Can Machine Learning Personalize Cardiovascular Therapy in Sepsis?

Large randomized trials in sepsis have generally failed to find effective novel treatments. This is increasingly attributed to patient heterogeneity, including heterogeneous cardiovascular changes in septic shock. We discuss the potential for machine learning systems to personalize cardiovascular resuscitation in sepsis. While the literature is replete with proofs of concept, the technological readiness of current systems is low, with a paucity of clinical trials and proven patient benefit. Systems may be vulnerable to confounding and poor generalization to new patient populations or contemporary patterns of care. Typical electronic health records do not capture rich enough data, at sufficient temporal resolution, to produce systems that make actionable treatment suggestions. To resolve these issues, we recommend a simultaneous focus on technical challenges and removing barriers to translation. This will involve improving data quality, adopting causally grounded models, prioritizing safety assessment and integration into healthcare workflows, conducting randomized clinical trials and aligning with regulatory requirements.

How far back do we need to look to capture diagnoses in electronic health records? A retrospective observational study of hospital electronic health record data.

OBJECTIVES: Analysis of routinely collected electronic health data is a key tool for long-term condition research and practice for hospitalised patients. This requires accurate and complete ascertainment of a broad range of diagnoses, something not always recorded on an admission document at a single point in time. This study aimed to ascertain how far back in time electronic hospital records need to be interrogated to capture long-term condition diagnoses. DESIGN: Retrospective observational study of routinely collected hospital electronic health record data. SETTING: Queen Elizabeth Hospital Birmingham (UK)-linked data held by the PIONEER acute care data hub. PARTICIPANTS: Patients whose first recorded admission for chronic obstructive pulmonary disease (COPD) exacerbation (n=560) or acute stroke (n=2142) was between January and December 2018 and who had a minimum of 10 years of data prior to the index date. OUTCOME MEASURES: We identified the most common International Classification of Diseases version 10-coded diagnoses received by patients with COPD and acute stroke separately. For each diagnosis, we derived the number of patients with the diagnosis recorded at least once over the full 10-year lookback period, and then compared this with shorter lookback periods from 1 year to 9 years prior to the index admission. RESULTS: Seven of the top 10 most common diagnoses in the COPD dataset reached >90% completeness by 6 years of lookback. Atrial fibrillation and diabetes were >90% coded with 2-3 years of lookback, but hypertension and asthma completeness continued to rise all the way out to 10 years of lookback. For stroke, 4 of the top 10 reached 90% completeness by 5 years of lookback; angina pectoris was >90% coded at 7 years and previous transient ischaemic attack completeness continued to rise out to 10 years of lookback. CONCLUSION: A 7-year lookback captures most, but not all, common diagnoses. Lookback duration should be tailored to the conditions being studied.

Age-specific Reference Intervals of Abbott Intact PTH-Potential Impacts on Clinical Care.

Background: PTH assays are not standardized; therefore, method-specific PTH reference intervals are required for interpretation of results. PTH increases with age in adults but age-related reference intervals for the Abbott intact PTH (iPTH) assay are not available. Methods: Deidentified serum PTH results from September 2015 to November 2022 were retrieved from the laboratory information system of a laboratory serving a cosmopolitan population in central-west England for individuals aged 18 years and older if the estimated glomerular filtration rate was ≥60 mL/min, serum 25-hydroxyvitamin D was >50 nmol/L, and serum albumin-adjusted calcium and serum phosphate were within reference intervals. Age-specific reference intervals for Abbott iPTH were derived by an indirect method using the refineR algorithm. Results: PTH increased with age and correlated with age when controlled for 25-hydroxyvitamin D, estimated glomerular filtration rate, and adjusted calcium (r = 0.093, P < .001). The iPTH age-specific reference intervals for 4 age partitions of 18 to 45 years, 46 to 60 years, 61 to 80 years, and 81 to 95 years were 1.6 to 8.6 pmol/L, 1.8 to 9.5 pmol/L, 2.0 to 11.3 pmol/L, and 2.3 to 12.3 pmol/L, respectively. PTH was higher in women compared with men (P < .001). Sex-specific age-related reference intervals could not be derived because of the limited sample size. Conclusion: Age-specific Abbott iPTH reference intervals were derived. Application of age-specific reference intervals will impact the diagnosis and management of normocalcemic hyperparathyroidism, based on current definitions, and secondary hyperparathyroidism. Additional studies are required to clarify the effect of sex and ethnicity on PTH.

