About

Lisa J. Harnack, MPH, DrPH, is a Professor in the Department of Epidemiology & Community Health at the University of Minnesota. She serves as the Head of the Epidemiology & Community Health department and is the Director of the Nutrition Coordinating Center. Her professional work is closely aligned with cancer prevention and control. Dr. Harnack's research extensively addresses nutrition, dietary intake, and public health, with a focus on evaluating food policies and interventions aimed at improving diet quality and reducing food insecurity. She leads and collaborates on numerous research projects, including evaluations of food restrictions in SNAP programs, mobile food markets, and telemedicine-delivered dietary interventions for obesity and type 2 diabetes. Her work contributes to multiple United Nations Sustainable Development Goals, including those related to poverty reduction, zero hunger, good health and well-being, and sustainable communities. Dr. Harnack's expertise encompasses nutrition assessment, dietary habits, and community-based nutrition assistance programs, reflecting her commitment to advancing public health nutrition through rigorous research and policy evaluation.

Research topics

• Computer Science
• Medicine
• Biology
• Sociology
• Political Science
• Environmental health
• Statistics
• Database
• Psychology
• Business
• Food science
• Marketing
• Geography
• Mathematics
• Internal medicine
• Demographic economics
• Environmental science
• Chemistry
• Animal science
• Ecology
• Nuclear medicine
• Demography
• Labour economics
• Pathology

Selected publications

• NutriRAG: unleashing the power of large language models for food identification and classification through retrieval methods

  Journal of the American Medical Informatics Association · 2026-01-09 · 1 citations

  articleOpen access

  OBJECTIVES: This study explores the use of advanced natural language processing (NLP) techniques to enhance food classification and dietary analysis using raw text input from a diet tracking app. MATERIALS AND METHODS: The study was conducted in 3 stages: data collection, framework development, and application. Data were collected from a 12-week randomized controlled trial (RCT: NCT04259632), in which participants recorded their meals in free-text format using the myCircadianClock app. Only de-identified data were used. We developed nutrition-focused retrieval-augmented generation (NutriRAG), an NLP framework that uses a retrieval-augmented generation approach to enhance food classification from free-text inputs. The framework retrieves relevant examples from a curated database and then leverages large language models, such as GPT-4, to classify user-recorded food items into predefined categories without fine-tuning. NutriRAG was then applied to data from the RCT, which included 77 adults with obesity recruited from the Twin Cities metro area and randomized into 3 intervention groups: time-restricted eating (TRE, 8-hs eating window), caloric restriction (CR, 15% reduction), and unrestricted eating. RESULTS: NutriRAG significantly enhanced classification accuracy and helped to analyze dietary habits, as noted by the retrieval-augmented GPT-4 model achieving a micro-F1 score of 82.24. Both interventions showed dietary alterations: CR participants ate fewer snacks and sugary foods, while TRE participants reduced nighttime eating. CONCLUSION: By using artificial intelligence, NutriRAG marks a substantial advancement in food classification and dietary analysis of nutritional assessments. The findings highlight NLP's potential to personalize nutrition and manage diet-related health issues, suggesting further research to expand these models for wider use.

  PublisherOA PDFDOI

• Relative Effects of Time-Restricted Eating, Energy-Restricted Eating, and Unrestricted Eating on Eating Patterns and Dietary Intake: Results From a Randomized Controlled Trial

  Journal of the Academy of Nutrition and Dietetics · 2026-03-01

  articleOpen access1st authorCorresponding
  PublisherOA PDFDOI

• Feasibility of a New Dietary Recall Method: Augmenting Interviewer-Administered 24-Hour Dietary Recalls with Photo-Based Mobile Food Records

