Obesity Related Mortality in the Next Generation: Projections Based on Machine Learning for Young Americans (1999–2035)

Abstract

ABSTRACT

Introduction

Obesity is a growing public health crisis, particularly among young adults. This study projected obesity related mortality rates in the United States, providing an outlook on future trends.

Methods

Data were obtained from the CDC Wide‐ranging Online Data for Epidemiologic Research (CDC WONDER) multiple causes of death database, focusing on the underlying cause of death as obesity (ICD‐10 codes E66.0‐E66.9) for young adults aged 18–39. Temporal trends in crude mortality rate (CMR) were assessed using Joinpoint software. Future mortality rates were forecasted using an optimal Autoregressive Integrated Moving Average (ARIMA) model.

Results

Between 1999 and 2023, there were 19,451 deaths with obesity as the underlying cause of death (males 60.58%) in young adults aged 18–39. Overall, the CMR showed an annual percent change (APC) of 3.13% (95% CI 2.45%–3.85%) for 1999–2023. The forecasted APC for 2024–2035 was 1.30% (95% CI 0.35%–2.31%). For females, CMR increased from 0.4/100,000 in 1999 to 0.8/100,000 in 2023, showing an APC of 2.44% (95% CI 1.71%–3.18%) with a forecasted APC of 2.81% (95% CI 2.74%–2.87%). For males, CMR increased from 0.6/100,000 in 1999 to 1.4/100,000 in 2023, showing an APC of 3.59% (95% CI 2.78%–4.42%) with a forecasted APC of 3.16% (95% CI 3.12%–3.20%).

Conclusion

This study demonstrated the efficacy of machine learning in projecting public health trends, providing critical insights into the future burden of obesity‐related mortality among young adults in the United States.

Introduction

Obesity has emerged as one of the most pressing public health challenges of the 21st century, as the global prevalence of obesity has nearly tripled since 1975 and continues to escalate at an exponential rate, emerging as the leading lifestyle-related risk factor for premature death [1]. Data from the National Health and Nutrition Examination Survey (NHANES) indicate that obesity prevalence in 2007–2008 was 33.8%, which doubled during 1976–1980 and increased by 50% from 1988 to 1994 [2]. In 2017–2018, the prevalence of obesity was 42.8%, which equates to 76 million American adults at risk for serious and costly chronic conditions [3]. Projections based on linear time trends indicate that by 2030, 51% of the population will be obese [4].

In the United States, obesity rates have risen sharply, particularly among young adults, who now face increasing health risks associated with this condition [2]. Among young adults aged 18–39, the impact of obesity is particularly concerning due to the potential for long-term health complications and reduced life expectancy. The Center for Disease Control and Prevention (CDC) reports that the prevalence of obesity among young adults aged 20–39 in the United States was 40% in 2017–2018, up from 30% in 1999–2000 with an increase in the prevalence of severe obesity from 4.7% to 9.2% [5].

Obesity has been linked to higher incidence and mortality rates of numerous diseases, such as certain cancers, cardiovascular disease, and diabetes, with obesity itself increasingly recognized as a disease [6–8]. In 1999, it was estimated that the mean number of annual deaths attributable to obesity in the United States was 280,184, with over 80% of these deaths occurring in individuals with a BMI over 30 kg/m² [9]. However, in 2016, excess weight in the US was responsible for over 1300 deaths per day, accounting for nearly 500,000 deaths annually, surpassing the mortality impact of smoking. This led to an average life expectancy loss of nearly 2.4 years [10]. Although the impact of excess weight on overall mortality has been estimated for the general population, little is known about how excess-weight related mortality varies by age group within the USA [9, 11–13].

Machine learning has garnered substantial attention due to its impressive capability in delivering reliable predictive analyses [14]. Unlike traditional methods, recent studies have demonstrated that machine learning excels in handling high-dimensional datasets and deciphering complex relationships among multiple variables [15]. However, to date, most machine learning models have concentrated on predicting mortality trends in young adults with obesity in America. This study, by examining temporal trends in mortality rates, aims to project future mortality rates up to 2035 among young adults for obesity. As obesity continues to rise among young adults, the associated mortality rates are expected to follow, highlighting the urgency for effective prevention and intervention measures.

Methods

The study adhered to STROBE reporting standards and did not require informed consent or institutional board approval as it used anonymized public data, following the Common Rule.

We retrieved de-identified data from the CDC WONDER multiple causes of death database (1999–2023) focusing on obesity-related mortality among individuals aged 18–39 years. The study population included patients classified under the International Classification of Diseases-10 codes (E66.0-E66.9) for obesity [16].

All statistical analyses were performed using Joinpoint Trend Analysis Software and Python with the PyCharm Integrated Development Environment (IDE) and Generative Pre-trained Transformer 4 (GPT-4) [17]. These tools were chosen for their advanced statistical capabilities, enabling more complex analyses and data visualization. We examined crude mortality rates (CMRs) per 100,000 population for individuals aged 18–39 years, standardized to the 2000 US population. Joinpoint software assessed Annual Percent Change (APC) temporal trends with 95% confidence intervals (CI), representing the change in mortality during the study period. To ensure reliable analysis, we employed Joinpoint's Empirical Quantile Confidence Interval method, providing robust 95% confidence intervals for APC [17]. This method offers a clearer understanding of the result reliability and significance without relying on p values.

