Article Text

Download PDFPDF

Development and implementation of a COVID-19 near real-time traffic light system in an acute hospital setting
Free
  1. Marcela P Vizcaychipi1,
  2. Claire L Shovlin2,
  3. Alex McCarthy3,
  4. Alice Howard1,
  5. Alexander Brown3,
  6. Michelle Hayes1,
  7. Suveer Singh1,
  8. Linsey Christie1,
  9. Alice Sisson1,
  10. Roger Davies1,
  11. Christopher Lockie1,
  12. Monica Popescu1,
  13. Amandeep Gupta1,
  14. James Armstrong1,
  15. Hisham Said1,
  16. Timothy Peters1,
  17. Richard T Keays1
  18. ChelWest COVID-19 Consortium
    1. 1 Department of Anaesthesia and Intensive Care, Chelsea and Westminster Hospital NHS Foundation Trust, London, UK
    2. 2 National Heart and Lung Institute, Imperial College London, London, UK
    3. 3 Department of Information, Data Quality and Clinical Coding, Chelsea and Westminster Hospital NHS Foundation Trust, London, UK
    1. Correspondence to Dr Marcela P Vizcaychipi, Chelsea and Westminster Hospital NHS Trust, London, UK; Marcela.Vizcaychipi{at}chelwest.nhs.uk

    Abstract

    Common causes of death in COVID-19 due to SARS-CoV-2 include thromboembolic disease, cytokine storm and adult respiratory distress syndrome (ARDS). Our aim was to develop a system for early detection of disease pattern in the emergency department (ED) that would enhance opportunities for personalised accelerated care to prevent disease progression. A single Trust’s COVID-19 response control command was established, and a reporting team with bioinformaticians was deployed to develop a real-time traffic light system to support clinical and operational teams. An attempt was made to identify predictive elements for thromboembolism, cytokine storm and ARDS based on physiological measurements and blood tests, and to communicate to clinicians managing the patient, initially via single consultants. The input variables were age, sex, and first recorded blood pressure, respiratory rate, temperature, heart rate, indices of oxygenation and C-reactive protein. Early admissions were used to refine the predictors used in the traffic lights. Of 923 consecutive patients who tested COVID-19 positive, 592 (64%) flagged at risk for thromboembolism, 241/923 (26%) for cytokine storm and 361/923 (39%) for ARDS. Thromboembolism and cytokine storm flags were met in the ED for 342 (37.1%) patients. Of the 318 (34.5%) patients receiving thromboembolism flags, 49 (5.3% of all patients) were for suspected thromboembolism, 103 (11.1%) were high-risk and 166 (18.0%) were medium-risk. Of the 89 (9.6%) who received a cytokine storm flag from the ED, 18 (2.0% of all patients) were for suspected cytokine storm, 13 (1.4%) were high-risk and 58 (6.3%) were medium-risk. Males were more likely to receive a specific traffic light flag. In conclusion, ED predictors were used to identify high proportions of COVID-19 admissions at risk of clinical deterioration due to severity of disease, enabling accelerated care targeted to those more likely to benefit. Larger prospective studies are encouraged.

    • emergency care systems
    • ventilation
    • thrombo-embolic disease
    • management
    • resuscitation
    • clinical care
    • infectious diseases
    • SARS

    This article is made freely available for use in accordance with BMJ’s website terms and conditions for the duration of the covid-19 pandemic or until otherwise determined by BMJ. You may use, download and print the article for any lawful, non-commercial purpose (including text and data mining) provided that all copyright notices and trade marks are retained.

    https://bmj.com/coronavirus/usage
    View Full Text

    Statistics from Altmetric.com

    Introduction

    Human infection due to the novel coronavirus SARS-CoV-2 (commonly referred to as COVID-19) was first reported at the end of 2019.1 The broad spectrum of disease severity ranges from asymptomatic and mild cases to severe multiorgan failure and death. As in the earliest reports,2 survival for patients developing critical disease remains poor, with adult intensive care units (AICUs) reporting mortality exceeding 50% for patients with completed encounters requiring invasive mechanical ventilation.3 4 It is imperative to try to understand progression of disease and whether any critical disease outcomes could be prevented.

