Critical Bias in Critical Care Devices

Critical care data contain information about the most physiologically fragile patients in the hospital, who require a significant level of monitoring. However, medical devices used for patient monitoring suffer from measurement biases that have been largely underreported. This article explores sources of bias in commonly used clinical devices, including pulse oximeters, thermometers, and sphygmomanometers. Further, it provides a framework for mitigating these biases and key principles to achieve more equitable health care delivery.

Going beyond the means: Exploring the role of bias from digital determinants of health in technologies

BACKGROUND

In light of recent retrospective studies revealing evidence of disparities in access to medical technology and of bias in measurements, this narrative review assesses digital determinants of health (DDoH) in both technologies and medical formulae that demonstrate either evidence of bias or suboptimal performance, identifies potential mechanisms behind such bias, and proposes potential methods or avenues that can guide future efforts to address these disparities.

APPROACH

Mechanisms are broadly grouped into physical and biological biases (e.g., pulse oximetry, non-contact infrared thermometry [NCIT]), interaction of human factors and cultural practices (e.g., electroencephalography [EEG]), and interpretation bias (e.g, pulmonary function tests [PFT], optical coherence tomography [OCT], and Humphrey visual field [HVF] testing). This review scope specifically excludes technologies incorporating artificial intelligence and machine learning. For each technology, we identify both clinical and research recommendations.

CONCLUSIONS

Many of the DDoH mechanisms encountered in medical technologies and formulae result in lower accuracy or lower validity when applied to patients outside the initial scope of development or validation. Our clinical recommendations caution clinical users in completely trusting result validity and suggest correlating with other measurement modalities robust to the DDoH mechanism (e.g., arterial blood gas for pulse oximetry, core temperatures for NCIT). Our research recommendations suggest not only increasing diversity in development and validation, but also awareness in the modalities of diversity required (e.g., skin pigmentation for pulse oximetry but skin pigmentation and sex/hormonal variation for NCIT). By increasing diversity that better reflects patients in all scenarios of use, we can mitigate DDoH mechanisms and increase trust and validity in clinical practice and research.

The High Price of Equity in Pulse Oximetry: A cost evaluation and need for interim solutions

Importance

Disparities in pulse oximetry accuracy, disproportionately affecting patients of color, have been associated with serious clinical outcomes. Although many have called for pulse oximetry hardware replacement, the cost associated with this replacement is not known.

Objective

To estimate the cost of replacing all pulse oximetry hardware throughout a hospital system.

Design

Single-center survey, 2023.

Setting

Single center.

Participants

One academic medical center with three hospitals.

Main Outcomes and Measures

Cost of fleet replacement as identified by current day prices for hardware.

Results

New and used prices for 5,079/5,678 (89.5%) across three hospitals for pulse oximetry devices were found. The average equipment cost to replace pulse oximetry hardware is $15,704.12 per bed. Replacement and integration costs are estimated at $28.5-31.8 million for the entire medical system. Extrapolating these costs to 5,564 hospitals in the United States results in an estimated cost of $14.1 billion.

Conclusions and Relevance

"Simply replacing" pulse oximetry hardware to address disparities may be neither simple, cheap, or timely. Solutions for addressing pulse oximetry accuracy disparities leveraging current technology may be necessary.

Trial Registration

Pro00113724, exempt.

Sepsis-Induced Coagulopathy and Disseminated Intravascular Coagulation: What We Need to Know and How to Manage for Prolonged Casualty Care

Coagulopathy can occur in trauma, and it can affect septic patients as a host tries to respond to infection. Sometimes, it can lead to disseminated intravascular coagulopathy (DIC) with a high potential for mortality. New research has delineated risk factors that include neutrophil extracellular traps and endothelial glycocalyx shedding. Managing DIC in septic patients focuses on first treating the underlying cause of sepsis. Further, the International Society on Thrombolysis and Haemostasis (ISTH) has DIC diagnostic criteria. "Sepsis-induced coagulopathy" (SIC) is a new category. Therapy of SIC focuses on treating the underlying infection and the ensuing coagulopathy. Most therapeutic approaches to SIC have focused on anticoagulant therapy. This review will discuss SIC and DIC and how they are relevant to prolonged casualty care (PCC).

