31
Who is taking benefit from advances in cancer treatment? Evidence from the French labor market Luis Sagaon-Teyssier UMR912 SESSTIM September 28 th 2015 Research funded by: National Cancer Institute (INCa) and Cancer Research Association (ARC), Programme Cancer: maintien dans l’emploi et retour au travail

Who is taking benefit from advances in cancer treatment?lesdonnees.e-cancer.fr/index.php/content/download/2469/176713/file... · Who is taking benefit from advances in cancer treatment?

Embed Size (px)

Citation preview

Who is taking benefit from

advances in cancer treatment?

Evidence from the French labor market

Luis Sagaon-Teyssier

UMR912 SESSTIM

September 28th 2015

Research funded by:

National Cancer Institute (INCa) and Cancer

Research Association (ARC), Programme Cancer:

maintien dans l’emploi et retour au travail

Outline

• Introduction

• Motivation

• Objectives

• Datasets and data description

• Methods

• Results

• Concluding remarks

IntroductionCancer, transitory or permanent shock on occupational status?

Major advances in cancer treatments• Survival rates and quality-of-life (QoL) improve• Half of adult cancer survivors < 65 years [de Boer et al, JAMA, 2009]

• In France (2005), mean age (men/women) at diagnosis: 67/64 years [INCa, 2010]

New opportunities to participate to the labor forceReturn to work/job tenure, component of cancer survivors QoL

Cancer survivors are more likely to be unemployed Rate of job tenure after cancer diagnosis

From 24% (3 months) to 75% (5 years)Varies with clinical (cancer site, treatment), demographic (gender, age), but

also social (occupational status, educational level, SES) characteristics[Bradley et al, JHE, 2005; Bradley et al, Psycho-Onco, 2002; de Boer et al, JAMA, 2009]

Impact of cancer on job tenure depends onProductivity loss (limitations in functional/psychological abilities)Adjustment of workstation and general working conditions

[van der Wouden et al, J Occup Med, 1992]

Social representations (discrimination, stigma, self-esteem)[Rothstein et al, Oncology, 1995; Paraponaris et al, Health Policy, 2011]

Motivation• Results from the first survey (ALD1)

– Deleterious impact of cancer and usual events in the job market to be disentangled

In what way does cancer induce job insecurity?

Most often, employment to employment transitions only considered

Non-employment transitions (unemployed, retired, other non-working)

neglected

[Joutard et al, Ann. Eco. Stat., 2012]

– Possible confounding effects between SES, cancer prognosis and treatment after-

effects

In France, gender inequalities in job tenure due to individual characteristics (age,

marital status) rather than clinical characteristics of cancer and its treatment

[Marino et al, JCO, 2013]

• Need of understanding the evolution of the labor market

situation among cancer survivors

Objectives

• To investigate how mobility in the labor marketevolved among cancer survivors two years afterdiagnosis in 2002 and 2010

• To investigate how situation in the labor marketevolved with respect to general population

• To disentangle sociodemographic, socioeconomicand cancer-related effects

Datasets

• ALD1: retrospective survey of cancer survivors two years

after diagnosis in 2002 (InCA, SESSTIM-Inserm, CNAM, RSI, MSA)

• VICAN2: retrospective survey of cancer survivors two

years after diagnosis in 2010 (InCA, SESSTIM-Inserm, CNAM, RSI, MSA)

• Contemporaneous French Labor Force Surveys (Insee)

Methods

1. Double Propensity Score Matching (PSM)

– ALD1 (treatment) and VICAN2 (Control)

– ALD1 (treatment) and LFS 2002 (Control)

– VICAN2 (treatment) and LFS 2010 (Control)

• Probit(cas=1) on socioeconomic, sociodémographic and cancer prognosis at diagnosis

2. Continuous-time Markov matrices

– Transition probabilities between three states in the labor market:

• employment,

• non-employment (unemployment and inactivity),

• retirement

Subsample selection

ALD1 database

N= 1,304

VICAN2 database

N= 2,166

ALD1 database

N= 4270

VICAN2 database

N= 4349

Excluded: >57 y.o. at diagnosis, cancer sites not appearing in one of the

datasets, and long-term sick-leave

Not comprable subsamples!

