HOW DOES THE FOOD ENVIRONMENT INFLUENCE HOUSEHOLD FOOD PURCHASE - - PowerPoint PPT Presentation

HOW DOES THE FOOD ENVIRONMENT INFLUENCE HOUSEHOLD FOOD PURCHASE PATTERNS AND NUTRITIONAL STATUS? EMPIRICAL EVIDENCE FROM FOOD VENDOR MAPPING IN PERI-URBAN DAR ES SALAAM, TANZANIA RAMYA AMBIKAPATHI PURDUE UNIVERSITY, JULY 1ST 2020, #ANH2020

http://www.anh-academy/ANH2020

CO-AUTHORS

Mary Mwanyika-Sando Dominic Mosha Crystal Patil Germana Leyna Japhet Killewo Nilupa Gunaratna Ramya Ambikapathi Morgan Boncyk Savannah Froese Cristiana Edwards Alli Mangara Patrick Kazonda Gerald Shively

¡In Africa, majority of peri-urban population relies on purchased foods. Food environment contains a high density

• f informal vendors, creating challenges to characterizing the FE.

BACKGROUND AND AIM

Characterize food environment and create summary metrics ¡In Africa, majority of peri-urban population relies on purchased foods. Food environment contains a high density

• f informal vendors, creating challenges to characterizing the FE.

BACKGROUND AND AIM

Characterize food environment and create summary metrics Household food purchase ¡In Africa, majority of peri-urban population relies on purchased foods. Food environment contains a high density

• f informal vendors, creating challenges to characterizing the FE.

BACKGROUND AND AIM

Characterize food environment and create summary metrics Diets Household food purchase ¡In Africa, majority of peri-urban population relies on purchased foods. Food environment contains a high density

• f informal vendors, creating challenges to characterizing the FE.

BACKGROUND AND AIM

Characterize food environment and create summary metrics Diets Household food purchase Nutritional Status ¡In Africa, majority of peri-urban population relies on purchased foods. Food environment contains a high density

• f informal vendors, creating challenges to characterizing the FE.

BACKGROUND AND AIM

Characterize food environment and create summary metrics Diets Household food purchase Nutritional Status ¡In Africa, majority of peri-urban population relies on purchased foods. Food environment contains a high density

• f informal vendors, creating challenges to characterizing the FE.

¡Nested within Diet, Choice, and Positive living (DECIDE) study: mixed-methods cohort set in peri-urban Dar

es Salaam, Tanzania.

¡Aims to characterize food choice and environment among families with persons living with human

immunodeficiency virus (PLHIV) using qualitative, geo-spatial and quantitative methods.

¡IRB approval from Purdue University and Tanzania's National Institute for Medical Research.

BACKGROUND AND AIM

GEOCODING A DYNAMIC FOOD ENVIRONMENT

Formal Semi-Formal Informal food vendors

• Fixed structures (super-market,

wet market, shops)

• Fixed location
• Semi-permanent structures

(umbrella, pallets)

• Consistent location daily
• Baskets/Bicycles
• Mobile through space and time

Example of formal food vendor Example of semi-formal food vendor Example of informal food vendor *Tool and protocol available

GEOCODING A DYNAMIC FOOD ENVIRONMENT

Formal Semi-Formal Informal food vendors

• Fixed structures (super-market,

wet market, shops)

• Fixed location
• Semi-permanent structures

(umbrella, pallets)

• Consistent location daily
• Baskets/Bicycles
• Mobile through space and time
• GPS & Gender
• Vendor typologies
• 8 food groups & 58+ food items
• Survey length: 1-2 mins
• GPS & Gender
• Vendor typologies
• 8 food groups & 58+ food items
• Survey length: 1-2 mins
• GPS & Gender
• 31+ food items
• Survey length: <1 min

Example of formal food vendor Example of semi-formal food vendor Example of informal food vendor *Tool and protocol available

FOOD ENVIRONMENT: CENSUS OF 6,627 VENDORS

17% Informal vendor

39% Formal vendor (restaurants, shops)

