A RANDOMIZED CONTROLLED TRIAL TO MOTIVATE AND SUSTAIN PHYSICAL - - PowerPoint PPT Presentation

A RANDOMIZED CONTROLLED TRIAL TO MOTIVATE AND SUSTAIN PHYSICAL ACTIVITY AMONG TAXI DRIVERS USING FINANCIAL INCENTIVES – TAKSI STUDY

D R E R I C A F I N K E L S T E I N ( P I )

• Professor, Health Services & Systems Research, Duke-NUS Singapore

D R H O T E C K H U A ( C O - I )

• Deputy President and Tan Chin Tuan Centennial Professor, National University of

Singapore

• Professor, Haas School of Business, University of California, Berkeley

D R U R I G N E E Z Y ( C O - I )

• Professor, Rady School of Management, University of California, San Diego

D R M A R C E L B I L G E R ( C O - I )

• Assistant Professor, Health Services & Systems Research, Duke-NUS Singapore

Funding: NUS Institute to Improve Health in Asia (NIHA) June 25th, 2015

Outline

q Background q Study design q Testable hypotheses q Preliminary Results q Discussion

A sedentary population group

q There are above 28,000 registered taxis in Singapore q Taxi drivers are an especially sedentary group due to their long shifts and working conditions q Taxi drivers are also unlikely to engage in structured physical activity programs due to their unusual shifts.

Source: www.taxi-photos.com

q Lack of physical activity puts taxi drivers at risk to a wide range of Non-Communicable Diseases (e.g. Pate et al., 1995)

Financial incentives to promote physical activity

q Neo-classical economics predicts that

q financial incentives can motivate drivers to increase their physical activity q cash incentives would work best as cash is assumed to be fully fungible

q Research in behavioral economics showed that

q individuals partition their budget according to expense types, which is a form of mental accounting (Thaler, 1980) q individuals prefer avoiding losses to acquiring gains (Kahneman and Tversky, 1984)

Our approach is to incentivize drivers with taxi rental credits to capitalize on both mental accounting and loss aversion

Study design

q The TAKSI study measures the effect of Cash and Taxi Rental incentives on physical activity and health among taxi drivers aged 50 to 75 q 316 eligible taxi drivers were randomized into one of two equisized parallel study arms: Arm 1: Drivers receive Cash incentives if they achieve activity goals Arm 2: Drivers receive Taxi Rental credits if they achieve activity goals q The only difference between the two arms is how the reward is framed ü the incentive amounts are the same in both groups. q In both arms, the incentive is reinforced by means of study messaging and monthly reports on physical activity and target attainment

Study design, goals and timeline

Baseline (Month 0) Intervention period (Months ¡1 ¡to ¡4) ¡ Post-intervention period (Months ¡5 ¡to ¡7) ¡ Timeline ¡ 21 days ¡ 28 days (per month) ¡ 28 days (per month) ¡ Step target ¡ None ¡ At least 7,000 steps in 20 or more days (per month) ¡ None ¡ Financial Incentive ¡ None ¡ One day Taxi rental credit (per month) ¡ None ¡ Data Collected

• Daily steps activity
• Daily driving data – mileage, earnings and log-on

hours

• Biometrics – height , weight, resting heart rate

and blood pressure (at Months 0, 4 and 7)

• Self-reported health and work productivity (at

Months 0, 4 and 7)

q The program takes 8 months to

complete, including baseline, intervention and post-intervention period q In the intervention period, drivers can earn incentives by meeting specified step targets – 7,000 or more steps daily in at least 20 out

• f 28 days in a month

§ Incentives are removed after Month 4 q All participants receive a Fitbit Zip™ wireless pedometer to monitor daily step activity q Assessments are done at Months 0, 4 and 7 to collect info on biometrics and self-reported health status

Testable Hypotheses

q Financial incentives increase physical activity and improve health outcomes

H1: Drivers in both arms will have a higher number of steps at Month 4 compared to baseline

q Rental Credit incentives are more effective than cash incentives

H2: Month 4 Improvements will be greater in the Taxi Rental arm compared to the Cash arm

Financial incentives will lead to habit formation even when they are withdrawn and this effect will be larger in the Taxi Rental arm

Same comparisons as above between Month 7 and baseline, that is 3 months after financial incentives have been discontinued

Study participants (baseline)

