Assessment of skin pigmentation-related bias in pulse oximetry readings among adults.

PURPOSE: Recent reports that pulse oximeters may overestimate oxygen saturation in individuals with darker skin pigmentation have prompted concerns from regulatory authorities regarding racial bias. We investigated the performance of TruSignal SpO2 sensors (GE Healthcare, Helsinki, Finland) in adults with varying skin pigmentation. METHODS: A retrospective study was conducted using a set of pooled assessments of SpO2/SaO2 measurements from nine studies to assess bias, accuracy (Arms), and precision of TruSignal sensors in healthy adults under induced hypoxia. Subgroup analyses were performed based on oxygen saturation levels (band 1, ≥ 70 and ≤ 80%; band 2, > 80 and ≤ 90%; band 3, > 90 and ≤ 100%). RESULTS: Of the 10,800 data points from 131 individuals, 8,202 (75.9%) and 2,598 (24.1%) were assigned to the light and dark pigment groups, respectively. Bias was 0.14% overall and less than 1% across oxygenation bands. The difference in bias between dark and light pigment groups was statistically significant at the low oxygenation band with SpO2 ≥ 70 and ≤ 80% (+ 0.58% and + 0.30% respectively; p = 0.0035). Throughout the saturation range, Arms was 1.64% in the light and 1.71% in the dark pigment group, within device specifications and regulatory requirements. Oxygenation was the dominating factor in stepwise ANOVA modeling. The mixed model also showed that bias was strongly affected by the oxygenation range. CONCLUSION: TruSignal sensors demonstrated higher bias at lower oxygen saturation, with less than 0.5% difference between pigment groups. These findings raise new questions, such as ways to improve pulse oximetry measurements during challenging clinical conditions, including low perfusion.

Automated Electronic Frailty Index-Identified Frailty Status and Associated Postsurgical Adverse Events.

Importance: Electronic frailty index (eFI) is an automated electronic health record (EHR)-based tool that uses a combination of clinical encounters, diagnosis codes, laboratory workups, medications, and Medicare annual wellness visit data as markers of frailty status. The association of eFI with postanesthesia adverse outcomes has not been evaluated. Objective: To examine the association of frailty, calculated as eFI at the time of the surgical procedure and categorized as fit, prefrail, or frail, with adverse events after elective noncardiac surgery. Design, Setting, and Participants: This cohort study was conducted at a tertiary care academic medical center in Winston-Salem, North Carolina. The cohort included patients 55 years or older who underwent noncardiac surgery of at least 1 hour in duration between October 1, 2017, and June 30, 2021. Exposure: Frailty calculated by the eFI tool. Preoperative eFI scores were calculated based on available data 1 day prior to the procedure and categorized as fit (eFI score: ≤0.10), prefrail (eFI score: >0.10 to ≤0.21), or frail (eFI score: >0.21). Main Outcomes and Measures: The primary outcome was a composite of the following 8 adverse component events: 90-item Patient Safety Indicators (PSI 90) score, hospital-acquired conditions, in-hospital mortality, 30-day mortality, 30-day readmission, 30-day emergency department visit after surgery, transfer to a skilled nursing facility after surgery, or unexpected intensive care unit admission after surgery. Secondary outcomes were each of the component events of the composite. Results: Of the 33 449 patients (median [IQR] age, 67 [61-74] years; 17 618 females [52.7%]) included, 11 563 (34.6%) were classified as fit, 15 928 (47.6%) as prefrail, and 5958 (17.8%) as frail. Using logistic regression models that were adjusted for age, sex, race and ethnicity, and comorbidity burden, patients with prefrail (odds ratio [OR], 1.24; 95% CI, 1.18-1.30; P < .001) and frail (OR, 1.71; 95% CI, 1.58-1.82; P < .001) statuses were more likely to experience postoperative adverse events compared with patients with a fit status. Subsequent adjustment for all other potential confounders or covariates did not alter this association. For every increase in eFI of 0.03 units, the odds of a composite of postoperative adverse events increased by 1.06 (95% CI, 1.03-1.13; P < .001). Conclusions and Relevance: This cohort study found that frailty, as measured by an automatically calculated index integrated within the EHR, was associated with increased risk of adverse events after noncardiac surgery. Deployment of eFI tools may support screening and possible risk modification, especially in patients who undergo high-risk surgery.

