SBStuti Bagri

Stuti Bagri

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Stuti Bagri has a total of 38 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Economic Evaluation of Artificial Intelligence Deployment in Lung Cancer Chest Diagnostic Imaging

    by Kevin Herbert, Stuti Bagri, Malina Bodea, Holly Elphinstone, Naomi Fulop, Fiona Gilbert, Amanda Halliday, Rachel Lawrence, Joanne Lloyd, Raj Mehta, et al.

    Background: Lung cancer imposes a heavy burden on the NHS, where rising imaging demand and workforce shortages delay early diagnosis. Despite early promise, real-world cost-effectiveness evidence for artificial intelligence (AI) deployment remains limited.Methods: We evaluated the cost-effectiveness of AI deployment in chest diagnostic imaging versus standard practice for patients referred with potential lung cancer (LC). A short-term decision-tree model evaluated the cost-utility from an NHS and Personal Social Services perspective.Costs and health outcomes were estimated using data from ten NHS England radiology services and published literature, while chest X-ray sensitivity and specificity estimates were derived from a single NHS trust. Cost-effectiveness was assessed using incremental cost-effectiveness ratios, and incremental net monetary benefit (INMBs).Findings: AI deployment showed no evidence of improved health outcomes versus standard practice (incremental quality-adjusted life years [QALYs], <0.001), but reduced costs by £2.05 per patient. The INMB was positive at thresholds £15,000/QALY (£3.62), £20,000/QALY (£4.14) and £30,000/QALY (£5.19) respectively.Chest X-ray specificity was the main driver of incremental costs and QALYs in sensitivity analyses. AI remained less costly and equally effective in all scenarios assessed including hypothetical autonomous AI reporting of high confidence normal images.Interpretation: This evaluation provides early, real‑world evidence consistent with AI deployment in chest imaging for suspected LC being cost-effective through modest cost savings rather than health gains. However, the uncontrolled pre-post design limits causal inference, and concurrent system changes may have influenced results. Scenario analyses suggest that autonomous reporting and shared licensing could deliver additional efficiencies and cost savings.

  • Open Access

    What Is the Impact of Introducing AI for Chest X-Ray Diagnostics on the Processing of Results and Their Follow-Up? A Real-World, Mixed Methods Evaluation

    by Stuti Bagri, Malina Bodea, Rachel Lawrence, Chris Sherlaw-Johnson, Holly Elphinstone, Naomi Fulop, Fiona Gilbert, Amanda Halliday, Nina Hemmings, Kevin Herbert, et al.

    Background: The English NHS has prioritized introducing Artificial Intelligence (AI) for chest radiology diagnostics. However, there is little ‘real-world’ evidence on how AI tools work in clinical practice.Methods: Observational, mixed-methods rapid evaluation of impact of AI on processing and follow up of chest x-ray (CXR) results. We triangulated quantitative data from eight NHS sites’ radiology information systems (n=460,663 scans), and qualitative interview data (n=27, including clinicians, IT staff, project managers, and AI suppliers) from four of these eight sites.Findings: Prioritising CXR images interpreted as abnormal by AI saw faster clinical reporting of suspected cancers than clinically normal or mildly abnormal cases (site-level odds ratios ranged from 1.34 to 6.09, all p≤0.01). Impact on proportions of abnormal x-rays needing follow-up varied: 2/8 sites showed significant increases, while 3/8 showed significant decreases. There was no significant increase in numbers of CT scans conducted within three days of CXR reports. Qualitative analysis supported quantitative findings on impact and identified potential contributing factors, e.g. preexisting activity levels and backlogs, additional investments in diagnostic pathways, and readers’ experience levels.Interpretation: Through worklist prioritisation and clinical decision support, AI may contribute to improved radiology workflows. However, varied impact of AI tools across sites suggests that clinical and organisational context and implementation approaches can influence the extent of impact.

