Addressing the flaws of current critical alarms: a fuzzy constraint satisfaction approach

Accuracy and availability of automated urine output monitoring in the operating room using a smart scale

PURPOSE

Urine output (UO) is an important intraoperative parameter that is not yet electronically monitored. We compared an automatic urinometer (AU) based on a smart scale with a manual urinometer (MU).

PATIENTS AND METHODS

This prospective study investigated the hourly UO of 35 preoperative patients with an indwelling urinary catheter using AU, MU, and cylinder measurements. Data were analyzed using the Bland-Altman method. A questionnaire related to the use of the AU was completed by medical staff (n=25).

RESULTS

Compared to the cylinder measurements, the differences in measurements by the AU and the MU were -6.31 ​± ​15.03 ​mL/h (p=0.018) and 20.26 ​± ​26.81 ​mL/h (p=0.001), respectively. The r values for the comparison of cylinder measurements with AU and MU values were 0.985 (p<0.001) and 0.968 (p<0.001), respectively. Bland-Altman analyses showed that cylinder measurements had better agreement with the AU measurements than with the MU measurements. Also, the medical staff reported that the use of the AU was easier to learn than the use of the MU (p<0.001).

CONCLUSIONS

Compared to the MU values, AU values were noninferior; they had significantly less bias and temporal deviation. Additionally, the medical staff reported that the use of the AU was easier to learn than the use of the MU.

Faster clinical response to the onset of adverse events: A wearable metacognitive attention aid for nurse triage of clinical alarms

Objective

This study evaluates the potential for improving patient safety by introducing a metacognitive attention aid that enables clinicians to more easily access and use existing alarm/alert information. It is hypothesized that this introduction will enable clinicians to easily triage alarm/alert events and quickly recognize emergent opportunities to adapt care delivery. The resulting faster response to clinically important alarms/alerts has the potential to prevent adverse events and reduce healthcare costs.

Materials and methods

A randomized within-subjects single-factor clinical experiment was conducted in a high-fidelity 20-bed simulated acute care hospital unit. Sixteen registered nurses, four at a time, cared for five simulated patients each. A two-part highly realistic clinical scenario was used that included representative: tasking; information; and alarms/alerts. The treatment condition introduced an integrated wearable attention aid that leveraged metacognition methods from proven military systems. The primary metric was time for nurses to respond to important alarms/alerts.

Results

Use of the wearable attention aid resulted in a median relative within-subject improvement for individual nurses of 118% (W = 183, p = 0.006). The top quarter of relative improvement was 3,303% faster (mean; 17.76 minutes reduced to 1.33). For all unit sessions, there was an overall 148% median faster response time to important alarms (8.12 minutes reduced to 3.27; U = 2.401, p = 0.016), with 153% median improvement in consistency across nurses (F = 11.670, p = 0.001).

Discussion and conclusion

Existing device-centric alarm/alert notification solutions can require too much time and effort for nurses to access and understand. As a result, nurses may ignore alarms/alerts as they focus on other important work. There has been extensive research on reducing alarm frequency in healthcare. However, alarm safety remains a top problem. Empirical observations reported here highlight the potential of improving patient safety by supporting the meta-work of checking alarms.

Using Supervised Machine Learning to Classify Real Alerts and Artifact in Online Multisignal Vital Sign Monitoring Data

OBJECTIVE

The use of machine-learning algorithms to classify alerts as real or artifacts in online noninvasive vital sign data streams to reduce alarm fatigue and missed true instability.

DESIGN

Observational cohort study.

SETTING

Twenty-four-bed trauma step-down unit.

PATIENTS

Two thousand one hundred fifty-three patients.

INTERVENTION

Noninvasive vital sign monitoring data (heart rate, respiratory rate, peripheral oximetry) recorded on all admissions at 1/20 Hz, and noninvasive blood pressure less frequently, and partitioned data into training/validation (294 admissions; 22,980 monitoring hours) and test sets (2,057 admissions; 156,177 monitoring hours). Alerts were vital sign deviations beyond stability thresholds. A four-member expert committee annotated a subset of alerts (576 in training/validation set, 397 in test set) as real or artifact selected by active learning, upon which we trained machine-learning algorithms. The best model was evaluated on test set alerts to enact online alert classification over time.

MEASUREMENTS AND MAIN RESULTS

The Random Forest model discriminated between real and artifact as the alerts evolved online in the test set with area under the curve performance of 0.79 (95% CI, 0.67-0.93) for peripheral oximetry at the instant the vital sign first crossed threshold and increased to 0.87 (95% CI, 0.71-0.95) at 3 minutes into the alerting period. Blood pressure area under the curve started at 0.77 (95% CI, 0.64-0.95) and increased to 0.87 (95% CI, 0.71-0.98), whereas respiratory rate area under the curve started at 0.85 (95% CI, 0.77-0.95) and increased to 0.97 (95% CI, 0.94-1.00). Heart rate alerts were too few for model development.