Mixed methods study protocol for combining stakeholder-led rapid evaluation with near real-time continuous registry data to facilitate evaluations of quality of care in intensive care units.

Background: Improved access to healthcare in low- and middle-income countries (LMICs) has not equated to improved health outcomes. Absence or unsustained quality of care is partly to blame. Improving outcomes in intensive care units (ICUs) requires delivery of complex interventions by multiple specialties working in concert, and the simultaneous prevention of avoidable harms associated with the illness and the treatment interventions. Therefore, successful design and implementation of improvement interventions requires understanding of the behavioural, organisational, and external factors that determine care delivery and the likelihood of achieving sustained improvement. We aim to identify care processes that contribute to suboptimal clinical outcomes in ICUs located in LMICs and to establish barriers and enablers for improving the care processes. Methods: Using rapid evaluation methods, we will use four data collection methods: 1) registry embedded indicators to assess quality of care processes and their associated outcomes; 2) process mapping to provide a preliminary framework to understand gaps between current and desired care practices; 3) structured observations of processes of interest identified from the process mapping and; 4) focus group discussions with stakeholders to identify barriers and enablers influencing the gap between current and desired care practices. We will also collect self-assessments of readiness for quality improvement. Data collection and analysis will be led by local stakeholders, performed in parallel and through an iterative process across eight countries: Kenya, India, Malaysia, Nepal, Pakistan, South Africa, Uganda and Vietnam. Conclusions: The results of our study will provide essential information on where and how care processes can be improved to facilitate better quality of care to critically ill patients in LMICs; thus, reduce preventable mortality and morbidity in ICUs. Furthermore, understanding the rapid evaluation methods that will be used for this study will allow other researchers and healthcare professionals to carry out similar research in ICUs and other health services.

Integrated rapid-cycle comparative effectiveness trials using flexible point of care randomisation in electronic health record systems.

Whilst the Randomised Controlled Trial remains the gold standard for deriving robust causal estimates of treatment efficacy, too often a traditional design proves prohibitively expensive or cumbersome when it comes to assessing questions regarding the comparative effectiveness of routinely used treatments. As a result, patients experience variation in practice as clinicians lack the evidence needed to personalise treatments effectively. This variation may be classified as unwarranted, where existing evidence is ignored, or legitimate where in the absence of evidence, clinicians rely on experience, expert opinion, and inferred principles from basic science to make decisions. We argue that within the right ethical and technological framework, legitimate variation can be transformed into a mechanism for evidence generation and learning. Learning Health Systems which harness existing variation in practice, represent a novel approach for generating evidence from everyday clinical practice. The development of these systems has gained traction due to the increased availability of modern Electronic Health Record Systems. However, despite their promise, overcoming hurdles to successfully integrating clinical trials within Learning Health Systems has proven challenging. This article describes the origins of integrated clinical trials and explores two main barriers to their further implementation - how best to obtain informed consent from patients to participate in routine comparative effectiveness research, and how to automate and integrate randomisation into a clinical workflow. Having described these barriers, we present a potential solution in the form of a research pipeline using a novel form of flexible point-of-care randomisation to allow clinicians and patients to participate in studies where there is clinical equipoise.

Pathways to interoperable electronic patient records in health and social care, Part 2: introduction for commissioners, chief clinical information officers and senior medical and social care leaders involved in health IT commissioning and strategy.

Since the start of the 2020 Coronavirus 2019 (COVID-19) pandemic, new models of care have rapidly emerged in both health and social care in the UK. The sharing of structured and unstructured data across care organisations has become increasingly important, especially in transfer of care situations and other services, such as hospital at home. At the same time synchronous and asynchronous communication between professionals, patients and their carers, which integrates with patients' records, is optimising care pathways, improving access to care and enhancing self-management of care, particularly for people with long-term conditions. Interoperability and integration of healthcare records is a complex undertaking with various technical, regulatory and organisational challenges. It requires a long-term commitment, collaboration, and investment for all stakeholders to create a comprehensive, interoperable health and social care information ecosystem across the UK. Engaged understanding of these building blocks by clinical and social care leaders will help ensure today's solutions do not become tomorrow's problems.

