Sub-Theme 2: Novel Analyses and Linkages of Multi-Morbidity and Long-Term Conditions in Retrospective Routine Clinical Big Datasets

Rationale for sub-theme and summary of activity

Custom made solutions to specific medical problems/questions are required to avoid research wastage while producing best evidence that will impact on patient care. The Novel analyses and linkages sub-theme in the Multi-Morbidity and Long-Term Conditions Theme of the Oxford BRC initiated in 2012 and has spent the last four years developing custom made solutions.

The sub-theme benefits from the collaboration between three world-leading institutions at the University of Oxford, the Nuffield Department of Primary Care Health Sciences (NDPCHS), The Health Economics Research Centre (HERC) and the Centre for Statistics in Medicine (CSM). This has allowed the sub-theme to carry on work in three focus areas:

a) Statistical methods used in monitoring/diagnosis/prognosis,
b) Epidemiological methods for big data and
c) Health economics with a focus in Primary Care.

Observations of plasma tNMN in 191 controls (open symbols) and 10 patients with neuroblastoma (closed symbols) along with normal percentiles according to age. Left: males, right: females.

Focus area a) statistical methods in monitoring / prognosis / diagnosis

Personalised reference ranges: Current uses of reference ranges are based around the idea of an average ‘normal’ population. This prevents their direct use in individuals with even a single condition. As the population ages, this tends to exclude the majority of the population. This project aims to investigate the impact/potential of including an unhealthy population on the reference range estimations including these conditions as ‘reference range’ modifiers.

Impact of measurement error in prognostic models: Prognostic models are used to enhance/support clinical decision-making. Nevertheless these models do not take into consideration the impact of measurement error (e.g. inaccuracies in the reading of Systolic Blood Pressure) on the model. We have carried out some preliminary evaluations of this impact and as part of the next stage we plan two potential extensions: a) evaluating the interaction of the different types of missing data and the amount of measurement error in the performance of prognostic models; and b) considering measurement error (independent and correlated, systematic and random, for instance) at the development and validation stages.

Focus area b) epidemiological methods for big data

Epidemiological analyses of large routinely collected health data have been focussing on heart failure, chronic kidney disease, consultation and morbidity patterns in primary care, and measuring complexity in primary care consultations.

Heart failure: Analysis of almost 56,000 adults aged 45 and over, diagnosed with heart failure between 2000 and 2017, showed that survival rates had modestly improved over time with 10-year survival increasing from 19.8% for those diagnosed in 2000 to 26.2% for those diagnosed in 2007. We observed a deprivation gap in median survival between the most and least deprived (median survival 4.1 years in most deprived compared to 4.6 years in the least deprived). Improvements in timely diagnosis and treatment initiation across all socioeconomic groups should be future research and policy priorities.

Chronic kidney disease: Analysis of data from 1.4 million adults showed that being overweight increases the risk of advanced CKD, and substantially for those who are obese. This was independent of pre-existing diabetes, hypertension or cardiovascular disease.⁽¹⁾ This has clear implications about the importance of maintaining a healthy body weight.

Our analysis ⁽²⁾ and a collaborative meta-analysis, including our data, ⁽³⁾ of the relationship between trends in urinary albumin-to-creatinine ratio (ACR) and advanced kidney disease has shown that change in ACR is consistently associated with future advanced kidney disease.

An example where database analysis can lead to direct patient benefit is a project that aimed to develop a tool to enhance communication between primary care practitioners and patients about age-appropriate renal function and the health risks associated with decreasing levels of kidney function.

The first stage was to produce estimates of the five-year risk of incident cardiovascular disease and end-stage renal disease by levels of renal function. This clearly showed that the risk of heart disease far outweighs the risks of renal failure at every level of kidney function. We integrated this information into a communication aid (paper-based and web-based) to facilitate GP-patient discussions about kidney age and the associated wider implications. This has undergone pilot testing with GPs and we are currently developing a project to understand the patient acceptability of new clinical language for decreased kidney function.