Efficacy of Transcendental Meditation to Reduce Stress Among Health Care Workers: A Randomized Clinical Trial

IMPORTANCE

Health care workers (HCWs) have been experiencing substantial stress and burnout, and evidence-based mitigation strategies are needed. Transcendental Meditation (TM) is a mantra meditation practice with potential efficacy in reducing stress.

OBJECTIVE

To assess the efficacy of TM practice in reducing stress among HCWs over a 3-month period.

DESIGN, SETTING, AND PARTICIPANTS

This single-center open-label randomized clinical trial was conducted among HCWs at an academic medical center from November 19, 2020, to August 31, 2021. Inclusion criteria comprised a score of 6 points or greater on the Subjective Units of Distress Scale and an increase of 5% or greater in baseline heart rate or an increase of 33% or greater in galvanic skin response after exposure to a stressful script. Exclusion criteria included the use of antipsychotic or β blocker medications, current suicidal ideation, or previous TM training. Of 213 HCWs who participated in prescreening, 95 attended in-person visits, resulting in 80 eligible participants who were randomized to receive a TM intervention (TM group) or usual treatment (control group).

INTERVENTIONS

The TM group practiced TM for 20 minutes twice daily over a 3-month period. The control group received usual treatment, which consisted of access to wellness resources.

MAIN OUTCOMES AND MEASURES

The primary outcome was change in acute psychological distress measured by the Global Severity Index. Secondary outcomes included changes in burnout (measured by the Maslach Burnout Inventory), insomnia (measured by the Insomnia Severity Index), and anxiety (measured by the Generalized Anxiety Disorder-7 scale).

RESULTS

Among 80 participants, 66 (82.5%) were women, with a mean (SD) age of 40 (11) years. One participant (1.3%) was American Indian or Alaska Native, 5 (6.3%) were Asian, 12 (15.0%) were Black, 59 (73.8%) were White, and 3 (3.8%) were of unknown or unreported race; 4 participants (5.0%) were Hispanic, and 76 (95.0%) were non-Hispanic. A total of 41 participants were randomized to the TM group, and 39 were randomized to the control group. Participants in the TM group did not show a statistically significant decrease in psychological distress on the Global Severity Index compared with those in the control group (-5.6 points vs -3.8 points; between-group difference, -1.8 points; 95% CI, -4.2 to 0.6 points; P = .13). Compared with the control group, the TM group had significantly greater reductions in the secondary end points of emotional exhaustion (Maslach Burnout Inventory subscore: -8.0 points vs -2.6 points; between-group difference, -5.4 points; 95% CI, -9.2 to -1.6 points; P = .006), insomnia (Insomnia Severity Scale score: -4.1 points vs -1.9 points; between-group difference, -2.2 points; 95% CI, -4.4 to 0 points; P = .05), and anxiety (Generalized Anxiety Disorder-7 score: -3.1 points vs -0.9 points; between-group difference, -2.2 points; 95% CI, -3.8 to -0.5; P = .01) at 3 months. A total of 38 participants (92.7%) in the TM group adhered to home practice.

CONCLUSIONS AND RELEVANCE

In this randomized clinical trial, TM practice among HCWs over a 3-month period did not result in a statistically significant reduction in the primary outcome of acute psychological distress compared with usual treatment but significantly improved the secondary outcomes of burnout, anxiety, and insomnia. These findings suggest that TM may be a safe and effective strategy to alleviate chronic stress among HCWs.

TRIAL REGISTRATION

ClinicalTrials.gov identifier: NCT04632368.

