TL;DR:
- Researchers utilized machine learning to predict adulthood obesity based on risk factors and BMI values in the initial 1,000 days of life.
- Modifiable risk factors include maternal BMI, weight gain during pregnancy, socioeconomic status, neonatal weight, and neighborhood variables.
- The study developed a dynamic BMI tracker using ML algorithms, enabling the identification of children at risk of obesity and informing prevention strategies.
- The model incorporated various variables related to height, weight, BMI, and time differences during specific age periods.
- The tracker accurately estimated childhood BMI within three six-month intervals, supporting clinicians and population-level efforts in obesity prevention.
Main AI News:
In a recent study featured in Scientific Reports Journal, researchers have leveraged a machine learning (ML) approach to forecast adult obesity by assessing risk factors and monitoring body mass index (BMI) values within the initial 1,000 days (between two and four years old) of an individual’s life.
The Growing Concern of Obesity
Across the globe, obesity prevalence has significantly surged among both adults and children. The early presence of excess adiposity during childhood serves as an indicator of adult obesity, cardiometabolic risks, and pediatric morbidities.
The Challenges of Addressing Obesity
Once established, obesity becomes challenging to treat and tends to persist. Consequently, research efforts have prioritized obesity prevention, with a focus on identifying individuals at a higher risk of developing adiposity in adulthood, thereby enhancing prevention strategies.
Uncovering Modifiable Risk Factors
Identifying modifiable risk factors is crucial in combatting obesity. These factors include higher BMI values for mothers prior to pregnancy, weight gain during pregnancy, low socioeconomic status, high neonatal weight, and neighborhood-level variables such as crime rates and food accessibility. However, data regarding the combined risk estimation potential of these variables is limited.
The Gap in Pediatric Obesity Estimation
Despite numerous studies highlighting the critical importance of antenatal and initial neonatal periods in determining obesity risks, existing efforts to estimate pediatric obesity are scarce. The age range of two to four years old presents a crucial developmental phase with increased flexibility and opportunities to influence health behaviors.
Insights into the Study
The researchers in this study employed ML algorithms to identify children at a heightened risk of obesity. The objective was to provide valuable insights into obesity prevention policies and the development of effective strategies. Furthermore, they designed a dynamic, predictive BMI tracker that can be utilized during childhood to identify the risk of obesity in adulthood.
Refining the Estimation Models
The team employed the least absolute shrinkage and selection operator (LASSO) regression to select features with the most significant coefficients and relevance to pediatric obesity, excluding height, weight, and BMI. They developed estimation models using support vector regression (SVR) with fivefold cross-validation to estimate BMI within the periods of 30 to 36 months (4,204 individuals), 36 to 42 months (4,130 individuals), and 42 to 48 months (2,880 individuals). Individuals without at least one clinical encounter within all the specified periods were excluded.
The Development Process The model development process involved gathering and integrating raw data, preprocessing the data, conducting feature engineering, training the models, and validating the tracker. The tracker was trained using 80.0% of the individuals’ data from all periods, serving as the training dataset.
Data Sources and Outcome Measures
The researchers retrieved electronic health records (EHRs), birth certificates, and geocoded data from the Obesity Prediction in Early Life (OPEL) registry spanning from 2004 to 2019. The study outcome was BMI based on participant age and gender, aligning with the recommendations of the Centers for Disease Control and Prevention (CDC).
Promising Results
The OPEL registry consisted of 149,625 visits involving 19,724 individuals aged between 0.0 and 48.0 months. Among these individuals, 10,348 were analyzed, with 4,204, 4,130, and 2,880 falling into the age ranges of 30.0 to 36.0 months, 36.0 to 42.0 months, and 42.0 to 48.0 months, respectively.
Following the elimination of erroneous records, imputation of missing values, and scaling of exposure variables, 50 variables were selected for analysis. After applying LASSO regression, data augmentation, and univariate tests, 19 variables were further analyzed.
Key Variables in the Model
The model incorporated variables such as mean height, BMI, and weight during the periods of 0.0 to 8.0 months, 8.0 to 16 months, and 16 to 24 months. It also considered the time differences between the final encounter within each period and the encounter prior to two years of age. Additionally, it incorporated variables such as mean age, weight, height, BMI, weight and height percentiles at two years, and estimation time differences between the final visit before two years and the target visit during any of the specified periods.
Validation of the Tracker
The accuracy of the tracker was assessed using a validation dataset consisting of 20.0% of the patients. The results indicated that the tracker provided precise estimations of childhood BMI, with a mean error of 1.0 within the periods of 30.0 to 36.0 months, 36.0 to 42.0 months, and 42.0 to 48.0 months.
Significant Correlations and Insights
Most variables included in the model exhibited significant correlations with pediatric BMI across all estimation ranges. These findings signify that the tracker could assist clinicians and population-level initiatives in preventing obesity during the initial stages of life. The study also revealed modifiable factors associated with higher childhood BMI during the prenatal and initial infancy stages, such as maternal risk factors during pregnancy, C-section delivery, higher infant birth weight, and sleep patterns. Furthermore, the study identified protective factors against elevated BMI, including the proportion of individuals residing in food deserts and Hispanic ethnicity.
Concluding Remarks
In conclusion, this study showcases the potential of ML and modifiable risk factors during early childhood to assess pediatric BMI trajectories. By intervening before the onset of unhealthy adiposity, healthcare providers can effectively reduce the burden of obesity. Factors such as maternal health, the quality of a child’s sleep, and socioeconomic factors were identified as influential in shaping weight trajectories during later childhood. Unlike existing models, the BMI tracker developed in this study can predict BMI within three future six-month intervals, presenting an extended observation period for pediatric providers to monitor changes in BMI.
Conclusion:
This study demonstrates the potential of machine learning in predicting and addressing adulthood obesity by analyzing risk factors and BMI values in early childhood. The findings highlight the significance of modifiable factors such as maternal health, socioeconomic status, and neonatal characteristics in shaping weight trajectories. The developed BMI tracker provides healthcare providers with a valuable tool to monitor BMI changes over extended periods, enabling timely interventions and reducing the burden of obesity.
This research offers important insights for the market, as it emphasizes the importance of early prevention strategies and the role of data-driven approaches in tackling obesity-related challenges. Businesses in the healthcare sector can leverage these findings to develop innovative solutions and contribute to obesity prevention efforts, ultimately improving public health outcomes.