    Initial attention was focused towards multilobar viral pneumonia, frequent requirement for supplementary oxygen/ventilatory support, and the development of adult respiratory distress syndrome (ARDS) in more than half of critically ill patients.2 Early risk factors for in-hospital death included older age, higher sequential organ failure assessment (SOFA) score and D-dimer greater than 750 ng/mL on admission.2 By February/March 2020, cytokine patterns were recognised as compatible with a ‘cytokine storm’ of dysregulated inflammation and hyperpyrexia.5

    At the same time, postmortem results were being shared from Italy (Drs Puya Dehgani-Mobaraki and Andrea Gianatti, personal communication 1 April 2020), the USA6 and reaching peer-reviewed publication.7 8 These demonstrated that morbid pathology is dominated by thrombosed small vessels, unsuspected venous thromboses, pulmonary emboli and features of catastrophic microvascular injury, with evidence of a procoagulant state. Thromboembolic events have become one of the most common causes of death following SARS-CoV-2 infection, and multiple advisory bodies are now issuing recommendations for thromboprophylaxis in patients with COVID-19.9–18 Guidelines are conflicting, but observational studies now suggest that benefit can be obtained from therapeutic anticoagulation,19 20 including in ventilated patients.20 Recognising potential hazards of full anticoagulation, the questions on how to risk stratify patients appropriately are taking on further importance.

    In our Trust, we created a near real-time traffic light system and accompanying clinical pathways that would allow us to accelerate the administration of personalised care as early as the point of admission into the ED. The objective of a dynamic stratification tool was twofold: (1) early identification of patients at risk of preventable complications and (2) modification of clinical management based on local clinical policies and emerging evidence. The overall intention was for this to result in an institution-wide approach that facilitated early expert input to the ED to have clear guidelines on disposition and what interventions might be needed while still in the ED.

    Methods

    Institution details

    The Chelsea and Westminster NHS Foundation Trust comprises 2 hospitals and 12 community-based clinics in North West London, delivering acute care to a diverse21 population of over 1.5 million people. The EDs at Chelsea and Westminster Hospital and West Middlesex University Hospital treat over 300 000 patients a year. As part of the Trust’s COVID-19 response, a reporting and bioinformatics team was deployed to support clinical and operational teams.

    Full traffic light system integrating with electronic patient records (EPRs)

    A new cloud database was created in Microsoft Azure with a near real-time pipeline from EPR, which was dedicated to COVID-19 reporting from an operational, nationally mandated and clinical perspective. The database was kept in a permanent development cycle to facilitate rapid additions and changes. The dashboard was created for use with COVID-19 patients in March 2020, went live on 20 March 2020 and is under continual development, capturing symptoms, observations, blood tests and elements of therapeutics.

    COVID-19 patient classification system

    Our goal was to identify predictive elements in the initial ED assessments for thromboembolism, cytokine storm and ARDS. More than 40 patient-specific variables were identified and examined initially in the first 697 patients admitted with either a positive COVID-19 result/diagnosis or had a pathology order that subsequently came back with a positive result (table 1). Early patterns led to the selection of 11 input variables for the models. These were age, sex, first recorded systolic blood pressure (BP), diastolic BP, respiratory rate, temperature, oxygen saturation (SpO2), fraction of inspired oxygen, heart rate and C-reactive protein (CRP). This dataset was passed through a Neural Network algorithm contained in SQL Server Analysis Services (SSAS) Data Mining Package.

    Table 1

    Initial elements examined for COVID-19 traffic lights

    Traffic light system

    Thromboembolic events, cytokine storm and ARDS/secondary infection risk were identified with flags using predictive markers from the SSAS data, with updated flags triggered by a change in clinical severity. The SSAS pathology results that were predictive of disease progression were CRP, D-dimer, ferritin and pro-calcitonin levels. Progression of disease was defined by these surrogates, as no flag, mild-risk flag, medium-risk flag (green), high-risk flag (amber) and suspected event flag (red, table 2). During this period, ‘no flag’ and ‘mild-risk’ were excluded from the traffic light system and managed by standard protocols. The near real-time electronic clinical dashboard was promoted across the organisation, and a link to the traffic lights was shared via internal institutional communication pathways.

