Time discounting and inflammation in later life

Report
Time discounting and inflammation in later life
Michael Daly, Liam Delaney
Behavioural Science Centre, University of Stirling
Psychology literature
• Established link between conscientiousness & subsequent physiological
function and mortality (Friedman et al., 1993; Hampson et al., 2013)
• Self-discipline facet of conscientiousness particularly predictive of later
health (Costa et al., 2014)
• Self-control and subsequent overweight, poor health, substance use,
physiological functioning including metabolic abnormalities and
inflammation gauged by CRP levels (Duckworth et al., 2010; Moffitt et al.,
2011)
Behavioural Economics literature
• Time discounting has been linked to bmi (Borghans & Golsteyn, 2006; Chabris et al.,
2008; Courtemanche et al., 2011; Dodd, 2008; Golsteyn et al., 2014; Ikeda et al., 2010;
Takada et al., 2011; Weller et al., 2008)
• Similarly, smokers discount the value of delayed money more than comparison groups
(Bickel et al., 1999; Chabris et al., 2008; Fuchs, 1982; Goto et al., 2009)
• Smoking example: Scharff & Viscusi (2011). Current smokers have an implied discount
rate of 13.9%, as compared to 8.1% for non-smokers
• Some evidence of links to other health behaviours such as physical activity and
substance use (Adams, 2009; Chabris et al., 2008; Kirby et al., 1999)
Is time discounting linked to health & biological functioning?
• Proposed that those with higher discount rates invest less in their health (Grossman,
1972)
• Evidence that early life discount rates predicts later health including early mortality
(Chabris et al., 2008; Goldsteyn et al., 2014)
• Opens the question as to the biological variables that mediate the association between
time preferences and health outcome
• Greater discounting of delayed rewards has been shown to correlate positively with
health-related biological variables including systolic blood pressure and cortisol reactivity
(Daly, Delaney, & Harmon, 2009; Lu et al., 2014)
Why inflammation?
• Inflammation is typically known as the body’s protective response to invasion and trauma
such as pathogens or damaged cells resulting from infection, burns, injury
• We are interested in low-grade chronic inflammation implicated in endothelial dysfunction,
plaques/lesions, atherosclerosis and cardiovascular disease-related mortality (Benson et
al., 2008; Libby et al., 2002)
• Cardiovascular disease (e.g. stroke, angina, heart attack, congestive heart failure) is the
leading cause of death worldwide (30% of deaths according to WHO)
• Strongest biological predictors of cardiovascular disease across 61 meta-analyses are Creactive protein (top vs. bottom tertile) (RR: 2.43, 95% confidence interval (ci): 2.10–2.83)
and fibrinogen (1 mg/L increase) (HR: 2.33, 95%ci: 1.91–2.84) (van Holten et al., 2013).
Stronger evidence for a causal role of fibrinogen, CRP may capture inflammation generally
Goal of the current study
• 1) Examine whether time preferences predict inflammation levels
• 2) Test whether a potential link between discounting and health is independent of a
range of important potential confounding variables (e.g. income, wealth, cognitive ability,
conscientiousness)
• 3) Test how much of the discounting – inflammation link is explained by health
behaviours
• 4) Consider what other pathways might explain the relationship
Methodology
• Participants from ELSA (wave 5 in 2010-2011) completed questions relating
to preference for monetary rewards over different time horizons
• The current study utilizes data drawn from three waves of ELSA (wave 4,
2008-2009; wave 5, 2010-2011; wave 6, 2012-2013)
• 1063 participants (50-75y, 55% female, 96.1% response rate) completed the
preference elicitation module
• 389 participants had complete biological data & time preference data
Discounting Elicitation
• Time discounting questions were presented on a laptop computer as part of
series of financial decision-making games involving real rewards
• The delay discounting game was called the ‘rectangle’ game and consisted of
12 binary choice options presented in adjacent rectangles
• Instructions: (1) start the financial module with £10 (2) would win money from
one of the 22 choices (3) randomly selected choice selected by the computer
at the end of the section
• Participants were instructed that the largest loss they could incur was £5 and
max gain £70. They would receive winnings per post at the chosen time.
