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.

Measuring enhanced recovery in obstetrics: a narrative review.

Enhanced recovery after cesarean delivery is a protocolized approach to perioperative care, with the aim to optimize maternal recovery after surgery. It is associated with improved maternal and neonatal outcomes, including decreased length of hospital stay, opioid consumption, pain scores, complications, increased maternal satisfaction, and increased breastfeeding success. However, the pace and enthusiasm of adoption of enhanced recovery after cesarean delivery internationally has not yet been matched with high-quality evidence demonstrating its benefit, and current studies provide low- to very low-quality evidence in support of enhanced recovery after cesarean delivery. This article provides a summary of current measures of enhanced recovery after cesarean delivery success, and optimal measures of inpatient and outpatient postpartum recovery. We summarize outcomes from 22 published enhanced recovery after cesarean delivery implementation studies and 2 meta-analyses. A variety of disparate metrics have been used to measure enhanced recovery after cesarean delivery success, including process measures (length of hospital stay, bundle compliance, preoperative fasting time, time to first mobilization, time to urinary catheter removal), maternal outcomes (patient-reported outcome measures, complications, opioid consumption, satisfaction), neonatal outcomes (breastfeeding success, Apgar scores, maternal-neonatal bonding), cost savings, and complication rates (maternal readmission rate, urinary recatheterization rate, neonatal readmission rate). A core outcome set for use in enhanced recovery after cesarean delivery studies has been developed through Delphi consensus, involving stakeholders including obstetricians, anesthesiologists, patients, and a midwife. Fifteen measures covering key aspects of enhanced recovery after cesarean delivery adoption are recommended for use in future enhanced recovery after cesarean delivery implementation studies. The use of these outcome measures could improve the quality of evidence surrounding enhanced recovery after cesarean delivery. Using evidence-based evaluation guidelines developed by the COSMIN (COnsensus-based Standards for the selection of health Measurement INstruments) group, the Obstetric Quality of Recovery score (ObsQoR) was identified as the best patient-reported outcome measure for inpatient postpartum recovery. Advances in our understanding of postpartum recovery as a multidimensional and dynamic construct have opened new avenues for the identification of optimum patient-reported outcome measures in this context. The use of standardized measures such as these will facilitate pooling of data in future studies and improve overall levels of evidence surrounding enhanced recovery after cesarean delivery. Larger studies with optimal study designs, using recommended outcomes including patient-reported outcome measures, will reduce variation and improve data quality to help guide future recommendations.

Incidence, prevalence, and timing of postpartum complications and mortality in Canada and the United States: a systematic review and meta-analysis protocol.

OBJECTIVE: The goal of this systematic review is to assess the incidence, prevalence, and timing of common postpartum (up to 1 year after delivery) medical, surgical/procedural, and psychosocial complications and mortality. INTRODUCTION: Childbirth is the most common cause for hospitalization, and cesarean delivery is the most commonly performed inpatient surgery. After delivery, mothers are at risk of short- and long-term complications that can impact their well-being. The results of this review will inform evidence-based recommendations for patient education, monitoring, and follow-up. INCLUSION CRITERIA: We will include studies performed in Canada and/or the United States that report the incidence or prevalence of medical, procedural/surgical, and psychosocial complications within 1 year postpartum. Observational studies (analytical cross-sectional studies, retrospective and prospective cohorts), randomized or non-randomized controlled trials with a control or standard of care group, systematic reviews, and meta-analyses will be included. Studies with fewer than 100 patients, participants younger than 18 years, no reporting of duration, or focus on patients with a specific condition rather than a general postpartum population will be excluded. METHODS: The search strategy was codeveloped with a medical librarian and included full-text English-language articles published within the past 10 years (2011-2021) in PubMed, CINHAL, Web of Science, and Cochrane Database of Systematic Reviews. Screening, critical appraisal, and data extraction will be performed by two independent reviewers using Covidence, standardized JBI tools, and a standardized form, respectively. For each complication, the incidence or prevalence, timing of the frequency measurement, and duration of follow-up from individual studies will be determined. Meta-analysis will be performed if feasible. SYSTEMATIC REVIEW REGISTRATION NUMBER: PROSPERO CRD42022303047.