    Table 2

    Final elements used in COVID-19 traffic lights

    Implementation

    From initial admission in the ED, all suspected COVID-19 patients data were autopopulated into the online COVID-19 interactive analytics clinical dashboard created in qlikView by the bioinformatic team (AM, AH and AB). The COVID-19 dashboard was updated every 10 min based on new patient data, facilitating recommended dynamic reassessments,13 and accelerated clinical management of potential thromboembolic events and cytokine storm throughout the hospital stay. In the first phase of implementation, all patient flags on the board were collated in a Microsoft Excel form and subsequently sent to an experienced critical care clinician (MPV) for review, interpretation and feedback to the clinical teams. The clinician (MPV) was also alerted by WhatsApp by the bioinformatician (AM). From 12 April, after the initial period with direct clinician oversight and communication, the flags were available to all clinical practitioners and managers in the organisation: clinicians reviewed on the clinical dashboard directly in real-time, without waiting for the end of day Excel review (figure 1).

    Figure 1

    Real-time decision support tool. The three patterns of disease that were recognised early in the outbreak to be causing mortality. In pattern 1, the disease remained single organ, usually respiratory but progressed from a viral pneumonia to a severe adult respiratory disress syndrome (ARDS)±superimposed bacterial pneumonia. Pattern 2 was characterised by circulatory collapse (cytokine storm), and pattern 3 was characterised by thromboembolism. To facilitate review across all flags, the decision support is set up as a ‘favourite’ tool in computers shared in communal areas such as nurses’ stations in the ED, acute admission unit and adult intensive care unit (AICU).

    Importantly, due to the nature of COVID-19 disease, and the priority to enhance patient care at all times, the flag system was always interpreted with caution, with anticoagulant changes reviewed at an individual level, particularly in pregnant patients, patients with severe liver disease and patients admitted with primary neurological events.

    COVID-19 pathways of care pre and post traffic lights

    Thromboembolism

    At the outset of the study, the usual institutional prophylactic dose was enoxaparin 40 mg once daily for prophylaxis of venous thromboemboli (VTE), and the usual standard of care for patients with proven or suspected VTE was full anticoagulation, with adjustments to implementation and conventional 6-month duration according to results from imaging and broader patient evaluations. The thromboembolism traffic light flags were developed to allow a broader group of patients earlier in their admission to benefit from enhanced anticoagulation. Thromboembolic events confirmed radiologically and/or D-dimer >10 000 ng/mL were accepted as best clinical markers for management of the disease and therefore used to create a predicting model to identify patients at risk of developing such complications. These dashboard markers resulted in urgent feedback to the clinical team.

    COVID-19 positive patients with ‘suspected thromboembolism’ were directed towards the usual computerised tomographic pulmonary angiogram (CTPA) to screen for an overt pulmonary embolus, and a non-contrast CT of the brain (CTB) using established institutional protocols, to exclude intracranial haemorrhage or an acute cerebral infarct, prior to initiating systemic anticoagulation. If CTB identified an acute cerebral infarct that could be at risk of haemorrhagic transformation, anticoagulation was maintained at prophylactic dose. In the absence of visible thrombus on CTPA, a microvascular thrombosis was assumed as previously reported,2 6 and therapeutic anticoagulation commenced with an intended shorter duration.

    Thromboembolism traffic light flags were used to identify a further group of ‘suspected thromboembolism’ patients (red traffic light) who received therapeutic anticoagulation unless a contraindication was present; ‘high-risk’ patients (amber traffic light) to be managed as for the suspected thromboembolism cases (therapeutic anticoagulation unless contraindication present); and a ‘medium-risk’ group (green traffic light) who would be managed by increasing thromboprophylaxis to 40 mg enoxaparin twice daily, recognising the data in reference.2

    Hyperinflammatory/’cytokine storm’

    In the earliest stage of the pandemic, this pattern was observed with sudden haemodynamic collapse after relative stability. Initial management consisted of aggressive fluid resuscitation, vasopressors and inotropic support on AICU.

    The cytokine storm traffic light flags were developed to allow a broader group of patients, even at the point of admission through the ED to benefit from early enhanced circulatory support. These dashboard markers resulted in urgent feedback to the clinical team. Patients meeting the criteria were monitored continuously and fluid management adjusted to replace insensible losses (1 mL/kg/h for every degree above 38°). If there was no response to initial fluid management and diastolic BP remained <35 mm Hg, or oxygen requirement increased, then the patients were referred to the AICU for further management.

