High school whiz kid retracts PLoS ONE herd immunity paper
It’s pretty impressive to publish two peer-reviewed papers on complicated vaccination models while you’re still in high school. So it’s not surprising that Nathan Georgette, who grew up outside of Jacksonville, Florida, earned a prestigious fellowship from the Davidson Institute for Talent Development.
But perhaps even more impressive is realizing you’ve made a fundamental error in one of those studies, and retracting it while you’re still a college senior at Harvard.
Georgette’s first paper, published in the Internet Journal of Epidemiology in 2007, the year he turned 16, was titled “The Quantification Of The Effects Of Changes In Population Parameters On The Herd Immunity Threshold.” Here’s the abstract:
The purpose was to develop a novel population parameter based (PPB) equation for the herd immunity threshold that incorporates the effects of population dynamics and immunization on the infectiousness of a disease and to analyze these effects. Previous research has not attempted this specific method. The researcher sought to improve cost effectiveness of outbreak response in resource-poor areas. This was achieved by solving for a PPB equation for the basic reproductive number and developing the threshold equation. The researcher applied this equation to three actual measles outbreaks. The PPB equation demonstrates that, using data from the 2003 Marshall Islands measles outbreak, gradual immunization decreases both the effective and the basic reproductive numbers when compared to pulse immunization (from 3.48 to 3.05 and 18.38 to 17.47 respectively). This decreases the potency of the outbreak, thus reducing the associated morbidities, mortalities, and costs.
The paper, Georgette tells Retraction Watch, was “a key part” of his submission to the Davidson Fellowships program.
So how did a 16-year-old kid in Florida develop an interest in outbreaks in poor countries? In 2009, in an interview about being valedictorian of his high school class, he told the Florida Times-Union:
I’ve always had an interest in global issues. In 10th grade, I found on the Internet that hundreds of thousands of children are dying from these diseases that can be cured for $1. … I wanted to see if I could learn more about it. I thought maybe I could somehow help in some small way.
Two years after publishing that paper, and having already applied for the Davidson fellowship, Georgette — now a high school senior — published another one, this time in PLoS ONE, in January 2009. Here’s the abstract of “Predicting the Herd Immunity Threshold during an Outbreak: A Recursive Approach:“
The objective was to develop a novel algorithm that can predict, based on field survey data, the minimum vaccination coverage required to reduce the mean number of infections per infectious individual to less than one (the Outbreak Response Immunization Threshold or ORIT) from up to six days in the advance.
First, the relationship between the rate of immunization and the ORIT was analyzed to establish a link. This relationship served as the basis for the development of a recursive algorithm that predicts the ORIT using survey data from two consecutive days. The algorithm was tested using data from two actual measles outbreaks. The prediction day difference (PDD) was defined as the number of days between the second day of data input and the day of the prediction. The effects of different PDDs on the prediction error were analyzed, and it was found that a PDD of 5 minimized the error in the prediction. In addition, I developed a model demonstrating the relationship between changes in the vaccination coverage and changes in the individual reproduction number.
The predictive algorithm for the ORIT generates a viable prediction of the minimum number of vaccines required to stop an outbreak in real time. With this knowledge, the outbreak control agency may plan to expend the lowest amount of funds required stop an outbreak, allowing the diversion of the funds saved to other areas of medical need.
About a month after the paper went online, Georgette realized there were typos in two equations, so the journal ran a correction.
Fast forward to this past summer, by which point the paper had been cited once. Georgette tells us:
…I had time to perform an in-depth (and more enlightened) review of my high school publications following a course in Ordinary and Partial Differential Equations. Once I had concluded that the PLoS ONE paper’s flaw was fundamental (in Aug 2012), I contacted the editorial board of the journal. From Aug to Oct, I was in correspondence with the PLoS ONE board to produce and finalize the retraction.
The retraction, Georgette notes, was unrelated to the earlier corrections. Here’s what happened. The Internet Journal of Epidemiology paper, he says,
…made an assumption that I did not recognize at the time. Any mathematical model of a natural phenomenon makes assumptions; these assumptions become errors when they are unknowingly violated (as is what happened with the PLoS ONE paper). The Internet J Epi paper didn’t violate its own assumption in its mathematical development and implementation. That said, I am planning to correct the Discussion section of the Internet J Epi paper to clarify the scope and possible applications of the model developed therein.
Here’s the retraction notice:
The author wishes to retract this publication due to a mathematical flaw that undermines the article’s methods and conclusions.
This paper builds upon the author’s previous publication in Internet J Epidemiology . In that previous paper, the author derived a mathematical model for the basic reproduction number from a modified SIR system of differential equations. During this derivation, an integration step was performed that implicitly assumed the per-susceptible-person rate of immunization, ρ, is constant for a given epidemic. At the time, the author did not recognize the existence of this assumption and unknowingly violated it while building upon  to develop this current paper: the author founded this work on a description of ρ as changing through time. This violation of an underlying assumption renders invalid the results of the current paper.
The author apologizes to the readers and editors of PLOS ONE for this error and the delay in its recognition.
 Georgette N (2007) The Quantification Of The Effects Of Changes In Population Parameters On The Herd Immunity Threshold. Internet J Epidemiology 5(1).
Note that this fundamental flaw wasn’t caught by peer reviewers.
In 2009, Georgette told the Florida Times-Union:
I have three main ideas: a medical researcher for infectious diseases, practice medicine in the field or go into human rights law with a focus on health care. I believe everyone in the world should have equal access to health care.
Based on the kind of rigor and transparency with which he handled these errors, we’d say he has a good chance of success in any of those fields.