A Spatial Microsimulation model of Comorbidity

The University of Exeter has undertaken a project to develop a new way of producing the spatially referenced health data required so that health services can be proactively planned to meet the local health needs. Specifically, the work on the project addressed the current need for a dataset that presents patterns of comorbidity at the small area level in England. Without this spatially referenced data health practitioners and policymakers currently do not have a spatial map of comorbidity for England and therefore areas with very high or very low incidences of comorbidity are not identifiable to clinicians or policymakers.

The University of Exeter previously produced baseline small-area population estimates of co-morbidity outcomes (Cardiovascular Disease (CVD), diabetes and obesity) at the Lower Super Output Area (LSOA) level. These have been simulated for England using spatial microsimulation techniques to combine information from the Census of Population and the Health and Safety Executive (HSE) 2008-2010. To ensure that the data produced by the spatial microsimulation model is robust, the newly produced data needed to be compared to real service use data. This is a standard validation method for area data estimation.

The University of Exeter obtained aggregated admissions data on patients with a diagnosis of CVD, diabetes and obesity from the Hospital Episodes Statistics (HES) Admitted Patient Care data at the LSOA level. Data on admissions for diabetes, CVD and obesity or any combination of these diseases (i.e. patient presenting with both diabetes and CVD) were obtained at the LSOA level and were also broken down by age bands of five years and gender at the LSOA level. The unsuppressed small number data was required to ensure that all individual cases of admissions were captured. As the data was only required for validation, the original data request did not cover further use of the data for other analysis, such as an econometrical analysis of differences in spatial use patterns.

However, on using the HES data for validation, clear spatial differences in expected rates of admission for the diseases of interest and the admissions reported by HES were observed. Mapping these it was increasingly obvious that these differences were particularly apparent in rural areas and more deprived areas, where actual admissions were much lower than predicted admissions. As such, it is believed that rurality and area level deprivation is a barrier to patients with CVD, diabetes, obesity or comorbidity from accessing hospital services.

Based on this hypothesis this Agreement is to extend the use of the already disseminated HES data-set already disseminated to include a further econometrical analysis of the impact of rurality, and area level deprivation on hospital admissions for the diseases noted above. No further data is required.

Funding for the original analysis was provided by the Economic and Social Research Council (ESRC) Secondary Data Analysis Initiative. Moving forward there will be no funders or commissioners involved in this study.

Data will be analysed using standard count data econometrical data and only standard coefficients and confidence intervals will be used in future publications. If maps produced in a Geographical Information System (GIS) are produced the HES data will be only presented at the Medium Super Output Area (MSOA) level in map form. It is expected that many of the requested cells will have small data numbers. However, researchers will not attempt to re-identify the patient data and these data will not be released as part of the research. These small number cells will just be used in the analysis, with results presented across all LSOAs.

East Kent Hospital Trust (EKHT) are interested in the results as their catchment area includes some of the most deprived areas in England. Thus, the EKHT are interested in understanding if area level deprivation is a barrier to hospital use. EKHT are not involved in designing the new research or processing the data and are therefore not a data controller. The University of Exeter is the only organisation processing the data for the purposes described here.

The University of Exeter are processing the data being accessed under this agreement as part of their public task around research under Article 6(1)(e) and 9(2)(j) of the GDPR.

Article 6(1)(e) - processing is necessary for the performance of a task carried out in the public interest.

Article 9(2)(j) - processing is necessary for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes.

The hypothesis is that while people living in the most deprived areas will have the highest health care service needs, in fact they will have lower admittance rates to hospital. This will be linked to factors such as access to hospital services, access to GP services, individual education levels and income (although the NHS is free, getting there, taking time off work, getting someone to look after children is difficult/expensive).

Additional hypothesis is that rural areas will have particularly low level of admittance relative to their actual need.

The aggregated HES data will be stored on the University of Exeter’s server located at the University of Exeter medical school, St. Luke’s Campus and accessed remotely by the sole applicant at the University’s Medical School located at the Royal Cornwall Hospital. Only one individual will have access to the unsuppressed HES data, and that person is substantively employed by the University of Exeter. The dataset has been used to validate the simulated data on CVD, diabetes, obesity and their comorbidity created by a spatial microsimulation model. Within the spatial microsimulation literature this process is referred to as external validation. This process has been successfully completed.

However based on the spatial patterns of admissions observed during the validation process, permission to undertake additional analysis, is being sought to explore whether deprivation and location-based inequalities exist in accessing hospital care for people with diabetes, CVD, obesity or a combination of these diseases in England.