Patient Characteristics and Clinical and Economic Outcomes Associated with Unplanned Medical and Surgical Intensive Care Unit Admissions: A Retrospective Analysis.

Purpose: To characterize medical and surgical patient characteristics, as well as clinical and economic outcomes, associated with unplanned intensive care unit (ICU) admissions. Patients and Methods: This was a retrospective matched cohort analysis that utilized the PINC AITM Healthcare Database, which collects deidentified data from 25% of United States (US) hospital admissions. Discharge records were assessed for medical and surgical admissions in 2021. An unplanned ICU admission was defined as direct transfer from a medical, surgical, or telemetry unit to the ICU. Patients with and without an unplanned ICU admission were 1:1 propensity score matched. Differences between patients with and without unplanned ICU admissions were assessed using two-sample t-tests for continuous measures and Chi-square tests for categorical measures. Results: A total of 3,807,124 qualifying admissions were identified. Medical admissions with unplanned ICU transfers were more likely to be urgent/emergent (odds ratio [OR] 2.9, 95% confidence interval [CI 2.7-3.0], p<0.0001), with patient characteristics including male sex (1.4, [1.4-1.4], p<0.0001), obesity (1.7, [1.6-1.7], p<0.0001), and increased Charlson Comorbidity Index (CCI=1: 1.8, [1.8-1.9], p<0.0001; CCI≥5: 3.2, [3.1-3.3], p<0.0001). Surgical admissions with unplanned ICU transfers were more likely to be urgent/emergent (3.1, [2.9-3.2], p<0.0001) and with patients of higher CCI (2.5, [2.3-2.6], p<0.0001 to a CCI of≥5 (7.9, [7.4-8.4], p<0.0001). Between matched medical patients, mean differences in length of stay, cost, and mortality were 4.1 days (p<0.0001), $13,424 (p<0.0001), and 21% (p<0.0001), respectively. Between matched surgical patients, mean differences in these outcomes were 6.4 days (p<0.0001), $21,448 (p<0.0001), and 14% (p<0.0001), respectively. Conclusion: Emergency care in patients with a higher co-morbid burden is more likely to lead to unplanned ICU admission, putting patients at a significantly increased chance of mortality, longer length of stay, and increased costs. Improving care and monitoring of patients outside the ICU may help detect early changes in pathophysiology and enable early intervention.

Bias in artificial intelligence algorithms and recommendations for mitigation.

The adoption of artificial intelligence (AI) algorithms is rapidly increasing in healthcare. Such algorithms may be shaped by various factors such as social determinants of health that can influence health outcomes. While AI algorithms have been proposed as a tool to expand the reach of quality healthcare to underserved communities and improve health equity, recent literature has raised concerns about the propagation of biases and healthcare disparities through implementation of these algorithms. Thus, it is critical to understand the sources of bias inherent in AI-based algorithms. This review aims to highlight the potential sources of bias within each step of developing AI algorithms in healthcare, starting from framing the problem, data collection, preprocessing, development, and validation, as well as their full implementation. For each of these steps, we also discuss strategies to mitigate the bias and disparities. A checklist was developed with recommendations for reducing bias during the development and implementation stages. It is important for developers and users of AI-based algorithms to keep these important considerations in mind to advance health equity for all populations.

Modeling the Cost Savings of Continuous Pulse Oximetry and Capnography Monitoring of United States General Care Floor Patients Receiving Opioids Based on the PRODIGY Trial.

INTRODUCTION: Despite the high incidence of respiratory depression on the general care floor and evidence that continuous monitoring improves patient outcomes, the cost-benefit of continuous pulse oximetry and capnography monitoring of general care floor patients remains unknown. This study modeled the cost and length of stay savings, investment break-even point, and likelihood of cost savings for continuous pulse oximetry and capnography monitoring of general care floor patients at risk for respiratory depression. METHODS: A decision tree model was created to compare intermittent pulse oximetry versus continuous pulse oximetry and capnography monitoring. The model utilized costs and outcomes from the PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY (PRODIGY) trial, and was applied to a modeled cohort of 2447 patients receiving opioids per median-sized United States general care floor annually. RESULTS: Continuous pulse oximetry and capnography monitoring of high-risk patients is projected to reduce annual hospital cost by $535,531 and cumulative patient length of stay by 103 days. A 1.5% reduction in respiratory depression would achieve a break-even investment point and justify the investment cost. The probability of cost saving is ≥ 80% if respiratory depression is decreased by ≥ 17%. Expansion of continuous monitoring to high- and intermediate-risk patients, or to all patients, is projected to reach a break-even point when respiratory depression is reduced by 2.5% and 3.5%, respectively, with a ≥ 80% probability of cost savings when respiratory depression decreases by ≥ 27% and ≥ 31%, respectively. CONCLUSION: Compared to intermittent pulse oximetry, continuous pulse oximetry and capnography monitoring of general care floor patients receiving opioids has a high chance of being cost-effective. TRIAL REGISTRATION: www.clinicaltrials.gov , Registration ID: NCT02811302.

