{
  "abstract": "Introduction Asthma is driven by inflammation and variance in lung function. While triggers including allergens worsen the condition, treatment with inhaled steroids (ICS) and beta-agonists dampen these responses. As treatment adherence is non-linear, utilizing point-in-time test results to measure inflammation and airflow obstruction, without the context of recent treatment use, makes it difficult to distinguish severe asthma from difficult to treat asthma. To address this, we developed a multi-modal model, incorporating serial lung function measures and treatment use to create a summary metric of asthma severity over time.Hypothesis Digital monitoring of trends in lung function, in combination with treatment adherence, may provide a clinically useful diagnostic tool for asthmaMethods The model was developed with data from 200 patients with difficult to treat asthma, who, over an 8-month period performed twice-daily home peak flow (PEF) readings (ASMA-1 digital PEF) while both reliver and preventer bronchodilator and ICS treatment use was digitally-assessed (INCA device). A composite score for each month was created. Model performance was assessed through assessing between group differences in inflammation and requirement for biologic therapy.Results The model, named the Peak flow Asthma Composite Test (PACT), demonstrates higher scores in those who remain persistently inflamed, compared to those who are only intermittently inflamed ( figure 1A), with a significant positive correlation between PACT and both eosinophils and FeNO (p<0.01). The score also demonstrated that those who were referred for biologic also had statistically significantly higher scores at each time point (figure 1B). Meanwhile, a negative correlation was seen with each month’s Asthma Control Test (ACT) (r = -0.26 to -0.18, all p <0.001) (figure 1C).The model was subsequently refined, utilizing machine learning tools, and incorporating steroid exposure and T2 inflammation. Linear regression demonstrated that the refined model was significantly associated with total number of respiratory exacerbations (p<0.001).Abstract P193 Figure 1A) illustrates relatively higher scores in those who were persistently inflamed. B) demonstrates higher scores seen in those who ultimately required biologic referrals. C) demonstrates an inverse relationship between ACT and PACT scoreConclusion The model reliably synthesizes clinically relevant data into a single score that predicts need for biologic therapy, persistence of T2 inflammation, and patient reported outcomes. This tool has potential to streamline asthma diagnosis and monitoring by providing clinicians with a quantitative, longitudinal assessment of both disease presence and control.",
  "authors": [
    {
      "affiliations": [
        "Department of Medicine, Royal College of Surgeons in Ireland, Dublin, Ireland"
      ],
      "name": "C Ottewill"
    },
    {
      "affiliations": [
        "Department of Medicine, Royal College of Surgeons in Ireland, Dublin, Ireland"
      ],
      "name": "C Gill"
    },
    {
      "affiliations": [
        "Department of Medicine, Royal College of Surgeons in Ireland, Dublin, Ireland"
      ],
      "name": "P Kerr"
    },
    {
      "affiliations": [
        "Department of Medicine, Royal College of Surgeons in Ireland, Dublin, Ireland"
      ],
      "name": "E MacHale"
    },
    {
      "affiliations": [
        "Department of Medicine, Royal College of Surgeons in Ireland, Dublin, Ireland"
      ],
      "name": "O Smith"
    },
    {
      "affiliations": [
        "Department of Medicine, Royal College of Surgeons in Ireland, Dublin, Ireland"
      ],
      "name": "H Doherty"
    },
    {
      "affiliations": [
        "Department of Respiratory Medicine, Beaumont Hospital, Dublin, Ireland"
      ],
      "name": "V Brennan"
    },
    {
      "affiliations": [
        "University College Dublin, Department of Mathematics and Statistics, Dublin, Ireland"
      ],
      "name": "G Greene"
    },
    {
      "affiliations": [
        "Department of Medicine, Royal College of Surgeons in Ireland, Dublin, Ireland"
      ],
      "name": "RW Costello"
    }
  ],
  "title": "P193 Development of a multi-modal model to predict asthma outcomes using digital monitoring tools",
  "uid": "3f912bb8-aec9-5288-bef6-bc8b66ebeb6b"
}