Developing an evidence-based system to facilitate the predictive assessment and optimisation of older adults with cancer

Introduction: Older adults with cancer have worse outcomes than their younger counterparts, including higher postoperative complications, chemotherapy toxicity and treatment allocation to best supportive care. Oncogeriatric assessment (OGA) can provide predictive information and optimisation targets to improve these outcomes. OGA has multiple implementation barriers, including uncertainty in delivery, health economic concerns and siloed data. The aim of this thesis was therefore to develop an evidence-based system to facilitate the predictive assessment and optimisation of older adults with cancer.Methods: Multiple methods were used, including i) a systematic realist review to understand implementation factors; ii) a decision-analytic health economic evaluation; iii) the design, implementation and delivery of a digital-first OGA service; iv) quantitative survey evaluation of a digitalised, patient reported OGA; v) the development and analysis of a complex model of an oncogeriatric population using machine learning.Results: A whole system approach is required to improve the implementation of OGA in cancer settings, including utilisation of technology, leveraging non-specialist staff skills and cancer MDT, insurer, payer and regulator consensus. OGA has additional costs over standard care alone of between £390 and £576, dependent upon implementation configuration. However, when major assumptions about the effectiveness of OGA were modelled or OGA is used before chemotherapy, with minimal healthcare staffing inputs and technological assistance, it was cost-effective. A new digital-first OGA service was implemented successfully, and patient-reporting was feasible for older adults with suspected or confirmed cancer. A complex model of an oncogeriatric population using synthetic individual patient data showed high fidelity to real world data and generated a sandbox environment for predictive algorithms for OGA selection and treatment outcome risk profiling.Conclusion: A digital-first OGA system is feasible and usable and may be cost-effective with careful implementation context considerations. The use of artificially intelligent systems may enhance patient selection and risk prediction but requires future validation.

Name dc:type.qualificationname

PhD

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

The University of Hull and the University of York

Year dc:date.issued

2021

Author dc:creator

• McKenzie, Gordon Arthur George

Language dc:language

en

Identifier

oai:hull-repository.worktribe.com:4192792

Author Identifier

0000-0003-0338-3513

OAI identifier oai:identifier

oai:hull-repository.worktribe.com:4192792