Origins of the Project

The ONQI project traces its origins to July, 2003. At that time, the U.S. Secretary of Health convened a group of 15 academic thought leaders to share ideas for improving dietary intake patterns in the United States, and specifically for curtailing the spread of epidemic obesity, with him, FDA Commissioner Mark McClellan, and the Directors of the NIH and CDC.

Specifically, the following recommendations were made:

Develop a universal guidance system based on the relative nutritiousness of foods within and across all categories.
To do so, convene a multidisciplinary group of leaders in nutrition and its implications of public health under the auspices of the US FDA or the Institute of Medicine (IOM).
Provide the group the financial support and time required to convert their collective knowledge and judgement into a universally applicable algorithm for nutrition scoring.
Convert nutrition quality scores into a set of symbols, reflecting relative nutritional quality, interpretable at a glance.
Apply those symbols to the entire food supply.
Provide the requisite support for on-going oversight of the scoring system, and its timely revisions based on changes in nutrition science, epidemiology, and/or the food supply.

In late 2005, a confluence of circumstances resulted in the willingness of Griffin Hospital, a Yale-affiliated, non-profit community hospital in Derby, CT, and home to Yale University’s Prevention Research Center, to provide financial and material support for the very project proposed to the Secretary of Health in 2003.

The ONQI project was thus initiated in late 2005, and the first draft of the ONQI algorithm was created in February, 2006.

Methods

The ONQI project was initiated in an explicit effort to satisfy the criteria proposed to the US Secretary of Health in 2003, and to replicate, to the extent possible, the very process that the FDA or IOM “would have” applied. The salient elements in such a process include:

Engagement of a multidisciplinary team of nutrition and public health scientists
Avoidance of conflicts of interest
Isolation of the science-based development effort from any outside or commercial influence
Reliance on objective sources of data to the extent possible
Transparency of methods
A fully iterative process, with consensus-based decision-making

The sequence of methodologic steps was approximately:

A review of literature pertaining to nutrition guidance and nutrient profiling from both the US and abroad by staff at the Prevention Research Center.
Concurrent with #1, a delineation of a master list of
potential members of the ONQI Scientific Expert Panel
based on a need to include thought leaders in nutrition science and public health nutrition; diverse disciplines; and the representation of leading institutions and organizations.
A winnowing of the list to achieve a panel of between 10 and 15 participating consultants.
Invitations extended to a total of 16 nutrition and public health scientists.
Fourteen of those invited to join the ONQI development panel did so; 3 of these subsequently stepped down. One participant stepped down after recognizing a potential conflict of interest; one did so after recognizing a personal conflict with the goal of the project; and one simply did not participate in the information exchange process and was lost to attrition. The final panel of 11 consultants represented fields from nutritional biochemistry to chronic disease epidemiology, pediatric endocrinology to behavior modification.
The identities of the consulting scientists were kept in confidence to avoid any and all outside influence.
The consultants were compensated for their time but did not have any financial interest in the final product; the details of the agreement between the panel members and Griffin Hospital were addressed in a letter of understanding.
Beginning in early 2006, a draft of the ONQI algorithm and its justifications were circulated to panel members for review and critique. The basic formula is based on the Dietary Reference Intakes (DRIs); Food and Drug Administration (FDA) Nutrition Facts Panel; US Department of Agriculture (USDA) MyPyramid and the Dietary Guidelines for Americans, 2005; and relevant international standards.
An iterative process of editing took place based on feedback from panel members via e-mail and at regularly scheduled conference calls, with process oversight and logistical support rendered by staff at the Prevention Research Center.
1. Over a period of 18 months, starting in March 2006, 13 conference calls were held to make revisions to the ONQI based on consensus, and to address validation/testing.
2. Work was managed in between calls primarily via e-mail exchanges.
Once an initial, consensus-based version of the ONQI formula was in circulation, validation testing was initiated. Consensus approval of the form of the algorithm was deemed commensurate with face validity.
Construct validity testing was conducted in the following manner:
1. Members of the ONQI panel were given foods lists of varying length and variety and asked to rank-order the foods on the basis of overall nutritional quality across food categories (lists of approximately 20 foods) and both within and across food categories (lists of approximately 100 foods).
2. A mean ranking for the ONQI panel was established.
3. The ONQI algorithm was used to rank order the same foods, and correlation analysis was performed.
4. This process was terminated with a food list of roughly 100-120 items, and with a correlation coefficient between ONQI and expert panel rankings of between R= 0.88 and R = 0.92, depending on certain fine adjustments to the algorithm.
5. At that point, the process was reversed: the ONQI was used to rank order over 1000 foods, and the relevant rank-ordered lists were circulated to the expert panel members for review. Any apparent anomalies identified in the food rankings resulted in scrutiny of the nutrient data entered, and a determination to either acknowledge that the ONQI ranking was correct, or to adjust the algorithm. This process continued until no further adjustments to the algorithm were deemed necessary.
6. Plans for criterion validity testing have been developed, and planning for this activity is now under way.
The initial working version of the ONQI algorithm was finalized in July 2007.