  • Open Access

    Electronic health record data to develop indicators of end-of-life care quality

    by Stuti Bagri, Sophie Julian, Miranda Davies, Sarah Scobie, Andrea L Schaffer

    Objectives: An understated disruption to health services from the COVID-19 pandemic was the increase in deaths occurring outside a hospital. Measures to monitor end-of-life care focused on community settings are needed. In this study, we explore the recording and face validity of patient-centric measures of care quality at the end of life which could be derived from primary care electronic records. Methods: With the approval of National Health Service England, we analysed the OpenSAFELY-TPP electronic healthcare records of over 970 000 patients who died between March 2019 and August 2023, covering periods before, during and after the COVID-19 pandemic. After reviewing potential measures, we developed two new measures of end-of-life care provision (specialist palliative care team contacts and advance care planning) and tracked the proportion of patients with these records, categorised by place and cause of death, along with an existing measure indicating palliative care needs. Results: The proportion of people with a general practice (GP) record of specialist palliative care was 4%–5% on average and was higher for those who died of cancer or died in a hospice. Advance care planning records increased from 19% to 27%, driven in large part by increases for patients who died in care homes. Conclusions: Advance care planning and recording of palliative care needs were plausible measures to track changes in care, unlike the specialist palliative care measure, where sparse recorded use rendered it ineffectual for service monitoring. Improved coding in primary care records would enhance the reliability of measures.

  • Open Access

    Implementing artificial intelligence in chest diagnostics for lung disease: A mixed-methods evaluation

    by Angus Ig Ramsay, Kevin Herbert, Rachel Lawrence, Chris Sherlaw-Johnson, Stuti Bagri, Nadia Crellin, Emma Dodsworth, Holly Elphinstone, Amanda Halliday, Joanne Lloyd, et al.

    Background: Artificial intelligence tools simulate aspects of human intelligence. Policy and research highlight artificial intelligence’s potential to support delivery of radiology pathways. In 2023, National Health Service (NHS) England invested £21M to deploy artificial intelligence diagnostic tools for chest X-ray and chest computed tomography in 66 NHS Trusts. Little is known about how artificial intelligence tools are implemented in practice, staff experience of these tools, and their effectiveness and cost. Objectives: Evaluate evidence on artificial intelligence tools within radiology internationally. Evaluate implementation of artificial intelligence for chest diagnostics in England. Investigate how effectiveness and cost-effectiveness of artificial intelligence for chest diagnostics can be measured. Methods: Ten-month mixed-methods study (rapid scoping review and empirical study comprising staff interviews, observations and documentary analysis). Findings were analysed using rapid assessment procedures, drawing on qualitative, quantitative and health economic approaches. Our evaluation was also designed to inform phase 2 of our study. Findings: The review included 114 articles on artificial intelligence use in radiology, internationally. Empirical work included 51 staff interviews, 57 observations and 166 documents from 10/11 of the networks and 6/66 trusts implementing artificial intelligence tools. Our review found evidence gaps, including real-world implementation of artificial intelligence tools; patient and carer experiences; impact on inequalities, sustainability and wider systems; and cost-effectiveness. Artificial intelligence for chest diagnostics was implemented in various ways, with different aims, pathways and approaches. Implementation takes time due to multiple activities related to planning, procurement, preparation for deployment, monitoring and evaluation. These tasks – technical and social – required time and resource, including a wide range of stakeholders and expertise. As of November 2024, 24/66 trusts had implemented artificial intelligence tools in practice. Implementation barriers included time, resources, challenges navigating processes and adapting these to local contexts. Facilitators included stakeholder engagement and support. Network and trust ability to evaluate service impact was influenced by factors such as data availability, data linkage, resources and capacity. Factors varied across implementation stages. Our findings indicated multiple data sources that may support measurement of effectiveness and cost-effectiveness within the English National Health Service. However, limitations to data availability need to be addressed. Limitations: Our rapid timeline meant we could not interview patients, carers, and several staff groups at trust and national levels. Delayed deployment meant we could not study implementation in practice or staff experiences. Future work: Should address real-world implementation, adaptation, and sustainability of artificial intelligence tools; impact of artificial intelligence on care, outcomes, and cost-effectiveness; and staff, patient, and carer experiences of artificial intelligence in practice. We will study these issues in phase 2 of our evaluation. Conclusions: Artificial intelligence tools may support effective, efficient chest diagnostics services. However, several factors should be considered when implementing and monitoring artificial intelligence tools. Implementation and monitoring may be improved through allowing sufficient time for procurement and preparation for deployment or extending capacity to speed completion of these tasks, early and ongoing stakeholder engagement, sufficient resourcing, dedicated expertise and clinical champions, simplifying governance processes, and improving data capacity. Parallels with learning from implementing other innovations suggests that artificial intelligence tools may not offer straightforward solutions anticipated by services and policy-makers. Funding: This article presents independent research funded by the National Institute for Health and Care Research (NIHR) Health and Social Care Delivery Research programme as award number NIHR167339.