We focused on the 18–39 age range to capture the emerging trend of rising obesity-related mortality in younger populations, a demographic increasingly affected by obesity but often overlooked.

For predictive time series analysis, we used the autoregressive integrated moving average (ARIMA) model for non-stationary data to forecast mortality rates until 2035 [18]. The ARIMA model was selected for its robustness in handling non-stationary time series data and its widespread use in healthcare forecasting. Compared to other models, ARIMA provides a more nuanced understanding of time-dependent patterns. The optimal ARIMA model, identified using the auto ARIMA function based on the Bayesian Information Criterion (BIC), was fitted to the data. The model's residuals were evaluated for white noise using the Ljung-Box test. The model's robustness was validated through time series cross-validation (n = 3), with Root Mean Squared Error (RMSE) reported for accuracy [18].

Results

Between 1999 and 2023, there were 19, 451 deaths with obesity as the underlying cause of death (males n = 11, 784 60.58%) for young adults aged 18–39. Overall, the CMR increased from 0.5/100,000 in 1999 to 1.1/100,000 in 2023, showing an APC of 3.13% (95% CI 2.45%–3.85%) (Figure 1).

[IMAGE OMITTED. SEE PDF]

For females, CMR increased from 0.4/100,000 in 1999 to 0.8/100,000 in 2023, showing an AAPC of 2.44% (95% CI 1.71%–3.18%). For males, CMR increased from 0.6/100,000 in 1999 to 1.4/100,000 in 2023, showing an APC of 3.59% (95% CI 2.78%–4.42%).

The ARIMA model was used to forecast mortality rates until 2035. For all forecasting, the best ARIMA models were selected because they had the lowest BIC and passed the Ljung-Box test, which shows that the residuals are independently distributed. This approach was chosen because it reduces information loss while effectively capturing the main trend in the data. The forecasted rate for 2024 was 1.3/100,000 (95% CI: 1.09–1.5) with a projected increase to 1.51/100,000 (95% CI: 1.21–1.81) by 2035 with an APC of 1.30% (95% CI 0.35%–2.31%); ARIMA model [1, 1, 1] with RMSE = 0.122).

For males, the forecasted rate for 2024 was 1.28/100,000 (95% CI: 1.12–1.44), with a projected increase to 1.80/100,000 (95% CI: 1.43–2.08) by 2035 with an APC of 3.16% (95% CI 3.12%–3.20%; ARIMA model [1, 1], RMSE = 1.121; Figure 2).

[IMAGE OMITTED. SEE PDF]

For females, the forecasted rate for 2024 was 0.83/100,000 (95% CI: 0.69–0.97), with a projected increase to 1.13/100,000 (95% CI: 0.89–1.30) by 2035 with an AAPC of 2.81% (95% CI 2.74%–2.87%; ARIMA model [1], RMSE = 0.082; Figure 2).

Discussion

In our study, the analysis of data from 1999 to 2023 revealed a concerning upward trend in obesity-related mortality among young adults aged 18–39. The overall crude mortality rate increased from 0.5 per 100,000 in 1999 to 1.1 per 100,000 in 2023, reflecting an annual percent change (APC) of 3.13%. The ARIMA model employed to project future mortality rates until 2035, predicted a continued rise in obesity-related mortality, with an expected rate of 1.3 per 100,000 in 2024 and a projected increase to 1.51 per 100,000 by 2035, reflecting an APC of 1.30%. Gender-specific predictions indicate that by 2035, males are projected to have a CMR of 1.80 per 100,000, with an APC of 3.16%, while females are projected to reach a CMR of 1.13 per 100,000 with an AAPC of 2.81%. These predictions emphasized the necessity for targeted interventions to address the distinct needs and risk factors of different gender groups.

Obesity is associated with reduced life expectancy. In 2014, Kitara et al. in their study that extreme obesity could reduce lifespan by up to 14 years; specifically, individuals with a BMI of 40–44.9 lost an average of 6.5 years, while those with a BMI of 55–59.9 lost about 13.7 years [19]. Ward et al. reported that excess weight in the United States leads to an average life expectancy loss of nearly 2.4 years, where individuals with a BMI of 40–59 kg/m² lose an estimated 6.5–13.7 years of life compared to those with a BMI of 18.5–24.9 kg/m² [10]. The life expectancy loss continued to increase beyond a BMI of 50 kg/m², with a 9.8-year reduction, surpassing the 8.9-year reduction observed for current versus never smokers [10].

