2. Who we are
Established in 2005,The NHS Information Centre is the central
authoritative source of health and social care information, for
England acting as a ‘hub’ for high-quality, national and local,
comparative data for all ‘secondary uses’
3. Our role
Quality & Access Delivering
Standards Improving access to solutions
Ensuring the right and interpretation of Work collaboratively
information quality, data through better alongside customers
governance and presentation and and partners to best
standards in data and reporting utilise our information
data collections Ensuring fair and to tailor make
equal access to the solutions
information
Data and Data Collection
4. Key Data Sources
Sources of health and social care data managed by the
IC include:
• Hospital Episode Statistics (HES)
• General Practice Extraction Service (GPES)
• Electronic Staff Record (ESR – NHS workforce data)
• Clinical Audit
5. Secondary uses of operational data
These are all examples of the secondary use of data derived from
operational management systems
• The benefits to using this kind of data is that it relieves burden on the
frontline and it is more likely to be of robust quality
• Together these data can form a holistic view of the English health and
social care system
• Much potential for research and development
• Key programme for the IC: Establishment of an ‘honest broker’ and ‘safe
haven’ capable of managing the authorised disclosure of information to
users
6. What are the information challenges
facing the NHS?
Source: Independent research 2008
7. Clinical Indicators
What do we know?
• The demand for the effective use of
clinical indicators within the NHS is
growing
• There is significant risk that the
plethora of different methodologies
currently in use by (all kinds of)
suppliers serve to diminish the value of
this use in the NHS
• The NHS IC has a role to ensure these
methodologies become standardised for
the NHS and available for all to access,
use and comment upon
• And there will be comments!!!
8. Clinical indicators and risk
Clinical indicators such as the HSMR are risk-adjusted outcome measures
The risk adjustment takes into account process factors such as patient
casemix (age, gender, co-morbidity levels) to estimate risk of individual
patient mortality
Typically logistic regression is used to provide a national baseline model,
which factors in casemix and other significant variables to provide the
expected number of mortalities for an NHS provider
These expected mortalities are then compared to the observed mortalities for
the provider, and the clinical indicator is derived
9. Visualising indicator variation and change
Monitoring quality and performance cross-sectionally and over time
Clinical indicators can be used to monitor and compare outcomes, both
cross-sectionally and over time
Cross-sectional comparison: comparing outcomes across NHS providers
Comparison over time: monitoring outcomes from individual providers over
time
Statistical process control (SPC) techniques are utilised to visualise and test
differences both cross-sectionally and over time
11. Monitoring over time
SMRs over time
130
120
110
100
90
SMR
80
70
60
50
40
1 2 3 4 5 6 7 8 9 10 11 12 13
Year
12. Utilisation of prior risk estimation to predict
outcomes
Monitoring outcomes by comparing to a baseline
Example of deriving a prior risk model to predict future outcomes
Comparison of expected outcomes (priors) to observed
‘Today’s posterior is tomorrow’s prior’ – how and when should we recalibrate
our baseline models to incorporate change into the baseline model?
Clearly a link between prior-based estimation and actual events – a link
between previously observed, expected and observed events – and in the
real world is that link the need to monitor and regulate?
13. When is monitoring using risk estimation
a risk in itself?
Can it provide a perverse incentive?
The majority of healthcare treatment is delivered following robust and well-
used standards and processes
As these standards and processes should deliver the same quality of care
across the board, and have minimal variation in outcome, it is viable to
monitor and compare
What about high risk treatment? What about innovative pioneering treatment
that is unproven, carries potential for great benefit but also carries more risk?
It is important that any risk based estimation and measures of outcomes
either incorporates this higher risk or estimates these events separately,
across peer groups
14. There are known unknowns...
How do we know what to monitor, to estimate the risk of and then compare
expected to observed events?
Shipman, Bristol and Mid-Staffordshire - key examples of where post-event
reviews have stated the importance of monitoring to help reduce the risk of
repeat occurrences
Monitoring mortality rates, emergency readmission rates, incidence of
hospital superbugs are all important and useful examples of utilising
outcome measures as a proxy for quality of care
What else is there? How do we know how to look for it? Clinical engagement
is fundamental and so is the utilisation of modern business intelligence
techniques
15. Forming a business intelligence function
Data and
systems
Analysis
Business
and
knowledge
statistics
16. Data mining strategy
Developing a data mining strategy
Clinical indicators developed within the NHS IC are currently principally
based on secondary care data (HES)
Combining the key datasets available to us and looking holistically for
unusual patterns and trends will help to identify ‘known unknowns’
Risk estimation and prediction forms a key way in which outcomes can be
monitored and unusual patterns of behaviour identified
Strategy will be implemented efficiently and effectively, working in
partnership with numerous stakeholders, utilising business intelligence
17. In summary
Risk estimation and comparing predicted to actual events forms an integral
part of the generation of clinical indicators
Extending our work to look across the system is an essential element of our
developing data mining strategy
This strategy will utilise existing and emerging methods used to both
visualise and estimate outcomes in healthcare, and will be delivered in
partnership with others
We need to learn and work with experts to ensure the strategy is developed
and delivered effectively. Any ideas???