Opioid-induced respiratory depression increases hospital costs and length of stay in patients recovering on the general care floor.

BACKGROUND: Opioid-induced respiratory depression is common on the general care floor. However, the clinical and economic burden of respiratory depression is not well-described. The PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY (PRODIGY) trial created a prediction tool to identify patients at risk of respiratory depression. The purpose of this retrospective sub-analysis was to examine healthcare utilization and hospital cost associated with respiratory depression. METHODS: One thousand three hundred thirty-five patients (N = 769 United States patients) enrolled in the PRODIGY trial received parenteral opioids and underwent continuous capnography and pulse oximetry monitoring. Cost data was retrospectively collected for 420 United States patients. Differences in healthcare utilization and costs between patients with and without ≥1 respiratory depression episode were determined. The impact of respiratory depression on hospital cost per patient was evaluated using a propensity weighted generalized linear model. RESULTS: Patients with ≥1 respiratory depression episode had a longer length of stay (6.4 ± 7.8 days vs 5.0 ± 4.3 days, p = 0.009) and higher hospital cost ($21,892 ± $11,540 vs $18,206 ± $10,864, p = 0.002) compared to patients without respiratory depression. Patients at high risk for respiratory depression, determined using the PRODIGY risk prediction tool, who had ≥1 respiratory depression episode had higher hospital costs compared to high risk patients without respiratory depression ($21,948 ± $9128 vs $18,474 ± $9767, p = 0.0495). Propensity weighted analysis identified 17% higher costs for patients with ≥1 respiratory depression episode (p = 0.007). Length of stay significantly increased total cost, with cost increasing exponentially for patients with ≥1 respiratory depression episode as length of stay increased. CONCLUSIONS: Respiratory depression on the general care floor is associated with a significantly longer length of stay and increased hospital costs. Early identification of patients at risk for respiratory depression, along with early proactive intervention, may reduce the incidence of respiratory depression and its associated clinical and economic burden. TRIAL REGISTRATION: ClinicalTrials.gov , NCT02811302 .

Cost-effectiveness of second-line vasopressors for the treatment of septic shock.

PURPOSE: To determine the cost-effectiveness of escalating doses of norepinephrine or norepinephrine plus the adjunctive use of vasopressin or angiotensin II as a second-line vasopressor for septic shock. MATERIALS AND METHODS: Decision tree analysis was performed to compare costs and outcomes associated with norepinephrine monotherapy or the two adjunctive second-line vasopressors. Short- and long-term outcomes modeled included ICU survival and lifetime quality-adjusted-life-years (QALY) gained. Costs were modeled from a payer's perspective, with a willingness-to-pay threshold set at $100,000/unit gained. One-way (tornado diagrams) and probabilistic sensitivity analyses were performed. RESULTS: Adjunctive vasopressin was the most cost-effective therapy, and dominated both norepinephrine monotherapy and adjunctive angiotensin II by producing higher ICU survival at less cost. For the lifetime horizon, while norepinephrine monotherapy was least expensive, adjunctive vasopressin was the most cost-effective with an incremental cost-effectiveness ratio of $19,762 / QALY gained. Although adjunctive angiotensin II produced more QALYs compared to norepinephrine monotherapy, it was dominated in the long-term evaluation by second-line vasopressin. Sensitivity analyses demonstrated model robustness and medication costs were not significant drivers of model results. CONCLUSIONS: Vasopressin is the most cost-effective second-line vasopressor in both the short- and long-term evaluations. Vasopressor price is a minor contributor to overall cost.