Consultation patterns in primary care: Our analysis of Clinical Practice Research Datalink (CPRD) data highlighted that GP workload increased by 16% over a seven-year period, which was due to both an increase in consultation rates and the duration of consultations (ref Lancet 2016) ⁽⁴⁾. Further analyses explored patient-level and practice-level factors that were associated with consultation rates ⁽⁵⁾ and duration ⁽⁶⁾. Finally, the association between consultation rates and clinical outcomes were examined and practice workload was found not to be associated with mortality or hospital admission rates⁽⁷⁾.

Patterns of morbidity in primary care: We have described the content of a 10% sample of all consultations that were included in the CPRD in 2013/14 and 2017/18 (over 2 million consultations). We have shown that many consultations include general activities that cannot be attributed to any specific health condition and a significant proportion of consultations are about respiratory, skin and musculoskeletal problems. This has implications for medical education since the latter two types of condition receive relatively little attention in many curricula. This is currently being prepared for publication.

Development of an index of complexity in primary care consultations: Previously we conducted a Delphi survey of GPs to identify patient and consultation factors that may contribute to a complex consultation. This exercise identified 34 factors and we have been conducting statistical modelling of these factors to identify how they are correlated and cluster together, and to estimate the proportion of consultations that can be defined as complex. Our next steps are to derive the complexity index over two other periods of time (2007/8 and 2017/18) and then examine any potential temporal changes. Further work to model the impact of these complexity indices on clinical outcomes such as reduced quality of life, significant events and mortality estimates and to predict health service utilisation and clinical outcomes will be conducted, supported by the Thames Valley ARC.

References

⁽¹⁾ Herrington WG, Smith M, Bankhead C, Matsushita K, Stevens S, Holt T, et al. Body-mass index and risk of advanced chronic kidney disease: Prospective analyses from a primary care cohort of 1.4 million adults in England. PLoS One. 2017;12(3):e0173515.

⁽²⁾ Smith M, Herrington W, Weldegiorgis M, Hobbs FD, Bankhead C, Woodward M. Change in albuminuria and risk of renal and cardiovascular outcomes: natural variation should be taken into account. Kidney International Reports. 2018;3(4):939-49.

⁽³⁾ Coresh J, Heerspink HJL, Sang Y, Matsushita K, Arnlov J, Astor BC, et al. Change in albuminuria and subsequent risk of end-stage kidney disease: an individual participant-level consortium meta-analysis of observational studies. Lancet Diabetes Endocrinol. 2019;7(2):115-27.

⁽⁴⁾ Hobbs FDR, Bankhead C, Mukhtar T, Stevens S, Perera-Salazar R, Holt T, et al. Clinical workload in UK primary care: a retrospective analysis of 100 million consultations in England, 2007-14. Lancet. 2016;387(10035):2323-30.

⁽⁵⁾ Mukhtar TK, Bankhead C, Stevens S, Perera R, Holt TA, Salisbury C, et al. Factors associated with consultation rates in general practice in England, 2013-2014: a cross-sectional study. Br J Gen Pract. 2018;68(670):e370-e7.

⁽⁶⁾ Stevens S, Bankhead C, Mukhtar T, Perera-Salazar R, Holt TA, Salisbury C, et al. Patient-level and practice-level factors associated with consultation duration: a cross-sectional analysis of over one million consultations in English primary care. BMJ Open. 2017;7(11):e018261.

⁽⁷⁾ Lay-Flurrie S, Mathieu E, Bankhead C, Nicholson BD, Perera-Salazar R, Holt T, et al. Patient consultation rate and clinical and NHS outcomes: a cross-sectional analysis of English primary care data from 2.7 million patients in 238 practices. BMC Health Serv Res. 2019;19(1):219.

Focus area c) health economics with a focus in Primary Care

Application of economic evaluation techniques to provide evidence for aiding adaptation or implementation decision making of new health techniques or interventions at national and local level of NHS

Application of early economic decision analytical modelling technique to identify care pathways, treatment decision making and synthesise existing evidences to quantify the possible effects of new technology and associated uncertainties, and inform future prospective study design to collect evidence and price setting for the new technology.

