A New Metrics Toolbox to Assess the
Cost and Geographic Distribution of
Healthy Diets
Anju Aggarwal, PhD
Assistant Professor, Department of Epidemiology
Affiliated Faculty, Nutritional Sciences Program
University of Washington
World Food Center, University of California - Davis
June 3rd, 2017
Aligning food systems with dietary needs
There are multiple drivers of food choices at the consumer level:
• Economics
• Geographic
• Psychosocial
Each dimension has its metrics and measures.
The UW Food Environment Research Team
Seattle Obesity Study (SOS)
• NIDDK R01 076608-10
• SOS I 2008-2011
• SOS II 2011-2015
• SOS III 2015-2020
• King, Pierce and Yakima counties (total
N approx 3,500)
• Diets: FFQ and 24-hr recall
• Affordability: market basket
• Geolocation: GIS and GPS
• Attitudes: Questionnaire self-report
• Health outcomes: measured ht/wt
• Goal: Assess drivers of food choice; their
interactions and body weight trajectories.
Economic dimension of food access
Bureau of Labor
Statistics –
household-level
food expenditures
Are healthy foods affordable?
Do healthier diets cost more?
No data on cost of
diets at individual-
level
A novel way to estimate cost of the diet at the individual-
level by attaching prices to their diets!
• Used a standard dietary data
collection tool – Food
Frequency Questionnaires
(FFQ).
• Lowest price, non sale price
• Collected for each of the 384
foods and beverages
underlying FFQ.
Metric #1: Individual-level estimate of diet cost
Attached prices exactly the same way a nutrient vector is attached
to the FFQ
Metric #1: Individual-level estimate of diet costHow it works?
This process yields:
• Average daily intake of calories, grams, and 45 macro- and micro-nutrients for each
respondent.
• Estimated cost of the habitual diet for each respondent.
Metric #1: Individual-level estimate of diet cost
Dietary intake data
from NHANES
Nutrient composition
data FNDDS 2.0
Price data CNPP
National food price data
Energy intake
Nutrient intake
Diet cost
Diet cost
Socioeconomic status
Nutrient intakes
Food groups (HEI)
Overall diet quality (Energy density, Nutrient adequacy)
This technique brings nutrition economics to
the field of nutrition epidemiology
0
10
20
30
40
50
60
70
80
90
Q1 Q2 Q3 Q4 Q5
Pro
po
rtio
n o
f h
igh
er
inco
me a
nd
ed
ucati
on
(%
)
Quintiles of energy adjusted diet cost
Income ≥ 50KEducation ≥ college grads
Lower cost diets are more likely to be consumed by lower SES
Source: Aggarwal et al, Eur J Clin Nutr. 2011. https://www.ncbi.nlm.nih.gov/pubmed/21559042 10
* P trend < 0.0001
Diet cost linked to socioeconomic status
Source: Aggarwal et al. PLoS One 2012. http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0037533
Nutrients-to-limit cost less
70
80
90
100
110
120
130
140
150
Q1 Q2 Q3 Q4 Q5
Die
t c
os
t w
ith
Q 1
no
rma
lize
d a
t 1
00
%
Quintile of energy adjusted nutrient intakesVitamin A Vitamin D Calcium Folate
Fiber Added sugars Trans fats Total Saturated Fat
Vitamin C Magnesium Potassium Beta carotene
* P trend < 0.0001
Nutrients-to-encourage costs even more
Beneficial nutrients cost more
Nutrient rich diets tend to cost more
Diet cost linked to nutrient intakes
2.4 3.1 3.6 4.1 4.90.41.9 2.3 2.8
4.7
1.92.8
3.43.7
4.3
2
3.34.5
5
5
1.9
2
2.42.6
2.7
5.1
5.9
5.95.8
5.2
4.9
5
55
5
3.9
2.9
3.54.3
5
4.5
4.4
44.2
4.8
5.3
4
3.83.3
2.7
4.4
5.2
5.9
6.8
8.7
7.4
9.7
11.6
12.6
13.5
0
10
20
30
40
50
60
70
Q1 Q2 Q3 Q4 Q5
Adju
ste
d m
ean H
EI-
2010 s
core
Quintiles of diet cost ($/2000kcal)
Energy from SoFAAS
Refined grains
Sodium
Fatty acid ratio
Seafood & plant proteins
Total protein foods
Dairy
Whole grains
Whole fruit
Total fruit
Greens & beans
Total veg
Source: Rehm et al. Prev Med. 2015. https://www.ncbi.nlm.nih.gov/pubmed/25625693
Diets with higher HEI scores tend to cost more
Diet cost linked to healthy eating index (HEI-2010)
0
1
2
3
4
5
6
7
8
Q1 Q2 Q3 Q4 Q5
AD
JU
ST
ED
ME
AN
EN
ER
GY
DE
NS
ITY
QUINTILES OF DIET COST ($/1800KCAL)
