The Leeds Data Model Past, Present and Future Will Ridge and Chris Charlton October 3rd 2017

Will Ridge Chris Charlton Integrated Business Intelligence Manager, Leeds City Council and Leeds CCGs Chris Charlton Senior Information Manager, Leeds CCGs and Leeds City Council

Valuing Business Intelligence Senior Leadership and Vision within Informatics and recognising the value of BI to inform decision making and whole systems transformation Use of joined up information and analytics across the city to provide knowledge and insights into how effective health and care processes are Intelligent analysis to enable the health and care system to evaluate the impact of changes, identify opportunities for change and support outcome based commissioning

Leeds Health and Care Hub LA Chief Digital Lead CCG Director of Informatics Head of IM&T & Technology Chief Analyst – City wide CCG and GP IT Team Public Health Intelligence Team Integrated Bi Manager Business Partner and Data Quality and IG Team Information Manager – City wide CCG Partnership Business Intelligence Team Business Intelligence functions of the Leeds CCGs Partnership and Leeds City Council Public Health Intelligence Teams combined to form a single intelligence function.

Example of the Leeds Data Model Single line- data set

Example of the Leeds Data Model Single line- data set

Example of the Leeds Data Model All lines- linked data

Example of the Leeds Data Model

Caretrak

Caretrak

Existing LDM ASC Activity Files ASC Contact Referrals Assessments Reablement Service Provision Reviews DSCRO ASC Activity Files Outpatients A&E Attendances Leeds Community Health Community Mental Health CIC Beds GP Data Inpatients Leeds Data Model Various Risk Models Project Specific Data Population Health Management Data? Person Identifiable Data Pseudonymised Data Activity Data no PID Data Processor Project Specific Data

IG Information Governance challenges: Original model struck in 2012- now five years old; IG needs to be updated; GDPR. Work with NHS digital to develop a more sustainable model; Resulting model can be developed and rolled out- though IG challenges remain

Requirement for new IT infrastructure; Failure of existing infrastructure; Development of a new, more sustainable approaches- move away from laptop lead activity, toward server level.

Restructure of the service; Larger influence of public health and re-orientation of the data model.

Leeds One City – diverse outcomes (iii) Analytics to identify inequalities – deprived v non deprived ; profiles; complexities of Leeds; two populations Note ‘Deprived Leeds cohort’ also over x2 England average living in areas ranking in top 10% most deprived nationally Increasing polarisation of deprivation in Leeds Analyses by PH Intelligence and other Intelligence departments of the Council suggest that the latest IMD shows a polarisation of the deprivation in Leeds, which we might expect to increase the health inequalities gap. This appears to be happening (see next slide).

Leeds Health and Care System – Macro Population Segmentation Model Life course & disease progression 9th category of episodic care Holistic person centred care, in each cohort service delivery would need to address inequalities, social, mental and emotional wellbeing, mental and physical health

Leeds data model activity and cost by per capita COST to the system Leeds data model activity and cost by per capita End of life care then frailty Also “system shocks” due to frailty Note: missing data sets, Ambulance, MH (ex Community MH)

Further Developments Labouring the metaphor from earlier, we want to improve the Leeds Data Model to include more data about context, populations and the bigger picture; Including working with health and care professionals; Orientating with other data sets; Improved population health management- closing the gap between the deprived and none deprived communities

Kent Integrated Dataset (KID) Dr Abraham P George Consultant in Public Health

Context Changing population - impacts on all care sectors with more to come as the epidemic takes hold eg. obesity, diabetes, alcohol, frailty Huge NHS and public sector funding gap Public sector services expected to discharge statutory functions with ever shrinking budgets Growing need for ‘systems thinking’ - how money and resources are being utilised for population health and wellbeing Better understanding role of public authorities in new models of care

Importance of question setting Modelling and simulation for capacity planning Predictive modelling / risk stratification Population segmentation / capitation budgets Complex care evaluation - matched controlled analyses

