Dr Richard A Armstrong
Clinician-Scientist | Applied AI & Biomedical Machine Learning
I am a doctor, health data scientist and applied AI engineer. I am completing a PhD at the MRC Integrative Epidemiology Unit, University of Bristol, after more than ten years of NHS clinical practice in anaesthesia. My research combines statistical genomics, multiomic machine learning and causal inference to understand and predict complications after major surgery. Alongside it, I design, evaluate and deploy production LLM agents and ML pipelines, with an emphasis on reliability, auditability and open science.
Research Pillars
Statistical Genomics & Causal Inference
GWAS, polygenic risk scores, genetic correlation and Mendelian randomisation to uncover the biology of postoperative complications and test which drivers are causal.
Multiomic Machine Learning
Reproducible prediction pipelines integrating proteomic, metabolomic and clinical data in UK Biobank, judged by whether added biology improves real decisions.
Applied AI & Agent Engineering
Shipping production multi-agent LLM systems: tool calling, grounding in real data, evaluation benchmarks, guardrails and human-in-the-loop review.
Selected Publications
All Publications (Google Scholar) →Genetic risk factors for postoperative complications after major surgery and shared genetic aetiology with non-postoperative phenotypes
GWAS of five major postoperative complications in >140,000 UK Biobank surgical patients, identifying a risk locus for postoperative atrial fibrillation near PITX2 and showing that polygenic risk for chronic disease predicts the matching postoperative complication.
The genetic architecture of postoperative delirium after major surgery and its relationship with nonpostoperative neurocognitive conditions: A genome-wide association study
Large-scale UK Biobank analysis of >140,000 individuals uncovering genetic risk factors for postoperative delirium and its relationship to Alzheimer's Disease risk.
Inflammation and postoperative complications after major surgery: a bi-directional two-sample Mendelian randomization study
Bi-directional two-sample Mendelian randomisation study testing whether inflammation causally influences postoperative complications.
Using multiomic data to predict postoperative complications after major surgery in the UK Biobank cohort
Developed adaptive machine learning predictive pipelines using clinical and multiomic (metabolomics, proteomics) features to forecast postoperative complication risk.
Comparing human vs. machine-assisted analysis to develop a new approach for Big Qualitative Data Analysis
Compared machine-assisted and human-only analysis of 5,196 free-text responses from a national perioperative cardiac arrest survey: the hybrid approach cut analysis time substantially, but human expertise remained essential for nuance.
Outcomes from intensive care in patients with COVID-19: a systematic review and meta-analysis
A landmark meta-analysis showing COVID-19 ICU mortality over time. Over 700 citations to date and informed national and international clinical guidelines.