linkage between sscas data and mortality data
DESCRIPTION
Linkage between SSCAS data and mortality data. Patients’ outcome. Determined by: Prior health and personal characteristics Severity of illness Effectiveness of treatment Chance. Previous analyses by ISD. Used routinely collected hospital discharge data – SMR01 to identify cases - PowerPoint PPT PresentationTRANSCRIPT
Linkage between SSCAS data and mortality data
Patients’ outcome
Determined by:
• Prior health and personal characteristics
• Severity of illness
• Effectiveness of treatment
• Chance
Previous analyses by ISD
• Used routinely collected hospital discharge data – SMR01 to identify cases
• Linked these to death certificate data from General Register Office
• Focused on case fatality by 30 days from admission
• Was limited in ability to adjust for casemix (age, sex and deprivation by postcode)
Stroke Outcomes 1990-93
Scottish Stroke Outcomes Study
WGHVHK GRI LAW Falk
Unadjusted 6 month case fatality
20 30 40 50Case fatality % (95% CI)
VHK
GRI
Law
WGH
Falkirk
6 month case fatality
20 30 40 50Case fatality % (95% CI)
VHK
GRI
Law
WGH
Falkirk
Adjusted for •age•pre-stroke independence•can walk?•can talk & not confused?•can lift both arms?
20 30 40 50Case fatality % (95% CI)
VHK
GRI
Law
WGH
Falkirk
Unadjusted
Methods of current linkage
• All MCNs gave permission to export individual patient data
• All centres have exported individual patient data to Mike McDowall (ISD contract)
• Linked these records with those held by ISD• Preliminary analyses to look at
– Data completeness by MCN and hospital– 6 month case fatality by MCN and hospital– 6 month case fatality adjusted for casemix
Patient included in analysesAll cases on SSCAS (n= 18831)
Linked to existing patient in ISD data (n= 17344)
Data available for casemix adjustment & included in analyses (n= 10018)
Survival data available for 6 months post admission (11507)
Restricted to stroke patients only (14421)
% of casemix data missingAge Lived
Alone?
Indepen-dent
Before?
Can
talk?
Lift
both
arms?
Can
walk?
Overall 0.0 2.5 4.4 4.1 5.7 5.3
Lowest 0.0 0.1 0.0 0.0 0.0 0.0
Highest 0.7 6.5 9.1 6.6 19.8 8.0
Excluding Island HBs with very small numbers
Factors likely to influence % of missing data in SSCAS
• Completeness of medical records
• Use of proforma or ICP
• Explicit collection of casemix variables
• Training & expertise of data extractor
• ? Willingness to best guess
• Amount of clinical support available
• Frequency of missing data checks
% dead at 6 months by Health Board
101520253035404550
Ayr
shir
e &
Arr
an
Bor
ders
Arg
yl &
Cly
de
Fif
e
Gre
ater
Gla
sgow
Hig
hlan
d
Lan
arks
hire
Gra
mpi
an
Ork
ney
Lot
hian
Tay
side
For
th V
alle
y
Wes
tern
Isl
es
Dum
frie
s &
Gal
low
ay
Shet
land
% surviving at 6 months by Health Board
40
50
60
70
80
90
100
Ayr
shir
e &
Arr
an
Bor
ders
Arg
yl &
Cly
de
Fif
e
Gre
ater
Gla
sgow
Hig
hlan
d
Lan
arks
hire
Gra
mpi
an
Ork
ney
Lot
hian
Tay
side
For
th V
alle
y
Wes
tern
Isl
es
Dum
frie
s &
Gal
low
ay
Shet
land
But these crude data do not take account of casemix and chance
• Need to adjust for differences in factors which are associated with case fatality
• Need to produce 95% confidence intervals to indicate precision of estimate
• Adjusted survival data should minimise the affect that poor case ascertainment has on results e.g. if you missed all severe strokes then your casemix would be mild.