Baseline characteristics ¡ All (N=316) ¡ By Incentive Arm ¡ Rental arm (N=157) ¡ Cash arm (N=159) ¡ Average age ¡ 58.6 (sd=4.8 ) ¡ 58.2 (sd=4.6) ¡ 58.9 (sd=4.9) ¡ Gender ¡ ¡ ¡ ¡ Female ¡ 1.91% ¡ 1.92% ¡ 1.90% ¡ Male ¡ 98.09% ¡ 98.08% ¡ 98.10% ¡ Ethnicity ¡ ¡ ¡ ¡ Chinese ¡ 88.5% ¡ 87.8% ¡ 89.2% ¡ Malay ¡ 4.5% ¡ 5.8% ¡ 3.2% ¡ Indian ¡ 6.4% ¡ 5.8% ¡ 7.0% ¡ Others ¡ 0.6% ¡ 0.6% ¡ 0.6% ¡ Marital status ¡ ¡ ¡ ¡ Married ¡ 93.6% ¡ 95.5% ¡ 91.8% ¡ Widowed ¡ 0.6% ¡ 0.0% ¡ 1.3% ¡ Divorced ¡ 3.5% ¡ 2.6% ¡ 4.4% ¡ Never married ¡ 2.2% ¡ 1.9% ¡ 2.5% ¡ Highest educational attainment ¡ ¡ ¡ ¡ Primary or lower ¡ 13.4% ¡ 14.1% ¡ 12.7% ¡ Secondary ¡ 76.8% ¡ 75.6% ¡ 77.9% ¡ Diploma or University Degree ¡ 9.9% ¡ 10.3% ¡ 9.5% ¡

q 157 drivers are assigned to Rental arm while 159 drivers are in the Cash arm q The drivers are 59 years old

• n average

q Most of the drivers are Chinese (89%), married (94%), males (98%) with at least secondary education (87%)

Study participants (baseline)

q Almost half of the drivers have monthly household income below $2,000 q Majority of the drivers (40%) are inactive or do little activity than usual daily activities q A large proportion of drivers are overweight (44%) and obese (33%) q 28% of the drivers are hypertensive

Baseline characteristics ¡ All (N=316) ¡ By Incentive Arm ¡ Rental arm (N=157) ¡ Cash arm (N=159) ¡ Household monthly income ¡ ¡ ¡ ¡ Below 2,000 ¡ 47.2% ¡ 49.0% ¡ 45.4% ¡ 2,000-2,999 ¡ 21.5% ¡ 21.8% ¡ 21.3% ¡ 3,000-3,999 ¡ 11.5% ¡ 10.2% ¡ 12.8% ¡ 4,000 and over ¡ 19.8% ¡ 19.1% ¡ 20.6% ¡ Usual pattern of daily physical activities ¡ ¡ ¡ ¡ Inactive or little activity than usual daily activities ¡ 39.0% ¡ 39.0% ¡ 39.1% ¡ Regularly (≥5 days per week) participate in low exertion level physical activities ¡ 16.1% ¡ 16.9% ¡ 15.2% ¡ Participate in aerobic exercises for 20 to 60 minutes per week ¡ 15.1% ¡ 16.9% ¡ 13.3% ¡ Participate in aerobic exercises for 1 to 3 hours per week ¡ 16.7% ¡ 16.9% ¡ 16.6% ¡ Participate in aerobic exercises for

• ver 3 hours per week ¡

13.1% ¡ 10.4% ¡ 15.9% ¡ Hypertensive ¡ 28.8% ¡ 28.3% ¡ 29.3% ¡ BMI category ¡ ¡ ¡ ¡ Normal ¡ 23.4% ¡ 22.2% ¡ 22.4% ¡ Overweight ¡ 44.1% ¡ 44.0% ¡ 46.1% ¡ Obese ¡ 32.5% ¡ 33.9% ¡ 31.6% ¡

Effects on daily step activity – Intervention period at Month 4

q Drivers in both arms increased their daily step activity

§ Rental arm walks more than baseline by 2,283 steps § Cash arm walks more than baseline by 1,535 steps

q The increase in daily steps is larger in Rental arm

§ Incremental difference: 2,283 (Rental) –1,535 (Cash) = 748 steps

Framing of the incentives as rental credit is more effective than cash

1000 2000 3000 4000 5000 6000 7000 8000 9000 Rental arm Cash arm 6281.74 6907.69 2283.39 1535.39 Average steps at baseline Average increment at intervention period

Average steps (per day) ¡ Baseline (N=316) ¡ Intervention Period (N=316) ¡ Rental arm ¡ 6281.7 ¡ 8565.1 ¡ (185.7) ¡ (180.5) ¡ Cash arm ¡ 6907.7 ¡ 8443.1 ¡ (238.4) ¡ (170.5) ¡ First diff - Rental arm ¡ ¡ 2283.4*** ¡ (164.3) ¡ First diff - Cash arm ¡ 1535.4*** ¡ (152.0) ¡ Diff-in-diff ¡ 748.0*** ¡ ¡ (223.78) ¡

* p<0.10, ** p<0.05, *** p<0.01 ¡ Standard errors in parentheses ¡

q Drivers in both arms are walking less without the incentive but are still more active than baseline – habit formation

§ Rental arm walks more than baseline by 1,994 steps § Cash arm walks more than baseline by 1,169 steps

q The increment is bigger in Rental arm

§ Incremental difference: 1,994 (Rental) –1,169 (Cash) = 825 steps

Financial incentives create habit formation and the effect is stronger by framing it as rental credit