Application of economic evaluation techniques to identify budget plan for local NHS authorities as to where the future costs and savings of new technologies would happen.

Use routinely collected clinical data linkage (CPRD, HES, PROMs, QOF, GP survey, Household panel survey data) and econometric methods to assess impact of health policy on NICE guidance, or assess quality or health expenditure of health care system.

Use behaviour economic theory or decision making techniques to support design of public health interventions at primary care.

Health Economics

Achievements:

Continued promoting and supporting health economic related studies for NewKI and OxVALVE population cohorts. This included contribution to health economic study design for NewKI and Heart Foundation Programme application for OxVALVE, as well as potential industrial collaboration to explore clinical care pathway and interventions for the valve disease and health care burden in the community using the unique OxVALVE cohort.
Three studies with BRC support have recently been completed. They examined online tool to support people with prior-clinical anxiety, explored educational and training needs to implement NICE testing for childhood asthma in primary care, and assess new devices to collect urine samples for UTI lab tests. Economic evaluation methods were applied to these projects and provided robust cost effectiveness evidence for decision making.

Current Links

Current link within the NDPCHS

Statistics Group, Infectious Diseases group Oxford Diagnostic Evidence Cooperative, Monitoring of Renal Specific Endpoints Programme (MORSE), Oxford NIHR Collaboration for Leadership in Applied Research and Care (CLAHRC)

Current Links within the University of Oxford (focus on BRC members)

Oxford CSM, Department of Psychiatry, the Oxford Renal cross-sectional study of chronic kidney disease (OxRen), the Valvular Heart Disease Population Cohort Study (OxVALVE), Oxford HERC.

Current National and International links

National School of Primary Care Research, Clinical Practice Research Datalink (CPRD), EuroQol group, UK Health Economists’ study group, European Commission Joint Research Office, Echocardiographic Normal Ranges Meta-analysis of the Left Heart (EchoNormal) Collaboration (New Zealand), University Hospital (CHUV) Lausanne (Switzerland).

Principal Investigators:

Professor Rafael Perera, Dr Clare Bankhead, Professor Richard Hobbs

Contact: Rafael Perera, Nuffield Department of Primary Care Health Sciences, Radcliffe Primary Care Building, Radcliffe Observatory Quarter, Woodstock Road, Oxford OX2 6JJ

Sub-Theme 2 Grants funded

Making both ends neat: exploring the effects of modifying the TTO protocol on non-trading and all-in trading. Co-applicant. €91,000 01/12/2019 – 31/05/2021. EuroQol Foundation

Development and evaluation of an online FeNO-guided primary care asthma management intervention. NIHR Programme Grants for Applied Research (Reference number: RP-PG-0617-20002). Co-applicant. £2,607,600. 01/02/2020 – 31/01/2025.
Artificial Intelligence and Big Data for Early Lung Cancer Diagnosis Prospective Study (Phase 2). NIHR i4i. £1,200,000. Lead health economist. 01/08/2018 – 31/07/2020.
OxVALVE Sub-study (Health Economic Burden of Mitral & Tricuspid Regurgitation): A Health Economic Impact Analysis of Mitral and Tricuspid Regurgitation and their Mechanisms in a Large Prospective Echocardiographic Population Screening Study. Edwards Ltd. Co-investigator & lead health economist. £50,000. 01/09/2018 – 30/06/2019.
Urinary Tract Infection (UTI) pathway mapping. NHS England. Co-applicant & lead health economist. £50,000. 01/11/2018 – 30/01/2019.
Rapid diagnosis of Urinary Tract Infection in Primary Healthcare. NIHR i4i. £1,202,400, co-applicant, local leader for health economics. 01.06/2018 – 31/05/2021, Reference Number II-LB-0417-20004).
COPD Exacerbation Alert for patient stratification. Small Business Research Initiative (BRBI) Stratified Medicine (Phase II). £2,000,000, co-applicant, local leader for health economics. 01/07/2017 – 30/06/2020, Application number: 63624-4631992017.