Machine learning in perioperative medicine: a systematic review

Background Risk stratification plays a central role in anesthetic evaluation. The use of Big Data and machine learning (ML) offers considerable advantages for collection and evaluation of large amounts of complex health-care data. We conducted a systematic review to understand the role of ML in the development of predictive post-surgical outcome models and risk stratification. Methods Following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines, we selected the period of the research for studies from 1 January 2015 up to 30 March 2021. A systematic search in Scopus, CINAHL, the Cochrane Library, PubMed, and MeSH databases was performed; the strings of research included different combinations of keywords: “risk prediction,” “surgery,” “machine learning,” “intensive care unit (ICU),” and “anesthesia” “perioperative.” We identified 36 eligible studies. This study evaluates the quality of reporting of prediction models using the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) checklist. Results The most considered outcomes were mortality risk, systemic complications (pulmonary, cardiovascular, acute kidney injury (AKI), etc.), ICU admission, anesthesiologic risk and prolonged length of hospital stay. Not all the study completely followed the TRIPOD checklist, but the quality was overall acceptable with 75% of studies (Rev #2, comm #minor issue) showing an adherence rate to TRIPOD more than 60%. The most frequently used algorithms were gradient boosting (n = 13), random forest (n = 10), logistic regression (LR; n = 7), artificial neural networks (ANNs; n = 6), and support vector machines (SVM; n = 6). Models with best performance were random forest and gradient boosting, with AUC > 0.90. Conclusions The application of ML in medicine appears to have a great potential. From our analysis, depending on the input features considered and on the specific prediction task, ML algorithms seem effective in outcomes prediction more accurately than validated prognostic scores and traditional statistics. Thus, our review encourages the healthcare domain and artificial intelligence (AI) developers to adopt an interdisciplinary and systemic approach to evaluate the overall impact of AI on perioperative risk assessment and on further health care settings as well.

Care and compassion at time of pandemic: an ICU field experience in Italy

After the COVID-19 pandemic outbreak in March 2020, the majority of hospital policies have followed guidelines aimed to prevent the virus transmission and the families’ entry was denied in all hospital wards and intensive care units (ICU). Despite being necessary, such restrictions have been experienced with discomfort and sufferings by the general ICU staff of Treviso Hospital (Italy) and by families of patients. Therefore, from April 2020, a step-by-step project was developed to reactivate contact with COVID-19 patients’ families. The several requests and appeals of intensive care communities and organizations, both nationally and internationally, motivated the Treviso hospital initiative, leading to a model that might be potentially useful to other intensive care units worldwide.

Perioperative goal-directed therapy and postoperative complications in different kind of surgical procedures: an updated meta-analysis

Background Goal-directed therapy (GDT) aims to assure tissue perfusion, by optimizing doses and timing of fluids, inotropes, and vasopressors, through monitoring of cardiac output and other basic hemodynamic parameters. Several meta-analyses confirm that GDT can reduce postoperative complications. However, all recent evidences focused on high-risk patients and on major abdominal surgery. Objectives The aim of the present meta-analysis is to investigate the effect of GDT on postoperative complications (defined as number of patients with a least one postoperative complication) in different kind of surgical procedures. Data sources Randomized controlled trials (RCTs) on perioperative GDT in adult surgical patients were included. The primary outcome measure was complications, defined as number of patients with at least one postoperative complication. A subgroup-analysis was performed considering the kind of surgery: major abdominal (including also major vascular), only vascular, only orthopedic surgery. and so on. Study appraisal and synthesis methods Meta-analytic techniques (analysis software RevMan, version 5.3.5, Cochrane Collaboration, Oxford, England, UK) were used to combine studies using odds ratios (ORs) and 95% confidence intervals (CIs). Results In 52 RCTs, 6325 patients were enrolled. Of these, 3162 were randomized to perioperative GDT and 3153 were randomized to control. In the overall population, 2836 patients developed at least one complication: 1278 (40%) were randomized to perioperative GDT, and 1558 (49%) were randomized to control. Pooled OR was 0.60 and 95% CI was 0.49–0.72. The sensitivity analysis confirmed the main result. The analysis enrolling major abdominal patients showed a significant result (OR 0.72, 95% CI 0.59–0.87, p = 0.0007, 31 RCTs, 4203 patients), both in high- and low-risk patients. A significant effect was observed in those RCTs enrolling exclusively orthopedic procedures (OR 0.53, 95% CI 0.35–0.80, p = 0.002, 7 RCTs, 650 patients. Also neurosurgical procedures seemed to benefit from GDT (OR 0.40, 95% CI 0.21–0.78, p = 0.008, 2 RCTs, 208 patients). In both major abdominal and orthopedic surgery, a strategy adopting fluids and inotropes yielded significant results. The total volume of fluid was not significantly different between the GDT and the control group. Conclusions and implications of key findings The present meta-analysis, within the limits of the existing data, the clinical and statistical heterogeneity, suggests that GDT can reduce postoperative complication rate. Moreover, the beneficial effect of GDT on postoperative morbidity is significant on major abdominal, orthopedic and neurosurgical procedures. Several well-designed RCTs are needed to further explore the effect of GDT in different kind of surgeries.