44% Semi-formal vendor

30% sell vegetables 15% green leafy vegetables 40% sell vegetables 27% green leafy vegetables 30% sell vegetables 58% green leafy vegetables

Vegetables include: cabbage, bell peppers, tuber, lemon, onion, tomato, okra, green leafy vegetable, eggplant, carrots

1 km Data collection: April to June 2019

FOOD ENVIRONMENT: CENSUS OF 6,627 VENDORS

17% Informal vendor

39% Formal vendor (restaurants, shops)

44% Semi-formal vendor

30% sell vegetables 15% green leafy vegetables 40% sell vegetables 27% green leafy vegetables 30% sell vegetables 58% green leafy vegetables

Vegetables include: cabbage, bell peppers, tuber, lemon, onion, tomato, okra, green leafy vegetable, eggplant, carrots

1 km Data collection: April to June 2019

FOOD ENVIRONMENT: METRICS DEFINITION

Metric Name Definition Density Food environment typology Count Informal, semi-formal, formal and all vendors Vegetable vendors Count Vendors who sell any of 10 vegetables Green leafy vegetable vendor Count Vendors who sell green leafy vegetables Dispersion Vegetable vendor hotspots / cold spots Clusters Vegetable vendors Green leafy vendor hotspots / cold spots Clusters Green leafy vegetable vendors Diversity / Dominance Shannon diversity of vendor typology (standardized 0 to 1) Variety and evenness 6 vendor typology: restaurants, mobile vendors, shops, semi-formal food vendors, butchers, umbrella vendors Dominance of vendor typology (standardized 0 to 1) Variety and evenness Measure of one/few vendor dominating (1- diversity). Lack of variety and evenness.

FOOD ENVIRONMENT: METRICS DEFINITION

Metric Name Definition Density Food environment typology Count Informal, semi-formal, formal and all vendors Vegetable vendors Count Vendors who sell any of 10 vegetables Green leafy vegetable vendor Count Vendors who sell green leafy vegetables Dispersion Vegetable vendor hotspots / cold spots Clusters Vegetable vendors Green leafy vendor hotspots / cold spots Clusters Green leafy vegetable vendors Diversity / Dominance Shannon diversity of vendor typology (standardized 0 to 1) Variety and evenness 6 vendor typology: restaurants, mobile vendors, shops, semi-formal food vendors, butchers, umbrella vendors Dominance of vendor typology (standardized 0 to 1) Variety and evenness Measure of one/few vendor dominating (1- diversity). Lack of variety and evenness. FE metrics are correlated with each other.

DIVERSITY – RICHNESS AND EVENNESS PER AREA

Distance 100 meters 500 meters

FOOD ENVIRONMENT: DISTANCE TO HOUSEHOLD

1000 meters Density: Median (IQR) number of all green leafy vegetable vendors 0 (0, 1) 2 (1, 4) 6 (3, 11) 18 (15, 26) 38 (29, 49) 69 (57,88) 200 meters 300 meters 700 meters

BACKGROUND ON THE PARTICIPANTS (PLHIV)

Selected main outcomes Median (IQR); N=239

Bought any (10) vegetables in the last 7 days, Frequency, Main purchase location 71%, 8 times Mostly from semi-formal/informal vendors Energy intake (kcal) from 24-hour recall 2694 kcal (1874, 3659) Body Mass Index (Kg/m2 , measure of obesity) 23.1 (20.7, 27.2) 10% underweight 36% overweight/obese Waist to Hip Ratio (~ measure of central adiposity) 0.85 (0 81, 0.90) 26% above 0.90 cutoff (risk factor for diabetes)

Participant : 70% of women, 40 years old, 4 years since HIV diagnosis, half share toilets with neighbors, and almost all have cellphone.

