A presentation by Carolina Samuelsson at the 2017 meeting of the Scandinavian Society of Anaestesiology and Intensive Care Medicine.
All available content from SSAI2017: https://scanfoam.org/ssai2017/
Delivered in collaboration between scanFOAM, SSAI & SFAI.
mHealth Israel_Healthcare Finance and M&A- What Comes Next
Using quality registry data in research
1. Quality registry data in research
Carolina Samuelsson MD PhD
Head of Department
Intensive and perioperative Care
Skåne University Hospital
Malmö and Lund
Sweden
2. National ICU registries
• The usage of registries in current
research paradigms with
examples
• Potentials for future research
3. National ICU registries
• The usage of registries in current
research paradigms with
examples
• Potentials for future research
Data driven perioperative
quality improvement and
research
the Swedish Perioperative
Register
SSAI-congress – SFAI-veckan
Lecture Hall C
Wednesday 13.30-14.30
4. The Association of Faculties of Medicine of Canada (http://phprimer.afmc.ca/)
Quantitative research designs
5. “Do what you can, with what
you have, where you are”
Theodore Roosevelt
6. - providing data
- generation of hypotheses
- aiding in designing RCTs
- detecting RCT outcome end-points
- providing infrastructure
- performing studies
Registries useful in research for…
7. Type of studies
Validation, development and improvement of modelling
Descriptive studies
Observational comparative studies
”Piggy back research”, i.e. database linkage
”Novel” trial designs
8. Type of studies
Validation, development and improvement of modelling
Descriptive studies
Observational comparative studies
”Piggy back research”, i.e. database linkage
”Novel” trial designs
9. Aim:
recalibrate the original predictive SAPS3 model with 30-day, 90-day, or
180-day mortality as measures of outcome.
Conclusion:
•recalibrated model performed well on long-term outcomes.
•The 30-day mortality may be a good benchmark of ICU performance.
•Choice of “the best end-point” is a compromise…
10. Type of studies
Validation, development and improvement of modelling
Descriptive studies
Observational comparative studies
”Piggy back research”, i.e. database linkage
”Novel” trial designs
11. Question: What is the mortality and what predicts mortality
in patients with ICU-treated refractory status epilepticus?
Answer:
•25% die within a year.
•Super-refractoriness, dependence in ADL functions,
severity of ICU-admission organ dysfunction and age predict
mortality.
12. Type of studies
Validation, development and improvement of modelling
Descriptive studies
Observational comparative studies
”Piggy back research”, i.e. database linkage
”Novel” trial designs
13. Question:
Do women of premenopausal age have better
outcome following intensive care?
Is patient gender associated to ICU-resource usage?
Crit Care. 2015 Mar 30;19:129.
Gender differences in outcome and use of resources do exist in
Swedish intensive care, but to no advantage for women of
premenopausal age.
Samuelsson C, Sjöberg F, Karlström G, Nolin T, Walther SM.
14. When adjusting for age and severity of
illness/injury
•“females of premenopausal age” have
no survival advantage
•Male patients use more ICU resources
Results:
15. Type of studies
Validation, development and improvement of modelling
Descriptive studies
Observational comparative studies
”Piggy back research”, i.e. database linkage
”Novel” trial designs
16. • Case Mix Programme Database(s)
– e.g. sociodemographics, physiology, severity,
comorbidities, diagnoses, mortality, length of stay,
organ support…..
• Unique person identifier
– allow linkage
Piggyback research
17. Aim: describe long-term risk of Chronic Kidney Disease, End
Stage Renal Disease and mortality in ICU-patients with de
novo Acute Kidney Injury.
• Swedish intensive Care Registry
• Cause of death register
• National Patient Register
• Swedish Renal Register (SRR)
18. Conclusion:
de novo acute kidney injury is associated with:
•increased risk of death
•increased risk of chronic kidney disease and end stage renal
disease
Recommendation:
Survivors after severe de novo AKI should be routinely
followed-up and renal function monitored.
19. Type of studies
Validation, development and improvement of modelling
Descriptive studies
Observational comparative studies
”Piggy back research”, i.e. database linkage
”Novel” trial designs
20.
21. Current paradigm
Has served us well:
–Ethical / scientific necessity: RCT-results should
not be predictable
–translate into improved health and lifespan
Djulbegovic et al, Nature 2013
22. Current paradigm
Concerns:
– time to plan and complete
– selected populations
– selected study centres
– surrogate endpoints
– Expensive
24. Do myocardial infarct patients benefit
from routine coronary thrombi
aspiration before percutaneous
coronary interventions or not?
