The human race is more overweight today than it has ever been in its relatively short stay on Earth. The World Health Organization (WHO) reports in 2016 about 1.9 billion adults were overweight, a third of whom were obese – three times more than in 1975.
Obesity among children, however, is increasing at a more alarming rate. The number of obese children aged five to 19 has increased tenfold in the last 40 years, from 11 million in 1975 to around 124 million in 2016, The Imperial College London and WHO reported .
These figures, coupled with calls to curb the current trend, have led health practitioners and researchers to develop new and more creative measures.
A group of researchers in New York believe that part of the answer lies with new technologies, machine learning to be specific. They claim to have come up with an algorithm that uses health records to help predict possible obesity in children, with tests as accurate as 76% and 81% for boys and girls respectively.
When determining the likelihood of obesity among young children, health practitioners, consider a range of factors, including the health of both the mother and the fetus during pregnancy and during the child’s first years of life.
Other factors can also be considered. The group at the New York University School of Medicine cast their slightly somewhat wider, factoring in as many variables as possible. These range from demographic information, home addresses, vital signs, laboratory tests and results, medication and diagnosis codes, all of which identified in over 19,000 variables. These were categorized and broken into various time periods: pre-birth, post-pregnancy, and up to 11 intervals during the baby’s first two years.
In order to obtain all the necessary information, the researchers needed children and their mothers who had made all the necessary visits to the doctor in order for their electronic health records (EHR) to be complete.
The group managed to find 3,449 children, nearly exactly half or whom were boys.
When the algorithm worked through the data, it is not only managed to accurately determine which children are at a higher risk of becoming obese, but helped the team establish the most important factors in predicting the risk of obesity by the age of five. “We found that weight for length z-score (determining the weight and length deviation from the average), BMI between 19 and 24 months, and the last BMI measure recorded before age two were the most important features for prediction,” they wrote.
(According to WHO, obesity among adults is defined as a BMI great than or equal to 30 and among children under the age of five, obese children have a weight-for-height greater than 3 standard deviations above the WHO Child Growth Standards median ”)
This is not the first study to attempt to predict the risks of obesity in young children. Neither is prediction in itself a new concept. However, using an algorithm, health practitioners could have crucial data for their patients at their disposal. That is at least if their patients made every required visit in order for their data to be captured in full.
“Our results suggest that machine learning approaches for predicting future childhood obesity using EHR data could improve the ability of clinicians and researchers to drive future policy, intervention design, and decision-making process in a clinical setting,” the researchers concluded.
There is an understandable need to curb the epidemic, especially as the Imperial College London and WHO predicted that if not, obese children and adolescents will moderately or severely underweight children by 2022 – a serious problem on its own.
Given the serious health risks associated with obesity, which range from cardiovascular diseases, diabetes and some cancers, data can assist in managing eating and activity habits. Potentially, technology could play an increasingly important role in ensuring healthier lifestyles for future generations.