Variation in avoidable hospital admissions by mental health status

The overall objective for the study is to identify the extent to which various supply (e.g., availability and quality of health and care services) and demand factors (e.g., population morbidity) influence variation between clinical commissioning groups (CCGs) in A&E attendance rates and rates of potentially avoidable emergency hospital admissions among 'mental health' and 'non-mental health' groups.

*************************

Potentially avoidable emergency admissions include those conditions which could be effectively managed outside of hospital. Such admissions partly reflect deficiencies in quality or access to physical health care, are expensive to the NHS and can be harmful to patients. Potentially avoidable admissions have been defined in relation to a specific set of conditions for which hospital admission could be prevented through adequate management or earlier detection outside of a hospital setting. The list of conditions comprising avoidable admissions varies between studies. A list of fourteen conditions (excluding one which relates only to children) will be used, identified through consensus methods by Coleman & Nicholl (2010), for which exacerbations should be able to be managed in a well-performing system without hospital inpatient admission. Relevant to the current study, this broadened the scope of ‘avoidable’ or ‘preventable’ admissions to include not just those that could be managed in primary care, but also those for whom exacerbations could be managed by a range of alternative care systems (e.g., GP out of hours services, district nursing, urgent care walk-in services). Examples of conditions include acute mental crisis, COPD, epileptic fit, non-specific abdominal pain, and blocked urinary catheter.

*******************

The research team at University College London (UCL) will assess variation at the CCG level, as these are responsible for the planning and commissioning of health care for their local area.

This study was instigated by the research team at UCL upon identification of an information gap in the published literature. Further, based on previous work with a local authority partner it was identified that there was a need to better understand whether quality and availability of services beyond primary care (i.e. including information about social services and mental health services) might influence variation in potentially avoidable hospital admissions.

UCL are examining CCG-level variation in potentially avoidable emergency admissions and A&E attendance rates among those in contact with secondary mental health services and comparing it with those not in contact with mental health services. This is because research has demonstrated that people with a variety of mental health conditions face elevated risks of morbidity and mortality; and, a wide range of barriers to good management of health conditions (e.g., due to symptoms and treatment related issues, availability of social support, stigma, and access to/receipt of appropriate health care). Studies also find those with mental health conditions have more frequent A&E attendances and potentially avoidable admissions (mainly for physical conditions) than the general population, not solely accounted for by greater severity of these conditions.

Locating and quantifying reasons for variation in these rates between areas of responsibility for commissioning health and care will help highlight where investment would make most difference to conditions management and prevention of avoidable hospital use.

The data requested will only be used to:

1. Calculate national CCG-level age and sex adjusted rates over a two year period (2016-2018) of i) potentially avoidable emergency hospital admissions and ii) A&E attendances for two cohorts: the national population (aged 18+ years) (non-mental health cohort) and those identified as in contact with secondary mental health services (aged 18+ years) (mental health cohort, including a subgroup identified with serious mental illness). The numerator of these rates will be calculated using admissions/attendance data, denominators will be calculated from openly available data (ONS mid-year population estimates for the national cohort) and from data received as part of this application (population in contact with mental health services for the mental health cohort). CCG-level variation in rates of potentially avoidable emergency admissions and A&E attendances will then be estimated. Rates over a two period will be calculated to reduce the impact of annual fluctuations of these rates at a CCG level.

2. Create CCG level indicators which will be used in regression models as generally considered predictors of variation (e.g. duration of admission (from date of admission/discharge fields); % of admissions from GP referral (from admission source field).

3. Describe characteristics of potentially avoidable emergency admissions among mental health/non-mental health cohorts (admission/discharge method, disposal, referral source).

Openly available routinely collected anonymised data aggregated to CCG level from external sources will subsequently be used to examine which supply (e.g. availability and quality of primary care) and demand factors (e.g. population deprivation, morbidity) explain any observed CCG-level variation in admission/attendance rates among mental health and non-mental health cohorts using regression analyses. Outputs will be at CCG-level rather than record level. These openly available sources include ONS, Atlas of Variation and Quality and Outcomes Framework.

This is a non-commercial standalone research project carried out at University College London and does not involve any third parties. The findings will be disseminated widely and are intended to help planners and funders of health and care to decide how best to focus efforts and resources to reduce potentially avoidable emergency hospital visits overall and among people with poor mental health.

**********************

UCL request data for a two year period to calculate average rates of potentially avoidable admissions and A&E attendances to account for annual fluctuations in rates. As a national level study UCL are requesting HES data on all admission/attendances over that time period in order to calculate rates of admissions/attendances, specifically calculating rates for each Clinical Commissioning Group in England. The study will compare rates among people identified as in contact with mental health services to those not in contact with services, thus data are requested for all admissions/attendances rather than just those in contact with mental health services.