    Acute respiratory failure syndrome (ARDS)/AICU

    In the earliest stage of the pandemic, only clinically unwell patients were referred to the AICU by the ED team as per standard Trust guidelines. Patients admitted to medical wards were then referred to the AICU team when patients presented further clinical deterioration and or failed medical management. They were managed in ward base settings until they deteriorated to reach critical disease endpoints or were discharged. Under the traffic-light system, patients flagged at risk of ARDS and requiring AICU review were referred to the AICU Consultant by the ED team. SpO2 <90% on arrival to the ED was used to flag patients at risk of ARDS, followed by the addition of PCT as a surrogate of superimposed secondary infection.

    Patients flagged for ARDS received a 30 min trial of continuous positive airway pressure (CPAP) in the ED. Failure to respond to CPAP led to a call to the AICU intubating team to proceed with endotracheal intubation. The initial invasive ventilatory support settings aimed for a target SpO2 of 92% and generally PEEP of 8 cm of water (H2O), as lung compliance was observed to be normal/high during the early phase of the disease process. Patients were proned within the first 4 hours of admission to the AICU. Cardiovascular support was managed at an individual level by continuous assessment of intravascular volume by different methods, for example, passive leg raising, non-invasive cardiac output monitoring (Cheetah, Lidco) and an initial transthoracic and lung ultrasound performed in the ED. There was not a standarised fluid balance, but a daily even balance was targeted and adjusted at an individual level when indicated. Patients’ fluid balance was part of the overall clinical management in medical wards and in the AICU. Patients admitted to AICU and acute respiratory wards received ascorbic acid 1 g twice a day, given enterally or intravenously by protocol to reduce oxidant stress in the setting of shock/microvascular shutdown.

    Patient and public involvement statement

    As part of clinical care, patients with COVID-19 infections and next of kin were informed of the severity of the disease and background work in the institution to identify preventable complications.

    Data analyses

    Selected dashboard data were downloaded on 22 April 2020 for analysis of admission indices and on 22 May 2020 to evaluate progression against the traffic light flags for thromboembolism, cytokine storm and ARDS. Analyses of perceived and actual benefits will be the subject of future communications.

    Statistical analyses were performed in Stata IC 15.1 (StataCorp, Texas, USA) and GraphPad Prism 8.1.1 (GraphPad Software, San Diego, California USA). Comparisons were performed using the Kruskal-Wallis equality-of-populations rank test, with Dunn’s multiple comparison test used to derive the multiple comparison-adjusted p values.

    Results

    Traffic light categories

    During the implementation period, 923 patients with COVID-19 were admitted through the ED, and 134 (14.5%) received a thromboembolism flag in the ED that led to immediate consideration of full anticoagulation. The flag in the ED captured 59 (40%) of the 149 patients who at any stage during their admission flagged for ‘suspected thromboembolism’ and 75 (35.6%) of all 211 patients who ultimately flagged at ‘high-risk’ of thromboembolism. Therapeutic anticoagulation was initiated in all except two patients who also presented small acute cerebrovascular events.

    The overall number of cases in the categories flagged for enhanced care are illustrated in figure 2: 241/923 (26.1%) received a cytokine storm flag, 361/923 (39.1%) received a ARDS flag and 592/923 (64.1%) received a thromboembolism flag.

    Figure 2

    Distinctive patterns and overlaps of critically ill patients. (A) Relative proportions of patients falling into dashboard-flagged categories are drawn to scale by area. The secondary infection component is not fully to scale since due to delays in set up, only 122 patients had procalcitonin results and therefore the total number was estimated for a population of 923 cases. (B) Subcategory overlaps with thromboembolism for patients flagging at risk of cytokine storm . These document all patients reaching the category criteria, so that ‘suspected’ and ‘high-risk’ groups are subgroups of ’medium-risk’. ARDS, adult respiratory distress syndrome.

    Patients receiving a cytokine storm flag met a thromboembolism flag in 85% of cases (Figure 2). ‘Medium-risk’ thromboembolism flags overlapped in 35% of cases with a cytokine storm-flag, ‘high-risk’ thromboembolism flags overlapped in 41% of cases with a cytokine storm flag and ‘suspected thromboembolism’ flags overlapped in 49% of cases with a cytokine storm flag.

    Value of traffic light flags in and beyond the ED

    Accelerations to care were also observed for patients who flagged at risk of medium-risk of thromboembolism, cytokine storm or ARDS, both in the ED, and during later ward-based care as patients moved through categories during their admission (figure 3).