Current limitations in ‘systems’ planning Datasets are anonymised, cannot join one dataset to another Difficulty in understanding the ‘patient’ or ‘citizen’ journey when and how they are using services and their costs Exploring cause / effect relationship between one indicator and another difficult Many factors and services affecting population health. How can we estimate the relative impact of each? Modelling demand is not ‘linear’ - lots of assumptions are used. Are they correct? Limited benefit to predict growth and demand for health and care services Business cases for investing in new care models and services also difficult

The story so far Started 4 years ago as national pilot KCC Public Health works closely with local data warehouse team that collates and link data from >160 organisations. >50 million patient records to date Minimal cost Almost 20 analytical projects carried out supporting local health and care commissioning including Kent & Medway STP 5 times finalist for HSJ and LGC awards Interest from academic organisations who want to work with us for analyse data to help towards planning

Our vision Development & design of whole population person level linked dataset for Kent Working in partnership with CCGs to support local and national health and care service planning Map and link person level data not just from NHS and adult social but also other public authorities – Fire & Rescue, districts, education etc. Complex analytics to support evidence based investment and disinvestment Harness skills and expertise from local and national intelligence teams to improve quality of the data Develop and use innovative solutions around information technology, data protection and statistical and modelling tools

KENT INTEGRATED DATASET Flow of data into the Kent Integrated Dataset GP practice Community health Mental health Out of hours Acute hospital Public health Adult social care Ambulance service Hospice KID minimum dataset: data on activity, cost, service/treatment received, staffing, commissioning and providing organisation, patient diagnosis, demographics and location. Datasets linked on a common patient identifier (NHS number) and pseudonymised UPRN also used to link at household level KENT INTEGRATED DATASET Kent County Council Public Health and HISBI data warehouse Arrangements are in progress to link to data covering other services, including: Health and social care services: Children’s social care, child and adolescent mental health, improving access to psychological therapies, and non-SUS-reported acute care. Non-health and social care services: District council, HM Prisons, Fire and Rescue, Probation, and Education.

KENT INTEGRATED DATASET Data Quality Improvement Uses of the Kent Integrated Dataset C 1. Designing place based budgets 2. SYSTEM MODELLING 3. EVALUATION KENT INTEGRATED DATASET Data Quality Improvement Benefits are: better targeting of services integrating services to tackle complex problems stripping out the multiplicity of bodies and funding streams cuts out waste and duplication. What is the impact and benefit of individual services in the community versus all other services on population health? Generating better assumptions to help predict demand and plan service capacity better Activity Staffing Estates Finance Quality and safety

Some recent examples

For the population aged 75 and above, 67 For the population aged 75 and above, 67.4% had one or more QOF LTC, with 28.3% having 3 or more QOF LTCs.

Health Checks in Kent – Equity Audit Females were significantly more likely to complete a health check when compared to males and there was moderate evidence that this inequity increases with age. Mixed and multiple ethnic groups were significantly less likely to complete a health check when compared to the white ethnic category. This also the case in general terms for all other ethnic groups considered in the analysis. ACORN segmentation analyses Patients categorised as ‘Financially Stretched’ or ‘Urban Adversity’ were significantly less likely to complete a health check when compared to ‘Affluent Achievers’ and ‘Comfortable Communities’. Similar findings include lesser uptake also seen for ‘Anxious Adversity’, ‘Poorly Pensioners’, Hardship Heartland’, ’Perilous Futures’ and ‘Struggling Smokers’. - Very elevated smoking prevalence rates among patients completing a health check under ‘Urban Adversity’ and ‘Financially Stretched’.

Kent population model

The model interface and scenario generator Changes in population health needs in response to prevention strategies  impact on service utilization rates

VERY COMPLEX and RESOURCE INTENSIVE Key challenges faced VERY COMPLEX and RESOURCE INTENSIVE Information Governance – resource intensive Data quality Commissioner buy in Influencing national policy Limited capacity and resource availability Changing governance arrangements

Discussion Points: How can advanced analytical approaches and linked datasets be used to help improve population health outcomes?; How can we work with senior decision makers on the use of data in decision making?