We use simulations to infer knowledge regarding causal assumptions in competing risks scenarios (a subset of Multi-State Models) and time-dependent measures of model calibration. The causal assessment involves the investigation of multiple real-world scenarios where confounding factors may interfere with the standard way of measuring the effects of a treatment on an event-of-interest and a competing event. This is important in the field of Multi-State Models as the inaccurate interpretation of an effect on a competing event can lead to misconceptions in the causes of the event-of-interest. Further simulations are performed to analyse how traditional methods of assessing the Calibration-in-the-Large of a Clinical Prediction Model can be affected by the censoring of patients over time, in particular when this censoring is caused by a competing event related to the variable of interest. To combat this, we use the Inverse Probability of Censoring to weight patients based on their likelihood to still be present in the data at a certain time, to re-align the measurements with reality and avoid bias due any underlying relationship between the competing event and the attributes of a patient. This knowledge feeds into the design and implementation of metrics to assess other aspects of model validity, namely accuracy, discrimination and calibration, in a Multi-State Clinical Prediction Model. The Brier Score is extended to account for multiple outcomes, and the c-statistic is replaced by the Polytomous Discriminatory Index. Both of these extended measures are adjusted to fit into the scales of their traditional counterparts. We also extend the measures of calibration (i.e.~Intercept and Slope), and encode further information into these metrics by also analysing the traditionally held assumption of that state predictions are completely independent. All of these methods are augmented with the information garnered from the previous simulations to ensure that bias due to censoring is accounted for. Data from the Salford Kidney Study and the West of Scotland Electronic Renal Patient Record are used to develop and validate our own clinical prediction model. This model can predict a Chronic Kidney Disease patient's journey through Renal Replacement Therapy and on to Death, and through the application of our validity metrics, we can be confident that it can be accurate and effective in its predictions.