    Figure 3

    Flag subcategories and dynamic progression. Patients are classified by the most severe flag received for thromboembolism (upper panel) and cytokine storm (lower panel). Symbols are drawn to scale by area and indicate the final numbers and percentages within each of the three categories (bold black text); the number of cases where the flag was the first the patient received (red or grey numbers above horizontal white lines); and the number of cases where the flag was preceded by one of the lower risk flags (red or grey numbers below horizontal white lines). Arrows indicate the number of patients progressing from a lower risk category and indicate the median number of days (and IQR) spent in the lower risk category.

    Evaluation of accuracy of the traffic light system: preliminary mortality analysis

    The ‘suspected thromboembolism’ flag was associated with an interim mortality of 42% (62/149) compared with 11% (35/331) for patients with no thromboembolism flag. The ‘high-risk’ flag derived solely from blood test results was associated with similar interim mortality at 47% (98/211, figure 4).

    Figure 4

    Heat map of interim mortality rates according to dynamic progression between thromboembolism risk categories. The upper row indicates the cases reaching the stated column thromboembolism flag with no previous flag. The middle row indicates cases with the ‘high-risk’ and ‘suspected thromboembolism’ flags that were reached after a previous ‘medium-risk’ flag. The lower row indicates cases with the ‘suspected thromboembolism’ flag reached after a previous ‘high-risk’ flag. Dunn’s multiple comparison test was used to derive the multiple comparison-adjusted p values.

    Traffic lights by sex of patient

    Details of the 923 patients admitted to the two hospitals who had tested positive for COVID-19 by 22 April 2020 are presented in the online supplementary file 1. Notably, males were over-represented in the higher categories for thromboembolism (χ2 p<0.0001, 3 df), cytokine storm (χ2 p<0.0001, 3 df) and ARDS (χ2 p=0.075, 1 df). For males, once the ‘suspected’ or ‘high-risk’ thromboembolism categories were reached, interim mortality rates were no higher than in females (data not shown).

    Supplemental material

    Discussion

    We developed a traffic light system dashboard that enabled rapid adjustment of management protocols based on patients’ own risk at the point of admission to the ED. This enabled an institution-wide approach that facilitated early expert input with clear disposition and management guidelines.

    The main study limitation is that it was a single-centre study performed at two hospital sites. It is a report of protocols put in place recognising the severe status of patients on admission to our institution, published mortality rates and the urgent aim to secure better survival rates for those infected by COVID-19 and requiring hospital admission.22 The report is limited to descriptive elements precluding detailed analysis of whether outcomes are improved or whether protocols should be adopted. Similar non-randomised, observational studies are available,19 20 and prospective clinical studies looking at early enhanced thromboembolic prophylaxis with full anticoagulation based on risk stratification are clearly needed, as indicated by others.

    The major strength of this system is that it offers stratifications to assist care delivery particularly in the ED, and prospective clinical trial design. The system represents a real-time clinical decision support tool that enables early escalation of treatments. A further strength is the reporting of bundles of care that can be implemented rapidly, synthesising elements in keeping with management practice elsewhere, as it evolves. For instance, elevated D-dimer and other biomarkers are used elsewhere for risk stratification,9 11 13 15 16 and enhanced prophylactic anticoagulation is considered for selected patient groups.11 12 15–17 A recent survey of thrombosis/haemostasis experts reported that 30% would use higher dose thromboprophylaxis for COVID-19 positive inpatients, increasing to 50% for patients on intensive care.16 Additionally, there has been a trend towards including therapeutic dosages in certain thromboprophylaxis settings,13 16 17 supported by data emerging from clinical practice reports.19 20