Average expected payment median £35, mean £38. 13 mins median time
taken (Alan et al., 2012)
Discounting Questions
Participants make a choice between
a one-off payment of a smaller
sooner amount in two weeks or a
later larger in one month’s [or two
month’s] time
Choice
2 week
payment (£)
1 month
payment (£)
2 months
payment
(£)
1
25
26
2
25
28
3
25
30
4
25
32
5
25
35
6
25
38
7
25
26
8
25
30
9
25
35
10
25
37
11
25
40
12
25
45
Inclusion criteria: Discounting & inflammation data
•
•
•
•
•
•
•
•
Descriptive Statistics
Discount rate 2wks/1mth
Discount rate 2wks/1mth
C-reactive protein base
C-reactive protein follow
Fibrinogen baseline
Fibrinogen follow-up
• Valid N (listwise)
N
953
983
641
639
664
657
389
Min.
.02
.01
.20
.10
1.60
1.70
Max.
.54
.33
9.90
9.80
5.40
5.30
Mean
.15
.12
2.31
2.09
3.34
2.95
SD
.20
.12
2.10
2.00
.55
.53
Discount rate
• Use V = A/(1+ kD)
• Where V is present value, A is the delayed or future value,
k is the discount rate and D is the delay in weeks
• Provides estimates of k within a range of values
• Choice 1: £25 in 2 weeks and £35 in 4 weeks indicates a weekly k of 33%
• Choice 2: £25 in 2 weeks and £38 in 4 weeks indicates a k of 54%
• If the participant chooses the larger later option for 2 but not 1 this mean their
discount rate k is somewhere between 33% and 54% per week. A rate of 43.5%
is assigned
Discount rate values
• For participants
• 2 weeks vs. 1 month = 15.59% (SD = 20.13)
• 2 weeks vs. 2 months = 11.32% (SD = 11.91)
• Discount rate values correlate highly (r (389) = .7, p <.001)
• Combined rate, k = 13.45% (SD = 14.86%)
• Those not included in the study k = 14.04% (SD = 15.01%),
not significantly different (t (1023) = .6, p = .53)
Discount rate values
C-reactive protein (mg/L) & Fibrinogen (g/L) levels
Covariates
• Demographics: age, gender, race, marital status, retirement status,
education level, log income, log benefit-unit non-pension wealth
• Health: long-standing illness, arthritis, cardiovascular disease (stroke,
congestive heart failure, angina, heart attack)
• Health behaviour: smoking, physical activity (3 q’s frequency of mild,
moderate, vigorous activity), drinking (3 q’s spirit, wine, beer
measures/glasses/pints in last week), BMI (kg/m2)
• Cognition: 10 words recalled without delay, with delay & verbal fluency
(animal naming in 60 seconds)
• Personality: 25 item Big Five
• Missing data: 98.5% completion on average for covariates, largest
missing for single variable was 5.4% & missing data replaced with mean
Descriptives
Discount rate and CRP
Discount rate and fibrinogen
Delay discounting and CRP 2 years later:
adjusted for demographics & prior CRP levels
Delay discounting and CRP 2 years later:
adjusted for demographics, health/health behaviour & prior CRP levels
Delay discounting and CRP 2 years later:
adjusted for demographics, health/health behaviour, personality &
cognition & prior CRP levels
Delay discounting and fibrinogen 2 years later:
adjusted for demographics & prior fibrinogen levels
Delay discounting and fibrinogen 2 years later*:
adjusted for demographics, health/health behaviour & prior fibrinogen levels
* Decrease in coefficient attributable to a combination of smoking and BMI
Delay discounting and fibrinogen 2 years later:
adjusted for demographics, health/health behaviour, personality &
cognition & prior fibrinogen levels
Alternative explanations?
• Ferguson (2013) discusses potential
pathways that may apply including:
- Treatment seeking
- Communication with physicians
- Treatment compliance
- Stress and coping (e.g. Daly et al., 2014)
- Health behaviour (objective PA, diet
measurement, accurate adiposity)
• Both discounting and biological functioning
could be determined by genetic factors or
early conditions
• Discounting may shape trajectories of
health over the lifespan (earlier mediators)
Summary
• Experimentally elicited discount rate predicts two measures of
inflammation: C-reactive protein and fibrinogen
• Associations are larger than population cross-sectional associations between
socioeconomic status (e.g. wealth, education)
• Associations are robust to adjustment for a broad set of demographic factors,
income, wealth, personality, cognitive ability and prior inflammation levels
• BMI and smoking account for 20% of the link between k & fibrinogen
• Link between k & CRP is not explained by health, health behaviour, or
confounding by personality or cognitive ability
• Inflammation could be a key pathway between discounting and health
outcomes including mortality

similar documents