Analysis of Discrepancies Between Pulse Oximetry and Arterial Oxygen Saturation Measurements by Race and Ethnicity and Association With Organ Dysfunction and Mortality

IMPORTANCE

Discrepancies in oxygen saturation measured by pulse oximetry (Spo2), when compared with arterial oxygen saturation (Sao2) measured by arterial blood gas (ABG), may differentially affect patients according to race and ethnicity. However, the association of these disparities with health outcomes is unknown.

OBJECTIVE

To examine racial and ethnic discrepancies between Sao2 and Spo2 measures and their associations with clinical outcomes.

DESIGN, SETTING, AND PARTICIPANTS

This multicenter, retrospective, cross-sectional study included 3 publicly available electronic health record (EHR) databases (ie, the Electronic Intensive Care Unit-Clinical Research Database and Medical Information Mart for Intensive Care III and IV) as well as Emory Healthcare (2014-2021) and Grady Memorial (2014-2020) databases, spanning 215 hospitals and 382 ICUs. From 141 600 hospital encounters with recorded ABG measurements, 87 971 participants with first ABG measurements and an Spo2 of at least 88% within 5 minutes before the ABG test were included.

EXPOSURES

Patients with hidden hypoxemia (ie, Spo2 ≥88% but Sao2 <88%).

MAIN OUTCOMES AND MEASURES

Outcomes, stratified by race and ethnicity, were Sao2 for each Spo2, hidden hypoxemia prevalence, initial demographic characteristics (age, sex), clinical outcomes (in-hospital mortality, length of stay), organ dysfunction by scores (Sequential Organ Failure Assessment [SOFA]), and laboratory values (lactate and creatinine levels) before and 24 hours after the ABG measurement.

RESULTS

The first Spo2-Sao2 pairs from 87 971 patient encounters (27 713 [42.9%] women; mean [SE] age, 62.2 [17.0] years; 1919 [2.3%] Asian patients; 26 032 [29.6%] Black patients; 2397 [2.7%] Hispanic patients, and 57 632 [65.5%] White patients) were analyzed, with 4859 (5.5%) having hidden hypoxemia. Hidden hypoxemia was observed in all subgroups with varying incidence (Black: 1785 [6.8%]; Hispanic: 160 [6.0%]; Asian: 92 [4.8%]; White: 2822 [4.9%]) and was associated with greater organ dysfunction 24 hours after the ABG measurement, as evidenced by higher mean (SE) SOFA scores (7.2 [0.1] vs 6.29 [0.02]) and higher in-hospital mortality (eg, among Black patients: 369 [21.1%] vs 3557 [15.0%]; P < .001). Furthermore, patients with hidden hypoxemia had higher mean (SE) lactate levels before (3.15 [0.09] mg/dL vs 2.66 [0.02] mg/dL) and 24 hours after (2.83 [0.14] mg/dL vs 2.27 [0.02] mg/dL) the ABG test, with less lactate clearance (-0.54 [0.12] mg/dL vs -0.79 [0.03] mg/dL).

CONCLUSIONS AND RELEVANCE

In this study, there was greater variability in oxygen saturation levels for a given Spo2 level in patients who self-identified as Black, followed by Hispanic, Asian, and White. Patients with and without hidden hypoxemia were demographically and clinically similar at baseline ABG measurement by SOFA scores, but those with hidden hypoxemia subsequently experienced higher organ dysfunction scores and higher in-hospital mortality.

One-Stop Serum Assay Identifies COVID-19 Disease Severity and Vaccination Responses