  Dietetics · 2026-04-23

  articleOpen access

  Background: Assessing food and nutrient intake is an important yet challenging component of nutrition research, particularly in populations at higher risk for dietary underreporting. Objective: To evaluate the feasibility, acceptability, and preliminary measurement characteristics of augmenting interviewer-administered 24 h dietary recalls with a photo-based mobile food record application (mCC: my Circadian Clock). Design: This was a randomized cross-over feasibility study in which each participant completed two sets of three 24 h dietary recalls. One set consisted of standard interviewer-administered recalls, while the other incorporated dietary intake captured via the mCC app during the 24 h preceding the recall to guide the interview. Participants: Participants (n = 10) were adults aged 18–65 years with obesity (BMI > 30 kg/m2) and less than a college-level education, recruited from a general community setting. Main Outcome Measures: Primary feasibility outcomes included recall adherence, protocol completion, participant burden, and usability of the mobile application. Secondary and exploratory outcomes included average energy intake (kcal/day), number of food items and eating occasions reported, Healthy Eating Index (HEI)-2015 scores, and recall duration. Statistical Analyses: Descriptive statistics and paired t-tests were used to explore differences between methods; analyses were considered exploratory and hypothesis-generating. Results: All enrolled participants completed every scheduled recall, resulting in 100% adherence and protocol completion. Most participants (70%) rated the mCC app as easy or very easy to use, although 60% reported greater burden with the Augmented Recalls. Average energy intake was 274 kcal/day lower with the augmented method compared with Standard Recalls (95% CI: −597, 50; p = 0.09), with no clear differences observed in reported food items, eating occasions, HEI-2015 scores, or recall duration. Conclusions: Augmenting interviewer-administered 24 h dietary recalls with a photo-based mobile food record is feasible and acceptable in adults with obesity, though it did not demonstrate clear improvements in dietary intake capture in this small feasibility sample. These findings provide practical guidance for refining technology-assisted recall protocols and informing the design of future, adequately powered studies.

  PublisherOA PDFDOI

• NutriRAG: Unleashing the Power of Large Language Models for Food Identification and Classification through Retrieval Methods

  medRxiv · 2025-03-20 · 1 citations

  preprintOpen access

  Objective: This study explores the use of advanced Natural Language Processing (NLP) techniques to enhance food classification and dietary analysis using raw text input from a diet tracking app. Materials and Methods: The study was conducted in three stages: data collection, framework development, and application. Data were collected via the myCircadianClock app, where participants logged their meals in free-text format. Only de-identified food-related entries were used. We developed the NutriRAG framework, an NLP framework utilizing a Retrieval-Augmented Generation (RAG) approach to retrieve examples and incorporating large language models such as GPT-4 and Llama-2-70b. NutriRAG was designed to identify and classify user-recorded food items into predefined categories and analyzed dietary patterns from free-text entries in a 12-week randomized clinical trial (RCT: NCT04259632). The RCT compared three groups of obese participants: those following time-restricted eating (TRE, 8-hour eating window), caloric restriction (CR, 15% reduction), and unrestricted eating (UR). Results: NutriRAG significantly enhanced classification accuracy and effectively identified nutritional content and analyzed dietary patterns, as noted by the retrieval-augmented GPT-4 model achieving a Micro F1 score of 82.24. Both interventions showed dietary alterations: CR participants ate fewer snacks and sugary foods, while TRE participants reduced nighttime eating. Conclusion: By using AI, NutriRAG marks a substantial advancement in food classification and dietary analysis of nutritional assessments. The findings highlight NLP's potential to personalize nutrition and manage diet-related health issues, suggesting further research to expand these models for wider use.

  PublisherOA PDFDOI

• Food Access Matters: Quantifying the Price Differential Across Four Major Food Retailers and a Mobile Market Aiming to Improve Affordable Nutritious Food Access

  Journal of Nutrition Education and Behavior · 2025-06-24

  articleOpen access
  PublisherOA PDFDOI

• Pressure-Mediated Reflection Spectroscopy Criterion Validity as a Biomarker of Fruit and Vegetable Intake: A 2-Site Cross-Sectional Study of 4 Racial or Ethnic Groups

