Upload
justin-heath
View
215
Download
0
Embed Size (px)
Citation preview
Limitations on cross-correlations of single indicators: The case for
a WHA clinical outcomes (unit-record) database.
Peter BaghurstPublic Health Research Unit
Women’s and Children’s HospitalWomen’s and Children’s Healthcare Network
North Adelaide, South Australia
percent epidurals in all vaginal births (2009-10)
pe
rce
nt
inst
rum
en
tals
in v
ag
ina
l bir
ths
in a
ll w
om
en
10 20 30 40
10
15
20
25
22
47
11
33
49
2593
35
31
4319
27
307
18
52
42
55
29
56
32
5763
48
5962
Spearman's r = 0.72, p = 0.00024
percent inductions in selected primiparous women (2009-10)
pe
rce
nt
cae
sare
an
se
ctio
n in
se
lect
ed
pri
mip
aro
us
wo
me
n
15 20 25 30 35 40
20
25
30
22
19
43
33
31
9
11
27
35
3
18
49
7
25
30
5763
52
42
62
55
21
56
32
48 29
59Spearman's r = 0.17, p = 0.38
percent women requiring instrumental assistance for a vaginal birth (2009-10)
% w
om
en
giv
ing
bir
th b
y C
ae
sare
an
se
ctio
n
10 15 20 25
20
25
30
35
22
47
11
33
31 9
3
25
18
35
30
7
49
271943
55 56
485257
63
292142
59
62
32
Spearman's r = 0.66, p = 0.00065
percent epidurals in all vaginal births (2009-10)
Po
st-p
art
um
ha
em
orr
ha
ge
(1
50
0+
ml)
fo
llow
ing
va
gin
al b
irth
10 20 30 40
0.5
1.0
1.5
2.0
2.5
3.0
22
47
11
33
49
25
9
335
31
43
19
27
30
7
18
52
42
29
32
57
63
48
59
62
Spearman's r = 0.56, p = 0.0061
Epidurals and PPH within one hospital
13,778 vaginal births with PPH volumes at Women’s and Children’s Hospital
No severe PPH
PPH > 1500 mls
No epidural in labour
98.09% 1.91%
Epidural 98.06% 1.99%
No significant risk associated with epidural use (Odds ratio=1.05, p = 0.70) – and after adjusting for known risk factors, the Odds Ratio fell to 0.80, (p = 0.14).
No PPH PPH > 500 mls
No epidural in labour
86.2% 13.8%
Epidural 81.9% 18.1%
Highly significant increased risk,Odds ratio 1.38, (p < 0.001) – but after adjustment the Odds Ratio fell to 0.91, (p = 0.10).
% third and fourth degree tears in SABy parity and epidural use
Parity
Epidural use Nulliparous Parous
no 4.14 0.89
yes 6.02 1.30
By instrument assistance, parity and epidural use
Instrument assistance
Not required Forceps assistance Vacuum assistance
Epidural use Nulliparous Parous Nulliparous Parous Nulliparous Parous
No 3.1 0.8 15.2 6.8 7.0 3.2
Yes 2.8 0.7 12.5 5.4 5.3 2.3
Epidural use is associated with increased perineal trauma
Epidural use is associated with less perineal trauma in all groups!!(But we needed data on individual women to construct this table)
Is stratification (selected primiparas) enough?
• Recall a selected primip is a woman giving birth– For the first time– At term (with vertex presentation)– Aged 20-34 years
• We examine outcomes in this group under the assumption that these women are more homogeneous with res[pect to their risk of particular outcomes
• But…. If we look at, say, severe perineal tears in selected primips, the probability of a tear may vary from less than 1% in a Caucasian woman giving birth spontaneously to a 2200g baby; to 50% in a Chinese woman having a 4200g baby and requiring assistance with forceps
• Also……selected primips comprise only 25% of the obstetric population. Shouldn’t we be paying more attention to the others?
The message…
• Cross-correlation of single clinical indicators is interesting – but interpretation is severely restricted – because most clinical outcomes are determined by multiple ‘risk’ factors.
• In order to compare hospital-outcomes fairly, we need to adjust for major risk factors – and this requires access to data on individual women (some times referred to as “unit-record” data).
How might it work?• Member hospitals would send de-identified data files
(e.g., Excel spreadsheets, in which each row contained data-items on every individual in their care over the past twelve months) - to WHA. – If neonatal outcomes are held in separate file – two files would
have to be uploaded
• A data analyst at WHA, in consultation with each member hospital) would pre-process these files to extract and compute common items required for the construction of clinical indicators (i.e., there would be NO demand for the uploaded data to be in a common format!)
• Clinical indicators would be generated for all members at WHA. (No more having to do it yourself!)
Advantages
• Current WHA indicators could be automatically calculated from such a database (no more pestering from me!)
• The database would become an extremely useful research tool for studies approved by WHA members
• This could well attract research funding
Potential problems
• AIHW is currently considering establishing a national database. The list of data items currently proposed is huge – and it might take many years – and a lot of good will - for all hospitals to collect data in a standardised format.
• Overlap with local jurisdictional data collections?