SARS-CoV-2 has caused over 100,000,000 cases and almost 2,500,000 deaths globally. Comprehensive assessment of the multifaceted antiviral Ab response is critical for diagnosis, differentiation of severity, and characterization of long-term immunity, especially as COVID-19 vaccines become available. Severe disease is associated with early, massive plasmablast responses. We developed a multiplex immunoassay from serum/plasma of acutely infected and convalescent COVID-19 patients and prepandemic and postpandemic healthy adults. We measured IgA, IgG, and/or IgM against SARS-CoV-2 nucleocapsid (N), spike domain 1 (S1), S1-receptor binding domain (RBD) and S1-N-terminal domain. For diagnosis, the combined [IgA + IgG + IgM] or IgG levels measured for N, S1, and S1-RBD yielded area under the curve values ≥0.90. Virus-specific Ig levels were higher in patients with severe/critical compared with mild/moderate infections. A strong prozone effect was observed in sera from severe/critical patients-a possible source of underestimated Ab concentrations in previous studies. Mild/moderate patients displayed a slower rise and lower peak in anti-N and anti-S1 IgG levels compared with severe/critical patients, but anti-RBD IgG and neutralization responses reached similar levels at 2-4 mo after symptom onset. Measurement of the Ab responses in sera from 18 COVID-19-vaccinated patients revealed specific responses for the S1-RBD Ag and none against the N protein. This highly sensitive, SARS-CoV-2-specific, multiplex immunoassay measures the magnitude, complexity, and kinetics of the Ab response and can distinguish serum Ab responses from natural SARS-CoV-2 infections (mild or severe) and mRNA COVID-19 vaccines.

Classification of 12-lead ECGs: the PhysioNet/Computing in Cardiology Challenge 2020

OBJECTIVE

Vast 12-lead ECGs repositories provide opportunities to develop new machine learning approaches for creating accurate and automatic diagnostic systems for cardiac abnormalities. However, most 12-lead ECG classification studies are trained, tested, or developed in single, small, or relatively homogeneous datasets. In addition, most algorithms focus on identifying small numbers of cardiac arrhythmias that do not represent the complexity and difficulty of ECG interpretation. This work addresses these issues by providing a standard, multi-institutional database and a novel scoring metric through a public competition: the PhysioNet/Computing in Cardiology Challenge 2020.

APPROACH

A total of 66 361 12-lead ECG recordings were sourced from six hospital systems from four countries across three continents; 43 101 recordings were posted publicly with a focus on 27 diagnoses. For the first time in a public competition, we required teams to publish open-source code for both training and testing their algorithms, ensuring full scientific reproducibility.

MAIN RESULTS

A total of 217 teams submitted 1395 algorithms during the Challenge, representing a diversity of approaches for identifying cardiac abnormalities from both academia and industry. As with previous Challenges, high-performing algorithms exhibited significant drops ([Formula: see text]10%) in performance on the hidden test data.

SIGNIFICANCE

Data from diverse institutions allowed us to assess algorithmic generalizability. A novel evaluation metric considered different misclassification errors for different cardiac abnormalities, capturing the outcomes and risks of different diagnoses. Requiring both trained models and code for training models improved the generalizability of submissions, setting a new bar in reproducibility for public data science competitions.

Classification of 12-lead ECGs: the PhysioNet/Computing in Cardiology Challenge 2020

OBJECTIVE

Vast 12-lead ECGs repositories provide opportunities to develop new machine learning approaches for creating accurate and automatic diagnostic systems for cardiac abnormalities. However, most 12-lead ECG classification studies are trained, tested, or developed in single, small, or relatively homogeneous datasets. In addition, most algorithms focus on identifying small numbers of cardiac arrhythmias that do not represent the complexity and difficulty of ECG interpretation. This work addresses these issues by providing a standard, multi-institutional database and a novel scoring metric through a public competition: the PhysioNet/Computing in Cardiology Challenge 2020.

APPROACH

A total of 66 361 12-lead ECG recordings were sourced from six hospital systems from four countries across three continents; 43 101 recordings were posted publicly with a focus on 27 diagnoses. For the first time in a public competition, we required teams to publish open-source code for both training and testing their algorithms, ensuring full scientific reproducibility.

MAIN RESULTS

A total of 217 teams submitted 1395 algorithms during the Challenge, representing a diversity of approaches for identifying cardiac abnormalities from both academia and industry. As with previous Challenges, high-performing algorithms exhibited significant drops ([Formula: see text]10%) in performance on the hidden test data.

