Prediction of pain using machine learning algorithms is an emerging field in both computer science and clinical medicine. Several machine algorithms were developed and validated in recent years. However, the majority of studies in this topic was published on bioinformatics or computer science journals instead of medical journals. This tendency and preference led to a gap of knowledge and acknowledgment between computer scientists who invent the algorithm and medical researchers who may use the algorithms in practice. As a consequence, some of these prediction papers did not discuss the clinical utility aspects and were causally reported without following related professional guidelines (e.g., TRIPOD statement). The aim of this protocol is to systematically summarize the current evidences about performance and utility of different machine learning methods used for automatic pain assessments based on human facial expression. In addition, this study is aimed to demonstrate and fill the knowledge gap to promote interdisciplinary collaboration. We will search all English language literature in the following electronic databases: PubMed, Web of Science and IEEE Xplore. A systematic review and meta-analysis summarizing the accuracy, interpretability, generalizability, and computational efficiency of machine learning methods will be conducted. Subgroup analyses by machine learning method types will be conducted. The formal meta-analysis will start on Jan 15, 2019 and expected to finish by April 15, 2019. Ethical approval will be exempted or will not be required because the data collected and analyzed in this meta-analysis will not be on an individual level. The results will be disseminated in the form of an official publication in a peer-reviewed journal and/or presentation at relevant conferences. PROSPERO CRD42018103059.

Anesthesiologists at our hospital commonly administer spinal anesthesia for routine lumbar spine surgeries. Anecdotal impressions suggested that patients received fewer anesthesia–administered intravenous medications, including vasopressors, during spinal versus general anesthesia. We hypothesized that data review would confirm these impressions. The objective was to test this hypothesis by comparing specific elements of spinal versus general anesthesia for 1–2 level open lumbar spine procedures. Retrospective single institutional study. Academic medical center, operating rooms. Consecutive patients (144 spinal and 619 general anesthesia) identified by automatic structured query of our electronic anesthesia record undergoing lumbar decompression, foraminotomy or microdiscectomy by one surgeon under general or spinal anesthesia. Spinal or general anesthesia. Numbers of medications administered during the case. Anesthesiologists administered in the operating room a total of 10 ± 2 intravenous medications for general anesthetics and 5 ± 2 medications for spinal anesthetics (-5, 95% CI -5 to -4, p<0.001, univariate analysis). Multivariable analysis supported this finding (spinal versus general anesthesia: -4, 95% CI -5 to -4, p<0.001). Spinal anesthesia patients were less likely to receive ephedrine, or phenylephrine (by bolus or by infusion) (all p<0.001, Chi-squared test). Spinal anesthesia patients were also less likely to receive labetolol or esmolol (both p = 0.002, Fishers’ Exact test). No neurologic injuries were attributed to, or masked by, spinal anesthesia. Three spinal anesthetics failed. For routine lumbar surgery in our cohort, spinal compared to general anesthesia was associated with significantly fewer drugs administered during a case and less frequent use of vasoactive agents. Safety implications include greater hemodynamic stability with spinal anesthesia along with reduced risks for medication error and transmission of pathogens associated with medication administration.

The purpose of this study was to evaluate perioperative medication-related incidents (medication errors (MEs) and/or adverse medication events (AMEs)) identified by 2 different reporting methods (self-report and direct observation), and to compare the types and severity of incidents identified by each method. We compared perioperative medication-related incidents identified by direct observation in Nanji et al’s 2016 study^[1] to those identified by self-report via a facilitated incident reporting system at the same 1046-bed tertiary care academic medical center during the same 8-month period. Incidents, including MEs and AMEs were classified by type and severity. In 277 operations involving 3671 medication administrations, 193 MEs and/or AMEs were observed (5.3% incident rate). While none of the observed incidents were self-reported, 10 separate medication-related incidents were self-reported from different (unobserved) operations that occurred during the same time period, which involved a total of 21,576 operations and approximately 280,488 medication administrations (0.004% self-reported incident rate). The distribution of incidents (ME, AME, or both) did not differ by direct observation versus self-report methodology. The types of MEs identified by direct observation differed from those identified by self-report (P = .005). Specifically, the most frequent types of MEs identified by direct observation were labeling errors (N = 37; 24.2%), wrong dose errors (N = 35; 22.9%) and errors of omission (N = 27; 17.6%). The most frequent types of MEs identified by self-report were wrong dose (N = 5; 50%) and wrong medication (N = 4; 40%). The severity of incidents identified by direct observation and self-report differed, with self-reported incidents having a higher average severity (P < .001). The procedure types associated with medication-related incidents did not differ by direct observation versus self-report methodology. Direct observation captured many more perioperative medication-related incidents than self-report. The ME types identified and their severity differed between the 2 methods, with a higher average incident severity in the self-reported data.

Introduction Infants can experience pain similar to adults, and improperly controlled pain stimuli could have a long-term adverse impact on their cognitive and neurological function development. The biggest challenge of achieving good infant pain control is obtaining objective pain assessment when direct communication is lacking. For years, computer scientists have developed many different facial expression-centred machine learning (ML) methods for automatic infant pain assessment. Many of these ML algorithms showed rather satisfactory performance and have demonstrated good potential to be further enhanced for implementation in real-world clinical settings. To date, there is no prior research that has systematically summarised and compared the performance of these ML algorithms. Our proposed meta-analysis will provide the first comprehensive evidence on this topic to guide further ML algorithm development and clinical implementation. Methods and analysis We will search four major public electronic medical and computer science databases including Web of Science, PubMed, Embase and IEEE Xplore Digital Library from January 2008 to present. All the articles will be imported into the Covidence platform for study eligibility screening and inclusion. Study-level extracted data will be stored in the Systematic Review Data Repository online platform. The primary outcome will be the prediction accuracy of the ML model. The secondary outcomes will be model utility measures including generalisability, interpretability and computational efficiency. All extracted outcome data will be imported into RevMan V.5.2.1 software and R V3.3.2 for analysis. Risk of bias will be summarised using the latest Prediction Model Study Risk of Bias Assessment Tool. Ethics and dissemination This systematic review and meta-analysis will only use study-level data from public databases, thus formal ethical approval is not required. The results will be disseminated in the form of an official publication in a peer-reviewed journal and/or presentation at relevant conferences. PROSPERO registration number CRD42019118784.