Postoperative mortality risk prediction that incorporates intraoperative vital signs: development and internal validation in a historical cohort.

PURPOSE: Accurate risk reassessment after surgery is crucial for postoperative planning for monitoring and disposition. Existing postoperative mortality risk prediction models using preoperative features do not incorporate intraoperative hemodynamic derangements that may alter risk stratification. Intraoperative vital signs may provide an objective and readily available prognostic resource. Our primary objective was to derive and internally validate a logistic regression (LR) model by adding intraoperative features to established preoperative predictors to predict 30-day postoperative mortality. METHODS: Following Research Ethics Board approval, we analyzed a historical cohort that included patients aged ≥ 45 undergoing noncardiac surgery with an overnight stay at two tertiary hospitals (2013 to 2017). Features included intraoperative vital signs (blood pressure, heart rate, end-tidal carbon dioxide partial pressure, oxygen saturation, and temperature) by threshold and duration of exposure, as well as patient, surgical, and anesthetic factors. The cohort was divided temporally 75:25 into derivation and validation sets. We constructed a multivariable LR model with 30-day all-cause mortality as the outcome and evaluated performance metrics. RESULTS: There were 30,619 patients in the cohort (mean [standard deviation] age, 66 [11] yr; 50.2% female; 2.0% mortality). In the validation set, the primary LR model showed a c-statistic of 0.893 (99% confidence interval [CI], 0.853 to 0.927), a Nagelkerke R-squared of 0.269, a scaled Brier score of 0.082, and an area under precision-recall curve of 0.158 (baseline 0.017 for an uninformative model). The addition of intraoperative vital signs to preoperative factors minimally improved discrimination and calibration. CONCLUSION: We derived and internally validated a model that incorporated vital signs to improve risk stratification after surgery. Preoperative factors were strongly predictive of mortality risk, and intraoperative predictors only minimally improved discrimination. External and prospective validations are needed. STUDY REGISTRATION: www. CLINICALTRIALS: gov (NCT04014010); registered on 10 July 2019.

RéSUMé: OBJECTIF: Une réévaluation précise des risques après la chirurgie est cruciale pour la planification postopératoire du monitorage et du congé. Les modèles existants de prédiction du risque de mortalité postopératoire utilisant des caractéristiques préopératoires n'intègrent pas les perturbations hémodynamiques peropératoires, lesquelles pourraient modifier la stratification du risque. Les signes vitaux peropératoires peuvent fournir une ressource pronostique objective et facilement disponible. Notre objectif principal était de dériver et de valider en interne un modèle de régression logistique (RL) en ajoutant des caractéristiques peropératoires aux prédicteurs préopératoires établis pour prédire la mortalité postopératoire à 30 jours. MéTHODE: À la suite de l'approbation du Comité d'éthique de la recherche, nous avons analysé une cohorte historique qui comprenait des patients âgés de ≥ 45 ans bénéficiant d'une chirurgie non cardiaque avec un séjour d'une nuit dans deux hôpitaux tertiaires (2013 à 2017). Les caractéristiques comprenaient les signes vitaux peropératoires (tension artérielle, fréquence cardiaque, pression télé-expiratoire en CO2, saturation en oxygène et température) par seuil et durée d'exposition, ainsi que des facteurs propres au patient, chirurgicaux et anesthésiques. La cohorte a été divisée temporellement 75:25 en ensembles de dérivation et de validation. Nous avons élaboré un modèle de RL multivariée avec la mortalité toutes causes confondues à 30 jours comme critère, et évalué les mesures de performance. RéSULTATS: Il y avait 30 619 patients dans la cohorte (âge moyen [écart type], 66 [11] ans; 50,2 % de femmes; 2,0 % de mortalité). Dans l'ensemble de validation, le modèle de RL primaire a montré une statistique c de 0,893 (intervalle de confiance [IC] à 99 %, 0,853 à 0,927), un R carré de Nagelkerke de 0,269, un score de Brier mis à l'échelle de 0,082 et une aire sous la courbe de rappel et précision de 0,158 (ligne de base 0,017 pour un modèle non informatif). L'ajout de signes vitaux peropératoires aux facteurs préopératoires a amélioré de façon minimale la discrimination et l'étalonnage. CONCLUSION: Nous avons dérivé et validé en interne un modèle qui incorporait des signes vitaux pour améliorer la stratification des risques après la chirurgie. Les facteurs préopératoires étaient fortement prédictifs du risque de mortalité, et les prédicteurs peropératoires n'ont que que très peu amélioré la discrimination. Une validation externe et prospective est nécessaire. ENREGISTREMENT DE L'éTUDE: www.ClinicalTrials.gov (NCT04014010); enregistrée le 10 juillet 2019.