SIGNIFICANCE

Data from diverse institutions allowed us to assess algorithmic generalizability. A novel evaluation metric considered different misclassification errors for different cardiac abnormalities, capturing the outcomes and risks of different diagnoses. Requiring both trained models and code for training models improved the generalizability of submissions, setting a new bar in reproducibility for public data science competitions.

Elevated SARS-CoV-2 Antibodies Distinguish Severe Disease in Early COVID-19 Infection

Background

SARS-CoV-2 has caused over 36,000,000 cases and 1,000,000 deaths globally. Comprehensive assessment of the multifaceted anti-viral antibody response is critical for diagnosis, differentiation of severe disease, and characterization of long-term immunity. Initial observations suggest that severe disease is associated with higher antibody levels and greater B cell/plasmablast responses. A multi-antigen immunoassay to define the complex serological landscape and clinical associations is essential.

Methods

We developed a multiplex immunoassay and evaluated serum/plasma from adults with RT-PCR-confirmed SARS-CoV-2 infections during acute illness (N=52) and convalescence (N=69); and pre-pandemic (N=106) and post-pandemic (N=137) healthy adults. We measured IgA, IgG, and/or IgM against SARS-CoV-2 Nucleocapsid (N), Spike domain 1 (S1), receptor binding domain (S1-RBD) and S1-N-terminal domain (S1-NTD).

Results

To diagnose infection, the combined [IgA+IgG+IgM] or IgG for N, S1, and S1-RBD yielded AUC values -0.90 by ROC curves. From days 6-30 post-symptom onset, the levels of antigen-specific IgG, IgA or [IgA+IgG+IgM] were higher in patients with severe/critical compared to mild/moderate infections. Consistent with excessive concentrations of antibodies, a strong prozone effect was observed in sera from severe/critical patients. Notably, mild/moderate patients displayed a slower rise and lower peak in anti-N and anti-S1 IgG levels compared to severe/critical patients, but anti-RBD IgG and neutralization responses reached similar levels at 2-4 months.

Conclusion

This SARS-CoV-2 multiplex immunoassay measures the magnitude, complexity and kinetics of the antibody response against multiple viral antigens. The IgG and combined-isotype SARS-CoV-2 multiplex assay is highly diagnostic of acute and convalescent disease and may prognosticate severity early in illness.

One Sentence Summary

In contrast to patients with moderate infections, those with severe COVID-19 develop prominent, early antibody responses to S1 and N proteins.

Machine Learning Methods to Predict Acute Respiratory Failure and Acute Respiratory Distress Syndrome

Acute respiratory failure (ARF) is a common problem in medicine that utilizes significant healthcare resources and is associated with high morbidity and mortality. Classification of acute respiratory failure is complicated, and it is often determined by the level of mechanical support that is required, or the discrepancy between oxygen supply and uptake. These phenotypes make acute respiratory failure a continuum of syndromes, rather than one homogenous disease process. Early recognition of the risk factors for new or worsening acute respiratory failure may prevent that process from occurring. Predictive analytical methods using machine learning leverage clinical data to provide an early warning for impending acute respiratory failure or its sequelae. The aims of this review are to summarize the current literature on ARF prediction, to describe accepted procedures and common machine learning tools for predictive tasks through the lens of ARF prediction, and to demonstrate the challenges and potential solutions for ARF prediction that can improve patient outcomes.

Randomized Controlled Trial of a Novel Communication Device Assessed During Noninvasive Ventilation Therapy

BACKGROUND

Noninvasive ventilation (NIV), a form of positive airway pressure (PAP) therapy, is the standard of care for various forms of acute respiratory failure (ARF). Communication impairment is a side effect of NIV, impedes patient care, contributes to distress and intolerance, and potentially increases intubation rates. This study aimed to evaluate communication impairment during CPAP therapy and demonstrate communication device improvement with a standardized protocol.

RESEARCH QUESTION

How does an oronasal mask affect communication intelligibility? How does use of an NIV communication device change this communication intelligibility?

