BRIDgE Research Projects and Publications

Image
A decorative photo with features a silhouette of a human head surrounded by data science imagery and text.

COVID-19 and Palliative Care

Exploring engagement with advance and urgent care planning using Coordinate My Care Record during the COVID-19 pandemic - Dr Joanne Droney

This retrospective cohort study explored engagement with advance care planning (ACP) using Coordinate My Care (CMC), the pan-London Electronic Palliative Care Coordination System, during the COVID-19 pandemic. Advance care planning involves the discussion, documentation and communicating individual patient wishes and preferences for end-of-life care. 100,000 anonymised records were analysed in this study to identify what patients and their families wanted in terms of care at the end of life during the pandemic. This study demonstrated that even in a time of global crisis, advance care planning is feasible and impacts positively on outcomes for patients at end of life.

An analysis of prognostication using the Coordinate My Care Record - Dr Joanne Droney

A second study using the same datasets as above, i.e. Coordinate My Care (CMC), the pan-London Electronic Palliative Care Coordination System, looked at how accurate clinicians are at predicting clinical survival for patients with both cancer and non-cancer diagnoses. This has implications for services to be able to provide and coordinate the best care for each patient in the timeliest manner and in accordance with individual patient’s wishes and preferences.

Lung Cancer

Early Detection of Recurrence following Radical Radiotherapy using Machine Learning - Dr Sumeet Hindocha and Dr Richard Lee

Surveillance is universally recommended for non-small cell lung cancer (NSCLC) patients treated with curative-intent radiotherapy. High-quality evidence to inform optimal surveillance strategies is lacking. Machine learning demonstrate promise in accurate outcome prediction for a variety of health conditions. This study used routinely collected medical record data to classify patients into high and low risk groups for recurrence after having had radical radiotherapy for Non-Small Cell Lung Cancer with the aim of potentially diagnosing recurrence earlier and therefore improving outcomes.

A comparison of machine learning methods for predicting recurrence and death after curative-intent radiotherapy for non-small cell lung cancer: Development and validation of multivariable clinical prediction models

An Evaluation of Patients with Lung Nodules Using Automated Data Extraction - Dr Ben Hunter and Dr Richard Lee

A significant proportion of patients with cancer will have lung nodules identified on CT scans. Some of these nodules will be benign, but others may represent metastatic cancer or second primaries. Using advanced computing tools, such as natural language processing, on electronic health records, we identified patients with lung nodules in their scan reports to find out how many of these were diagnosed as metastatic disease, how the nodules were investigated, including the number and frequency of scans and the distribution of primary cancer types. We sought to evaluate our current service for diagnosing lung nodules and metastases in different cancer types, and whether our current routinely collected data is suitable for Structured Query Language (SQL) and Natural Language Processing analysis.

Development of a Structured Query Language and Natural Language Processing Algorithm to Identify Lung Nodules in a Cancer Centre

Breast Cancer

Using semi-automated curation of high volume routinely collected clinical data to document local moderate risk family history breast screening performance over ten years - Dr Richard Sidebottom

RMH has been conducting screening of moderate risk (of developing breast cancer) patients since 1986. Cancer detection rates, recall rates, and interval cancer information is collated for the general population national breast screening programme, however moderate risk screening is locally provided and this information has not been recorded prospectively. This study aimed to extract the information required to determine these measures from the hospital Electronic Patient Record using automated extraction methods whilst generating a tomosynthesis dataset suitable for use in AI projects, and to inform the development of larger scale data projects.

Gastrointestinal Cancer

The Impact of TP53 mutations on prognosis in metastatic colorectal adenocarcinoma - Dr Caroline Fong

Survival and responses to treatment in patients with advanced bowel cancer can vary greatly. It has been shown that tailoring treatment according to certain molecular or anatomical features associated with bowel cancer can affect treatment prescribed and patient outcomes. TP53 is a type of tumour suppressor gene, which in its normal form would slow down cell division, repair DNA mistakes or control cell death. When tumour suppressor genes do not work properly, cells grow out of control and can lead to cancer. TP53 mutations are found in up to 70% of bowel cancers. Although its role in survival or treatment response has yet to be confirmed in bowel cancer, it is associated with poorer survival for certain types of lymphoma. As there has been no definitive evidence linking TP53 mutations in bowel cancer to survival this study sought to explore this hypothesis using information already collected by the Royal Marsden Hospital, using automated data extraction techniques such as natural language processing.

Remote monitoring of Wearable Activity Trackers for detection of TOXicity in people receiving systemic anticancer treatment – Dr Olivia Curtis

WATTOX is a study that involves prospective monitoring of patients starting a new line of systemic anti-cancer treatment using a FitBit, a commercially available wearable activity tracker. The study intends to determine if this form of remote monitoring is feasible and acceptable, but also provides us with a wealth of activity and clinical data from the activity tracker, including step counts, heart rate and sleep data. We plan to use the BRIDgE platform to assist with the analysis of this large amount of data to determine if there are interesting and useful associations that could assist with the remote monitoring and treatment of patients.

Patient Reported Outcomes

Assessing the collection and reporting of PROMs data at the Royal Marsden: a Scoping Review Exercise – Emma Lidington and Linda Wadlake

A growing number of research studies are collecting patient-reported outcomes (PROs). PROs can be concepts such as health-related quality of life, symptom and functioning levels or health care satisfaction that are reported directly by patients without interpretation by anyone else often through questionnaires. Recent studies have shown, however, that methodologies outlined in randomised controlled trial protocols often lack details regarding PRO assessment and many PRO results are never published. This may cause misleading results and waste the time and effort of the patients and research teams. As the Royal Marsden (RM) is a leading cancer research centre, a number of studies collect PRO data. This review will assess the current practice of PRO data collection in clinical studies at RM and identify potential areas for improvement.