  • Open Access

    Using electronic health record data accessed via OpenSAFELY to develop indicators of end-of-life care quality

    by Stuti Bagri, Sophie Julian, Miranda Davies, Sarah Scobie, Andrea Schaffer, The OpenSAFELY Collaborative

    An understated disruption to health services brought about by the Covid-19 pandemic was the increase in deaths occurring outside a hospital. Since quality of end-of-life care is typically monitored through place of death and hospital activity, a new approach focused on care in community settings is needed. In this study, we aimed to test whether patient-centric measures of quality at the end of life can be derived from primary care electronic records. With the approval of NHS England, analysis was undertaken in OpenSAFELY-TPP using electronic health care records of over 970,000 patients who died between March 2019 and August 2023, covering periods before, during and after the pandemic. We developed two new measures of end-of-life care quality—specialist palliative care team contacts and advance care planning, and tracked the proportion of patients with these records, categorized by place and cause of death, along with an existing measure indicating palliative care needs. The proportion of people with a GP record of specialist palliative care was 4-5% on average, higher for those who died of cancer or died in a hospice. Advance care planning records increased from 19% to 27% (barring a decrease following the onset of the Covid-19 pandemic) driven in large part by increases for patients who died in care homes. Advance care planning and recording of palliative care needs were plausible measures to track changes in care, unlike the specialist palliative care measure where recorded use was sparse. Improved coding in primary care records would improve reliability of measures. Key messages: Quality of end-of-life care is traditionally measured by how patients use health services (for example emergency department attendances) We used routine GP health records to track aspects of end-of-life care quality which matter to patients and discuss the impact of the covid-19 pandemic on these quality measures A new measure of advance care planning and the existing palliative care needs measure could be used to track end of life care delivered in the community The measure of specialist palliative care was sparsely coded and unlikely to be useful unless coding and data linkage between GP and other systems improves

  • Open Access

    Implementation, Experiences, Impact, and Costs of Artificial Intelligence in Chest Diagnostics: Protocol for a Mixed Methods Evaluation

    by Angus I G Ramsay, Chris Sherlaw-Johnson, Kevin Herbert, Stuti Bagri, Malina Bodea, Nadia Crellin, Holly Elphinstone, Amanda Halliday, Nina Hemmings, Rachel Lawrence, et al.

