Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
NLM 2012 Nutritional Informatics Lecture
Transcript of NLM 2012 Nutritional Informatics Lecture
"The effective retrieval, organization, storage, and optimum use of information, data, and knowledge for food and nutrition related problem solving and decision making. Informatics is supported by the use of information standards, information processes, and information technology”. Adopted from the definition of biomedical informatics by Shortliffe & Cimino Everything that flows into your mouth -Food Frequency Questionnaires Nutritional Informatics defined... Starts with understanding the Food Stream How we measure what we eat John F. Hurdle, MD, PhD
Dept. Biomedical Informatics
University of Utah 2001 2002 R01
high-performance computing for advanced
clinical narrative preprocessing R21
POET: consolidated, comprehensive clinical text preprocessing F38
Statistical NLP Analysis of Cross-discipline Clinical Text 2003 2007 2011 VA HSR&D NLM NLM NLM R13
(J. Pestian, PI)
Developing and evaluating natural language processing methods that identify emotional language in suicide notes 2010 What is nutrition? USDA Economic Research Service estimates 77.7% of the food (as meals) streaming into our mouths comes from food eaten at home, presumably mostly from grocery stores "medical informatics"[mh] (nutritional OR nutrition) => 2,053 hits (May 2012)... The RastaPickle This is new balanced-diet metaphor
promulgated by the USDA Todd, Jessica E., Lisa Mancino, and Biing-Hwan Lin. 2010. “The Impact of Food Away From Home on Adult Diet Quality.” ERR-90, United States Department of Agriculture, Economic Research Service. -Dietary recall -Food diaries (gold standard) -Pros and Cons? Cons -FFQs and DRs rely on memory
(can you remember what you ate for lunch last Tuesday? Over past month?) -Force people to map from foods to types of food (e.g., bread=>grains) -People tend to cheat ("Oh that Snickers bar was so small...") Pros -None Okay, that's not fair. There is one: they are what we have to work with. -The percentage of children aged 6–11 years in the United States who were obese increased from 7% in 1980 to nearly 20% in 2008. The obesity pickle -Adult obesity rates: the South has the highest prevalence of obesity (29.4%) followed by the Midwest (28.7%), Northeast (24.9%) and the West (24.1%). -In 2008, medical costs associated with obesity were estimated at $147 billion; the medical costs paid by third-party payors for people who are obese were $1,429 higher than those of normal weight. No dark blue states No dark blue states! Okay, two: measure the whole food stream. Is there something better? VA HSR&D VA Informatics Post-doc Our informatics solution: make dietary pattern data as routine in the EHR/PHR as serum cholesterol. Few clinicans are trained to advise patients about diet and nutrition. But if we can't measure it, clinicans can't use it. So... use it to model dietary patterns at a resolution to detect clinically significant trends, and... If we could collect a family's point-of-purchase food sales data, and... integrate those data into the EHR and PHR, then maybe we can start shifting the BMI curve => a measure that clinicians can monitor and discuss. The current dietary measures are blunt tools, too hard to use in real world at scale. Painless data collection; no bias; no memory issues; virtually free data; collected at scale Sounds simple: map
UPC to nutrition... Zut alors, not so simple.
FDA mandates facts panel; manufacturers provide UPCs, but there is no non-commercial DB that links them (commercial fees as high as $1/UPC; typical grocery store has >30,000 items). I'd like to thank:
Kristina Brinkerhoff, PhD
Phil Brewster, PhD
Valli Chidambaram, MS
Ed Clark, MD (NCS Utah)
Kris Jordan, PhD
Patricia Guenther, PhD (USDA)
Susan Krebs-Smith, PhD (NCI)
NLM T15 Training Grant Our current work Our hypothesis: there is sufficient signal in the grocery item sales time series to model HEI and other dietary patterns at a resolution that is clinically significant. Our raw data:
Grocery item descriptors
Grocery database db schema NLM We work with a national grocery chain to collect PoP UPC data.
We have one year's data on 50 families for study, and a 2M item snapshot for testing.
We built regression models to predict BMI, sat fat, etc. from sales data on 50 families:, using FFQ; a proof of concept.
We are actively Web crawling UPC Web sites, to improve sales description mapping (fuzzy text matching). Healthy Eating Index and mortality in a nationally representative elderly cohort. Rathod AD, Bharadwaj AS, Badheka AO, Kizilbash M, Afonso L.
Arch Intern Med. 2012 Feb 13;172(3):275-7. PMID:22332163 Scale invariant HEI tracked in the EHR, like a vital or a lab test Some things to consider:
Giving clinicians something to measure gives them something to monitor and discuss.
Not eating enough green veggies? Clinicians knows from sales data what green veggies patient does eat (when they eat them...)
Easier to focus on components of HEI...maybe most of the problem is with SoFAAS or maybe too little grain=> informed advice is better than "Lose some weight..." T15
University of Utah Biomedical Informatics Training Grant 2012 NLM Now ye may suppose that this is afoolishness in me; but behold I say unto you, that by small and simple things are great things brought to pass; and small means in many instances doth confound the wise.
Alma 37:6 (BoM) http://www.extension.iastate.edu/foodsavings/plan/menuplanning/plate/ Calories: eat less, move more... Shifting public issues take time...