Association of anaesthesia technique with 30-day primary graft patency after open lower limb revascularization: retrospective cohort study.

BACKGROUND: The relationship between anaesthetic technique and graft patency after open lower limb revascularization is unclear. The aim of this study was to evaluate the association between 30-day graft patency after elective infrainguinal bypass and anaesthetic technique (regional anaesthesia (RA, i.e. neuraxial and/or peripheral nerve blockade) compared with general anaesthesia (GA)). METHODS: Patients who underwent elective infrainguinal bypass in the 2014-2019 National Surgical Quality Improvement Program Vascular Procedure Targeted Lower Extremity Open data set were included. Excluded patients were those under 18 years old, those who did not receive RA or GA, and/or had an international normalized ratio of 1.5 of greater, a partial thromboplastin time more than 35 s, or a platelet count less than 80 × 109/L. The primary outcome was primary graft patency without reintervention. The relationship between anaesthetic technique and patency was analysed with multivariable logistic regression. RESULTS: Included were 8893 patients with a mean(s.d.) age of 68(11) years and 31.5 per cent female. Within the cohort, 7.7 per cent (n = 688) patients received RA only, 90.4 per cent (n = 8039) GA only, and 1.9 per cent (n = 166) both GA and RA. In the RA-only group, 91.7 per cent (631 of 688) received neuraxial anaesthesia. The primary patency rate was 93.2 per cent (573 of 615) for RA only, and 91.5 per cent (6390 of 6983) for GA only (standardized mean difference, 0.063). RA was not associated with a higher rate of patency compared with GA (adjusted OR, 1.16; 95 per cent c.i., 0.83 to 1.63; P = 0.378). CONCLUSION: There was no association between anaesthetic technique and 30-day graft patency after elective infrainguinal bypass surgery. Further prospective studies would be useful to study the impact of anaesthesia technique on important patient-centred outcomes such as long-term patency and non-home discharge.

Outcomes of Coronary Artery Bypass Grafting in Patients With Poor Myocardial Viability: A Systematic Review and Meta-Analysis of the Last Decade.

OBJECTIVES: Our objective is to assess whether the presence of myocardial viability is a predictor of mortality among patients undergoing coronary artery bypasss grafting (CABG) through a systematic review meta-analysis. METHODS: Comprehensive review of EMBASE and PubMed in accordance with PRISMA guidelines, including studies of patients undergoing CABG with assessment of myocardial viability and recorded long-term mortality, age and sex. Studies were restricted to the last decade, and data were stratified by imaging modality (magnetic resonance imaging [MRI] or nuclear medicine). Random-effects model for assessing pooled effect, heterogeneity assessment using Chi-square and I2 statistics, publication bias assessed by funnel plots and Egger's test. RESULTS: Meta-analysis of contemporary data (January 2010 to October 2020) yielded 3,621 manuscripts of which 92 were relevant, and 6 appropriate for inclusion with 993 patients. Pooled analysis showed that patients with non-viable myocardium undergoing CABG are at 1.34 times the risk of mortality compared to those with viable myocardium (95% CI 1.01-1.79, p=0.05). Subgroup analysis of the MRI or nuclear medicine modalities was not statistically significant and there was no confounding by age or sex in meta-regression. There was significant heterogeneity in imaging modality and diagnostic criteria, but heterogeneity between study findings was low with an I2 statistic of 29%. The risk of publication bias was moderate on the Newcastle-Ottawa Scale), but not statistically significant (Egger's Test coefficient=1.3, 95%CI -0.35-2.61, p=0.10). CONCLUSIONS: There is a multitude of methods for assessing cardiac viability for coronary revascularisation surgery, making meta-analyses fraught with limitations. Our meta-analysis demonstrates that the finding of non-viable myocardium can not be used draw conclusions for risk assessment in coronary surgery.

