Evaluation of an electronic prescription platform: Clinicians' feedback on three distinct services aiming to facilitate clinical decision and safer e-prescription.

BACKGROUND: Health Care Professionals (HCPs) are the main end-users of digital clinical tools such as electronic prescription systems. For this reason, it is of high importance to include HCPs throughout the design, development and evaluation of a newly introduced system to ensure its usefulness, as well as confirm that it tends to their needs and can be integrated in their everyday clinical practice. METHODS: In the context of the PrescIT project, an electronic prescription platform with three services was developed (i.e., Prescription Check, Prescription Suggestion, Therapeutic Prescription Monitoring). To allow an iterative process of discovery through user feedback, design and implementation, a two-phase evaluation was carried out, with the participation of HCPs from three hospitals in Northern Greece. The two-phase evaluation included presentations of the platform, followed by think-aloud sessions, individual platform testing and the collection of qualitative as well as quantitative feedback, through standard questionnaires (e.g., SUS, PSSUQ). RESULTS: Twenty one HCPs (8 in the first, 18 in the second phase, and five present in both) participated in the two-phase evaluation. HCPs comprised clinicians varying in their specialty and one pharmacist. Clinicians' feedback during the first evaluation phase already deemed usability as "excellent" (with SUS scores ranging from 75 to 95/100, showing a mean value of 86.6 and SD of 9.2) but also provided additional user requirements, which further shaped and improved the services. In the second evaluation phase, clinicians explored the system's usability, and identified the services' strengths and weaknesses. Clinicians perceived the platform as useful, as it provides information on potential adverse drug reactions, drug-to-drug interactions and suggests medications that are compatible with patients' comorbidities and current medication. CONCLUSIONS: The developed PrescIT platform aims to increase overall safety and effectiveness of healthcare services. Therefore, including clinicians in a two-phase evaluation confirmed that the introduced system is useful, tends to the users' needs, does not create fatigue and can be integrated in their everyday clinical practice to support clinical decision and e-prescribing.

Use of Real-World Data to Support Adverse Drug Reactions Prevention During ePrescription.

Adverse Drug Reactions (ADRs) are a crucial public health issue due to the significant health and monetary burden that they can impose. Real-World Data (RWD), e.g., Electronic Health Records, claims data, etc., can support the identification of potentially unknown ADRs and thus, they could provide raw data to mine ADR prevention rules. The PrescIT project aims to create a Clinical Decision Support System (CDSS) for ADR prevention during ePrescription and uses OMOP-CDM as the main data model to mine ADR prevention rules, based on the software stack provided by the OHDSI initiative. This paper presents the deployment of OMOP-CDM infrastructure using the MIMIC-III as a testbed.

The PrescIT Knowledge Graph: Supporting ePrescription to Prevent Adverse Drug Reactions.

Adverse Drug Reactions (ADRs) are an important public health issue as they can impose significant health and monetary burdens. This paper presents the engineering and use case of a Knowledge Graph, supporting the prevention of ADRs as part of a Clinical Decision Support System (CDSS) developed in the context of the PrescIT project. The presented PrescIT Knowledge Graph is built upon Semantic Web technologies namely the Resource Description Framework (RDF), and integrates widely relevant data sources and ontologies, i.e., DrugBank, SemMedDB, OpenPVSignal Knowledge Graph and DINTO, resulting in a lightweight and self-contained data source for evidence-based ADRs identification.

Randomization of Clinical Trial Participants via an Integrated Web Service.

Randomization is an inherent part of Randomized Clinical Trials (RCTs), typically requiring the split of participants in intervention and control groups. We present a web service supporting randomized patient distribution, developed in the context of the MyPal project RCT. The randomization process is based on a block permutation approach to mitigate the risk of various kind of biases. The presented service can be used via its web user interface to produce randomized lists of patients distributed in the various study groups, with a variant block size. Alternatively, the presented service can be integrated as part of wider IT systems supporting clinical trials via a REST interface following a micro-service architectural pattern.

Serious Gaming Technology in Upper Extremity Rehabilitation: Scoping Review.

BACKGROUND: Serious gaming has increasingly gained attention as a potential new component in clinical practice. Specifically, its use in the rehabilitation of motor dysfunctions has been intensively researched during the past three decades. OBJECTIVE: The aim of this scoping review was to evaluate the current role of serious games in upper extremity rehabilitation, and to identify common methods and practice as well as technology patterns. This objective was approached via the exploration of published research efforts over time. METHODS: The literature search, using the PubMed and Scopus databases, included articles published from 1999 to 2019. The eligibility criteria were (i) any form of game-based arm rehabilitation; (ii) published in a peer-reviewed journal or conference; (iii) introduce a game in an electronic format; (iv) published in English; and (v) not a review, meta-analysis, or conference abstract. The search strategy identified 169 relevant articles. RESULTS: The results indicated an increasing research trend in the domain of serious gaming deployment in upper extremity rehabilitation. Furthermore, differences regarding the number of publications and the game approach were noted between studies that used commercial devices in their rehabilitation systems and those that proposed a custom-made robotic arm, glove, or other devices for the connection and interaction with the game platform. A particularly relevant observation concerns the evaluation of the introduced systems. Although one-third of the studies evaluated their implementations with patients, in most cases, there is the need for a larger number of participants and better testing of the rehabilitation scheme efficiency over time. Most of the studies that included some form of assessment for the introduced rehabilitation game mentioned user experience as one of the factors considered for evaluation of the system. Besides user experience assessment, the most common evaluation method involving patients was the use of standard medical tests. Finally, a few studies attempted to extract game features to introduce quantitative measurements for the evaluation of patient improvement. CONCLUSIONS: This paper presents an overview of a significant research topic and highlights the current state of the field. Despite extensive attempts for the development of gamified rehabilitation systems, there is no definite answer as to whether a serious game is a favorable means for upper extremity functionality improvement; however, this certainly constitutes a supplementary means for motivation. The development of a unified performance quantification framework and more extensive experiments could generate richer evidence and contribute toward this direction.

