BNOS: presenting our work on post-surgical outcomes

Now that the stress of presenting our work at a conference has passed and our time in the lovely city of Liverpool is coming to an end, it is a great opportunity to go back to our work and update those of you who could not make it to BNOS this year.

Although brain tumours are rare, they are the leading cause of cancer death in under 40’s. Nonetheless, national data on patient demographics, their treatment and care as well as variation in these has not been analysed and presented. GlioCova, is a project that uses linked national data, to understand brain tumour patient care and variation in these.

Neurosurgery is a key part of brain tumour patient treatment, whether it is to cure the cancer, diagnose or improve survival. Nonetheless, it is not without risks – neurosurgery is linked with complications such as venous thrombosis or hospital-acquired infections. Our previous work looked at surgeon volume and 30-day mortality.

In the GlioCova dataset, we have a big national dataset with all patients diagnosed with a primary brain tumour between 2013 and 2018. Having this data, we wanted to find out:

  • What are the rates of post-surgical outcomes in brain tumour patients in England?
  • How many patients are readmitted within 30-day or have complications?
  • Are there any patient level characteristics that lead to patients being more at risk for these unwanted outcomes?

We selected over 29,000 patients, from our total cohort of 51,775 brain tumour patients, that have undergone a first surgical intervention (either surgical resection or a biopsy). We extracted most common diagnosis codes during all surgical intervention admissions in our dataset and had a group of clinicians, clinical coders and other staff members select those that could indicate a post-surgical complication in our brain tumour patient cohort. We also calculated the scores for Elixhauer Comorbidity Index – an index used to assess other patient conditions besides cancer that could influence their outcomes such as diabetes, congestive heart failure or other neurological disorders. We then developed models to predict three post-surgical outcomes: 30-day mortality, 30-day readmission and risk of having one or more complication code and looked at which patient characteristics were influencing these outcomes.

Here are some of our key findings:

  • Age and comorbidity score increased risk of 30-day mortality and risk of complication, but age slightly decreased risk of 30-day readmission;
  • Females had lower risk of 30-day readmission and complication;
  • Biopsy was linked with lower risk of complication, but higher risk of 30-day mortality;
  • Number of complications increased 30-day mortality, but patients had lower 30-day readmission rates;
  • Patients that had other tumour types than cranial glioblastoma (acoustic neuroma, meningioma or other cranial tumours), had lower risk of 30-day mortality and readmission, but most groups had larger risk of complication;
  • Higher deprivation status linked with higher risk of 30-day readmission;

Just to note, we did have a disproportionate number of patients having surgical debulking (~28,000) versus biopsy (over 1,000), thus we need to be careful when interpreting the results for the intervention type. Also, the lower risk of readmission for patients with more complication could probably be explained by the fact that these patients are more likely to stay in hospital for longer during their index admission, and thus might be less likely to be then readmitted within 30-days.

As one conference ends, I am starting to think about what awaits. We are currently in the final stages of writing up a paper on post-surgical complications where we will discuss these results more broadly and link them with previous research. In the meantime, I am on my way to do some more work with the GlioCova dataset on surgeon and centre volume which we will be taking to an international conference (EANO) in Vienna in September.

More information on our other work can be found on the Computational Oncology website.

Leave a Reply

Your email address will not be published.