Fragility hip fractures in older adults are a major burden for individuals and society due to increased morbidity, mortality and substantial health care expenditure. The vast majority of hip fractures are caused by falls resulting in an impact on the hip, inducing a force that exceeds the femoral bone strength. At the same time, only 1 - 3% of all falls result in a fracture, suggesting that the fall dynamics defines its severity and, through that, the magnitude of the impact force. Additionally, trochanteric soft tissue thickness is known to play a crucial role in the energy-dissipating mechanism during impact. Thus, the fracture risk depends on the rate of falling, the fall-induced impact force, and the femoral bone strength.

This PhD thesis presents a novel hip fracture risk calculator that is based on a mechanistic stochastic framework that was presented in the literature. The original model uses a Poisson process characterised by the rate parameter λ to describe the occurrence of a fall. A stochastic distribution is then introduced to model the conditional probability that the fall-induced impact force exceeds the femoral bone strength. By combining these, the fall events resulting in a fracture are identified, and a fracture risk can be calculated. The novel model introduced a fall rate model to estimate a personalised λ, and refined the probability distribution representing the chances of the femoral bone to break upon impact with a mechanical impact force model.

The fall rate model to predict a personalised fall rate λ was developed by analysing three independent cohorts that assessed various risk factors of falls. Negative binomial regression models were fitted, and a variable selection algorithm was applied. Thereby, the prior number of falls treated as a categorical variable was the only predictor selected in all cohorts. A meta-analysis and validation of the models confirmed that the number of prior falls is a robust predictor for the prediction of a fall rate among different cohorts.

Furthermore, the personalised impact force model was developed using subject-specific parameters that can be extracted from quantitative computed tomography (QCT) images or substituted with anthropometric data. The model calculates the full range of possible impact forces of an individual, indirectly representing the variability in the dynamics of a fall. By introducing a stochastic distribution that describes the probability of the fall dynamics, a fracture risk can be calculated. With this approach, the stochastic aspects of the mechanism resulting in hip fracture are concentrated in the Poisson process describing the occurrence of a fall and the fall dynamics. Bone strength was estimated with QCT-based finite element analysis, but it can also be estimated with other densitometric measures. Thus, the required parameters for the model can be extracted from QCT images or substituted with anthropometric and densitometric data.

A sensitivity analysis was conducted using clinical data from the AFFIRM-CT cohort, confirming that the fall rate, the trochanteric soft tissue thickness and the bone strength are the dominating parameters influencing the risk of fragility hip fractures. Furthermore, output variables such as the predicted impact velocity and impact force aligned well with experimental data from the literature.

The work of this thesis resulted in a novel fragility hip fracture risk calculator that models the underlying process of hip fractures by combining the stochastic aspects of a fall with the mechanistic elements of an impact on the hip. Thereby, the model was shown to reflect key observations from empirical data, indicating that it can capture the intrinsic aspects that affect the risk of fragility hip fractures.