ALD1 & Vican2: unbalanced samples

ALD1/VICAN2: 28.5% / 27.6% ALD1/VICAN2: 71.5% / 72.4%

Patients in ALD1 are older than those in VICAN2 (48 vs 45)

90,9

82,5

4,3

9,4

4,8

8,1

0% 20% 40% 60% 80% 100%

ALD1

VICAN2

Health insurance regime

CNAM

MSA

RSI

61,9

54,5

35,5

40,2

2,6

5,4

0% 20% 40% 60% 80% 100%

ALD1

VICAN2

Socioeconomic status (SES)

CSP-

CSP+

Jamais travaillé

SES-

SES+

Never worked

ALD1 & Vican2: unbalanced samples

80,4

82,3

16,5

16,5

3,1

1,1

0% 20% 40% 60% 80% 100%

ALD1

VICAN2

Activity status at diagnosis

Emploi

Chômage

Inactivité

0 20 40 60

Sein

Poumon

Côlon-rectum

Prostate

VADS

Vessie

Rein

Thyroïde

Lymphome

Mélanome

Corps de l'utérus

Col de l'utérus Tumor site

VICAN2

ALD1

Metastasis:

ALD1/VICAN2: 24.3% / 25.3%

Employment

Unemploy.

Inactivity

Subsample selection: balancing

ALD1

N= 4270

VICAN2

N= 4349

Excluded: >57 y.o. at diagnosis, cancer sites not appearing in one of the

datasets, and long-term sick-leave

ALD1

N= 1084

VICAN2

N= 1084

ALD1

N= 1304

VICAN2

N= 2166

1st. Propensity score matching:

Gender, age, health insurance scheme, education level, SES, activity status,

urban/rural, tumor site, stage of tumor

Cancer subsamples are balanced (e.g. comparable)

Subsample selection: balancing cancer

and LFS surveys

ALD1

N= 1084

VICAN2

N= 1084

ALD1

N= 1081

LFS 2002

N= 1081

VICAN2

N= 1058

LFS 2010

N= 1058

2nd Propensity score matching:

Gender, age, health insurance scheme,

education level, SES, activity status, urban/rural

2nd Propensity score matching:

Gender, age, health insurance scheme, education

level, SES, activity status, urban/rural

Final sample

N=4220

ALD1-LFS and VICAN2-LFS (After PSM)

• 1081 ALD1 when matching with LFS (3 cases lost)

• 1058 VICAN2 when matching with LFS (26 cases lost)

Matched:

The structure obtained after the first matching (ALD1-VICAN2) ispreserved when matching with the respective LFS

Unmatched (ALD1): n=3

54 y.o., women, never worked, <high school, unemployed, urban

Unmached (VICAN2): n=26

50 y.o., women (85%), never worked (85%), < high school, inactive, urban

Results:

Continuous-time Markov

matrices:

Transition probabilities between

Employment, non-employment, retirement

Employment-Employment

(Employment tenure)