Smart alarms towards optimizing patient ventilation in intensive care: the driving pressure case.

OBJECTIVE: Alarms are a substantial part of clinical practice, warning clinicians of patient complications. In this paper, we focus on alarms in the intensive care unit and especially on the use of machine learning techniques for the creation of alarms for the ventilator support of patients. The aim is to study a method to enable timely interventions for intubated patients and prevent complications induced by high driving pressure (ΔP) and lung strain during mechanical ventilation. APPROACH: The relation between the ΔP and the total set of the ventilator parameters was examined and resulted in a predictive model with bimodal implementation for the short-term prediction of the ΔP level (high/low). The proposed method includes two sub-models for the prediction of future ΔP level based on the current level being high or low, named cH and cL, respectively. Based on this method, for both sub-models, an alarm will be triggered when the predicted ΔP level is considered to be high. In this vein, three classifiers (the random forest, linear support vector machine, and kernel support vector machine methods) were tested for each sub-model. To adjust the highly unbalanced classes, four different sampling methods were considered: downsampling, upsampling, synthetic minority over-sampling technique (SMOTE) sampling, and random over-sampling examples (ROSE) sampling. MAIN RESULTS: For the cL sub-model the combination of linear support vector machine with SMOTE sampling showed the best performance, resulting in accuracy of 93%, while the cH sub-model reached the best performance, with accuracy of 73%, with kernel support vector machine combined with the downsampling method. SIGNIFICANCE: The results are positive in terms of the generation of new alarms in mechanical ventilation. The technical and organizational possibility of integrating data from multiple modalities is expected to further advance this line of work.

Driving pressure during proportional assist ventilation: an observational study.

BACKGROUND: During passive mechanical ventilation, the driving pressure of the respiratory system is an important mediator of ventilator-induced lung injury. Monitoring of driving pressure during assisted ventilation, similar to controlled ventilation, could be a tool to identify patients at risk of ventilator-induced lung injury. The aim of this study was to describe driving pressure over time and to identify whether and when high driving pressure occurs in critically ill patients during assisted ventilation. METHODS: Sixty-two patients fulfilling criteria for assisted ventilation were prospectively studied. Patients were included when the treating physician selected proportional assist ventilation (PAV+), a mode that estimates respiratory system compliance. In these patients, continuous recordings of all ventilator parameters were obtained for up to 72 h. Driving pressure was calculated as tidal volume-to-respiratory system compliance ratio. The distribution of driving pressure and tidal volume values over time was examined, and periods of sustained high driving pressure (≥ 15 cmH2O) and of stable compliance were identified and analyzed. RESULTS: The analysis included 3200 h of ventilation, consisting of 8.8 million samples. For most (95%) of the time, driving pressure was < 15 cmH2O and tidal volume < 11 mL/kg (of ideal body weight). In most patients, high driving pressure was observed for short periods of time (median 2.5 min). Prolonged periods of high driving pressure were observed in five patients (8%). During the 661 periods of stable compliance, high driving pressure combined with a tidal volume ≥ 8 mL/kg was observed only in 11 cases (1.6%) pertaining to four patients. High driving pressure occurred almost exclusively when respiratory system compliance was low, and compliance above 30 mL/cmH2O excluded the presence of high driving pressure with 90% sensitivity and specificity. CONCLUSIONS: In critically ill patients fulfilling criteria for assisted ventilation, and ventilated in PAV+ mode, sustained high driving pressure occurred in a small, yet not negligible number of patients. The presence of sustained high driving pressure was not associated with high tidal volume, but occurred almost exclusively when compliance was below 30 mL/cmH2O.

Clusters of ineffective efforts during mechanical ventilation: impact on outcome.

PURPOSE: The aim of this study was to investigate the role of ineffective efforts (IEs), specifically clusters of IEs, during mechanical ventilation on the outcome of critically ill patients. METHODS: In a prospective observational study, 24-h recordings were obtained in 110 patients on the 1st day of assisted ventilation (pressure support or proportional assist), using a prototype monitor validated to identify IEs. Patients remaining on assisted ventilation were studied again on the 3rd day (n = 37) and on the 6th day (n = 13). To describe the clusters of IEs, the concept of an IEs event was developed, defined as a 3-min period of time containing more than 30 IEs. Along with all patient data, to minimize selection bias by time of recording, analysis was performed only on 1st day data of patients with ≥16 h of recording (1st day group). RESULTS: The analysis included 2931 h of assisted ventilation and 4,456,537 breaths. Neither the IEs index (IEs as a percentage of total breaths) in general nor a value above 10 % was correlated with patient outcome. Overall, IEs events were identified in 38 % of patients. In multivariate analysis, the presence of events in the 1st day group (n = 79) was associated with the risk of being on mechanical ventilation ≥8 days after first recording [odds ratio 6.4, 95 % confidence interval (1.1-38.30)] and hospital mortality [20 (2.3-175)]. Analysis of the data for all patients revealed similarly increased risks for prolonged ventilation [3.4 (1.1-10.7)] and mortality [4.9 (1.3-18)]. CONCLUSIONS: Clusters of IEs are often present in mechanically ventilated critically ill patients and are associated with prolonged mechanical ventilation and increased mortality. Studies to find ways of improving patient-ventilator interaction are warranted.