Obeki et al. (2018–2021) found that among different age groups, individuals aged 25–34 exhibited a higher rate of obesity-related mortality among males (62.74%) compared to females (37.26%). Additionally, within the 25 to 34 age group, there were 1943 reported deaths due to obesity-related complications, of which females accounted for 37.26%, whereas males accounted for 62.74% [20]. Notably in our study, gender-specific analyses also indicated that males have experienced a higher increase in mortality rates compared to females with an increase in the CMR rose from 0.6 per 100,000 in 1999 to 1.4 per 100,000 in 2023, with an APC of 3.59% in males. In contrast, females showed an increase from 0.4 to 0.8 per 100,000 during the same period, with an AAPC of 2.44%. These disparities highlight the need for gender-specific public health interventions and preventive strategies.

The interplay between obesity, gender, and mortality is intricate and can be affected by multiple factors, resulting in differences across various populations and time periods. Excess weight significantly increases the risk of nearly every major cause of death, suggesting a broad range of physiological impacts. Most of the excess mortality has been linked to heart disease, diabetes, cancer, and kidney and liver diseases [21, 22]. Obesity is closely associated with metabolic disorders such as insulin resistance, type 2 diabetes, hypertension, and dyslipidemia [23]. The severity of these metabolic abnormalities increases with higher BMI and tends to improve with weight loss through lifestyle changes or bariatric surgery [24]. Additionally, extreme obesity may cause other physiological impairments, such as reduced lung capacity and airway obstruction, leading to higher mortality from chronic lower respiratory disease [25–27].

Flegal et al. (2013) demonstrated that obesity is associated with increased all-cause mortality, particularly among younger adults. Their findings corroborate the rising trend in obesity-related mortality rates observed in our study [28]. Another study by Masters et al. (2013) reported similar trends between 1986 and 2006, where overweight and obesity were associated with 5.0% and 15.6% of adult deaths among Black and White men, and 26.8% and 21.7% among Black and White women, respectively [29].

Our study utilized robust statistical methodologies, including Joinpoint Trend Analysis Software and ARIMA modeling, to analyze and forecast mortality trends. The Joinpoint software effectively identified significant temporal changes in mortality rates, while the ARIMA model provided accurate forecasts, validated through rigorous time series cross-validation and residual diagnostics [30]. The use of these advanced tools ensures the reliability and validity of the findings.

The increasing trend in obesity-related mortality among young adults calls for urgent action to implement effective obesity prevention and intervention programs. Public health strategies should focus on promoting healthy lifestyles, increasing awareness of obesity-related health risks, and providing access to resources for weight management. Moreover, the gender-specific differences in mortality trends and projections highlight the need for tailored interventions that address the unique risk factors and health behaviors of males and females.

However, there were several limitations to our study. The use of ICD codes for obesity may have underestimated the true number of obesity-related deaths. These codes specifically attributed obesity as the primary cause of death, potentially overlooking cases where obesity significantly contributed to mortality without being listed as the primary cause. Also, this study did not account for other significant factors such as smoking, alcohol consumption, and physical inactivity, which may have interacted with obesity and affected mortality rates. While the ARIMA model was robust for forecasting non-stationary time series data, it relied on historical data trends and assumed that past patterns would continue. This may not have accounted for unexpected public health interventions or changes in obesity prevalence (AAPC/APC). The high RMSE relative to small CMR values may indicate prediction uncertainty. The study did not consider the potential variability in BMI thresholds and their association with mortality among different racial/ethnic groups. Also, there was a lack of consensus in the literature on the appropriateness of standard BMI categories for various subpopulations, which complicated the understanding of how obesity impacted different groups and might have affected the study's estimates. Despite these limitations, the study provided valuable insights into the future burden of obesity among young adults in the United States. Future research should aim to address these limitations by incorporating more detailed and diverse datasets and refining the modeling approaches to better capture the complexity of obesity-related mortality.

Conclusion

In conclusion, this study provided valuable insights into the growing burden of obesity-related mortality among young adults in the United States. The use of machine learning and advanced statistical techniques enabled accurate projections of future trends, emphasizing the critical need for sustained public health efforts and gender-specific strategies to address this pressing issue. By implementing comprehensive and targeted interventions, it was possible to mitigate the rising trend of obesity-related mortality and improve the health outcomes of young Americans.

Author Contributions

Hassam Ali: conceptualization, data curation, analyses, methodology, original draft, review and editing. Waqas Rasheed: conceptualization. Vishali Moond: data curation, original draft, review and editing. Dushyant Singh Dahiya: data curation, original draft, review and editing. Abdulazeez Swaiti: data curation. Manesh Kumar Gangwani: data curation. Rashmi Advani: methodology, review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

References

L. Abarca‐Gómez, Z. A. Abdeen, Z. A. Hamid, et al., “Worldwide Trends in Body‐Mass Index, Underweight, Overweight, and Obesity from 1975 to 2016: A Pooled Analysis of 2416 Population‐Based Measurement Studies in 128·9 Million Children, Adolescents, and Adults,” Lancet 390, no. 10113 (2017): 2627–2642, [DOI: https://dx.doi.org/10.1016/s0140-6736(17)32129-3].