An Adaptable Long-term Condition Workload Prediction Model for Primary Health Care
The primary health care (PHC) system must manage the growing demand for care due to patients with long-term conditions (LTCs) such as diabetes, hypertension, and asthma. Population-based care can help address this workload management problem, specifically by enabling a shift from a reactive to a proactive patient management approach. However, current PHC systems lack the ability to provide population-based care. This thesis presents a tool to predict the future workload generated by a population of patients. We use a rule-based system, for its modularity, flexibility and the automated modifiability behaviour, to develop the care pathways as rules that, when given the patient data, would simulate the patient visits for the upcoming year (since some start date). It is assumed that the GPs follow best practice and their patients adhere to their plan-of-care, making visits to the medical practice on their scheduled LTC appointments. Then, these visits are aggregated to a population-level as a count of appointments per week from these LTC patients, referred to as the workload to be managed within the capacity of the practice. Knowing this predicted workload, the PHC organisation can then plan and deliver care accordingly. In this thesis, we also explore using seven what-if scenarios the impacts of alternatives in practices and evaluate the strategies to address them. We then propose the use of Bayesian inference in our workload prediction model, in order to incorporate the variation in patient visits due to the impact of renewal of their LTC prescriptions. This work is done in collaboration with BPAC, a non-profit organisation that promotes best practice for primary care within New Zealand.The collaborations on this work were in the form of 1. health data from a medical practice; 2. the knowledge base to understand the primary health care domain, and the practical issues at a medical practice. Approval for this research using the anonymised patient data provided by BPAC has been given by the University of Otago Human Ethics Committee (Health). We follow the design science research (DSR) methodology to develop our adaptable best practice based workload prediction model for a PHC due to its LTC patients. From a DSR perspective, we developed a construct called the three-layer LTC PHC construct, and a process called the encounter-based unfolding plan-of-care process, which are used to build our artefact: the adaptable best practice based workload prediction model. In DSR, much emphasis is on communicating the developed artefact or model to a wider community and the feedback guides further improvement of the model or the artefact developed. We followed seven iterative cycles to incrementally build the rule base, and the feedback served as a guide to improve the simulation capability of our ABP-WPM. Apart from the feedback from the collaborator on this work, feedback was also collected through informal meetings with the care providers of a medical practice at Mosgiel, the executive members of two different PHOs (one in the North Island and the other in the South Island of New Zealand) at various stages of development of this artefact. The artefact developed was also communicated to the research community through two publications. In the process of developing this population-level workload prediction model, we identified shortcomings (for example, the LTC status of a patient is not explicit) in the current health data models and in the PHC data shared with us, which is needed to support a population-level workload analysis. We, therefore, developed a patient information data model that makes this information explicit.
Advisor: Cranefield, Stephen; Winikoff, Michael; Lloyd, Hywel
Degree Name: Doctor of Philosophy
Degree Discipline: Information Science
Publisher: University of Otago
Keywords: Workload; long-term conditions; primary health care; prediction; Bayesian inference; Patient information model; Rule based system
Research Type: Thesis