Global, regional, national trends of femur fracture and machine learning prediction: Comprehensive findings and questions from global burden of disease 1990-2019

Abstract

Background: Femur fracture is a type of fracture with high disability and mortality. There is no comprehensive analysis and prediction of the global distribution of femur fractures, so we conducted this study.

Methods: Age-standardized incidence rate (ASIR), age-standardized prevalence rate (ASPR), and years living with disability (YLDs) of femur fractures (excluding femoral neck) were downloaded from the Global burden of disease database. Trend analysis was performed, and 6 time-series machine learning algorithms were applied to predict the global ASIR, ASPR, and YLDs.

Results: ASPR for femur fracture had been increasing in most countries worldwide from 1990 to 2019, with the highest in East Asia (AAPC = 1.25 95%Confidence Interval (1.2, 1.3)) and lowest in Central Latin America (AAPC = -0.74 95%CI (-0.81, -0.67)). However, ASIR showed a significant downward trend worldwide, with East Saharan Africa decreasing the most (AAPC = -4.04 95%CI (-5.56, -2.47)), and East Asia elevating the most (AAPC = 1.11 95%CI (0.87, 1.42)). YLDs were increasing over the world, with East Asia still elevating the most AAPC= (3.9 95%CI (3.85, 3.95)), with the only region of decrease being Eastern Europe (AAPC = -0.28 95%CI (-0.3, -0.26)). Both ASPR and ASIR were higher in women than in men in the >75 year group, whereas YLDs was lower in women than in men in the >60 year group. Globally, the ARIMA model was optimal in the prediction of ASPR, the PROPHET model effected in the prediction of ASIR, and the PROPHET WITH XGBOOST model was the best in the prediction of YLDs. The projections showed increase in both ASPR and YLDs, except for ASIR decreasing by 2030.

Conclusions: Our study found a rise in femur fracture ASPR and ASIR from 1990 to 2019 in war conflict areas and East Asia, meanwhile, the YLDs of femur fracture increased in populous countries. In both 1990 and 2019, both ASPR and ASIR were higher in women over 75 years than that in men, but YLDs was higher in men over 60 years than that in women. In 2020-2030, while global femur fracture ASIR might decline, both ASPR and YLDs might rise.

The translational potential of this article: Femur fracture is a high-energy injury due to direct violence, and in war, conflicting and underdeveloped regions such as East Asia. Accidental injuries may occur due to the rapid development of industry and the frequent traffic accidents. This study suggests that we should focus on elderly women (≥75 years) in the above regions in the future. For older men (>60 years old), more attention should be paid to post-fracture functional rehabilitation and early reintegration into society to reduce the disability rate and lower the socio-economic burden.

Keywords: Femur fracture; GBD; Machine learning; Prediction; Visualization.

Graphical abstract

Figure 1

Femur fracture age standardized prevalence rate (ASPR) in 1990 (A); Femur fracture ASPR in 2019 (B).

Figure 2

Age-standardized incidence rate (ASIR) in 1990 (A); Femur fracture ASIR in 2019 (B).

Figure 3

Femur fracture years lived with disability (YLDs) in 1990 (A); Femur fracture YLDs in 2019 (B).

figs1

Femur fracture average annual percent change (AAPC) of ASPR from 1990 to 2019.

figs2

Femur fracture AAPC of ASIR from 1990 to 2019.

figs3

Femur fracture AAPC of YLDs from 1990 to 2019.

figs4

Femur fracture population pyramid of ASPR in 1990 (A); Femur fracture population pyramid of ASPR in 2019 (B).

figs5

Femur fracture population pyramid of ASIR in 1990 (A); Femur fracture population pyramid of ASIR in 2019 (B).

figs6

Femur fracture population pyramid of YLDs in 1990 (A); Femur fracture population pyramid of YLDs in 2019 (B).

figs7

Time-series machine learning prediction of global femur fracture ASPR to 2030.

figs8

Time-series machine learning prediction of global femur fracture ASIR to 2030.

figs9

Time-series machine learning prediction of global femur fracture YLDs to 2030.