Transcript Document

Analysing mortality data derived
from Secondary User Services
Purpose Of Mortality Indicators
•Highlight unexpected variation and areas of concern for
further investigation
•Enable the Trust to make more informed decisions to drive
change and improvement
•Demonstrate progress towards a reduction in avoidable
deaths
•Understanding variation in mortality rates leads to the
spread of best practice
A suite of indicators should always be used when analysing
and interpreting mortality data
National Benchmarking
•Enables the Trust to compare itself against peers
and provides expected rates
•Uses data from a full financial year
•Currently based on 2012/13 Secondary User
Services (SUS) data
Hospital Standardised Mortality Ratio
(HSMR)
•Is the ratio of observed deaths to expected deaths for 56
diagnosis groups, often expressed as a percentage - if
greater than 100% then mortality has exceeded the
expected level
•These groups represent 80% of inpatient deaths
•Expected deaths are calculated using crude mortality data
adjusted for the profile of a hospital’s patients
•Factors influencing that adjustment include primary
diagnosis, age, sex co-morbidity, deprivation and method of
admission
Hospital Standardised Mortality Ratio
(HSMR)
It is important not to use HSMRs in isolation and confidence
intervals are vital
HSMRs can be distorted by changes in coding practice i.e coding
of:
•Primary diagnosis
•Inclusion or exclusion of palliative care codes
•Depth of coding (co-morbidities)
•Place of death (more patients may die in hospital if community
alternatives are limited)
HSMR December 2013 – November 2014
•For the rolling 12 month period HSMR is 86.45 (significantly
lower than expected)
•Shows a reducing trend
•WHHT is one of 9 Trusts (out of 17) across the region with
lower than expected HSMR
•Significant difference between the weekday and weekend
HSMR for emergency admissions (84.99 and 91.86), but
neither is higher than expected
•However the HSMR for septicaemia (except in labour) is
137.74 and is significantly higher than expected (75 deaths vs
54.45 expected), but analysis has shown this is due to coding
mistakes
HSMR Trend December 2010 – November
2014
HSMR Trend December 2013 – November 2014
Standardised Mortality Rate (SMR)
•Ratio of observed to expected deaths
•Expected deaths are calculated for a typical area
with the same case mix adjustment
•May be quoted as a percentage
•If higher than 100%, then observed deaths are higher
than expected
SMR December 2013 – November 2014
All diagnosis SMR is 83.63 which is lower than expected
There are however two diagnosis groups with a higher than
expected SMR:
•Septicaemia (except in labour)
•Rehabilitation care, fitting of prostheses and adjustment of
devices
SMR by Division
Division
SMR
Statistical Status
Medical
86.18
Lower than
expected
Surgical
90.82
As expected
Women’s and
children’s
64.14
Lower than
expected
SMR Divisional Diagnosis Group Outliers
Diagnosis
Group
Division
Observed
Expected SMR
Respiratory
failure,
insufficiency,
arrest (adult)
Medical
49
36.01
136.06
Acute
myocardial
infarction
Surgical
(once again
coding as
these are
ITU)
5
0.68
733.57
Septicaemia
(except in
labour)
Surgical (
reviewed
and due to
coding)
10
3.93
254.71
Pneumonia
Women’s
and
children’s
2
0.18
1106.15
All Diagnosis SMR Trend December 2010 –
November 14
All Diagnosis SMR Trend December 2013 –
November 2014
All Diagnosis SMR Peer Comparison December
2010- November 2011
All Diagnosis SMR Peer Comparison December
2011 to November 2012
All Diagnosis SMR December 2012 to November
2013
All Diagnosis SMR Peer Comparison December
2013 – November 2014
Summary Hospital Level Mortality
Indicator (SHMI)
•Mortality at Trust level across NHS England
•Published quarterly by the Health & Social Care
Information Centre (HSCIC) since October 2011
•It is the ratio between the number of patients who die
following hospitalisation at the Trust and the number of
patients expected to die
•All deaths in hospital or within 30 days post discharge are
counted
Summary Hospital Level Mortality
Indicator
•Expected deaths are based on the England average given
the characteristics of the patients
•SHMI may also be expressed as a percentage and if
