final weight.

2) Calculate B estimates using the B bootstrap

weights.

3) Calculate the variance of the B estimates.

Assessing the Fit of the Model- Hosmer-Lemeshow goodness of fit test

To calculate the test statistic Order the fitted values Group the fitted values into g classes of

roughly equal size Calculate the observed and expected number

in each group Perform a goodness of fit test

)exp(1

)exp()(:

)exp(1

)exp()(: 10

x

x

x

x

YEHYEH

2

22

1

1

0

2

2 ~

g

g

j k jk

jkjk

E

EOX

Work Stressors (0 – 48)

Mean=19.4 (sd=5.0)

4239363330272421181512963

Perc

en

t

10

8

6

4

2

0

Mean=19.1 (sd=5.2)

39363330272421181512963

Perc

en

t

10

8

6

4

2

0

CCHS1.1CCHS1.1 CCHS1.2CCHS1.2

Work stressors in different depression groups

Work stressors (Depression=1)

42.5

40.0

37.5

35.0

32.5

30.0

27.5

25.0

22.5

20.0

17.5

15.0

12.5

10.0

7.5

5.0

CCHS1.1

Frequency

400

300

200

100

0

Std. Dev = 5.58

Mean = 21.6

N = 1676.82

Work stressors (Depression=0)

50.045.040.035.030.025.020.015.010.05.00.0

CCHS1.1

Fre

quency

10000

8000

6000

4000

2000

0

Std. Dev = 4.92

Mean = 19.2

N = 21127.72

Work Stressors by MDSD

depression mean mean diff. t Sig.

work stressors

Yes 21.58 2.35 16.74 <.01 No 19.23

AMDSD mean mean diff. t Sig.

work stressors

Yes 21.51 2.69 14.94 <.01 No 18.82

CCHS1.1

CCHS1.2

Other variables by MDSD

Depression AMDSD C G C G

Age 0.07 -0.20 0.13 -0.32Sex 0.10 - 0.03 -Marital status 0.09 - 0.18 -Education 0.03 -0.08 0.11 -0.16Income 0.06 -0.16 0.02 -0.01Race 0.03 - 0.07 -BMI 0.03 (-0.00) 0.05 (0.04)Type of smoker 0.09 -

All values are significant at 0.05 level except those in parentheses.

An Example: Logistic Regression Model for AMDSD

Variable Coeff. S.E. OR95% CI for OR

Sig. Lower Upper

Constant -4.70 0.23 0.01 <.01

Work Stressors 0.09 0.01 1.09 1.07 1.10

<.01

15-29 years old 0 1

<.0130-44 years old -0.02 0.11 0.98 0.79 1.22

45-64 years old -0.38 0.13 0.69 0.53 0.88

65 years old or more -0.68 0.38 0.51 0.24 1.06

Less than s.s. grad. 0 1

<.01S.s grad, no post-sec 0.02 0.12 1.02 0.81 1.28

Some post-sec edu 0.53 0.13 1.69 1.33 2.17

Post-sec deg/diploma

-0.04 0.10 0.97 0.79 1.18

Variable Coeff. S.E. OR95% CI for OR

Sig. Lower Upper

Constant -4.70 0.23 0.01 <.01

Work Stressors 0.09 0.01 1.09 1.07 1.10

<.01

15-29 years old 0 1

<.0130-44 years old -0.02 0.11 0.98 0.79 1.22

45-64 years old -0.38 0.13 0.69 0.53 0.88

65 years old or more -0.68 0.38 0.51 0.24 1.06

Less than s.s. grad. 0 1

<.01S.s grad, no post-sec 0.02 0.12 1.02 0.81 1.28

Some post-sec edu 0.53 0.13 1.69 1.33 2.17

Post-sec deg/diploma

-0.04 0.10 0.97 0.79 1.18

Variable Coeff.

S.E. OR95% CI for OR

Sig. Lower Upper

White 0 1 <.01Non-white 0.64 0.11 1.89 1.54 2.32

Married 0 1

<.01Common-law 0.98 0.13 2.68 2.08 3.45

Wid./Sep./Div. 0.98 0.13 2.67 2.07 3.45

Single 0.85 0.11 1.89 1.54 2.32

An Example Model for AMDSD (Continued)

Goodness-of-fit: Hosmer and Lemeshow x2=3.79 on 8 d.f., P=0.88

Comparison of Variance Estimations

Variable SE (1) (unadj.)

SE (2) (bootstrap)

Ratio of SE’s SE(2)/SE(1)

Work stressors

15-29 years old (baseline)

30-44 years old

45-64 years old

65 years old or more

Less than s.s grad.(baseline)

S.s.grad, no post-sec.edu.

Some post-sec.grad.

Post-sec. degree/diploma

0.007

0.11

0.13

0.38

0.12

0.13

0.10

0.009

0.12

0.18

0.71

0.14

0.16

0.15

1.29

1.09

1.38

1.87

1.17

1.23

1.50

Comparison of Variance Estimations (Continued)

Variable SE (1) (unadj.)

SE (2) (bootstrap)

Ratio of SE’s SE(2)/SE(1)

White (baseline)

Non-white

Married (baseline)

Common-law

Wid./Sep./Div.

Single

0.11

0.13

0.13

0.11

0.20

0.23

0.15

0.14

1.82

1.77

1.15

1.27

Summary of Findings

MDSD

Work stressors highly predicted MDSD

Sub-groups at higher risk of MDSD: non-white, women, younger workers,

smokers, higher BMI, not married

Summary of Findings (Cont.)

Mental health care use CCHS1.1 - More likely to consult with a

mental health professional: older worker, women, whites, education

at least to high school graduation CCHS1.2 - More likely to use any

resource: older worker, women, non-whites, lower

work stressors

Conclusion

Work stressors were confirmed as predictive of MDSD.

Identified sub-groups that do not use the health care service for their mental health problems.

Thank you!

Recommended