  UNC Libraries · 2025-02-28

  articleOpen access
  PublisherDOI

• A scoping review of community-based summer interventions with a nutrition assistance component aiming to improve children’s weight-related outcomes

  BMC Public Health · 2025-03-22 · 1 citations

  reviewOpen access

  BACKGROUND: Children, particularly those from low-income families, experience increased weight gain and food insecurity in the summer months. Summer interventions that include nutrition assistance through free or subsidized food, or money for food, are well-positioned to address food insecurity and obesity. However, there is no comprehensive review of the characteristics and findings of summer interventions that aimed to improve children's weight-related outcomes, including food security, dietary intake, physical activity, and body weight. This study aimed to describe the characteristics and findings of summer interventions with a nutrition component that included child weight-related outcomes. METHODS: For this scoping review, CINAHL, ERIC, Ovid Medline, and Scopus databases were searched using the terms, "summer," "out of school," "food," "nutrition," "meal," "lunch," or "insecurity." Three independent reviewers screened manuscripts for eligibility. RESULTS: Thirteen manuscripts were identified. The majority (n = 10, 77%) of summer interventions offered activities for nutrition education and/or physical activity engagement or education in addition to nutrition assistance. Most interventions (69%) were provided through summer camps or school, and 60% provided nutrition assistance in the form of free meals or snacks through the Summer Food Service Program. Food insecurity was the least studied outcome. The associations between these summer interventions and children's weight-related outcomes were examined using various measures and study designs, with only three randomized controlled studies, two of which had inadequately powered samples. Some quasi-experimental studies documented the positive associations between intervention participation and fruit and vegetable intake, moderate to vigorous physical activity, and BMI z-scores or percentiles, but the findings were inconsistent. CONCLUSIONS: Further studies with more rigorous designs and adequately powered samples are needed to evaluate the effects of multicomponent summer interventions with nutrition assistance to maximize the intervention benefits for children's weight-related health and equity. CLINICAL TRIAL NUMBER: Not applicable.

  PublisherOA PDFDOI

• Evaluation of the Feasibility and Construct Validity of a Novel Method to Measure Household Fruit and Vegetable Procurement in Low-Income Community Settings

  Journal of the Academy of Nutrition and Dietetics · 2025-10-15

  articleOpen accessSenior author
  PublisherOA PDFDOI

• Evaluating the impact of a full-service mobile food market on food security, diet quality and food purchases: a cluster randomised trial protocol and design paper

  BMJ Open · 2025-03-01 · 2 citations

  articleOpen accessSenior author

  INTRODUCTION: Mobile food markets may help to mitigate diet-related and weight-related inequities by bringing low-cost, nutritious food directly to underserved populations. By stocking foods to meet a range of dietary needs, full-service mobile markets may improve multiple aspects of diet, food security and fruit and vegetable procurement with a convenient one-stop shop. METHODS AND ANALYSIS: This cluster randomised trial is evaluating the impact of a full-service mobile market, the Twin Cities Mobile Market (TCMM). The TCMM sells staple foods at affordable prices from a retrofitted bus that regularly visits communities experiencing low incomes. The trial's primary outcome is participant diet quality. Secondary outcomes include intake of specific foods and nutrients, food security and servings of fruits and vegetables procured for the home.Together with our partners, we enrolled four subsidised community housing sites in three waves (12 total sites), aimed to recruit 22 participants per site (N=264) and collected baseline data. Sites were then randomised to either receive the full-service TCMM intervention or serve as a waitlist control, and the full-service TCMM began implementing at intervention sites. Follow-up data collection is occurring at 6 months post-implementation. After follow-up data collection for each wave, the full-service TCMM intervention is being implemented at the waitlist control sites. Waves 1 and 2 are complete and Wave 3 is in progress.At baseline and follow-up data collection, dietary quality and intake are being assessed through three, interviewer-administered, 24-hour dietary recalls, food insecurity is being assessed by the 18-item Food Security Screening Module and fruit and vegetable procurement is being measured by collecting one month of food procurement tracking forms.We will use intent-to-treat analyses to determine if participant diet quality, food security and procurement of fruits and vegetables improve in the sites that received the full-service TCMM intervention relative to the participants in the waitlist control condition. ETHICS AND DISSEMINATION: Trial procedures have been approved by the University of Minnesota Institutional Review Board. We plan to disseminate main outcomes in Grant Year 5 in both scientific and community spaces. TRIAL REGISTRATION NUMBER: ClinicalTrials.gov: NCT05672186.