Intraoperative cardiac function assessment by transesophageal echocardiography versus FloTrac/Vigileo™ system during pectus excavatum surgical repair

Background Pectus excavatum (PE), a congenital deformity of the chest wall, can lead to cardiac compression and related symptoms. PE surgical repair can improve cardiac function. Intraoperative transesophageal echocardiography (TEE) has been successfully employed to assess intraoperative hemodynamic variations in patients undergoing PE repair. FloTrac/Vigileo™ system (Edwards Life-sciences Irvine, CA) (FT/V) is a minimally invasive cardiac output monitoring system. This retrospective study aimed to assess hemodynamic changes in surgical repair of PE using FT/V and concordance with parameters measured by TEE. Results N=19 patients submitted to PE repair via Ravitch or Nuss technique were enrolled. Intraoperative cardiac assessments simultaneously obtained via TEE and FT/V system were investigated. The agreement between TEE-derived cardiac output (CO-TEE) and FT/V system parameter (COAP) was evaluated. The relationship between COTEE and COAP was analyzed for all data using linear regression analysis. A significant correlation between COAP and COTEE values (R = 0.65, p < 0.001) was found. Bland-Altman analysis of COAP and COTEE showed a bias of 0.13 L/min and a limit of agreement of − 2.33 to 2.58 L/min, with a percentage error of 48%. Intraoperative measurements by TEE and FT/V both showed a significant increase in CO after surgical correction of PE (p < 0.005). Conclusions FT/V system compared to TEE in hemodynamic monitoring during PE surgery yielded clinically unacceptable results due to a high percentage error. After surgical correction of PE, CO, measured by TEE and FT/V, significantly improved.

Life-threatening complications of streptococcal sepsis: a PICU contemporary series

Background Life-threatening streptococcal sepsis nowadays represents an uncommon event in previously healthy infants and children. Critically ill patients suffering from severe streptococcal sepsis complications may present with pre-antibiotic era clinical pictures and require a timely clinical approach to achieve restitutio ad integrum. Results We report a series of four patient groups affected by an uncommon life-threatening streptococcal sepsis, each of them exhibiting some distinct features. Streptococcus Agalactiae sepsis was associated with cerebral thrombotic/ischaemic lesions, whereas severe cardiogenic shock was prominent in the Streptococcus Viridans group; Streptococcus Faecalis and β-hemolytic group A Streptococcus patients mostly reported lung complications. Conclusions Previous antibiotic treatments should not delay aggressive treatment in the intensive care setting. Early diagnostic suspicion, as well as appropriate and aggressive treatment provided within an intensive care setting are crucial for the clinical outcome.

Dexamethasone and postoperative analgesia in minimally invasive thoracic surgery: a retrospective cohort study