Clinician preference instrumental variable analysis of the effectiveness of magnesium supplementation for atrial fibrillation prophylaxis in critical care.

Atrial fibrillation is a frequently encountered condition in critical illness and causes adverse effects including haemodynamic decompensation, stroke and prolonged hospital stay. It is a common practice in critical care to supplement serum magnesium for the purpose of preventing episodes of atrial fibrillation. However, no randomised studies support this practice in the non-cardiac surgery critical care population, and the effectiveness of magnesium supplementation is unclear. We sought to investigate the effectiveness of magnesium supplementation in preventing the onset of atrial fibrillation in a mixed critical care population. We conducted a single centre retrospective observational study of adult critical care patients. We utilised a natural experiment design, using the supplementation preference of the bedside critical care nurse as an instrumental variable. Using routinely collected electronic patient data, magnesium supplementation opportunities were defined and linked to the bedside nurse. Nurse preference for administering magnesium was obtained using multilevel modelling. The results were used to define "liberal" and "restrictive" supplementation groups, which were inputted into an instrumental variable regression to obtain an estimate of the effect of magnesium supplementation. 9114 magnesium supplementation opportunities were analysed, representing 2137 critical care admissions for 1914 patients. There was significant variation in magnesium supplementation practices attributable to the individual nurse, after accounting for covariates. The instrumental variable analysis showed magnesium supplementation was associated with a 3% decreased relative risk of experiencing an atrial fibrillation event (95% CI - 0.06 to - 0.004, p = 0.03). This study supports the strategy of routine supplementation, but further work is required to identify optimal serum magnesium targets for atrial fibrillation prophylaxis.

Embedded point of care randomisation for evaluating comparative effectiveness questions: PROSPECTOR-critical care feasibility study protocol.

INTRODUCTION: Many routinely administered treatments lack evidence as to their effectiveness. When treatments lack evidence, patients receive varying care based on the preferences of clinicians. Standard randomised controlled trials are unsuited to comparisons of different routine treatment strategies, and there remains little economic incentive for change.Integrating clinical trial infrastructure into electronic health record systems offers the potential for routine treatment comparisons at scale, through reduced trial costs. To date, embedded trials have automated data collection, participant identification and eligibility screening, but randomisation and consent remain manual and therefore costly tasks.This study will investigate the feasibility of using computer prompts to allow flexible randomisation at the point of clinical decision making. It will compare the effectiveness of two prompt designs through the lens of a candidate research question-comparing liberal or restrictive magnesium supplementation practices for critical care patients. It will also explore the acceptability of two consent models for conducting comparative effectiveness research. METHODS AND ANALYSIS: We will conduct a single centre, mixed-methods feasibility study, aiming to recruit 50 patients undergoing elective surgery requiring postoperative critical care admission. Participants will be randomised to either 'Nudge' or 'Preference' designs of electronic point-of-care randomisation prompt, and liberal or restrictive magnesium supplementation.We will judge feasibility through a combination of study outcomes. The primary outcome will be the proportion of prompts displayed resulting in successful randomisation events (compliance with the allocated magnesium strategy). Secondary outcomes will evaluate the acceptability of both prompt designs to clinicians and ascertain the acceptability of pre-emptive and opt-out consent models to patients. ETHICS AND DISSEMINATION: This study was approved by Riverside Research Ethics Committee (Ref: 21/LO/0785) and will be published on completion. TRIAL REGISTRATION NUMBER: NCT05149820.

Clinical deployment environments: Five pillars of translational machine learning for health.