Coagulopathy Management of an Acute Type A Aortic Dissection in a Patient Taking Apixaban.

This paper reports the successful management of a patient with acute type A Penn B thoracic aortic dissection who was on apixaban therapy for atrial fibrillation. Emergency surgery was performed due to the patient's clinical deterioration, with innominate artery compromise and severe aortic valve regurgitation. The anesthesia team used point-of-care rotational thromboelastometry-guided coagulation replacement therapy consisting of prothrombin concentrate, fibrinogen, and platelets. The surgical team used a complementary approach with topical hemostatic agents and a pericardial patch. No additional blood products were required. The patient recovered fully and was discharged home.

Perioperative mobile application for mothers undergoing Cesarean delivery: a prospective cohort study on patient engagement.

PURPOSE: Giving birth is the most common reason for hospital admission, with Cesarean delivery being the most frequently performed inpatient surgery. Through a needs assessment and iterative design process involving patients and obstetric anesthesiologists, we previously developed a mobile application, C-Care, for patients undergoing Cesarean delivery. The focus of C-Care is perioperative education and self-monitoring of potential anesthetic complications. This study aimed to obtain feedback on patient engagement with C-Care. METHOD: We conducted a prospective cohort study of patients ≥ 18 yr (n = 36) undergoing elective Cesarean delivery. Anonymous usage data were recorded for 30 days. On postoperative days 1-5, participants received daily self-monitoring questionnaires within C-Care. Fourteen days after surgery, participants received an online survey regarding satisfaction and use of C-Care. RESULTS: Thirty-five out of 36 participants visited the application after orientation, with a median [interquartile range (IQR)] age of 32 [31-36] yr. Each participant visited the application a median of 15 [9-31] times over 30 days and completed a median of 3 [2-4] out of five self-monitoring questionnaires. Each participant viewed a median of 4 [2-7] out of eight education topics, with the most viewed patient education topics being "Controlling Pain" and "The First Few Days". Visits to the application were highest in the first week postpartum. Of the 18 respondents who completed the day 14 survey, 83% (n = 15) participants would recommend C-Care to other women, and the median participant satisfaction score was 7.5 out of 10 (range, 2-10). CONCLUSION: Most participants used this mobile application for patient education and self-monitoring after elective Cesarean delivery. Insights into patient engagement with C-Care after Cesarean delivery could help design more effective perioperative mobile telehealth programs. TRIAL REGISTRATION: www.ClinicalTrials.gov (NCT03746678); registered 5 November 2018.

