{
  "abstract": "Background Biologic therapies targeting eosinophilic inflammation hold promise for COPD management. Realising their benefits will require effective patient identification and pathway development to improve access. AI-based risk prediction models offer a novel approach to stratify patients and optimise treatment delivery.Methods Using de-identified routine clinical data from Glasgow SafeHaven, we established a cohort of c38,000 patients with a coded COPD diagnosis. AI-based models were applied to the 2021 dataset, to identify 3,639 patients at highest risk of hospital admission (6 months) or mortality (12 months). Among these, 382 patients had an eosinophil count >300 cells/µL in the prior 12 months despite triple inhaler therapy, suggesting eligibility for biologic treatment.Results The high-risk group’s adverse deprivation demographics mirrored COPD burden in the wider population. Most biologic-eligible high-risk patients were aged >60 years and resided >5 km from central hospital sites where biologic therapy is typically initiated. However, a high proportion live <5 km from community vaccination hubs, presenting an opportunity to adapt treatment initiation locations. Based on RCT data, a projected reduction of 520 hospital admissions/year could be achieved in our organisation if biologic therapy could be provided to this highest risk cohort.Conclusion AI-driven risk prediction enables targeted identification of COPD patients who may benefit from biologic therapy. Model derived insights can support pathway reconfiguration to improve access and equality, particularly via decentralised treatment initiation, facilitating timely intervention and better outcomes.",
  "authors": [
    {
      "affiliations": [
        "NHS Greater Glasgow and Clyde, Glasgow, UK"
      ],
      "name": "C Carlin"
    },
    {
      "affiliations": [
        "NHS Greater Glasgow and Clyde, Glasgow, UK"
      ],
      "name": "S Burns"
    },
    {
      "affiliations": [
        "NHS Greater Glasgow and Clyde, Glasgow, UK"
      ],
      "name": "A Cushing"
    },
    {
      "affiliations": [
        "NHS Greater Glasgow and Clyde, Glasgow, UK"
      ],
      "name": "A Taylor"
    },
    {
      "affiliations": [
        "NHS Greater Glasgow and Clyde, Glasgow, UK"
      ],
      "name": "E Walker"
    },
    {
      "affiliations": [
        "NHS Greater Glasgow and Clyde, Glasgow, UK"
      ],
      "name": "D Anderson"
    },
    {
      "affiliations": [
        "NHS Greater Glasgow and Clyde, Glasgow, UK"
      ],
      "name": "DJ Lowe"
    }
  ],
  "title": "P148 AI-driven identification of high-risk COPD patients for biologic therapy: pathway development opportunities",
  "uid": "b4f10d4e-c860-5531-90f6-ed1a7c61aa5d"
}