    Background: The ability to perform complex tasks has seen artificial intelligence (AI) used to support radiology in clinical settings, including lung cancer detection and diagnosis. Evidence suggests that AI can contribute to accurate diagnosis, reduce errors, and improve efficiency. The National Health Service England (NHSE)–funded Artificial Intelligence Diagnostic Fund (AIDF) is currently supporting 12 National Health Service (NHS) networks to implement AI for chest diagnostic imaging. There is, however, limited evidence on real-world AI implementation and use, including staff, patient, and caregiver experience, and costs and cost-effectiveness. A National Institute for Health and Care Research Rapid Service Evaluation Team Phase 1 evaluation provided insights into the early implementation of these tools and developed a framework for monitoring and evaluation of AI tools for chest diagnostic imaging in practice. Objective: This mixed methods evaluation of AI tools for chest diagnostic imaging aims to address previous research gaps by exploring the implementation of AI tools for chest diagnostic imaging, the impact and costs of implementing these service models, and the experiences of patients, caregivers, and staff. Methods: This study will be a mixed method evaluation of implementation, experiences, impact, and costs of AI for chest diagnostic imaging in NHS services in England, with the evaluation informed by the Major System Change Framework. Trust-level case studies (3 in-depth and up to 9 light-touch) will be performed, including staff member, patient, and caregiver; NHSE AIDF team interviews; meeting observations; and analysis of key relevant documentation. Qualitative data will be analyzed using Rapid Assessment Procedures and inductive thematic analysis, supplemented by in-depth deductive thematic analysis. Data from case study sites and other relevant sources will be used to assess outcomes at the other sites and for comparators. A pragmatic economic model of the chest diagnostic imaging pathway will be developed to estimate key costs and resource use associated with AI tool deployment. Together with input from national stakeholders and staff workshops, the study findings will then be finalized for reporting. Results: As of September 2025, trust-level research and development approvals with participating sites are complete, and data collection has commenced. Results are expected to be reported by the end of February 2026. Conclusions: The study will provide new insights into the facilitators and barriers to the adoption of AI technology in health care and the perceptions of both the general public and health care staff on its use. It will also inform best practices in approaches for service performance evaluation, for the implementation of AI into existing care pathways, and for the development of models to best support evidence-based decision-making. It will thus establish a framework upon which the greatest benefits of the use of AI in health care can be realized. International Registered Report Identifier (IRRID): DERR1-10.2196/81421


  • Open Access

    Implementation, Experiences, Impact, and Costs of Artificial Intelligence in Chest Diagnostics: Protocol for a Mixed Methods Evaluation (Preprint)

    by Angus I G Ramsay, Chris Sherlaw-Johnson, Kevin Herbert, Stuti Bagri, Malina Bodea, Nadia Crellin, Holly Elphinstone, Amanda Halliday, Nina Hemmings, Rachel Lawrence, et al.

    Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet #Preprint #PeerReviewMe: Warning: This is a unreviewed preprint. Readers are warned that the document has not been peer-reviewed by expert/patient reviewers or an academic editor, may contain misleading claims, and is likely to undergo changes before final publication, if accepted, or may have been rejected/withdrawn. Readers with interest and expertise are encouraged to sign up as peer-reviewer, if the paper is within an open peer-review period. Please cite this preprint only for review purposes or for grant applications and CVs (if you are the author). Background: The ability to perform complex tasks has seen artificial intelligence (AI) used to support radiology in clinical settings, including lung cancer detection and diagnosis. Evidence suggests that AI can contribute to accurate diagnosis, reduce errors, and improve efficiency. The National Health Service England (NHSE)–funded Artificial Intelligence Diagnostic Fund (AIDF) is currently supporting 12 National Health Service (NHS) networks to implement AI for chest diagnostic imaging. There is, however, limited evidence on real-world AI implementation and use, including staff, patient, and caregiver experience, and costs and cost-effectiveness. A National Institute for Health and Care Research Rapid Service Evaluation Team Phase 1 evaluation provided insights into the early implementation of these tools and developed a framework for monitoring and evaluation of AI tools for chest diagnostic imaging in practice. Objective: This mixed methods evaluation of AI tools for chest diagnostic imaging aims to address previous research gaps by exploring the implementation of AI tools for chest diagnostic imaging, the impact and costs of implementing these service models, and the experiences of patients, caregivers, and staff. Methods: This study will be a mixed method evaluation of implementation, experiences, impact, and costs of AI for chest diagnostic imaging in NHS services in England, with the evaluation informed by the Major System Change Framework. Trust-level case studies (3 in-depth and up to 9 light-touch) will be performed, including staff member, patient, and caregiver; NHSE AIDF team interviews; meeting observations; and analysis of key relevant documentation. Qualitative data will be analyzed using Rapid Assessment Procedures and inductive thematic analysis, supplemented by in-depth deductive thematic analysis. Data from case study sites and other relevant sources will be used to assess outcomes at the other sites and for comparators. A pragmatic economic model of the chest diagnostic imaging pathway will be developed to estimate key costs and resource use associated with AI tool deployment. Together with input from national stakeholders and staff workshops, the study findings will then be finalized for reporting. Results: As of September 2025, trust-level research and development approvals with participating sites are complete, and data collection has commenced. Results are expected to be reported by the end of February 2026. Conclusions: The study will provide new insights into the facilitators and barriers to the adoption of AI technology in health care and the perceptions of both the general public and health care staff on its use. It will also inform best practices in approaches for service performance evaluation, for the implementation of AI into existing care pathways, and for the development of models to best support evidence-based decision-making. It will thus establish a framework upon which the greatest benefits of the use of AI in health care can be realized. Trial Registration:

  • Open Access

    Investigating innovations in outpatient services: a mixed-methods rapid evaluation

    by Chris Sherlaw-Johnson, Theo Georghiou, Sarah Reed, Rachel Hutchings, John Appleby, Stuti Bagri, Nadia Crellin, Stephanie Kumpunen, Cyril Lobont, Jenny Negus, et al.

    Background: Within outpatient services, a broad range of innovations are being pursued to better manage care and reduce unnecessary appointments. One of the least-studied innovations is Patient-Initiated Follow-Up, which allows patients to book appointments if and when they need them, rather than follow a standard schedule. Objectives: To use routine national hospital data to identify innovations in outpatient services implemented, in recent years, within the National Health Service in England. To carry out a rapid mixed-methods evaluation of the implementation and impact of Patient-Initiated Follow-Up. Methods: The project was carried out in four sequential workstreams: (1) a rapid scoping review of outpatient innovations; (2) the application of indicator saturation methodology for scanning national patient-level data to identify potentially successful local interventions; (3) interviews with hospitals identified in workstream 2; and (4) a rapid mixed-methods evaluation of Patient-Initiated Follow-Up. The evaluation of Patient-Initiated Follow-Up comprised an evidence review, interviews with 36 clinical and operational staff at 5 National Health Service acute trusts, a workshop with staff from 13 National Health Service acute trusts, interviews with four patients, analysis of national and local data, and development of an evaluation guide. Results: Using indicator saturation, we identified nine services with notable changes in follow-up to first attendance ratios. Of three sites interviewed, two queried the data findings and one attributed the change to a clinical assessment service. Models of Patient-Initiated Follow-Up varied widely between hospital and clinical specialty, with a significant degree of variation in the approach to patient selection, patient monitoring and discharge. The success of implementation was dependent on several factors, for example, clinical condition, staff capacity and information technology systems. From the analysis of national data, we found evidence of an association between greater use of Patient-Initiated Follow-Up and a lower frequency of outpatient attendance within 15 out of 29 specialties and higher frequency of outpatient attendance within 7 specialties. Four specialties had less frequent emergency department visits associated with increasing Patient-Initiated Follow-Up rates. Patient-Initiated Follow-Up was viewed by staff and the few patients we interviewed as a positive intervention, although there was varied impact on individual staff roles and workload. It is important that sites and services undertake their own evaluations of Patient-Initiated Follow-Up. To this end we have developed an evaluation guide to support trusts with data collection and methods. Limitations: The Patient-Initiated Follow-Up evaluation was affected by a lack of patient-level data showing who is on a Patient-Initiated Follow-Up pathway. Engagement with local services was also challenging, given the pressures facing sites and staff. Patient recruitment was low, which affected the ability to understand experiences of patients directly. Conclusions: The study provides useful insights into the evolving national outpatient transformation policy and for local practice. Patient-Initiated Follow-Up is often perceived as a positive intervention for staff and patients, but the impact on individual outcomes, health inequalities, wider patient experience, workload and capacity is still uncertain. Future research: Further research should include patient-level analysis to determine clinical outcomes for individual patients on Patient-Initiated Follow-Up and health inequalities, and more extensive investigation of patient experiences. Study registration: This study is registered with the Research Registry (UIN: researchregistry8864). Funding: This award was funded by the National Institute for Health and Care Research (NIHR) Health and Social Care Delivery Research programme (NIHR award ref: 16/138/17) and is published in full in Health and Social Care Delivery Research; Vol. 12, No. 38. See the NIHR Funding and Awards website for further award information.