Employment tenure: survey

Survey

.93

ALD1

LFS

.85

.94

.81

LFS

VICAN2

2002-2004

2010-2012

Δ=-.12

Δ=-.09

Δ=.04

Employment tenure: gender

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

Δ=-.12

Δ=-.09

Δ=.04

Gender

.96 men

.80 men

.83 men

.97 men

Δ=-.16

Δ=-.14

Δ=.03

Employment tenure: gender

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

Δ=-.12

Δ=-.09

Δ=.04

Gender

.92women

.81women

.86 women

.93women

Δ=-.11

Δ=-.07

Δ=.05

Employment tenure: SES

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

Δ=-.12

Δ=-.09

Δ=.04

SES

.93 Blue collar

.80 Blue collar

.81 Blue collar

.93 Blue collar

Δ=-.13

Δ=-.12

Δ=.01

Employment tenure: SES

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

Δ=-.12

Δ=-.09

Δ=.04

SES

.94

.82

.91

.95

Δ=-.12

Δ=-.04

White collar

White collar

White collar

White collar

Δ=.09

Employment tenure: SES & Prognosis

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

Δ=-.12

Δ=-.09

Δ=.04

SES

.93 Blue collar

.80 Blue collar

.81 Blue collar

.93 Blue collar

Δ=-.13

Δ=-.12

Δ=.01

.82 good

bad

.87 good

bad

.73

.66

Prognosis

Δ=-.11

Δ=-.09

Good: Δ=.05

Bad: Δ=-.07

Employment tenure: SES & Prognosis

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

Δ=-.12

Δ=-.09

Δ=.04

SES

.94

.82

.91

.95

Δ=-.12

Δ=-.04

White collar

White collar

White collar

White collar

Δ=.09

.85 good

bad

.92 good

bad

.82

.90

Prognosis

Δ=-.02

Δ=-.03

Good: Δ=.07

Bad: Δ=.08

Non-employment to employment

(Access into employment)

Access to employment: survey

Survey

.17

ALD1

LFS

.20

.26

.10

LFS

VICAN2

2002-2004

2010-2012

Δ=-.07

Δ=-.06

Δ=.10

Access to employment: gender

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

Δ=-.07

Δ=-.06

Δ=.10

Gender

.16 men

.09 men

.18 men

.20 men

Δ=-.05

Δ=-.02

Δ=.09

Access to employment: gender

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

Gender

.15women

.09women

.20 women

.24women

Δ=-.06

Δ=-.04

Δ=.11

Δ=-.07

Δ=-.06

Δ=.10

Access to employment: SES

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

SES

.17 Blue collar

.11 Blue collar

.12 Blue collar

.24 Blue collar

Δ=-.06

Δ=-.12

Δ=.01

Δ=-.07

Δ=-.06

Δ=.10

Access to employment: SES

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

SES

.16

.07

.39

.31

Δ=-.09

Δ=.08

White collar

White collar

White collar

White collar

Δ=. 32

Δ=-.07

Δ=-.06

Δ=.10

Access to employment: SES & Prognosis

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

SES

.17 Blue collar

.11 Blue collar

.12 Blue collar

.24 Blue collar

Δ=-.06

Δ=-.12

Δ=.01

Δ=-.07

Δ=-.06

Δ=.10

.15 good

bad

.22 good

bad

.07

.002

Prognosis

Δ=-.21

Δ=-.08

Good: Δ=.07

Bad: Δ=-.06

Access to employment: SES & Prognosis

Survey

ALD1

LFS

LFS

VICAN2

2002-2004

2010-2012

SES

.16

.07

.39

.31

Δ=-.09

Δ=.08

White collar

White collar

White collar

White collar

Δ=. 32

Δ=-.07

Δ=-.06

Δ=.10

.25 good

bad

.50 good

bad

.04

.40

Δ=-.10

Δ=-.21

Good: Δ=.25

Bad: Δ=.36

Concluding remarksIn general, better situation for cancer survivors: gaps with respect to general population remain similar

No gender differences in terms of employment tenure,

Persistent occupation differences accentuated by prognosis: unfavorable for blue collar

– Employment tenure depends strongly on the employment type

– Cancer severity accentuate the differences in terms of return-to-work between SES

– Acces to employment and employment tenure remains possible for white collareven with bad prognosis

– Big firms are protective of patients already employed, but barriers to access intoemployment exist in these kind of firms.

Strengths:

• Markov matrices estimation is a useful first step in developping suitable models

• Offers a landscape of the labor market situation of cancer survivors

Limits:

• Estimations conditional to survival

• Lost of information about situations between diagnosis and interview