Digitisation of Postoperative Care to Reduce Avoidable Mortality

Postoperative mortality is the third leading cause of death worldwide, but has seen little advancement in the modern digital era. Previous work from the group has shown that the persistently high mortality is largely driven by variations in the health system’s ability to identify and respond to deteriorating patients, with worse outcomes for Māori.¹ Key contributors to improving postoperative outcomes include early warning systems, use of antibiotics, and wider use of therapeutic endoscopy and radiology.²

In New Zealand, the current EWS system is based on legacy 4-hourly manual vital sign monitoring by nurses and forms the basis of criteria-led escalation of care. The team believes the future is in continuously monitoring patients using wearable sensor technologies and advanced data analysis techniques.

“We need to transform our legacy system of care to a digital paradigm that enables the timely detection of patient deterioration, and escalation to the clinical team who can respond.”  - Prof O’Grady

Dr Wells explains, “previously, low-fidelity sensors were limited by poor validation accuracy, and vast amounts of continuous signal data could not be efficiently operationalised. However, advances in sensor technologies, neural network architectures, and graphics processor unit capacities have made it possible to implement digital systems.”

Dr Varghese adds, “this is about creating solutions that enable systems-level change. We want to design technology that works for nurses and doctors, reducing their workload and providing solutions that support the seamless analysis and automatic information relay. This is about empowering clinicians, giving them tools to support decision making, not merely giving them more data.”

“We’ll train models to understand what features contribute to adverse outcomes and what patterns precede critical clinical events and use this to help guide decision making. This has the potential to return hours of time spent collating and documenting vital signs, to clinicians who can instead focus on caring for patients.”

The team were successful in the latest HRC AI in Healthcare funding round with their 24-month proposal receiving $700K. Partnering with experts across the motu and world-leading hospitals in the UK & USA, they have set out to modernise our national EWS and establish Aotearoa New Zealand’s (AoNZ) as a frontier location for digital innovation in surgical care. Designing systems and platforms with robust data sovereignty, aligned with Te Tiriti principles and privacy guardrails is important to the team. They want their work to set an example for building systems that align with the values of AoNZ.

Their proposal will be split into five projects:

Project 1: Evaluate the current EWS across multiple hospitals to identify the optimal approach to patient monitoring and deterioration detection. It will include detailed analytics of real-world EWS data and exploratory biomarker discovery. The team’s central hypothesis is that comprehensive profiling of EWS trends preceding clinical deterioration, interventions, and safe discharge will support reliable feature engineering for machine learning models.

Project 2: Map out the health systems needs and promote the digitised surgical recovery model. Including gathering feedback on wearable technology, pathways for care, key design elements for digital solutions, and implementation strategies.

Project 3: Validate and define the value of continuous postoperative monitoring systems toward earlier

detection of clinical deterioration: a) evaluate off-the-shelf wearable cardiorespiratory sensors (UK and

AoNZ) for continuous monitoring; and b) evaluate live deterioration indices that are dynamically

generated from electronic medical records (USA). Apply AI tools to characterise dynamic

signatures indicating patient deterioration, toward automated pipelines for early detection and prediction.

Project 4: Pioneer scalable software platforms for collation, signal processing, and AI analytics. As the volume of data increases in this emerging paradigm of sensor-based patient monitoring, a robust platform to collate, synthesise, visualise, and communicate data is essential. Explainable AI will be leveraged to promote transparent modelling, and qualitative inputs will guide patient and clinician-centred design. The intention is to develop an AoNZ platform technology that will not become obsolete due to advances in AI and sensor technologies.

Project 5: Establish the ‘Global Digital Postoperative Monitoring Consortium’ to promote equity and access, and open science in AI-driven postoperative care whilst developing standardised practices that prioritise patient data security and sovereignty. A multi-institutional, server-based federated learning platform would enable AI models can be trained on local data across institutions without transferring sensitive patient information.  

Evidence shows that algorithm-based care, particularly in the postoperative setting, has the capacity to reduce mortality.³ The team believes their proposed digitised paradigm can be scaled up over time and could ultimately reduce postoperative mortality by 50% in AoNZ. Dr Wells also wants to see digital tools designed to tackle inequities. His previous research on ‘failure to rescue’ patients from post-operative complications has shown the recognition and management of deterioration is a major factor driving poorer outcomes for Māori. ⁴  Objective digital approaches to monitoring may help address subjectivity and bias.