******************************

All outputs/findings will be non-identifiable and aggregated with small numbers suppressed in line with the HES analysis guide. Data will only be accessed, processed and analysed by substantive employees of University College London.

All organisations party to this agreement must comply with the Data Sharing Framework Contract requirements, including those regarding the use (and purposes of that use) by “Personnel” (as defined within the Data Sharing Framework Contract ie: employees, agents and contractors of the Data Recipient who may have access to that data)

Pseudonymised data will be transferred from NHS Digital to the Data Safe Haven within University College London (UCL) via secure File Transfer. Data will be stored, handled and analysed within the UCL Data Safe Haven, no data will be shared with third parties. The UCL Data Safe Haven has been certified to conform to the NHS Information Governance Toolkit. It is built using a ‘walled garden’ approach and a file transfer mechanism enables information to be transferred into the ‘walled garden’ simply and securely.

Data will be processed by substantive employees of University College London, all with up to date Information Governance training, and only those members of the research team who have been granted access to the secure data storage area within the UCL Data Safe Haven will be able to access the data for the purposes outlined above.

The research team at UCL will use HES admitted patient care and A&E datasets to calculate CCG-level rates of potentially avoidable emergency admissions, and A&E attendances as outlined above. The bridging file will be used to identify which patients have been in contact with mental health services within each year requested to calculate the above rates for the mental health cohort. As those in contact with mental health services are a heterogeneous group, the MHSDS data will be used to identify a subgroup mental health cohort diagnosed with serious mental illness.

Diagnosis data and information relating to contact with mental health services is sensitive data thus the research team have kept information requested to a minimum (e.g. primary diagnosis for first finished consultant episode rather than all diagnosis codes). The research team at UCL will minimise the risk of re-identification by:

- including minimum necessary diagnosis data (e.g. first 3 or 4 characters of primary diagnosis of first

finished consultant episode rather than all diagnosis codes),

- retaining information about those first finished episodes for which the primary diagnoses pertain to one of

a list of conditions for which admissions have been identified as 'potentially avoidable' (e.g., non-specific

chest pains, COPD, blocked urinary catheter). This list of conditions does not include any rare conditions

(affecting fewer than 5 in 10,000 individuals), minimising the risk of re-identification from the data items

requested.

- conducting analyses at a CCG-level & producing CCG level statistics as outputs (i.e. calculating age/sex

standardised CCG-level rate of emergency admissions for potentially avoidable conditions).

The research team at UCL will use openly available routinely collected anonymised data aggregated to CCG level from external sources (e.g. Office for National Statistics, Atlas of Variation, Quality and Outcomes Framework) to examine which supply (e.g. availability and quality of primary care) and demand factors (e.g. population deprivation, morbidity) explain any observed CCG-level variation in admission/attendance rates among mental health and non-mental health cohorts using regression analyses. The CCG will be the unit of analysis and data from these other sources will be linked at a CCG level only.

No data will be linked to record patient level data.

No data directly provided from NHS Digital will be stored, processed, or accessible to or by any third party organisations not listed in this agreement.

Data will only be accessed, processed and analysed by substantive employees of University College London.

******************

For data from the Mental Health (MHSDS, MHLDDS, MHMDS) data sets, and any Mental Health data linked to HES or SUS, the following disclosure control rules must be applied:

• National-level figures only may be presented unrounded, without small number suppression

• Suppress all numbers between 0 and 7

• Round all other numbers to the nearest 7

• Percentages can be calculated based on unrounded values, but need to be rounded to the nearest integer in any outputs

• In addition for Learning Disability data in Mental Health (MHSDS, MHLDDS, MHMDS), the England-level data also must apply the suppression of all numbers between 0 and 7, and rounding of other numbers to the nearest 7

The university will disseminate findings widely among NHS Trusts, CCGs and Local Authority partners and through Collaborations for Leadership in Applied Health and Care (CLAHRC) networks. The research team will produce full and brief reports that will be e-mailed to CLAHRC partners through the research team organisation's Research Updates newsletter, made available on the website, and highlighted through the organisation's social media platforms (specifically the organisation's Twitter feed). The research team at UCL will also submit a paper for peer reviewed publication in a Journal (e.g., BMJ Quality & Safety) and conference presentation (e.g., Health Services Research UK).