The University of Exeter is intending to analyse this data using standard statistical analysis, specifically regression analysis. If GIS based maps are produced, the HES data will be visualised at the MSOA geographical level which will prevent anybody from being identified. If GIS based maps are presented as a means of visualising the spatial differences in admission, the HES data will be only mapped at the MSOA level in map form, in range form (10-15%), rather than point estimates. Thus, if mapping of the data does occur LSOAs with small numbers will be automatically aggregated to a higher geography. Only data obtained from the simulated data will be released as a map within a data range rather than point results.

HES data from NHS Digital will be linked to the Index of Multiple Deprivation for 2015, which is publicly available data.

To mitigate the risk of re-identification, data will not be presented at the individual level but as average results across respondents. If GIS based maps are produced, the HES data will be visualised at the MSOA geographical level which will prevent anybody from being identified.

No automated decision-making algorithm will be used on this data.

The HES data was originally sought for validation, to check the rates of comorbidity simulated through the spatial microsimulation model against rates observed in the HES. This work has been completed. Based on the spatial pattern of admissions that was observed during the validation process, an extension to use the data is being sought to examine and explore the impact of deprivation and rurality in accessing hospital care for people with diabetes, CVD, obesity or a combination of these diseases.

From this research one academic paper will be published discussing the impact of deprivation and rurality on admission rates for diabetes, CVD, obesity or a combination of these diseases to the Journal of Health & Place by December 2020. This publication will only focus on the impact of deprivation and rurality on admission rates for diabetes, CVD, obesity or a combination of these diseases. No further analysis will be included.

The applicant's hypothesis of deprivation on hospital usage is that while people living in the most deprived areas will have the highest health care service needs, in fact they will have lower admittance rates to hospital. This will be linked to factors such as access to hospital services, access to GP services, individual education levels and income (although the NHS is free, getting there, taking time off work, getting someone to look after dependents may be difficult/expensive).

Shortly after completion of the analysis (late 2020), further presentations will be made to clinicians and policymakers from East Kent Hospital Trust, the Royal Cornwall Hospital Trust and the Royal Devon and Exeter Trust. It is also expected that the results of the analysis will be presented at a number of academic conferences, specifically the Social Science Medicine conference and the British and Irish Regional Science conference in 2021.

All outputs will be aggregated with small numbers suppressed in accordance with the HES analysis guide.

People are increasingly presenting with a number of chronic diseases such as diabetes and obesity or CVD and diabetes. With regard to the provision of health services, recent research (Morrissey et al., (2016) A Multinomial Model for Comorbidity in England of Long-standing Cardiovascular Disease, Diabetes, and Obesity. Health & Social Care in The Community) indicates that in England single disease management approach is no longer suitable for a large number of patients. Since comorbidity is significantly related to increased levels of mortality and decreased functional status and quality of life, health care should shift its focus from specific diseases, to multiple pathologies, worsening functional status, increasing dependence of care and the increased risk of mental and social problems.

Health practitioners and policymakers currently do not have a spatial map of comorbidity for England. This means that areas with very high or very low incidences of comorbidity are not identifiable to clinicians or policymakers. As such appropriate services, both preventative and disease management focused cannot be spatially targeted to the populations with the highest needs. Using the simulated data validated against the HES data, policymakers and clinicians will be given insight on what areas need to be targeted in terms of the rising incidence of comorbidity so that they can plan services to meet expected demand. The methodology, but not the data will be freely available from the authors so that other interested stakeholders may utilise the model for their specific research question. The method outlined is a well established method to examine hospital admissions data. It is envisioned that the methodology will be widely used by the academic community that currently undertake Small Area Estimation (University of Exeter, University of Leeds, University of Southampton, University of Liverpool, University of South East Anglia) and policymakers at both the national and regional level.

As the methodology will be freely available to be used by other groups, it is anticipated that this work will reduce the need for future flows of patient information.

As originally planned, the rates of comorbidity at the small area level produced by the spatial microsimulation model has been validated against the HES data. This work found that the simulated data is representative of the pattern of admissions observed in the HES. Given that the spatial distribution is correct, this means that the data controller can now use the outputs of the spatial microsimulation model to examine different policy questions, such as equitable access to health care services based on need.

Furthermore, based on the spatial patterns of admissions observed during the validation process, permission to undertake additional analysis is being sought to explore whether deprivation and location-based inequalities exist in accessing hospital care for people with diabetes, CVD, obesity or a combination of these diseases in England. The analysis of the data has not yet been completed due to the teaching commitments of the researcher.