25. Thrombus Aspiration in ST- Elevation myocardial
infarction in Scandinavia (the TASTE-trial)
Frobert et al, N Engl J Med 2013; 369:1587-1597
• the first Registry-based Randomized Clinical Trial
(RRCT)
• multicenter, prospective, open-label, randomized,
controlled clinical trial
• SWEDEHEART-registry used for:
- randomization
- collection of consent
- case record forms
- follow-up
26. Strengths
•Unselected populations improves external validity
•Large & complete cohorts, high-follow up, consecutive patients
•Quick
•Cheap
Weaknesses
•Data quality may not be as good
•Less complete monitoring
•Less opportunity for mechanistic sub-studies and follow-up on
secondary endpoints
”Registry-based randomized clinical trials”
James et al, Heart 2012; 98: 1329-31
31. RCT vs. RRCT
RCT RRCT
Device – CE approved, in use Yes
Device, ’first in man’ Yes
Drug in clinical practice Yes
Old drug, new indication Yes
New drug Yes
32. TASTE
Frobert et al, N Engl J Med 2013; 369:1587-1597
Advantages:
•broad inclusion criteria to ensure wide clinical applicability
•simple enrollment process
•maximize commitment/compliance of the participating
hospitals
•substantial reduction in cost
•high rates of follow-up both of patients who underwent
randomization and of those who did not.
33. NICE-National Intensive Care Evaluation Foundation
Control Theory predicts
If care provider rates own performance low by comparing it to a target they may
develop intentions to improve.
If discrepancy is great, or they lack the knowledge on how to improve, care
provider may however disregard the discrepancy or lower their target to make it
more achievable
34. Can electronic audit and feedback together with
an action implementation toolbox improve
intensive care?
cluster randomised controlled trial
12+12 Dutch ICUs
Question: Can structured electronic A&F together with a
suggested action plan improve ICU pain management?
Control: A&F only
Intervention: A&F + tool-box with suggested actions
Outcome: Pain quality indicators
Editor's Notes
Thank you for the invitation. Want to start of by thanking Mr Chariman Walther for inspirational support in preparing this lecture.
Touch upon on how we use them today.
I will also describe the future potential in using our databases in big clinical trials.
An area where our sister speciality cardiology is greatly ahead of us
Touch upon on how we use them today.
I will also describe the future potential in using our databases in clinical randomisation trials.
An area where our sister speciality cardiology is greatly ahead of us
This tree outlines different type of research designs.
Register have so far had its most important role in this area.
Less common to use registers for experimental studies or randomisation of exposure.
The amount of collected data has increased partly due automisation and digitalisation of health care.
The best, first and most important advice to anyone interested in performing registry research is : Start off by getting to know your data with respect to validity and completeness.
If we dont have the data or the data is not good enough– we should not bother asking the data scientific question.
I am a department director – as a such I have good use of our ICU-registry to descibe ”production and performance”, for billig, for managerial decisions and
To aqquire new knowledge we however can and we should ALSO use our registries for research
Here I specified different study designs where registry data is useful.
I will now run through this list and for each design I will provide ONE Scandinavian example for you to enjoy.
I will be extremely brief in this overview. My aim is only to provide ”a taste” of the type of questions that can be asked. For more details on the actual studies I, note down the references and look them up!
intensive care outcome research require good performing prediction models in order to account for confounding variables and describe our patients..
For these models to be useful they need to be updated, calibrated and validated and this process requires statistical expertise and large, complete data sets.
This is one Scandinavian example of modelling improvement research.
The authors set out to use the Simplifies Acute Physiology Score which originally predicts in-hospital mortality but they wanted….
And showed that that is possible.
And suggested that the “best endpoint” in comparing intensive care probably will be a compromise between three aspects - what is relevant to the patients, what is affected by ICU care quality, and what is practically feasible
Registries are excellent to describe – but bare in mind – they describe what has been registrered.
Straight forward descriptive observational study using the Finnish database
A question was asked.
The Answer was in the register
and the investigators also identified associations and suggested a causal relationships.
We can also use registries to compare groups for example with respect to outcome or demographics or received intervenations.
This study compared outcome.
Since animal experiments indicate a surival advantage of a female-hormonal profile this study was carried out to determine if human epidemiological data support that younger women have a survival advantage following critical illness or injury as compared to other groups.
This was a nation wide study involving more than 100 000 ICU admissions.
By comparing outcome in different groups and controllring for confounders registry data did not support the study hypothesis.
These are e.g. studies where we link case report forms to registry data for a specific patient and thereby can reduce the amount of data needed to be collected in the CRF.
PBR can also be when we, by linking several registries can improve and validate data quality and answer questions that could not be answered using one registry only.
where data inclusion follow strict definitions and is stored in a manner that prevents duplication and allows extraction.
Equally important is the existence of unique person identifier that allows correct linkage between different databases
Case Mix Programme Database (CMPD) –impose same criteria for collection of data with respect to definitions
maximise use of data/resources for data collection
maximise use of data for methodological research
avoid duplication of data and duplication of data collection..
