Observational data, causal inference & subarachnoid haemorrhage
Research type
Research Study
Full title
Observational data, causal inference & subarachnoid haemorrhage
IRAS ID
328791
Contact name
H Patel
Contact email
Sponsor organisation
Northern Care Alliance NHS Foundation Trust
Duration of Study in the UK
1 years, 0 months, 1 days
Research summary
Research Summary:
Evidence-based therapeutic decision-making uses the best empirical evidence about cause and effect relationships (causal effects) between treatments and outcome in the light of the patient's characteristics. It is widely accepted in the medical community that such evidence should ideally be generated via randomised experimental studies (RCTs)
Unfortunately, randomised experiments are not universally feasible or practical, especially in low volume subspecialties (e.g., Neurosurgery) where the randomisation of a sufficient number of participants may be problematic on practical and/or ethical grounds
aSAH is a type of brain haemorrhage that affects about 8000 patients in the UK a year. There have been significant improvements in processes of care but despite this 10-15% of the hospitalised patients die, and only 50% of patients are discharged home without physical disability.
It is well recognised that aSAH is a heterogeneous condition and that patients may experience a number of complications following admission There is currently a lack of evidence to inform a rationale therapeutic approach to these, and this is an important area of research. Of special interest to our collaboration are the research recommendations in the recently published NICE aSAH guidelines about the treatment of delayed cerebral ischaemia (DCI) and use of an external ventricular drain (EVD)
DCI is considered the most important, potentially treatable, cause of mortality and morbidity following aSAH. It is characterised by a reduction in cerebral blood flow (CBF) which can be progressive, causing cerebral infarction and long-lasting disability or death. There has been an attempt to compare the two therapeutic approaches via RCT. Due to difficulties of patient recruitment , the trial had insufficient power to show a treatment effect
We have performed a naïve (non-causal) analysis of the UK and Ireland SAH dataset (UK&ISAH) and showed that insertion of an EVD is associated with a poor outcome.
We propose a proof of principle study using causal inference methodology within a well-designed registry study to help shed light on clinical intervention effects that RCT’s have not been (nor will probably ever be) able to clarify
Summary of results:
A subarachnoid haemorrhage (aSAH) is a serious type of stroke caused by bleeding around the brain. Around 8,000 people in the UK are affected every year, and although treatments have improved, many patients still die or are left with lasting disability. Doctors use several treatments to manage this condition, but there is still uncertainty about which ones are most effective, because large clinical trials are difficult to carry out in emergency neurosurgery.This study explored whether it is possible to use routinely collected hospital data to answer important treatment questions using modern statistical techniques called causal inference. These methods aim to draw conclusions about cause and effect—similar to what a randomised trial would do—but using data from real patients instead of specially designed experiments.
The research focused on two key areas of uncertainty:
1. External Ventricular Drains (EVDs) – a tube placed into the brain to relieve pressure from fluid build-up.
2. Treatments for Delayed Cerebral Ischaemia (DCI) – a common complication where parts of the brain don’t get enough blood flow.
The study was carried out in four hospitals across England and Wales over one year, collecting data on 507 patients with aSAH. The research team examined patients’ scans, clinical signs, and treatments, and used advanced statistical models to see which factors influenced treatment decisions and outcomes.The analysis showed:
Data could be collected from several hospitals, but the process was uneven due to administrative delays and differences in data systems.
Doctors varied widely in how and when they used EVDs, even for patients with similar symptoms.
Around one in four EVDs appeared to be inserted in patients who had not shown clear deterioration, suggesting potential overuse in some cases.
For patients treated for DCI, about 60% improved after treatment, but there was no clear pattern linking improvement to the patient’s initial severity.
A pilot statistical analysis showed that, although causal methods can be applied to this kind of data, large differences in how doctors treat patients make it hard to compare treatments directly. However, when focusing on a smaller, well-defined group of patients (those with mild symptoms but certain brain scan features), the models performed much better. This suggests that with clearer definitions and larger datasets, future studies could meaningfully assess which treatments truly benefit patients.
What This Means
This project demonstrated that:
• It is feasible to collect detailed data on brain haemorrhage patients from multiple hospitals.
• Causal inference methods can be used to explore how everyday treatment decisions affect outcomes.
• To get reliable answers, future research must focus on carefully defined patient groups and improve hospital data collection processes.
Next Steps
The team recommends:
• Simplifying research approval processes between hospitals.
• Strengthening data collection systems and staff training.
• Using these findings to design a larger, more focused study to assess whether routine treatments like EVDs and induced hypertension genuinely improve recovery.REC name
Yorkshire & The Humber - Leeds East Research Ethics Committee
REC reference
23/YH/0125
Date of REC Opinion
1 Jun 2023
REC opinion
Favourable Opinion