When asked about challenges for the programme, the team acknowledged that the use of electronic medical records is not widespread on hospital wards, with paper records still being the norm. Despite this, they remain optimistic, believing that through collaboration with the health sector on this programme they can be a driver for much needed digital transformation. The potential benefits of solutions like theirs, adds strong support to the case for adopting electronic medical records.

This project captures the power of digital technologies to rethink how we deliver care and the potential of algorithms and agentic workflows to support our healthcare workforce. The team’s vision, co-design philosophy, and experience with FDA pathways position them well for success. The AI in Health Research Network looks forward to following their journey and sharing learnings and knowledge with the wider community. 

Research Team: 

Professor Greg O’Grady, Dr Chris Varghese & Dr Cameron Wells

References:

[1] Wells CI, Varghese C, Boyle LJ, McGuinness MJ, Keane C, O’Grady G, Gurney J, Koea J, Harmston C, Bissett IP.

“Failure to Rescue” following Colorectal Cancer Resection: Variation and Improvements in a National Study of Postoperative Mortality: Reducing Mortality after Colorectal Surgery. Ann Surg 2022. https://doi.org/10.1097/SLA.0000000000005650.

[2] Mohan C, Entezami P, John S, Hewitt J, Sylevych V, Psirides A. Comparison of the Aotearoa New Zealand Early Warning Score and National Early Warning Score to predict adverse inpatient events in a vital sign dataset. Anaesthesia 2023;78:830–9. https://doi.org/10.1111/anae.16007.

[3] Smits FJ, Henry AC, Besselink MG, Busch OR, van Eijck CH, Arntz M, Bollen TL, van Delden OM, van den Heuvel D, van der Leij C, van Lienden KP, Moelker A, Bonsing BA, Borel Rinkes IH, Bosscha K, van Dam RM, Derksen WJM, den Dulk M, Festen S, Groot Koerkamp B, de Haas RJ, Hagendoorn J, van der Harst E, de Hingh IH, Kazemier G, van der Kolk M, Liem M, Lips DJ, Luyer MD, de Meijer VE, Mieog JS, Nieuwenhuijs VB, Patijn GA, te Riele WW, Roos D, Schreinemakers JM, Stommel MWJ, Wit F, Zonderhuis BA, Daamen LA, van Werkhoven CH, Molenaar IQ, van Santvoort HC, Blomjous JG, de Boer MT, van den Boezem P, Bouwense S, Bruijnen R, Buis CI, del Chiaro M, Coene PP, Coolsen M, Daams F, Dejong K, Draaisma W, Eker HH, Elsen AH, Gerhards MF, Hartog H, Hoogwater FJ, Imani F, Jenniskens S, de Jong KP, Karsten TM, Klaase JM, de Kleine RHJ, van Laarhoven CJ, van der Lelij H, Manusama ER, Meerdink M, Meijerink M, Nederend J, Nijkamp MW, Nota CL, Porte RJ, Reef J, de Reuver P, van Rijswijk C, Romkens T, Rupert C, van der Schelling GP, Serafino JP, Vos LD, Vriens MR, Beers-Vural E, Wagtenberg JM, Wijsman JH, de Wilde RF, Wolfgang CL, Zeh HJ.

Algorithm-based careversus usual care for the early recognition and management of complications after pancreatic resection in the Netherlands: an open-label, nationwide, stepped-wedge cluster randomised trial. Lancet 2022. https://doi.org/10.1016/S0140-6736(22)00182-9.

[4] Wells CI, Wehipeihana E, Varghese C, Paterson L, O'Grady G, Harmston C, Gurney J, Bissett I, Koea J. Inequities in 'failure to rescue' for Indigenous Māori after gastrointestinal cancer surgery in New Zealand. Br J Surg. 2025 Aug 1;112(8). doi: 10.1093/bjs/znaf161.

Content Manager:

Nathan Baker, AI in Health Research Network

Nathan.baker@auckland.ac.nz