Machine Learning for Health (ML4H) has demonstrated efficacy in computer imaging and other self-contained digital workflows, but has failed to substantially impact routine clinical care. This is no longer because of poor adoption of Electronic Health Records Systems (EHRS), but because ML4H needs an infrastructure for development, deployment and evaluation within the healthcare institution. In this paper, we propose a design pattern called a Clinical Deployment Environment (CDE). We sketch the five pillars of the CDE: (1) real world development supported by live data where ML4H teams can iteratively build and test at the bedside (2) an ML-Ops platform that brings the rigour and standards of continuous deployment to ML4H (3) design and supervision by those with expertise in AI safety (4) the methods of implementation science that enable the algorithmic insights to influence the behaviour of clinicians and patients and (5) continuous evaluation that uses randomisation to avoid bias but in an agile manner. The CDE is intended to answer the same requirements that bio-medicine articulated in establishing the translational medicine domain. It envisions a transition from "real-world" data to "real-world" development.

Machine learning for real-time aggregated prediction of hospital admission for emergency patients.

Machine learning for hospital operations is under-studied. We present a prediction pipeline that uses live electronic health-records for patients in a UK teaching hospital's emergency department (ED) to generate short-term, probabilistic forecasts of emergency admissions. A set of XGBoost classifiers applied to 109,465 ED visits yielded AUROCs from 0.82 to 0.90 depending on elapsed visit-time at the point of prediction. Patient-level probabilities of admission were aggregated to forecast the number of admissions among current ED patients and, incorporating patients yet to arrive, total emergency admissions within specified time-windows. The pipeline gave a mean absolute error (MAE) of 4.0 admissions (mean percentage error of 17%) versus 6.5 (32%) for a benchmark metric. Models developed with 104,504 later visits during the Covid-19 pandemic gave AUROCs of 0.68-0.90 and MAE of 4.2 (30%) versus a 4.9 (33%) benchmark. We discuss how we surmounted challenges of designing and implementing models for real-time use, including temporal framing, data preparation, and changing operational conditions.

Establishing a colorectal cancer research database from routinely collected health data: the process and potential from a pilot study.

OBJECTIVE: Colorectal cancer is a common cause of death and morbidity. A significant amount of data are routinely collected during patient treatment, but they are not generally available for research. The National Institute for Health Research Health Informatics Collaborative in the UK is developing infrastructure to enable routinely collected data to be used for collaborative, cross-centre research. This paper presents an overview of the process for collating colorectal cancer data and explores the potential of using this data source. METHODS: Clinical data were collected from three pilot Trusts, standardised and collated. Not all data were collected in a readily extractable format for research. Natural language processing (NLP) was used to extract relevant information from pseudonymised imaging and histopathology reports. Combining data from many sources allowed reconstruction of longitudinal histories for each patient that could be presented graphically. RESULTS: Three pilot Trusts submitted data, covering 12 903 patients with a diagnosis of colorectal cancer since 2012, with NLP implemented for 4150 patients. Timelines showing individual patient longitudinal history can be grouped into common treatment patterns, visually presenting clusters and outliers for analysis. Difficulties and gaps in data sources have been identified and addressed. DISCUSSION: Algorithms for analysing routinely collected data from a wide range of sites and sources have been developed and refined to provide a rich data set that will be used to better understand the natural history, treatment variation and optimal management of colorectal cancer. CONCLUSION: The data set has great potential to facilitate research into colorectal cancer.

A Mass Spectrometric Approach to the Proteomic Profiling of the Canis lupus familiaris Acquired Enamel Pellicle on Hydroxyapatite Discs.

The acquired enamel pellicle (AEP) is a multi-protein film attached to the surface of teeth, which functions to lubricate the dental surface, form an anti-erosive barrier and exhibits antimicrobial properties. The initiation of AEP formation occurs within seconds of exposure to saliva, a biofluid rich in protein species. While there have been many publications on the formation of human AEP there is little research on the composition of canine AEP during its acquisition. The aim of these studies was to explore the composition of canine AEP formation, utilising hydroxyapatite (HA) discs as a tooth substitute matrix, over time. Qualitative and quantitative proteomics techniques using tandem mass tag labelled peptides and LC-MS/MS were used to follow the formation of canine AEP on hydroxyapatite discs over the course of an hour. Proteins adsorbed to the HA surface included highly abundant proteins in canine saliva, antimicrobial proteins, protease inhibitors and the buffering agent carbonic anhydrase. Greater understanding of the canine AEP deepens fundamental knowledge of the early processes driving bacterial colonisation of the tooth surface and subsequent plaque accumulation.