RéSUMé: OBJECTIF: L'accouchement est la raison la plus courante d'admission à l'hôpital, et l'accouchement par césarienne constitue la chirurgie hospitalière la plus fréquemment pratiquée. Grâce à une évaluation des besoins et un processus de conception itérative impliquant des patientes et des anesthésiologistes obstétricaux, nous avions mis au point une application mobile baptisée C-Care destinée aux patientes subissant un accouchement par césarienne. L'application C-Care a pour but d'être un outil d'éducation périopératoire et d'autosurveillance des complications anesthésiques potentielles. Cette étude visait à obtenir des commentaires de rétroaction sur l'intérêt des patientes pour l'application C-Care. MéTHODE: Nous avons mené une étude de cohorte prospective auprès de patientes ≥ 18 ans (n = 36) subissant un accouchement par césarienne programmée. Des données d'utilisation anonymes ont été enregistrées pendant 30 jours. Les jours postopératoires 1-5, les participantes ont reçu des questionnaires quotidiens d'autosurveillance dans l'application C-Care. Quatorze jours après la chirurgie, les participantes ont reçu un sondage en ligne sur leur satisfaction et leur utilisation de C-Care. RéSULTATS: Trente-cinq des 36 participantes, d'un âge médian [écart interquartile (ÉIQ)] de 32 ans [31-36], ont consulté l'application après leur formation initiale. Chaque participante a consulté l'application 15 [9-31] fois en moyenne sur 30 jours et a rempli en moyenne 3 [2-4] questionnaires d'autosurveillance sur cinq. Chaque participante a consulté en moyenne 4 [2-7] des huit sujets de formation, les sujets les plus consultés étant « Le contrôle de la douleur ¼ et « Les premiers jours ¼. Les consultations de l'application ont été les plus élevées au cours de la première semaine post-partum. Parmi les 18 répondantes qui ont complété le sondage du jour 14, 83 % (n = 15) des participantes recommanderaient C-Care à d'autres femmes, et le score médian de satisfaction des participantes était de 7,5 sur 10 (fourchette, 2-10). CONCLUSION: La plupart des participantes ont utilisé cette application mobile destinée à l'éducation des patientes et à l'autosurveillance après un accouchement par césarienne programmée. Une compréhension de l'intérêt des patientes pour C-Care après un accouchement par césarienne pourrait permettre la conception d'applications mobiles périopératoires plus efficaces. ENREGISTREMENT DE L'éTUDE: www.clinicaltrials.gov (NCT03746678); enregistrée le 5 novembre 2018.

Continuous peripheral nerve blocks compared to thoracic epidurals or multimodal analgesia for midline laparotomy: a systematic review and meta-analysis.

BACKGROUND: Continuous peripheral nerve blocks (CPNBs) have been investigated to control pain for abdominal surgery via midline laparotomy while avoiding the adverse events of opioid or epidural analgesia. The review compiles the evidence comparing CPNBs to multimodal and epidural analgesia. METHODS: We conducted a systematic review using broad search terms in MEDLINE, EMBASE, Cochrane. Primary outcomes were pain scores and cumulative opioid consumption at 48 hours. Secondary outcomes were length of stay and postoperative nausea and vomiting (PONV). We rated the quality of the evidence using Cochrane and GRADE recommendations. The results were synthesized by meta-analysis using Revman. RESULTS: Our final selection included 26 studies (1,646 patients). There was no statistically significant difference in pain control comparing CPNBs to either multimodal or epidural analgesia (low quality evidence). Less opioids were consumed when receiving epidural analgesia than CPNBs (mean difference [MD]: -16.13, 95% CI [-32.36, 0.10]), low quality evidence) and less when receiving CPNBs than multimodal analgesia (MD: -31.52, 95% CI [-42.81, -20.22], low quality evidence). The length of hospital stay was shorter when receiving epidural analgesia than CPNBs (MD: -0.78 days, 95% CI [-1.29, -0.27], low quality evidence) and shorter when receiving CPNBs than multimodal analgesia (MD: -1.41 days, 95% CI [-2.45, -0.36], low quality evidence). There was no statistically significant difference in PONV comparing CPNBs to multimodal (high quality evidence) or epidural analgesia (moderate quality evidence). CONCLUSIONS: CPNBs should be considered a viable alternative to epidural analgesia when contraindications to epidural placement exist for patients undergoing midline laparotomies.

Cannabis use and outcomes after aneurysmal subarachnoid hemorrhage: A nationwide retrospective cohort study.

BACKGROUND: Cannabis is the most consumed recreational drug in the world. It is possible that cannabis has an association with an increased risk of vasospasm-related strokes and delayed cerebral ischemia (DCI), which are major causes of morbidity and mortality in aneurysmal subarachnoid hemorrhage (aSAH). Hence, this study aimed to explore the independent relationship between cannabis use and outcomes after aSAH using the 2016 United States Nationwide Inpatient Sample. METHODS: This study was conducted using the 2016 National Inpatient Sample with ICD-10 codes. Multivariate logistic regression was used to examine the association between cannabis use, the primary (inpatient mortality) and secondary outcomes. RESULTS: There were 42,394 patients identified with aSAH, of whom 925 were identified as cannabis users. Cannabis users and non-users were similar in terms of severity of aSAH. Although the unadjusted mortality rate was lower among cannabis users (16%) than non-users (22%), (p = 0.04), both the age-adjusted odds ratio (OR) (0.83, 95% confidence interval (CI): 0.56; 1.24) and the multivariate-adjusted OR (0.87, 95% CI: 0.54; 1.42) did not reach statistical significance. Secondary outcomes did not reach statistical significance. CONCLUSION: In this nationwide cohort, cannabis users with aSAH had similar outcomes compared to nonusers. However, these results are likely limited by underreporting of cannabis use. Future prospective studies are needed to elucidate the pathophysiology and association between cannabis and outcomes following aSAH.