Source: Aggarwal et al. EJCN 2011. https://www.ncbi.nlm.nih.gov/pubmed/21559042
p-value <0.0001
Energy density defined as total calories over grams of foods consumed (Kcal/g)
Energy dense diets tend to cost less
Diet cost linked to dietary energy density
68
70
72
74
76
78
80
82
84
86
88
Q1 Q2 Q3 Q4 Q5
AD
JU
ST
ED
ME
AN
EN
ER
GY
DE
NS
ITY
QUINTILES OF DIET COST ($/1800KCAL)
Source: Aggarwal et al. EJCN 2011. https://www.ncbi.nlm.nih.gov/pubmed/21559042
p-value <0.0001
Mean adequacy ratio (MAR) was defined as the truncated index of the percent of daily recommended intakes for
key nutrients. Nutrients included Vit A, C, D, E, B12, Calcium, Iron, Magnesium, Potassium, Folate and Fiber
Nutrient adequate diets tend to cost more
Diet cost linked to nutrient adequacy (MAR)
Diet cost estimated from other dietary surveys
Attached
retail
prices
Diet
records
FFQ
24 hour
recalls
US, Spain,
Japan, France
Small-scale studies,
Nurses Health
Study, NHANES
Attached prices to FFQ,
diet records, year 2000
Diet cost and diet quality in French studies
SES, diet cost and diet quality studies from the US
Attached prices to FFQ,
diet records in the US,
year 2004
Linking diet costs with
diet quality and BMI
Diet cost and diet quality evidence from Spain and Japan
Willett group attached
prices to Nurses Health
Study FFQs
Evidence from Nurses Health Study
Evidence from US National level Surveys
National prices attached
to NHANES
Recent evidence from Mexico dietary data
National prices attached
to ENSANUT 2012
Recent article from Lancet
Local prices
attached to FFQ
Geographic dimension of food accessAre healthy foods available and physically accessible to the consumer?
Supermarket within 1 mile from home
Neighborhood-centric approach
Geographic dimension of food access
Are healthy foods available and physically accessible to the consumer?
Moved to person-centric approach
https://vimeo.com/67365274
Check out this video!
Moved out of neighborhood boundaries to define
food environment
Geolocated
food sources
Source: SOS I. Developed by UFL.
Moved out of neighborhood boundaries to define
food environment
Geolocated
places of food
purchase and
consumption
Source: SOS I. Developed by UFL
Moved out of neighborhood boundaries to define
food environment
Geolocated
respondent’s
home/ work/
primary activity
Source: SOS I. Developed by UFL
Connected the two!
Source: Aggarwal et al, Am J Pub Health 2014. https://www.ncbi.nlm.nih.gov/pubmed/24625173
Opens the door to consumer-centric metrics of
geographic access
• Distance to primary food stores
• Distance to fast-food restaurants
• Distance to restaurants
• Distance to convenience stores
Home
• Distance to primary food stores
• Distance to fast-food restaurants
• Distance to restaurants
• Distance to convenience stores
Work/ primary activity
Physical distance to food shopping destinations not
linked to diets or obesity in King County
Geographic access not a barrier in King County, huge disparities in obesity prevalence still exist
4
4
8
9
13
15
17
20
22
23
25
29
38
0 10 20 30 40 50 60 70 80 90 100
Whole Foods
Madison market
Trader Joe
Met Market
PCC
Red Apple Market
QFC HI
Costco
Fred Meyer
Top Foods
Safeway
Windo Foods
Albertsons
% Non obese % ObeseSource: Seattle Obesity Study. Unpublished data
Source: Drewnowski et al, Int J Obes 2014.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3955743/
Obesity prevalence
16-22%
Obesity prevalence
36-43%
Geographic distribution of obesity at census tract level
Introduced a spatial dimension to nutrition epidemiology
Source: Drewnowski et al, Int J Obes 2014.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3955743/
Moving from State/ County level indicators down to census to neighborhood level
Source: CDC, BRFSS data 2011
https://www.cdc.gov/obesity/data/prevalence-maps.html
Source: Obesity prevalence, females, age-standardized, 2011 IHME
(http://vizhub.healthdata.org/subnational/usa)
Source: Drewnowski et al, Prev Med. 2015.