Combining faecal immunochemical testing with blood test results for colorectal cancer risk stratification: a consecutive cohort of 16,604 patients presenting to primary care.

BACKGROUND: Faecal immunochemical tests (FITs) are used to triage primary care patients with symptoms that could be caused by colorectal cancer for referral to colonoscopy. The aim of this study was to determine whether combining FIT with routine blood test results could improve the performance of FIT in the primary care setting. METHODS: Results of all consecutive FITs requested by primary care providers between March 2017 and December 2020 were retrieved from the Oxford University Hospitals NHS Foundation Trust. Demographic factors (age, sex), reason for referral, and results of blood tests within 90 days were also retrieved. Patients were followed up for incident colorectal cancer in linked hospital records. The sensitivity, specificity, positive and negative predictive values of FIT alone, FIT paired with blood test results, and several multivariable FIT models, were compared. RESULTS: One hundred thirty-nine colorectal cancers were diagnosed (0.8%). Sensitivity and specificity of FIT alone at a threshold of 10 µg Hb/g were 92.1 and 91.5% respectively. Compared to FIT alone, blood test results did not improve the performance of FIT. Pairing blood test results with FIT increased specificity but decreased sensitivity. Multivariable models including blood tests performed similarly to FIT alone. CONCLUSIONS: FIT is a highly sensitive tool for identifying higher risk individuals presenting to primary care with lower risk symptoms. Combining blood test results with FIT does not appear to lead to better discrimination for colorectal cancer than using FIT alone.

Nudging within learning health systems: next generation decision support to improve cardiovascular care.

The increasing volume and richness of healthcare data collected during routine clinical practice have not yet translated into significant numbers of actionable insights that have systematically improved patient outcomes. An evidence-practice gap continues to exist in healthcare. We contest that this gap can be reduced by assessing the use of nudge theory as part of clinical decision support systems (CDSS). Deploying nudges to modify clinician behaviour and improve adherence to guideline-directed therapy represents an underused tool in bridging the evidence-practice gap. In conjunction with electronic health records (EHRs) and newer devices including artificial intelligence algorithms that are increasingly integrated within learning health systems, nudges such as CDSS alerts should be iteratively tested for all stakeholders involved in health decision-making: clinicians, researchers, and patients alike. Not only could they improve the implementation of known evidence, but the true value of nudging could lie in areas where traditional randomized controlled trials are lacking, and where clinical equipoise and variation dominate. The opportunity to test CDSS nudge alerts and their ability to standardize behaviour in the face of uncertainty may generate novel insights and improve patient outcomes in areas of clinical practice currently without a robust evidence base.

Impact of the COVID-19 pandemic on routine surveillance for adults with chronic hepatitis B virus (HBV) infection in the UK.

Background: To determine the impact of the COVID-19 pandemic on the population with chronic Hepatitis B virus (HBV) infection under hospital follow-up in the UK, we quantified the coverage and frequency of measurements of biomarkers used for routine surveillance (alanine transferase [ALT] and HBV viral load). Methods: We used anonymized electronic health record data from the National Institute for Health Research (NIHR) Health Informatics Collaborative (HIC) pipeline representing five UK National Health Service (NHS) Trusts. Results: We report significant reductions in surveillance of both biomarkers during the pandemic compared to pre-COVID-19 years, both in terms of the proportion of patients who had ≥1 measurement annually, and the mean number of measurements per patient. Conclusions: These results demonstrate the real-time utility of HIC data in monitoring health-care provision, and support interventions to provide catch-up services to minimise the impact of the pandemic. Further investigation is required to determine whether these disruptions will be associated with increased rates of adverse chronic HBV outcomes.

An informatics consult approach for generating clinical evidence for treatment decisions.