Developing a broad programme of research of relevance to those commissioning, managing, practising and receiving critical care
Improving the quality of audit and research in critical care
Promoting the use of evidence in critical care practice and policy
Building research capacity in critical care
My example of piggy back research is this one.
Authors wanted to know more about a specific patient group. Those who needed ICU care due to acute kidney failure but had NO previous history of renal problems.
To isolate this group and follow them – the authors had to combined 4 different registries.
NPR – determine and exclude those with renal disease diagnoses prior to ICU and to determine thos who developed renal failure after ICU.
premorbid status was ascertained and classified to identifiy Patients
with pre-existing renal disease and those developing CKD (ICD-10 codes N18 and N19)
post ICU admission
Could answered their question – and provide a recommendation for our friends in the nephrology department.
WELL DONE
In the ideal world – all our care is based on high quality RCT.
To assess new treatments and old treatments in a new setting - RCT are fundamentally important.
We perform them because we CAN NOT INTUITIVELY know what is the best option.
And thanks to them there has been massive health care improvements during the last 50 yrs
reputation of RCT has however suffered lately – there are concerns about complexity, expense, and time
And what about those we do not study - the oldest, the youngest and those who speak the wrong language?
Can our registries help us out?
Registry data is collected prospectively AND can EITHER be analysed in real time OR in hindsight.
There are novel designs that may allow us to use registries and not have to perform or to repeat classical RCT
PAD is when an intervention is evaluated by comparing changes in outcomes in several different centers before and after the intervention becomes available.
When clusters of patients, instead of individuals are randomized. A cluster can be a hospital, unit, center. And each cluster receives each intervention at least once during separate periods of time
“Adaptive trials allows changes in trial characteristics … in response to information that accumulates during the trial.
An adaption can be that inclusion is stopped as soon as a significant result is observed
OR that more patients are assigned to the superior treatment as the trial progresses.
This can increase speed, reduce costs and sometimes also reduce ethical concerns
My last minutes will be spend on RRCT. Term introduced 4 yrs ago.
previous studies had been inconclusive, contradictive or underpowered.
that used the infrastructure of an existing population-based registry to facilitate patient enrollment and data collection.
6000 we
50 USD per enrolled patient
re randomized – no lost to follow ups
Inclusion was speedy 2 years
And it seems to be a winning concept. Last week this got out in print – AGAIN - a major registry-based, randomized, controlled trial using the SWEDEHEART infra-structure with an add-on randomisation module.
Large, fast, cheap and conlusive study - show that quality of PCI-care can be maintained and costs can be massively reduced by choosing heparin instead of the equally efficient, modern and expensive alternative.
I print 27 augusti.
Early on ICNARC established a clinical trials unit.
A strategic choice that has led to many studies and results that guide us all in everyday practice
that used the infrastructure of a population-based registry to facilitate patient enrollment and data collection.
In the TASTE trial, we introduced a randomization module in an online, comprehensive, national clinical registry, thus combining the benefits of randomized treatment assignment with the best features of a large-scale clinical registry.
The Thrombus Aspiration in ST-Elevation myocardial infarction in Scandinavia is a multicenter, prospective, randomized, controlled, clinical open-label trial based on the Swedish angiography and angioplasty registry (SCAAR) platform with blinded evaluation of end points. A total of 5,000 patients with STEMI undergoing primary percutaneous coronary intervention (PCI) will randomly be assigned either to conventional PCI or to thrombus aspiration followed by PCI. SCAAR will be used as the platform for randomization, allowing a broad population of all-comers in the registry network to be enrolled. All follow-up will also be done in SCAAR and other national registries. The primary end point is time to all-cause death at 30 days.
Intentions can translate to actions and behaviour
This study they used the registry infrastrucure as a platform around which randmisation was perform, data was collected and investigator will communicate with the participating ICUs
From previous research we know from before that there is a gap between knowing that things are not optimal and incentive/intention/action to do something about it.
This study is designed to investigate how this gap potentially can shrink. Reasons not to take action, not to try to improve may be the lack of practical realistic tools for improvement.
Neat small, feasable study that would have been virtually impossible to perform without the NICE-infrastructure and without pre-approved quality indicators for pain management.
Crude levels and improvement from baseline
Today Intensice care registries exists all over the world. This is N-I-C-E, the Dutch National Intensive Care Evaluation –
Sri Lanka – bed availability system
Italy Group for the Evaluation of Interventions in Intensive Care Medicine.
Australian and New Zealand Intensive Care Society –
Centre for Outcome and Resource Evaluation (ANZICS CORE) is a bi-national peer review and quality assurance program which has provided audit and analysis of the performance of Australian and New Zealand intensive care since 1992.
Finnish Intensive Care Consortium (FICC) database
Impressive brittish ICU audit system have provided us with excellent research