All outputs will contain only aggregate level data with small numbers suppressed in line with the HES analysis guide. For all outputs data will be presented as age/sex adjusted CCG-level rates of potentially avoidable emergency admissions/A&E attendances per 10,000 or 100,000 population. CCGs will be the unit of analysis and the age/sex adjusted rates will be the dependent variable of interest. Coefficients will be presented indicating the strength/direction of association between CCG-level supply (e.g. % of population satisfied with GP consultations) and demand factors (e.g. population deprivation), and the dependent variable. The extent to which various supply and demand factors influence rates of hospital admission/A&E attendances will be depicted through changes in adjusted R-squared scores (from regression analyses).

Specific outputs of the study will be:

• Full and brief report: Open Access (free) – made available electronically to CLAHRC North Thames partners (NHS Trusts, CCGs, Local Authorities) through the research team organisation's Research Updates e-newsletter. The research team will disseminate links to the report findings more broadly (e.g. to members of the public and mental health charities) through the organisation's website, and highlighted through the organisation's social media platforms - specifically through the associated Twitter account. The goal date of the dissemination of this report is currently October 2019.

• Peer reviewed article: Open Access (free) - research team submission to BMJ Quality & Safety. The intended customer group for this output are academics and clinicians. The goal date of the publication of the peer reviewed article is currently October 2019.

• Conference presentation: This will be a paid event – the research team will submit an abstract to the Health Services Research Network UK. It will be attended by clinicians and academics in the study field. The goal date for presentation at this conference is July 2019.

***************************

Findings will be shared with charities such as the Mental Health Foundation and the 'Mental Elf Service' to help disseminate to the general public. In particular, The Mental Elf shares and summarise information about evidence-based research findings in a lay format through blogs and Twitter. UCL will also contact the Royal College of Psychiatrists and the Royal College of General Practices to ask them to promote findings through their website and newsletters.

Through producing the outputs specified, the research team at UCL aims to deliver information to key decision-makers in an accessible format about how best to reduce potentially avoidable emergency admissions in general and among those in contact with mental health services. Specifically, the outputs will highlight which elements of the health and care system influence unnecessary variation in A&E visits and potentially avoidable hospital admissions among people with poor mental health. By doing so, the researchers will highlight where improvements to different elements of the health and care system may reduce avoidable emergency hospital use.

This is important as potentially avoidable admissions have been defined in relation to a specific set of physical and mental conditions. Admissions for these conditions are thought to be preventable through proper management outside of a hospital setting, such as in primary care (e.g., by a GP). Availability and quality of other services (social services, mental health services) may also influence such hospital use, but thus far there is limited information about the relative importance of these, compared to primary care.

The researchers believe that the following will benefit from the information created:

1) Potentially avoidable emergency admissions and A&E attendances are expensive to the NHS, thus benefits from the information provided include more efficient targeting of resources to reduce such hospital use, and savings from any reductions incurred.

2) Potentially avoidable emergency admissions and A&E attendances are also detrimental to patients, who may be better supported outside of the emergency care system. People with poor mental health may be particularly negatively affected by unplanned hospital visits whether these are for physical or psychological reasons. For example, long wait times can be stressful and exacerbate acute mental health crises, and non-mental health professionals treating patient’s health may not have the correct skills to provide appropriate care. There is also a risk of exposure to stigma from non-mental health professionals. Benefits of the information provided may therefore include reduced exposure to the detrimental sequelae of emergency care for patients.

3) High rates of avoidable hospital admissions are therefore thought to reflect poor primary care performance, poor access to primary care, and underlying population characteristics. However, the influence of quality and availability of other services (including social services and mental health) on potentially avoidable emergency admissions and A&E attendances has not yet been considered. Potential benefits of the information provided therefore also include a more holistic understanding of how such hospital use could be reduced which does not unfairly penalise primary care.

The research team at UCL will be responsible for sharing the findings in several formats in order to be accessible and usable by different end users (see specific outputs). Responsibility for implementing any changes informed by the findings would be borne by those planners and funders of relevant elements of health and care services (i.e. Clinical Commissioning Groups (e.g. mental health services, hospital services), NHS England (primary care), Local Authorities (social services)).

It is possible (and likely) that the research team at UCL will not be able to explain all the observed variation in emergency hospital use outcomes, due to lack of available information on all possible factors influencing such variation. It is unlikely, however, that the research team will not be able to explain any of the variation since previous work using a subset of the proposed indicators has been undertaken in which some of the variation could be explained by CCG-level differences in availability and quality of health care services (primary care).

Even before attempts to explain variation, benefits will also be accrued from information on the extent of national variation between CCGs in hospital use outcomes which will be used to highlight which CCGs are doing more or less well.

As yet there are not yielded benefits, but UCL anticipate these to accrue within the extension period.