Patient-centred perioperative mobile application in Cesarean delivery: needs assessment and development.

PURPOSE: Cesarean delivery (CD) is the most commonly performed inpatient surgery in Canada, with 103,425 performed in 2016-2017. Mobile technology can contribute to patient-centred perioperative care. Our aim was to involve patients and anesthesiologists in designing a mobile application to enhance the perioperative care of CD patients. METHOD: We completed an exploratory qualitative study involving three iterative design cycles. Individual structured phone or in-person interviews with CD patients (n = 15) and anesthesiologists (n = 9) were conducted. Thematic analysis was used to 1) assess gaps and opportunities in the anesthesiology care after CD, 2) identify roles and limitations of mobile application technology in bridging perioperative care, and 3) to solicit feedback on a prototype mobile application for the anesthesiology postoperative care of elective CD patients. RESULTS: Patients and anesthesiologists were in favour of adding mobile applications to the interdisciplinary perioperative management of patients undergoing CD, particularly for providing reliable information, identifying and following patients with complications, and research in perioperative outcomes. Medical-legal issues, privacy, workflow, and payment policy frameworks are barriers for mobile technology linking patients and anesthesiologists. Our final prototype was refined to focus on timely, concise education and self-monitoring. CONCLUSION: The participant-driven changes in direction of the prototype showed the importance of involving key stakeholders early. This study provides guidance on further iterative development and implementation of an interdisciplinary mobile platform for patient-centred perioperative care and outcomes research.

RéSUMé: OBJECTIF: L'accouchement par césarienne est la chirurgie avec hospitalisation la plus fréquemment réalisée au Canada, atteignant 103 425 cas en 2016­2017. La technologie mobile pourrait contribuer aux soins périopératoires centrés sur la patiente. Notre objectif était d'impliquer les patientes et les anesthésiologistes pour mettre au point une application mobile afin d'améliorer les soins périopératoires des femmes accouchant par césarienne. MéTHODE: Nous avons réalisé une étude qualitative exploratoire comprenant trois cycles itératifs de conception. Des entretiens individuels structurés par téléphone ou en personne avec des patientes subissant une césarienne (n = 15) et des anesthésiologistes (n = 9) ont été menés. Une analyse thématique a été utilisée pour 1) évaluer les écueils et les opportunités dans les soins anesthésiques après une césarienne, 2) identifier les rôles et les limites de la technologie d'application mobile pour combler les besoins des soins périopératoires, et 3) demander une rétroaction concernant un prototype d'application mobile pour les soins anesthésiques postopératoires des patientes subissant une césarienne planifiée. RéSULTATS: Les patientes et les anesthésiologistes étaient en faveur de l'ajout d'applications mobiles à la prise en charge périopératoire des patientes subissant une césarienne, particulièrement pour offrir des informations fiables, identifier et faire le suivi des patientes souffrant de complications, et effectuer de la recherche sur les pronostics périopératoires. Les questions médicolégales, le respect de la vie privée, la charge de travail et les cadres de politiques de paiement constituent des obstacles à la connexion des patientes et des anesthésiologistes via la technologie mobile. Notre prototype final a été amélioré afin de se concentrer sur des outils de formation concis et opportuns et sur l'auto-monitorage. CONCLUSION: Les changements proposés au prototype par les participants ont démontré l'importance d'impliquer les principaux intéressés tôt dans le processus. Cette étude offre des pistes quant à la mise au point et la mise en œuvre progressives d'une plateforme mobile interdisciplinaire pour des soins périopératoires centrées sur les patients et la recherche sur les pronostics périopératoires.