Vegetables include: cabbage, bell peppers, tuber, lemon, onion, tomato, okra, green leafy vegetable, eggplant, carrots

Round 1 Data collection: February to June 2019

p = 0.010 p = 0.002 p = 0.030 p = 0.016 p = 0.091 p = 0.092 p = 0.018 p = 0.013 p = 0.015 p = 0.029 p = 0.105 p = 0.066 vegdensity_100 vegdensity_200 vegdensity_300 vegdensity_500 vegdensity_700 vegdensity_1000 greenvegdensity_100 greenvegdensity_200 greenvegdensity_300 greenvegdensity_500 greenvegdensity_700 greenvegdensity_1000 .5 1 1.5 2

Odds ratio of buying vegetables in the last 7 days

REGRESSION RESULTS – HOUSEHOLD FOOD PURCHASE

Bought any vegetables last week?

*All models adjusted for age, gender, education, asset quartiles, years since HIV diagnosis, renting house, head of household status, morbidity; robust standard error

N=239 Vegetable vendor density Green leafy vegetable vendor density

100 meters 200 meters 300 meters 500 meters 700 meters 1000 meters 100 meters 200 meters 300 meters 500 meters 700 meters 1000 meters

p = 0.010 p = 0.002 p = 0.030 p = 0.016 p = 0.091 p = 0.092 p = 0.018 p = 0.013 p = 0.015 p = 0.029 p = 0.105 p = 0.066 vegdensity_100 vegdensity_200 vegdensity_300 vegdensity_500 vegdensity_700 vegdensity_1000 greenvegdensity_100 greenvegdensity_200 greenvegdensity_300 greenvegdensity_500 greenvegdensity_700 greenvegdensity_1000 .5 1 1.5 2

Odds ratio of buying vegetables in the last 7 days

REGRESSION RESULTS – HOUSEHOLD FOOD PURCHASE

Bought any vegetables last week?

*All models adjusted for age, gender, education, asset quartiles, years since HIV diagnosis, renting house, head of household status, morbidity; robust standard error

N=239 Vegetable vendor density Green leafy vegetable vendor density

100 meters 200 meters 300 meters 500 meters 700 meters 1000 meters 100 meters 200 meters 300 meters 500 meters 700 meters 1000 meters

p = 0.010 p = 0.002 p = 0.030 p = 0.016 p = 0.091 p = 0.092 p = 0.018 p = 0.013 p = 0.015 p = 0.029 p = 0.105 p = 0.066 vegdensity_100 vegdensity_200 vegdensity_300 vegdensity_500 vegdensity_700 vegdensity_1000 greenvegdensity_100 greenvegdensity_200 greenvegdensity_300 greenvegdensity_500 greenvegdensity_700 greenvegdensity_1000 .5 1 1.5 2

Odds ratio of buying vegetables in the last 7 days

REGRESSION RESULTS – HOUSEHOLD FOOD PURCHASE

• A greater density of vegetable vendors within 500 meters of

home increases the likelihood of purchasing vegetables in the last week.

• This effect increases as vendors are found closer to home.

Bought any vegetables last week?

*All models adjusted for age, gender, education, asset quartiles, years since HIV diagnosis, renting house, head of household status, morbidity; robust standard error

N=239 Vegetable vendor density Green leafy vegetable vendor density

100 meters 200 meters 300 meters 500 meters 700 meters 1000 meters 100 meters 200 meters 300 meters 500 meters 700 meters 1000 meters

p = 0.054 p = 0.047 p = 0.007 p = 0.055 p = 0.120 p = 0.140 p = 0.020 p = 0.034 p = 0.007 p = 0.079 p = 0.173 p = 0.156

• 250
• 200
• 150
• 100
• 50

50

Energy(Kcal)

p = 0.010 p = 0.002 p = 0.030 p = 0.016 p = 0.091 p = 0.092 p = 0.018 p = 0.013 p = 0.015 p = 0.029 p = 0.105 p = 0.066 vegdensity_100 vegdensity_200 vegdensity_300 vegdensity_500 vegdensity_700 vegdensity_1000 greenvegdensity_100 greenvegdensity_200 greenvegdensity_300 greenvegdensity_500 greenvegdensity_700 greenvegdensity_1000 .5 1 1.5 2

Odds ratio of buying vegetables in the last 7 days

REGRESSION RESULTS – HOUSEHOLD FOOD PURCHASE

• A greater density of vegetable vendors within 500 meters of

home increases the likelihood of purchasing vegetables in the last week.