Background Dexamethasone is commonly used for the prevention of postoperative nausea and vomiting (PONV), and recent reviews suggest a role for dexamethasone in postoperative analgesia. The aim of this study is to evaluate the efficacy of dexamethasone as an analgesic adjuvant in minimally invasive thoracic surgery. Primary outcome was morphine consumption 24 h after surgery; secondary outcomes were pain control, measured as numeric rating scale (NRS), glycemic changes, PONV, and surgical wound infection. Results We performed a retrospective cohort study considering 70 patients who underwent elective lobectomy, segmentectomy, or wedge resection surgery with a mini-thoracotomy approach or video-assisted thoracoscopic surgery (VATS). All patients received the same locoregional techniques and short-acting opioids during surgery; 46 patients received dexamethasone at induction. There were no significant differences in morphine consumption at 24 h (p = 0.09) and in postoperative pain scores. Nevertheless, a higher frequency of rescue therapy (p = 0.01) and a tendency for a higher attempted-PCA pushes count were observed in patients who did not receive dexamethasone. No cases of surgical wound infections were detected, and the incidence of PONV was similar in the two groups. Postoperative glycemia was transiently higher in the dexamethasone group (p = 0.004), but the need of hypoglycemic therapy was not significantly different. Conclusions Preoperative administration of dexamethasone did not cause a significant reduction in morphine consumption, but appears to be safe and plays a role in a multimodal anesthesia approach for patients undergoing elective minimally invasive thoracic surgery.

Echocardiographic assessment of left ventricular function in ex situ heart perfusion using pump-supported and passive afterload working mode: a pilot study

Ex situ heart perfusion (ESHP) has been developed to decrease cold ischemia time and allow metabolic assessment of donor hearts prior to transplantation. Current clinical ESHP systems preserve the heart in an unloaded condition and only evaluate the cardiac metabolic profile. In this pilot study we performed echocardiographic functional assessment using two alternative systems for left ventricular (LV) loading: pump supported afterload working mode (SAM) and passive afterload working modes (PAM). Six hearts were procured from male Yorkshire pigs. During cold ischemia, hearts were mounted on our custom made ESHP circuit and a 3D-printed enclosure for the performance of echocardiography with a standard TEE probe. Following perfusion with Langherdorf mode of the unloaded heart, the system was switched into different working modes to allow LV loading and functional assessment: pump supported (SAM) and passive (PAM). Echocardiographic assessment of left ventricular function in the donor hearts was performed in vivo and at 1 h of ESHP with SAM, after 4.5 h with PAM and after 5.5 h with SAM. We obtained good quality epicardial echocardiographic images at all time points allowing a comprehensive LV systolic assessment. All indices showed a decrease in LV systolic function throughout the trial with the biggest drop after heart harvesting. We demonstrated the feasibility of echocardiographic functional assessment during ESHP and two different working modes. The expected LV systolic dysfunction consisted of a reduction in EF, FAC, FS, and strain throughout the experiment with the most significant decrease after harvesting.

Predictors of intubation and mortality in COVID-19 patients: a retrospective study

Background Estimating the risk of intubation and mortality among COVID-19 patients can help clinicians triage these patients and allocate resources more efficiently. Thus, here we sought to identify the risk factors associated with intubation and intra-hospital mortality in a cohort of COVID-19 patients hospitalized due to hypoxemic acute respiratory failure (ARF). Results We included retrospectively a total of 187 patients admitted to the subintensive and intensive care units of the University Hospital “Maggiore della Carità” of Novara between March 1st and April 30th, 2020. Based on these patients’ demographic characteristics, early clinical and laboratory variables, and quantitative chest computerized tomography (CT) findings, we developed two random forest (RF) models able to predict intubation and intra-hospital mortality. Variables independently associated with intubation were C-reactive protein (p < 0.001), lactate dehydrogenase level (p = 0.018) and white blood cell count (p = 0.026), while variables independently associated with mortality were age (p < 0.001), other cardiovascular diseases (p = 0.029), C-reactive protein (p = 0.002), lactate dehydrogenase level (p = 0.018), and invasive mechanical ventilation (p = 0.001). On quantitative chest CT analysis, ground glass opacity, consolidation, and fibrosis resulted significantly associated with patient intubation and mortality. The major predictors for both models were the ratio between partial pressure of arterial oxygen and fraction of inspired oxygen, age, lactate dehydrogenase, C-reactive protein, glycemia, CT quantitative parameters, lymphocyte count, and symptom onset. Conclusions Altogether, our findings confirm previously reported demographic, clinical, hemato-chemical, and radiologic predictors of adverse outcome among COVID-19-associated hypoxemic ARF patients. The two newly developed RF models herein described show an overall good level of accuracy in predicting intra-hospital mortality and intubation in our study population. Thus, their future development and implementation may help not only identify patients at higher risk of deterioration more effectively but also rebalance the disproportion between resources and demand.