Mean Age
68
70
72
74
76
78
80
Mea
n A
ge
Health Board Scotland
Younger than average
% Cases Independent
0
20
40
60
80
100
120
% C
ases
Health Board Scotland
Odd
% Cases Living Alone
01020304050607080
% C
ases
Health Board Scotland
% Cases Able to Lift Arms
01020304050607080
% C
ases
Health Board Scotland
More severe Milder
% Cases Able to Walk
0
10
20
30
40
50
60
% C
ases
Health Board Scotland
% Cases Able to Talk
0102030405060708090
% C
ases
Health Board Scotland
% of total who showed haemorrhage on scan
0
5
1015
20
25
3035
40
45
% C
ases
Health Board Scotland
Why might casemix vary between Health Boards?
• Different populations• Different admission criteria – e.g. do patients with
minor stroke in Fife and D & G stay at home or are treated in clinic?
• Were mild or severe cases missed by SSCAS?• Was casemix data missing for particular severity
of stroke patient in some places and therefore excluded from analyses?
W score explained
• Observed number of patients surviving at 6 months
• Predicted number of patients surviving at 6 months based on– Average survival for Scotland– Modelled using 6 casemix factors
• W is excess no. of survivors at 6 months per 100 admissions over that predicted (+ values good) with 95% confidence intervals
Unadjusted
-60
-50
-40
-30
-20
-10
0
10
20
30A
&A
Bor
ders
A&
C
Fife
GG
Hig
hlan
d
Lana
rk.
Gra
mp.
Ork
ney
Loth
ian
Tay
side
FV
W Is
les
D&
G
She
tland
W s
core
Good
Bad
Good
Bad
Adjusted with 6 variable model
-60
-50
-40
-30
-20
-10
0
10
20
30A
&A
Bor
ders
A&
C
Fife
GG
Hig
hlan
d
Lana
rk.
Gra
mp.
Ork
ney
Loth
ian
Tay
side
FV
W Is
les
D&
G
She
tland
W s
core
Good
Bad
Unadjusted
-30
-20
-10
0
10
20
30
40
50
AR
I
WG
H
ER
I
ST
J
DG
RI
New
ER
I
Wis
haw
RA
H
Mon
klan
ds
Fal
kirk
RI
Cro
ssho
use
PR
I
Ayr
Inve
rcly
de
VH
K
SR
I
Vic
t In
Rai
gmor
e
QM
H
Wes
tern
/Gar
t
Sou
th G
en
Nin
ewel
ls
Bor
ders
GR
I
Sto
bhill
Wes
t Isl
es
Lorn
Val
e Le
ven
Roy
al V
ic
Hai
rmyr
es
Inch
keith
Gilb
ert B
ain
W s
core
Good
Bad
Adjusted with 6 Variable model
-30
-20
-10
0
10
20
30
40
50A
RI
WG
H
ER
I
ST
J
DG
RI
New
ER
I
Wis
haw
RA
H
Mon
klan
ds
Fal
kirk
RI
Cro
ssho
use
PR
I
Ayr
Inve
rcly
de
VH
K
SR
I
Vic
t In
Rai
gmor
e
QM
H
Wes
tern
/Gar
t
Sou
th G
en
Nin
ewel
ls
Bor
ders
GR
I
Sto
bhill
Wes
t Isl
es
Lorn
Val
e Le
ven
Roy
al V
ic
Hai
rmyr
es
Inch
keith
Gilb
ert B
ain
W s
core
Stroke Unit Trialist CollaborationMeta-analysis of trials of
stroke unit care
Absolute outcomes
Organised (SU) care
Conventional care
Risk Diff (95%CI)
Dead 22% 26% -3 (-6,-1)
Conclusions
• There are large variations in crude 6 month survival between health boards
• Most of these are due to variation in age and severity of stroke patient admitted
• Having adjusted for casemix and having taken chance into account, differences are small
Planned analyses
• Explore relationship between case fatality and process of care– Admission to stroke unit– Brain scanning– Aspirin– Discharge on secondary prevention
• Look at agreement between diagnostic codes in SSCAS and SMR01 by hospital
Discussion
• Are you happy to include these sorts of data in National Report?
• Is the process of pooling data from each Health Board satisfactory?
• Should we be making more use of these data in research?
• How could efforts of contributors be appropriately acknowledged?