Predicting ABPM status using machine learning.

  • Research type

    Research Study

  • Full title

    INFERRING TRUE HYPERTENSION STATUS USING MACHINE LEARNING CLASSIFICATION ANALYSIS

  • IRAS ID

    284649

  • Contact name

    Sandosh Padmanabhan

  • Contact email

    sandosh.padmanabhan@glasgow.ac.uk

  • Sponsor organisation

    Research Facilitator Proportionate Review Team Research and Innovation

  • Duration of Study in the UK

    3 years, 0 months, 1 days

  • Research summary

    Hypertension is a common risk factor for cardiovascular mortality with a vast majority of patients treated in primary care. Blood pressure (BP) is inherently variable. White-coat (WCH) and masked hypertension (MH) are two manifestations of this variability that have direct implications on diagnosis and management of hypertension. WCH is when an individual has high BP in a clinical setting but normal BP readings at home and masked hypertension is the opposite where the individual has normal BP in a clinical setting but high BP readings at home. While the cause of this phenomenon is not fully known, predicting whether an individual has WCH or MH will improve efficiencies in the primary care management of hypertension. Ambulatory blood pressure monitoring (ABPM) is when the BP is measured during usual daily activities, over the course of 24 hours. ABPM is the gold standard method used to establish whether a patient has true hypertension, WCH or MH which is essential for treatment decisions. In NHS Greater Glasgow & Clyde, all primary care referrals for patients with new or complex hypertension are through a nurse led virtual clinic where all patients will get an ABPM to establish the diagnosis of hypertension. Baseline demographics, medical history, primary care BP measurements from primary care letters and ABPM measurements will be collected. Machine learning (ML) is where computer algorithms are used to identify patterns in large datasets with a multitude of variables and predict various outcomes. To date, the benefit of utilising ML with primary care data to predict hypertension status from ABPM has not been evaluated. Therefore, this study aims to assess the feasibility and accuracy of ML to predict WCH, MH and true hypertension from primary care referrals to the Glasgow BP Clinic.

  • REC name

    London - Chelsea Research Ethics Committee

  • REC reference

    20/PR/0229

  • Date of REC Opinion

    20 Aug 2020

  • REC opinion

    Favourable Opinion