Machine learning models demonstrate remarkable potential to correct self-reported anthropometric data and improve the accuracy of obesity prevalence estimates.

In our latest work, “Building Machine Learning Models to Correct Self-reported Anthropometric Measures,” we take on the challenge of combating biases in self-reported data prevalent in obesity research, leveraging the advanced capabilities of Machine Learning (ML).

Obesity is an alarming health crisis in the US, impacting more than 42% of adults. This condition is a risk factor for severe health complications, including diabetes, heart disease, and certain types of cancer. It also contributes to significant economic burdens borne by both individuals and health care systems. The accurate monitoring of obesity trends is critical in informing and shaping successful public health strategies that could potentially save countless lives and significant financial resources.

Historically, health surveys have been the primary tool for tracking these trends. However, they encounter significant hurdles due to their dependence on self-reported data. Essential anthropometric measurements such as height and weight can be significantly distorted due to recall errors or respondents’ desire to conform to societal expectations. Under or over-reporting these metrics inevitably clouds the accuracy of obesity prevalence estimates.

Our research takes aim at this issue by harnessing ML, a field that is rapidly growing within the public health sector. This growth is fueled by tremendous advances in computing power, vast amounts of health-related data, and the democratization of data science that has made ML tools widely accessible.

We utilized data from the National Health and Nutrition Examination Survey (NHANES), spanning the years 1999 to 2020, to develop nine ML models. These models were designed to predict height, weight, and body mass index based on their self-reported counterparts. We then compared these models using the root-mean-square error, a commonly used measure in prediction models. The results revealed that the Extreme Gradient Boosting model performed exceptionally well, significantly reducing the discrepancy between self-reported and objectively measured obesity prevalence by over 99%.

Our findings suggest that ML models can effectively mitigate biases in self-reported anthropometric data, providing health researchers and practitioners with a reliable tool for improving the accuracy of obesity prevalence estimates. This holds tremendous implications for public health practice and policy making, offering a more accurate picture of obesity prevalence and the underlying risk factors.

Harnessing the Power of AI in Public Health

The emergence of artificial intelligence (AI) has created a plethora of opportunities in health research, a fact our study reinforces by showcasing the significant potential of ML in addressing complex problems, including obesity research. By utilizing the advanced capabilities of ML and deep learning (DL), we can refine our understanding of obesity prevalence and its risk factors, and drive more targeted, data-driven public health interventions.

Read our article in JPHMP

AI is revolutionizing health care, laying the foundation for a new era of predictive, personalized, and efficient care. AI can efficiently process large data sets to identify disease trends, predict potential outbreaks, and enable a more proactive, rather than reactive, approach to public health management. It opens the door to personalized health care by analyzing electronic health records, genetic data, and lifestyle factors, assisting in predicting individual health outcomes. Furthermore, AI can tailor health promotion campaigns to resonate with individuals based on their unique characteristics and health profiles. AI can optimize the allocation of health care resources, predicting where assistance will be most needed.

In summary, AI holds the potential to be an invaluable partner in public health research and practice. Our study provides further evidence of its capacity to tackle complex health issues and its ability to transform our understanding and management of major public health challenges like obesity.

Ruopeng An, PhD, MPP, FACE. Dr. An is an Associated Professor at Brown School, Washington University in St. Louis. His research aims to develop a well-rounded knowledge base and policy recommendations that can inform decision-making and the allocation of resources to combat obesity.

Mengmeng Ji, MBBS, PhD, is an epidemiologist and public health professional.  Her research interests include obesity, obesity-related cancers, and health disparities. She is currently a postdoctoral research associate at Washington University School of Medicine in St. Louis.

jphmpdirect