Modelling patient pathway progression and high resource individuals

Abstract

Healthcare provision across the UK is facing unprecedented financial challenges. One key challenge is that a small number of individuals are responsible for a disproportionate cost. The thesis mainly addresses the topic of high resources individuals (HRIs) and is based on confidential data that are provided by Macmillan Cancer Support UK and also studied open data published on line by National Health Services (NHS) Scotland (PHS) as well as rich confidential pathways data also provided by National Health Services (NHS) Scotland. The thesis used methodological directions from both major data sets to address two Research Questions (RQs): (a) To identify characteristics that discriminate among People affected by cancer (PABCs) who complete the Macmillan Cancer Support UK pathway, (b) to identify the High Resource Individuals (HRIs) in the PABC data. These two RQ’s can be addressed in both data sets from the NHSS(National Health Services (NHS) Scotland) and from Macmillan Cancer Support UK. With the NHSS data the thesis processed confidential pathways categorical data of 7 Mn persons while with the MCS confidential data the thesis coped with approximately 1 Mn categorical survey data from the eHNA tests that MCS conducted. The thesis also analysed numerical open data from public services delivery. The thesis used a range of methods to analyse the data. At first, the thesis linked open data services demand to other services demand. For the PABC data, the thesis explored counts of co-occurrences and CC or χ2 tests to study how often PABCs with specific attributes engaged through the whole pathway. The attributes who were mostly seen in PABCs engaging until the end of the eHNA pathway were more likely drivers whereas those that were likely attached to PABCs who did not continue till the end were more likely non drivers than drivers.

Contributions/Findings: The thesis offers a thorough study of cases that can be used as a reference by interested researchers thus contributing to HC cost reduction also by spotting cases or persons or pathways that incurred high cost or complexity. Furthermore, this can contribute to a fairer distribution of the limited HC resources insuring thus equal access to HC facilities by the rationalization of the resources. This is expected to add to better planning of social policies and to limiting the frustration of the society in regards to equality in accessing HC facilities. Another important contribution is that traditional persons demographics attributes like age and intensity of need (score) do not seem to play a crucial role to predict complete engagement. Also, the thesis found that on average the percentage of engagement was (25%) which may reflect the value of appreciation by persons. The variation of engagement across the UK and the high 80% in Scotland was also an important finding. The thesis found close to 200K registered to take the eHNA test(PABCs), and 54K completed it. The thesis did not find, though, absolutely clear attributes that are completion drivers. On top of that, the thesis found that most completing PABCs prefer the ‘email‘ as a way of communication and this can help to predict engagement, as it says that the completing PABCs are likely computer literate. The thesis found that the complexity of the treatment (the attribute is ‘Diagnosis.2‘ or ‘Diagnosis.3‘ that stand for comorbidities) causes full engagement (driver for completing PABCs) and also flags high-cost persons in the context defined by MCS . That is, the HRIs are more likely those who get referrals or medications. 

Methods: To link open demand data (public services to other services) the thesis used prediction methods such as (LR, ARMA , CART, RF) and also used classification methods such as Mutual Information Exchange, Cross Correlation, K-Means, and χ2 test to find similarity across demand patterns or to group them. Finally, the thesis used PCA to derive major public services (alcohol related admissions were the major ones). The thesis extended the use classification methods to MCS data in order to spot completing PABCs and HRIs and also was challenged with combining two MCS data sets (people affected by cancer and People living with cancer).

Date of Award	24 Nov 2023
Original language	English
Awarding Institution	Abertay University
Sponsors	Scottish Government & Northwood Charitable Trust
Supervisor	James Bown (Supervisor) & Euan Dempster (Supervisor)