STUDY DESIGN AND METHODS

A single-center randomized controlled trial (36 outpatients with OSA on CPAP therapy) assessed exposure to CPAP 10 cm H₂O and PAP communication devices (SPEAX, Ataia Medical). Communication impairment was evaluated by reading selected words and sentences for partners to record and were tabulated as %words correct. Each outpatient-partner pair performed three assessments: (1) baseline (conversing normally), (2) mask baseline (conversing with PAP), and (3) randomized to functioning device (conversing with PAP and device) or sham device. After each stage, both outpatients and partners completed Likert surveys regarding perceived intelligibility and comfort.

RESULTS

While conversing with PAP, word and sentence intelligibility decreased relatively by 52% (87% vs 41%) and relatively by 57% (94% vs 40%), respectively, compared with normal conversation. Word and sentence intelligibility in the intervention arm increased relatively by 75% (35% vs 61%; P < .001) and by 126% (33% vs 76%; P < .001) higher than the control arm, respectively. The device improved outpatient-perceived PAP comfort relatively by 233% (15% vs 50%, P = .042) and partner-perceived comfort by relatively 245% (20% vs 69%, P = .0074).

INTERPRETATION

Use of this PAP communication device significantly improves both intelligibility and comfort. This is one of the first studies quantifying communication impairment during PAP delivery.

TRIAL REGISTRY

ClinicalTrials.gov; No.: NCT03795753; URL: www.clinicaltrials.gov.

Computerized detection of adverse drug reactions in the medical intensive care unit

OBJECTIVE

Clinical event monitors are a type of active medication monitoring system that can use signals to alert clinicians to possible adverse drug reactions. The primary goal was to evaluate the positive predictive values of select signals used to automate the detection of ADRs in the medical intensive care unit.

METHOD

This is a prospective, case series of adult patients in the medical intensive care unit during a six-week period who had one of five signals presents: an elevated blood urea nitrogen, vancomycin, or quinidine concentration, or a low sodium or glucose concentration. Alerts were assessed using 3 objective published adverse drug reaction determination instruments. An event was considered an adverse drug reaction when 2 out of 3 instruments had agreement of possible, probable or definite. Positive predictive values were calculated as the proportion of alerts that occurred, divided by the number of times that alerts occurred and adverse drug reactions were confirmed.

RESULTS

145 patients were eligible for evaluation. For the 48 patients (50% male) having an alert, the mean±SD age was 62±19 years. A total of 253 alerts were generated. Positive predictive values were 1.0, 0.55, 0.38 and 0.33 for vancomycin, glucose, sodium, and blood urea nitrogen, respectively. A quinidine alert was not generated during the evaluation.

CONCLUSIONS

Computerized clinical event monitoring systems should be considered when developing methods to detect adverse drug reactions as part of intensive care unit patient safety surveillance systems, since they can automate the detection of these events using signals that have good performance characteristics by processing commonly available laboratory and medication information.

A Bayesian method for identifying genetic interactions

An important challenge in the analysis of single nucleotide polymorphism (SNP) data is the identification of SNPs that interact in a nonlinear fashion in their association with disease. Such epistatic interactions among genetic variants at multiple loci likely underlie the inheritance of common diseases. We have developed a novel method called the Bayesian combinatorial method (BCM) for detecting combination of genetic variants that are predictive of disease. When compared with the multifactor dimensionality reduction (MDR), a widely used combinatorial method, BCM has significantly greater power to detect interactions and is computationally more efficient.

Bayesian combinatorial partitioning for detecting interactions among genetic variants

Detecting epistatic (nolinear) interactions among single nucleotide polymorphisms (SNPs) at multiple loci is important in the analysis of genomic data in association studies. We developed a Bayesian combinatorial partitioning (BCP) for detecting such interactions among SNPs that are predictive of disease. When compared with multifactor dimensionality reduction (MDR), a widely used combinatorial partitioning method for detecting interactions, BCP has significantly greater power and is computationally more efficient.