• This effect increases as vendors are found closer to home.

Bought any vegetables last week?

*All models adjusted for age, gender, education, asset quartiles, years since HIV diagnosis, renting house, head of household status, morbidity; robust standard error

N=239 Vegetable vendor density Green leafy vegetable vendor density

100 meters 200 meters 300 meters 500 meters 700 meters 1000 meters 100 meters 200 meters 300 meters 500 meters 700 meters 1000 meters

Total Energy (Kcal)

• This effect translated into reduced intake of

total energy by 50-100 Kcal.

p = 0.656 p = 0.887 p = 0.410 p = 0.400 p = 0.344 p = 0.152 p = 0.057 p = 0.404 p = 0.404 p = 0.531 p = 0.038 p = 0.089 p = 0.064 p = 0.102 p = 0.075 p = 0.064 p = 0.042 p = 0.042 p = 0.202 p = 0.317 p = 0.168 p = 0.591 p = 0.603 p = 0.012 p = 0.746 p = 0.287 p = 0.287 p = 0.173 p = 0.540 p = 0.228 p = 0.373 p = 0.361 p = 0.514 p = 0.615 p = 0.164 p = 0.164 informal_200 semiformal_200 formal_200 vegdensity_200 greenvegdensity_200 veg_hotspot_200 grveg_hotspot_200 diversity200 dominance200 informal_1000 semiformal_1000 formal_1000 vegdensity_1000 greenvegdensity_1000 veg_hotspot_1000 grveg_hotspot_1000 diversity1000 dominance1000

• .5

.5

• 20

20 40

ind_whr ind_bmi

REGRESSION RESULTS–NUTRITIONAL STATUS

Waist to Hip Ratio Body mass index Different metrics of food environment at different distances to household have various association with nutritional status

*All models adjusted for age, gender, education, asset quartiles, years since HIV diagnosis, renting house, head of household status, morbidity; robust standard error

200 meters 1000 meters

Semi-formal vendors Informal vendors Density Dispersion Diversity Formal vendors Vegetable vendors Vegetable hotspots Green.veg vendors Green.veg hotspots Vendor diversity Vendor dominance Semi-formal vendors Informal vendors Density Dispersion Diversity Formal vendors Vegetable vendors Vegetable hotspots Green.veg vendors Green.veg hotspots Vendor diversity Vendor dominance

SUMMARY OF FINDINGS

Food environment metrics inspired by ecology are associated with food purchase patterns, diets, and nutritional status.

Food environment metrics Diets Household food purchase Nutritional Status

SUMMARY OF FINDINGS

Food environment metrics inspired by ecology are associated with food purchase patterns, diets, and nutritional status.

Food environment metrics

• Peri-urban setting: Having vendors closer to home is associated with increased purchase of vegetables and

reduced total energy intake. Diets Household food purchase Nutritional Status

SUMMARY OF FINDINGS

Food environment metrics inspired by ecology are associated with food purchase patterns, diets, and nutritional status.

Food environment metrics

• Peri-urban setting: Having vendors closer to home is associated with increased purchase of vegetables and

reduced total energy intake.

• Food purchasing behavior and consumption is complex. Need to align specific FE metrics with specific
• behaviors. Ex. vegetable vendor density is associated with vegetable purchase.
• Future work:
• Analyze other food purchase behavior (soda, prepared foods, packaged foods, fruits, recommended foods for PLHIV).
• Examine spatial and temporal variation of food environment using geo-spatial methods.
• Identify intervention points: Ex. optimize and target semi/informal vendors for healthy eating patterns.

Diets Household food purchase Nutritional Status

THANK YOU!

¡ Grateful for my team and

participants for giving me this

• pportunity to present on their

behalf and highlight these findings from this community.

¡ Funder: Drivers of Food Choice ¡ Questions/Comments, please

contact me at rambikap@purdue.edu