BACKGROUND: An Informatics Consult has been proposed in which clinicians request novel evidence from large scale health data resources, tailored to the treatment of a specific patient. However, the availability of such consultations is lacking. We seek to provide an Informatics Consult for a situation where a treatment indication and contraindication coexist in the same patient, i.e., anti-coagulation use for stroke prevention in a patient with both atrial fibrillation (AF) and liver cirrhosis. METHODS: We examined four sources of evidence for the effect of warfarin on stroke risk or all-cause mortality from: (1) randomised controlled trials (RCTs), (2) meta-analysis of prior observational studies, (3) trial emulation (using population electronic health records (N = 3,854,710) and (4) genetic evidence (Mendelian randomisation). We developed prototype forms to request an Informatics Consult and return of results in electronic health record systems. RESULTS: We found 0 RCT reports and 0 trials recruiting for patients with AF and cirrhosis. We found broad concordance across the three new sources of evidence we generated. Meta-analysis of prior observational studies showed that warfarin use was associated with lower stroke risk (hazard ratio [HR] = 0.71, CI 0.39-1.29). In a target trial emulation, warfarin was associated with lower all-cause mortality (HR = 0.61, CI 0.49-0.76) and ischaemic stroke (HR = 0.27, CI 0.08-0.91). Mendelian randomisation served as a drug target validation where we found that lower levels of vitamin K1 (warfarin is a vitamin K1 antagonist) are associated with lower stroke risk. A pilot survey with an independent sample of 34 clinicians revealed that 85% of clinicians found information on prognosis useful and that 79% thought that they should have access to the Informatics Consult as a service within their healthcare systems. We identified candidate steps for automation to scale evidence generation and to accelerate the return of results. CONCLUSION: We performed a proof-of-concept Informatics Consult for evidence generation, which may inform treatment decisions in situations where there is dearth of randomised trials. Patients are surprised to know that their clinicians are currently not able to learn in clinic from data on 'patients like me'. We identify the key challenges in offering such an Informatics Consult as a service.

Longitudinal Analysis of the Utility of Liver Biochemistry as Prognostic Markers in Hospitalized Patients With Corona Virus Disease 2019.

The association of liver biochemistry with clinical outcomes of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection is currently unclear, and the utility of longitudinally measured liver biochemistry as prognostic markers for mortality is unknown. We aimed to determine whether abnormal liver biochemistry, assessed at baseline and at repeat measures over time, was associated with death in hospitalized patients with COVID-19 compared to those without COVID-19, in a United Kingdom population. We extracted routinely collected clinical data from a large teaching hospital in the United Kingdom, matching 585 hospitalized patients who were SARS-CoV-2 real-time reverse transcription-polymerase chain reaction (RT-PCR) positive to 1,165 hospitalized patients who were RT-PCR negative for age, sex, ethnicity, and preexisting comorbidities. A total of 26.8% (157/585) of patients with COVID-19 died compared to 11.9% (139/1,165) in the group without COVID-19 (P < 0.001). At presentation, a significantly higher proportion of the group with COVID-19 had elevated alanine aminotransferase (20.7% vs. 14.6%, P = 0.004) and hypoalbuminemia (58.7% vs. 35.0%, P < 0.001) compared to the group without COVID-19. Within the group with COVID-19, those with hypoalbuminemia at presentation had 1.83-fold increased hazards of death compared to those with normal albumin (adjusted hazard ratio [HR], 1.83; 95% confidence interval [CI], 1.25-2.67), while the hazard of death was ~4-fold higher in those aged ≥75 years (adjusted HR, 3.96; 95% CI, 2.59-6.04) and ~3-fold higher in those with preexisting liver disease (adjusted HR, 3.37; 95% CI, 1.58-7.16). In the group with COVID-19, alkaline phosphatase (ALP) increased (R = 0.192, P < 0.0001) and albumin declined (R = -0.123, P = 0.0004) over time in patients who died. Conclusion: In this United Kingdom population, liver biochemistry is commonly deranged in patients with COVID-19. Baseline hypoalbuminemia and rising ALP over time could be prognostic markers for death, but investigation of larger cohorts is required to develop a better understanding of the relationship between liver biochemistry and disease outcome.