Data-driven Markov models and their application in the evaluation of adverse events in radiotherapy

Abstract

Decision-making processes in medicine rely increasingly on modelling and simulation techniques; they are especially useful when combining evidence from multiple sources. Markov models are frequently used to synthesize the available evidence for such simulation studies, by describing disease and treatment progress, as well as associated factors such as the treatment's effects on a patient's life and the costs to society. When the same decision problem is investigated by multiple stakeholders, differing modelling assumptions are often applied, making synthesis and interpretation of the results difficult. This paper proposes a standardized approach towards the creation of Markov models. It introduces the notion of ‘general Markov models’, providing a common definition of the Markov models that underlie many similar decision problems, and develops a language for their specification. We demonstrate the application of this language by developing a general Markov model for adverse event analysis in radiotherapy and argue that the proposed method can automate the creation of Markov models from existing data. The approach has the potential to support the radiotherapy community in conducting systematic analyses involving predictive modelling of existing and upcoming radiotherapy data. We expect it to facilitate the application of modelling techniques in medical decision problems beyond the field of radiotherapy, and to improve the comparability of their results.

Genomics models in radiotherapy: From mechanistic to machine learning

“…The robustness of data‐driven models relates to how often observations (data or variables) can be related to an outcome according to certain relationships (equations). As such, data‐driven NTCP models may be able to identify models that perform better than mechanistic counterparts, given sufficient data, or that depart significantly from standard therapeutic strategies, such as charged particle therapy …”

Classification of Microscopic Models with Respect to Aggregated System Behaviour

“…Clearly, any state-discrete model belongs to this class, though it might be possible, that models with hybrid state-space are output-discrete if the continuous parts of the state-space are not considered as model-output. The best examples for this feature are found in classic microsimulation models or, to be precise, Markov models for health technology assessment (HTA, see [Abler et al, 2013]), wherein discrete health-states are observed, while (potentially) continuous patient parameters like age or blood values may have an influence on the transition probabilities between the health-states, but are not focus of interest. Anyway, in case of state-/output-discrete microscopic models there is a bijective mapping of Γ/Γ onto a subset of N d .…”

“…In the current thesis only statecontinuous microscopic models without contacts were discussed, namely the groundwater pollution random-walk model 5.1 and the agent-based GEPOC model 6.1. Examples for output-/state-discrete microscopic models without interaction are the mentioned HTA focussed Markov models [Abler et al, 2013].…”

Generating Evidence for Chronic Obstructive Pulmonary Disease (COPD) Clinical Guidelines Using EHR Data

“…[16][17][18] They provide support for evaluating uncertainties in decision making by approximating patient transitions through a set of "health states", each of which corresponds to a clinical event. 19 While using Markov chains has been useful for chronic conditions such as COPD, directly using EHR data in Markov models presents certain challenges. In particular, Markov chains assume a one-step time-invariant transition.…”