greater than 100% then mortality has exceeded the
expected level
Differences Between SHMI and HSMR
SHMI includes deaths occurring outside of hospital
HSMR only includes in hospital deaths
SHMI includes deaths from all Clinical Classification System
(CCS) Groups
HSMR Includes deaths from 56 CCS Groups
Variables used in the statistical model to calculate estimated
deaths differ eg
SHMI does not include adjustment for palliative care codes or
deprivation
HSMR does include adjustment for palliative care codes and
deprivation
SHMI June 2013 – June 2014
SHMI
90.33 - within expected range
SHMI (in hospital) 88.75 - significantly lower than expected
SHMI (adjusted for palliative care) 90.51 significantly lower than
expected
3 diagnosis groups with a significantly higher than expected
SHMI:
Septicaemia
Cancer of breast
Leukaemias
178.59
193.32
194.67
Hence all deaths now coded by a consultant
SHMI Trend July 2010 – June 2014
SHMI
110
105
100
95
SHMI
90
85
80
July 10 to June 11
July 11 to June 12
July 12 to June 13
July 13 to June 14
SHMI Trend Quarter 2 2011 – Quarter 1
2014
SHMI and HSMR
By Peers For All
Admissions
July 2011 – June 2012
July 2010 – June 2011
SHMI AND HSMR
By Peers For All
Admissions
July 2013 – June 2014
July 2012 – June 2013
Cumulative Sum Analysis (CUSUM)
•The CUSUM chart provides an early warning
system for changing mortality rates
•Plots patients’ actual outcomes against their
expected outcomes sequentially over time. The
chart has upper and lower thresholds and breaching
these threshold triggers an alert
•Can reveal when a change occurred
•Is used by the CQC to monitor Trust performance
CUSUM December 2013 – November 2014
Diagnosis / Procedure Group
CUSUM
Alert
Observed
Expected
Relative
Risk
Diagnosis group
5 
80
57.7
139
HSMR basket of 56 diagnosis
group
6   4
1241
1433.8
87
Septicaemia (except in labour)
4
79
57.7
137
Sprains and strains
1
1
0.1
1828
Procedure Group
6
72
46.7
154
Chemotherapy
1
4
0.8
491
Compensation for renal failure
1
14
6.8
206
Coronary angioplasty
1
8
6.4
124
Rest of arteries and veins
1
43
30.7
140
Rest of joint
1
1
1
96
Rest of lower GI
1
2
0.9
211
Patient Safety Indicators
Currently two metrics are available through Dr
Foster
•Death in low risk diagnosis groups which is as
expected
•Death after surgery which is also as expected
Septicaemia Tracking
Relative risk remains significantly higher than expected
for the rolling 12 month period at 136.93
But
Relative risk is reducing month on month and the latest
rolling 6 month picture shows that relative risk is as
expected at 71.59
Correction of a coding error during the last 6 months has
led to an improvement on the previous 6 months
indicators
Amongst peers the Trust is improving its position
Septicaemia Trend December 2013 –
November 2014
Septicaemia Trend December 2010 –
November 2014
Septicaemia vs. Peers December 2010 –
November 2011
Septicaemia vs. Peers December 2011 –
November 2012
Septicaemia vs. Peers December 2012 –
November 2013
Septicaemia vs. Peers – Current 6 months vs.
Previous
December 2013 –
May 2014
June 2014 –
November 2014
Fractured Of Neck Of Femur Tracking
•#NOF relative risk is statistically as expected at 108.62
•An improvement from 2 months earlier when relative risk
was higher than expected at 118.57
•6 month data shows relative risk within the expected range
at 90.16
•Significantly higher than expected number of deaths on a
Sunday and for those admitted on a Sunday, though this is
not reflected across the 6 month data
•Amongst peers the Trust is improving its position
#NOF Trend December 2010 – November
2014
#NOF Trend December 2013 – November
2014
#NOF vs. Peers December 2010 –
November 2011
#NOF vs. Peers December 2011 –
November 2012
#NOF vs. Peers December 2012 –
November 2013
#NOF vs. Peers – Current 6 Months vs. Previous
June 2014 – November 2014
December 2013 – May 2014
In Summary
• The picture is one of general improvement
• The Trust is performing well within its peer group
• Several areas for further focus have been
highlighted, including the difference between
mortality on weekdays and weekend days
• Data highlights the importance of correct coding
and demonstrates the impact of coding errors on
performance analysis (scepticaemia)