    While the study cannot address at present whether implementation initiated post-traffic lights prevented patients deteriorating to more severe disease endpoints, there was evidence of more patients anticoagulated if suspected thromboembolism was flagged. Thromboembolism was the most common and survival-threatening critical scenario identified in the ED. It was recognised early that theoretically, the thrombotic process was most amenable to preventative and early treatments. Results from our work and elsewhere19 20 seems to suggest that early stratification of patients at risk of thromboembolism permitted targeted full anticoagulation, and less patients may therefore be exposed to potential risks of therapeutic anticoagulation. Our data demonstrate that D-dimer and/or CRP values contribute to the early identification of a large (~40%) at-risk subgroup who may potentially benefit from early anticoagulation that was initiated in the ED. Further prospective clinical studies are warranted, but in light of this prospective real-time analysis of our clinical practice, and postmortem reports,6 ,7 ,8 ,23, it is difficult for us to concur with the endorsed recent guidelines14 18 that effectively increase the threshold for anticoagulation in patients with COVID-19 by discounting D-dimer and other biomarkers. Additional questions are whether the 25% of patients denoted at ‘medium-risk’ of thromboembolism would benefit from full anticoagulation and whether the ‘low-risk’ group would have benefited from additional intervention: these would be useful prospective studies for the future.

    A striking feature was that males were over-represented in the higher risk categories. Others have reported male susceptibility to severe COVID-19,1–4 8 and we have recently reported genomic evidence24 25 that adds to the compelling clinical data that emphasise males are biologically more at risk of severe complications if infected by COVID-19. The traffic light system appears to be able to capture males with excess risk at the time of admission through the ED.

    In conclusion, we developed and implemented a traffic light system that enabled the detection, in near real time, of the three major patterns of COVID-19 from the point of admission to the ED, where implementation enabled early ascertainment of disease progression, and accelerated enhanced treatments. We report specific clinical bundles of care that could be extended as the place for new elements of COVID-19 care become identified through randomised controlled trials.

    Acknowledgments

    The authors wish to thank the Chelsea & Westminster NHS Foundation Trust personnel and especially the adult intensive care unit nurses for the delivery of personalised care to all patients admitted with COVID-19 infection. The authors would also wish to thank the CW Plus charity for invaluable support throughout the COVID-19 outbreak. A special thanks to Dr Simon M Greenfield for revising the manuscript for grammar and syntax. The first version of the manuscript was submitted on 9 May 2020 and a preprint posted at medrxiv on 11 May 2020.