https://www.ncbi.nlm.nih.gov/pubmed/26657348
Geographic distribution of diet quality by census blocks -1st time ever!
Psychosocial dimension of food access
• Positive attitude towards eating healthy/ inexpensive/
convenient (5- point Likert scale)
• Cooking-at-home frequency per week
• Time spent cooking and cleaning after meals per day
Using standard previously validated questions from national level
surveys
Psychosocial dimension of food access
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Strongly agree
Somewhat agree
Neutral/ disagrees
Proportion meeting 5-A-Day recommendation
It is im
po
rta
nt
to b
e t
ha
t fo
od
s I
ea
t are
healthy…
Yes NoSource: Aggarwal et al. J Acad Nutr Diet 2014.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3947012/pdf/main.pdf
Positive attitude towards healthy foods was associated with much
higher intakes of fruits and vegetables
Positive attitude towards healthy foods associated with higher quality
diets irrespective of where you shop!
Source: Aggarwal et al. J Acad Nutr Diet 2014.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3947012/pdf/main.pdf
Frequent cooking-at-home associated with better compliance with dietary guidelines at no extra cost
Source: Tiwari et al, AM J Prev Med. 2017.
https://www.ncbi.nlm.nih.gov/pubmed/28256283
Good news! Healthier diets need not cost more. There is a huge variability in diet quality at every
level of cost
Source: Aggarwal et al, Prev Med. 2016.
https://www.ncbi.nlm.nih.gov/pubmed/27374943
Evidence from SOS data
The concept of nutrition resilience
Source: Aggarwal et al, 2017. http://www.fasebj.org/content/31/1_Supplement/45.8.short
Median = 72.0
Median = 8.8
Resilients!H
EI-
20
10
The concept of nutrition resilience
Evidence from NHANES
Source: Aggarwal et al, Prev Med. 2016.
https://www.ncbi.nlm.nih.gov/pubmed/27374943
Having the right attitude is linked with higher quality diets at all levels of income and education, NHANES 2007-10
40
45
50
55
60
65
<High school High school Some college College
Me
an
HE
I-2
01
0 s
co
re
Education level
Not important Somewhat important Very important
40
45
50
55
60
65
<1.3 1.3 - 1.849 1.85-3.99 ≥4.0
Me
an
HE
I-2
01
0 s
co
re
Family income-to-poverty ratio
Not important Somewhat important Very important
“When you buy foods from grocery store or supermarket, how important is nutrition”
Source: Aggarwal et al, Prev Med. 2016.
https://www.ncbi.nlm.nih.gov/pubmed/27374943
Summary
1. Toolbox of metrics provided in-depth understanding of the food environment in King County
• Individual-level estimates of diet cost (economics)
• Physical distances to food shopping destinations (geographic access)
• Food-related attitudes and practices (psychosocial access)
• Nutrition resilience (interactions)
2. Deployed the toolbox in suburban and rural Counties, WA State.
Next steps
1. Scale up the toolbox of metrics to low and middle income countries.
Source: http://www.thelancet.com/infographics/obesity-food-policy
The food system is an interconnected
network of producers, industry and
institutions. But its heart is the individual.
Lancet Obesity 2015 series. explored
Our metrics toolbox aligns with the Lancet 2015 scheme
✓ Activities in the food system
(production, distribution, storage,
marketing)
✓ Food prices and economic barriers
✓ Geographic barriers
✓ Importance of food preferences and
psychosocial barriers
✓ Interactions among all