  PublisherDOI

• Abstract P1105: The Planetary Health Diet Score is Inversely Associated with Sodium Intake in NHANES, Pre-Pandemic (2017-2020)

  Circulation · 2025-03-11

  articleSenior author

  Introduction: In 2019, the EAT-Lancet Commission proposed a reference diet as a way to mitigate the impact of what we eat on the planet by lowering agricultural greenhouse gas emissions, and land and water use. Overconsumption of sodium has been well described to increase CVD risk. Little is known about how following this recommended eating pattern relates to sodium intake in the United States. Hypothesis: We hypothesized that closer alignment with the Eat-Lancet Commission reference diet would be inversely associated with sodium intake. Methods: Dietary recalls from the 2017-2020 Pre-Pandemic dataset from the National Health and Nutrition Examination Survey (NHANES) were used for this analysis. Participants without two 24-hour dietary recalls, those younger than 3, or with implausible average energy intake (<500 or >8000 kcals/day) were excluded, giving us a sample size of N=9800. The Planetary Health Diet Index (PHDI), based on caloric density, has been created as a measure of alignment with the EAT-Lancet reference diet. The PHDI has 14 components, each scored proportionally between 0 and 10; sodium is not one of the scored components. The components are grouped into adequacy, optimum, and moderation categories. PHDI is scored 0-140, with a higher score more closely aligned with the EAT-Lancet reference diet. We calculated the PHDI score and its component scores for each participant using the 2-day averages for 24-hour dietary recall. We then assessed the associations between PHDI score and the 2-day averages of sodium using survey-weighted correlations and multivariate linear regression. Results: In our sample, the average age was 40 years, with 51% identifying as women. The mean PHDI score was 68 (27-124), mean sodium intake was 3250 (298-14435) mg/day, and mean energy intake was 2026 kcals/day. With unadjusted analysis, an inverse correlation between PHDI score and sodium was observed (r = -0.18). After adjusting for gender, age, education, and income the significant inverse association persisted (r=-0.17). For every 10 point increase in PHDI, sodium intake decreased by 140 mg. Conclusions: While the EAT-Lancet recommended diet does not explicitly include a recommended limit on sodium intake, these findings suggest that if someone adheres to the recommended diet they are apt to have a lower sodium diet in comparison to someone who does not. These results are specific to the American context and we cannot be sure they would extrapolate to other countries.

  PublisherDOI

Recent grants

• NIH Grant R01DK067599

  NIH · $654k · 2009

• Designing a Food Benefit Program to Optimize Diet Quality for Obesity Prevention

  NIH · $1.5M · 2013–2018

• Food and Nutrition System for Research

  NIH · $9.0M · 1998–2014

• NIH Grant R03HL064592

  NIH · $142k · 2002

Frequent coauthors

• Yori Gidron

  University of Haifa

  460 shared

• Steven Gambert

  University of Maryland, Baltimore

  190 shared

• Deborah Lee Young-Hyman

  National Institutes of Health

  186 shared

• Jennifer Pellowski

  John Brown University

  182 shared

• Mustafa Al’Absi

  University of Minnesota

  182 shared

• Barbara Resnick

  University of Maryland, Baltimore

  180 shared

• J. Rick Turner

  IQVIA (United States)

  178 shared

• Kristin E. Anderson

  136 shared

Education

• Other, Public Health

  University of Minnesota

  2015

• Other, Public Health

  University of Minnesota

  2009

• B.S., Nutrition

  University of Minnesota

  2005