    References

    View Abstract

    Footnotes

    • Handling editor Edward Carlton

    • Collaborators Christopher Abela, MD dip Aesth Surg, FRCS Plast; Nisha Abraham-Thomas, MD; Ahmed Al-Hindawi, MD; Joanna Allam, MD FRCA; Mauro Arrica, MD FRCA; Christelle At, BSc; Javier Bargados, DipHE; Madeleine Beach, MD; Ian Beveridge, MD MRCP ; Neil Bodagh, MD; Peter Brooks, MD FRCA; Tom Browning, MD; Charo Bruce, MD; Kiran Chima, MD; James Cofie, City and Guilds; Harriet Collier, DipHE; Jonathan Collier, MD MA MFDS FRCS (OMFS) PhD; Declan Collins, MD FRCS Ed (Plast) PhD; Karen Collins, MD; Deirdre Conway, MD FRCA; Victoria Cordrey, MD; Caroline Cormack, MD FRCA; Alona Courtney, MD MSc MRCS; Mark Cox, MD FRCA; Sarah Cox, MD MRCP; Joshua Cuddihy, MD; Aleck Dalrymple, City and Guilds; Paramjeet Deol, MD FRCEM; Daryl Dob, MD FRCA; Juliet Dunn, MD FRCA; Jackeline Durbridge, MD FRCA; Simon Eccles, MD BDS FRCS (Plast); Jana Elbadaoui, DipHE; Fouad El-Hibri, MD; Muna Elsawahli, MD FRCA; Niveen El-Wahab, MD FRCA; Philippa Evans, MD; Noel Fee, DipHE; Emma Forman, MD; Gabriela Frunza, MD FRCA; Susan Gallagher, MD; Seth Galton, MD FRCA; Rea Ganatra, MD; Ajay Gandhi, MD; Gary Davies, MD FRCP; Clare Glicksman, MD; Joseph Gonzales, BSc; Lisa Greaney, MD BDS MFDS RCS FRCS OMFS; Samuel Greenhalgh, MD; Samuel Gregson, MD; Kevin Haire, MD FRCA; Sofia Hanger, MD; Seleena Haque, MD FRCA; Alison Hare, MD FRCA; Charlie Hensher, MD; Maria Herincs, MD; Alfred Hill, MD; Martine Howard, DipHE; Isabel Jones, MD FRCS (Plast); Andrzej Jandziol, MD FRCA; Mo Jawad, MD; John Jeans, MD; Jo Jennings, RGN; Ma Julve, MD; Jacyntha Kaur Khera, MD; Ami Kotecha, MD FRCA; Manisha Kulkarni, MD FRCA; Holly Lamont, MD; Corina Lee, MD FRCA; Phillip Lee, MD MRCP ; William Lever, MD; Alex Li, MD FRCA; Leda Lignos, MD; Ganga Liyanage, MD FRCA; Samantha Luff, BSc; Wanda Lui, BSc; Georgios Malietzis, MD MRCS PhD; Georgina Margiotta, MD; Zuzanna Matasova, BSc; Daniel McNaughton, MD; Ayo Meduoye, MD; Hannah Mills, MD; Alex Milne, BSc; Marco Morosin, MD; Sarah Morton, MD; Kenneth Murray, MD; Quentin Nelson, City and Guilds; Saaman Neriman, MD; Lisa Newell, DipHE; Bernard Norman, MD FRCA; Emma Norton, City and Guilds; Ben Nurdin, MD; Catherine Onuorah, MD; Leyla Osman, MD; Catherine O'Sullivan, MD; Chandni Parikh, MD; Saqib Parwez, MD; Shashank Patil, MD FRCEM; Sherina Peroos, MD; Elspeth Pickering, MD FRCA; Kris Pillay, MD FRCEM; Rob Pilling, MD FRCA; Martin Porter-Moore, City and Guilds; Olivera Potparic, MD FRCA; Kate Richardson, MD FRCA; John Roa, BSc; Eleanor Roderick, MD; Katherine Russell, MD; Atika Sabharwal, MD FRCA; Amee Samani, MD FRCA; Aleksei Sedov, DipHE; Natalie Silvey, MD; Jonathan Simon, MD; James Smellie, MD FRCS(Gen); Rebecca-Lea Smith, MD FRCA; Andrew Snell, City and Guilds; Jagdish Sokhi, MD; Ewelina Szubert, DipHE; Ben Thomas, MD FRCA; John Thornton, MD FRCA; Jose Lopes Vieira, MD; Leon Villapalos Jorge, MD MSc DIC FRCS (Plast); Annett Volger, MD FRCA; Paul Waddell, DipHE; Josh Wall, MD; Kate Wannap, MD; Patrick Ward, MD FRCA; Ilhan Wardhere, BSc; Andrea Weigert, MD FRCA; Helen Westall, MD; Maria Wilk, BSc; Andrew Williams, MD FRACS (Plast); Jessica Williams, BSc; William Wynn-Jones, MD; Steve Yentis, MD FRCA; Noel Young, MD.

    • Contributors The study was conceived and overseen by MPV. Literature searches were performed by MPV, CLS, MH and RD. The traffic light bioinformatics system was designed and implemented by AH, AB and AM. The clinical study design was by MPV. Data analysis for real-time feedback was performed by MPV and AM. Real-time analytic data application was performed by MPV, MH, SS, LC, AS, RD, CL, MP, AG, JA, TP and RTK. Data analysis was conceived and performed by CLS, and data interpretation was performed by MPV and CLS. Figure 1 was generated by MPV and remaining figures were generated by CLS. The manuscript was written by CLS and revised by MPV, MH, RD, AM and RTK. All authors reviewed and approved the final manuscript.

    • Funding The study received funding support from Chelsea & Westminster NHS Foundation Trust, London, UK, and NHS England.

    • Competing interests None declared.

    • Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.

    • Patient consent for publication Not required.

    • Ethics approval This work was approved by the Chelsea & Westminster NHS Foundation Trust Clinical Governance team and the Data Protection Officer, Head of Information Governance. As we report on routinely collected non-identifiable clinical audit data, no ethical approval was required under the UK policy framework for health and social care.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data availability statement Fully anonymised data will be available post peer-review publication on reasonable request, in accordance with institutional protocols.

    • Addendum Note added in proof. Outcome data are now reported in Vizcaychipi et al, Increase in COVID-19 inpatient survival following detection of Thromboembolic and Cytokine storm risk from the point of admission to hospital by a near real time Traffic-light System (TraCe-Tic). Braz J Infect Dis 2020; S1413-8670(20)30109-4. doi:10.1016/j.bjid.2020.07.010 [online ahead of print]

      .

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.