Note: Estimates are still preliminary as data collection is still on going. Final data collection is in July. 1000 2000 3000 4000 5000 6000 7000 8000 9000 Rental arm Cash arm 6281.74 6907.69 1994.49 1169.78 Average steps at baseline Average increment at intervention period

Average steps (per day) ¡ Baseline (N=316) ¡ Post-­‑interven4on ¡ Period ¡(N=297) ¡ Rental arm ¡ 6281.74 ¡ 8276.23 ¡ (185.71) ¡ (236.35) ¡ Cash arm ¡ 6907.69 ¡ 8077.48 ¡ (238.37) ¡ (173.30) ¡ First diff – Rental arm ¡ ¡ 1994.49*** ¡ (223.77) ¡ First diff – Cash arm ¡ 1169.78*** ¡ (181.99) ¡ Diff-in-diff ¡ 824.71*** ¡ ¡ (288.43) ¡

* p<0.10, ** p<0.05, *** p<0.01 ¡ Standard errors in parentheses ¡

Effects on daily step activity – Post-intervention period at Month 7

q Rental arm drivers achieve their monthly step target 84% of the time (vs 81% in the cash arm)

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% Rental arm Cash arm 18% 27% 66% 54% Baseline Increment at intervention period

Predicted probabilities of achieving the target ¡ Baseline (N=316) ¡ Intervention Period (N=316) ¡ Rental arm ¡ 0.18 ¡ 0.84 ¡ (0.03) ¡ (0.02) ¡ Cash arm ¡ 0.27 ¡ 0.81 ¡ (0.04) ¡ (0.03) ¡ First diff – Rental arm ¡ ¡ 0.66*** ¡ (0.03) ¡ First diff – Cash arm ¡ 0.54*** ¡ (0.04) ¡ Diff-in-diff ¡ 0.12** ¡ ¡ (0.05) ¡

* p<0.10, ** p<0.05, *** p<0.01 ¡ Standard errors in parentheses ¡

Achieving the monthly step target during the 4-month intervention period

q At month 4, all drivers report being more active than at baseline:

q Rental arm’s sedentary time is lower than baseline by 2 hours q Cash arm’s sedentary time is higher than baseline by 0.28 hours

q The incremental difference is bigger for the Rental arm drivers by 2.19 hours

• 2

2 4 6 8 10 Rental arm Cash arm 9.56 8.35

• 1.92

0.28

Figure 4

Baseline Month 4 Increment

Sedentary hours (per day) ¡ Baseline (N=276) ¡ Month 4 (N=276) ¡ Rental arm ¡ 9.56 ¡ 7.64 ¡ (0.66) ¡ (0.38) ¡ Cash arm ¡ 8.35 ¡ 8.63 ¡ (0.39) ¡ (0.91) ¡ First diff – Rental arm ¡ ¡

• 1.92** ¡

(0.75) ¡ First diff - Cash arm ¡ 0.28 ¡ (0.93) ¡ Diff-in-diff ¡

• 2.19* ¡

¡ (1.19) ¡

* p<0.10, ** p<0.05, *** p<0.01 ¡ Standard errors in parentheses ¡

Effect on sedentary behaviors during the intervention period at Month 4 compared to baseline

0.84 0.86 0.88 0.9 0.92 Rental arm Cash arm 0.87 0.89 0.033 0.014

Figure 5. EQ5D Index

Baseline Month 4 Increment 76 78 80 82 84 Rental arm Cash arm 79.27 80.01 3.91 2.48

Figure 6. EQ-VAS

Baseline Month 4 Increment

At month 4, improvements in health measures (EQ5D, EQ- VAS, weight and blood pressure) are observed in both arms

§ Rental arm drivers have higher EQ5D and EQ-VAS index than baseline by 0.03*** and 3.9*** § Cash arm drivers increased EQ5D and EQ-VAS index by 0.014 and 2.5** § Rental arm drivers also have lower systolic blood pressure than baseline § Rental and cash arms have reduced weight by 0.86 and 0.34*** kilograms, respectively

* p<0.10, ** p<0.05, *** p<0.01

• 50

50 100 150 Rental arm Cash arm 131.86 132.03

• 2.52

0.26

Figure 7. Systolic BP

Baseline Month 4 Increment

• 20

20 40 60 80 Rental arm Cash arm 72.24 72.84

• 0.86
• 0.34

Figure 8. Weight

Baseline Month 4 Increment

Health effects during the intervention period at Month 4 compared to baseline

Summary and policy implications

q Financial incentives are found to increase step-based physical activity among older taxi drivers q Health improvements in terms of improved systolic blood pressure and quality of life have also been found q Framing the incentive as a taxi rental credit provides an additional benefit for free q Behavioral economics can improve the cost- effectiveness of public health interventions

Acknowledgment

q We are extremely grateful to ComfortDelGro for their help with screening and recruiting the participants and for their valuable logistical support.

www.duke-nus.com.sg