CONCLUSIONS

Machine-learning models can discern clinically relevant peripheral oximetry, blood pressure, and respiratory rate alerts from artifacts in an online monitoring dataset (area under the curve > 0.87).

Real alerts and artifact classification in archived multi-signal vital sign monitoring data: implications for mining big data

Huge hospital information system databases can be mined for knowledge discovery and decision support, but artifact in stored non-invasive vital sign (VS) high-frequency data streams limits its use. We used machine-learning (ML) algorithms trained on expert-labeled VS data streams to automatically classify VS alerts as real or artifact, thereby "cleaning" such data for future modeling. 634 admissions to a step-down unit had recorded continuous noninvasive VS monitoring data [heart rate (HR), respiratory rate (RR), peripheral arterial oxygen saturation (SpO₂) at 1/20 Hz, and noninvasive oscillometric blood pressure (BP)]. Time data were across stability thresholds defined VS event epochs. Data were divided Block 1 as the ML training/cross-validation set and Block 2 the test set. Expert clinicians annotated Block 1 events as perceived real or artifact. After feature extraction, ML algorithms were trained to create and validate models automatically classifying events as real or artifact. The models were then tested on Block 2. Block 1 yielded 812 VS events, with 214 (26 %) judged by experts as artifact (RR 43 %, SpO₂ 40 %, BP 15 %, HR 2 %). ML algorithms applied to the Block 1 training/cross-validation set (tenfold cross-validation) gave area under the curve (AUC) scores of 0.97 RR, 0.91 BP and 0.76 SpO₂. Performance when applied to Block 2 test data was AUC 0.94 RR, 0.84 BP and 0.72 SpO₂. ML-defined algorithms applied to archived multi-signal continuous VS monitoring data allowed accurate automated classification of VS alerts as real or artifact, and could support data mining for future model building.

A new device to automate the monitoring of critical patients' urine output

Urine output (UO) is usually measured manually each hour in acutely ill patients. This task consumes a substantial amount of time. Furthermore, in the literature there is evidence that more frequent (minute-by-minute) UO measurement could impact clinical decision making and improve patient outcomes. However, it is not feasible to manually take minute-by-minute UO measurements. A device capable of automatically monitoring UO could save precious time of the healthcare staff and improve patient outcomes through a more precise and continuous monitoring of this parameter. This paper presents a device capable of automatically monitoring UO. It provides minute by minute measures and it can generate alarms that warn of deviations from therapeutic goals. It uses a capacitive sensor for the measurement of the UO collected within a rigid container. When the container is full, it automatically empties without requiring any internal or external power supply or any intervention by the nursing staff. In vitro tests have been conducted to verify the proper operation and accuracy in the measures of the device. These tests confirm the viability of the device to automate the monitoring of UO.

On the minute by minute variations of urine output: a study in a porcine model

BACKGROUND

Urine output (UO) is usually measured hourly in acutely ill patients. Devices capable of more continuous (minute by minute urine output, UOm) measurements have become available recently. This paper aims to (1) analyze the minute by minute variations of UO, (2) analyze the impact of sepsis on those variations and (3) test if UO measured over periods shorter than 60 min provides information not available in hourly measurements.

METHODS

Fifteen male pigs were anesthetized, tracheostomized and mechanically ventilated. Sepsis was induced by the administration of live Escherichia coli. Three groups were studied: nonseptic (n = 7) and septic (n = 4), both receiving sodium chloride (NaCl) 0.9 % at 4 ml kg(-1) h(-1); and septic (n = 4) receiving NaCl 0.9 % at 17 ml kg(-1) h(-1). UOm was measured during 6 h.

RESULTS

There was a significant variation of UOm over time, as assessed by the coefficient of variation of the root-mean-squared error (CV(RMSE)), which was significantly more pronounced under conditions of sepsis than under control conditions. A UO production pattern in sepsis was identified, characterized by low UO production compared to baseline levels for approximately 30 min, followed by high UO production for approximately 30 min after initiation of the septic challenge. This pattern was noticeable if UO was measured every 10 min but not over longer periods of time.

CONCLUSIONS

UOm provides information not conveyed by hourly measurements, especially under the cardiovascular alterations associated to sepsis. This information could enable an early identification of sepsis.

A three stage ontology-driven solution to provide personalized care to chronic patients at home

PURPOSE

The goal of this work is to contribute to personalized clinical management in home-based telemonitoring scenarios by developing an ontology-driven solution that enables a wide range of remote chronic patients to be monitored at home.

METHODS

Through three stages, the challenges of integration and management were met through the ontology development and evaluation. The first stage dealt with the ontology design and implementation. The second stage dealt with the ontology application study in order to specifically address personalization issues. For both stages, interviews and working sessions were planned with clinicians. Clinical guidelines and MDs (medical device) interoperability were taken into account as well during these stages. Finally the third stage dealt with a software prototype implementation.

RESULTS

An ontology was developed as an outcome of the first stage. The structure, based on the autonomic computing paradigm, provides a clear and simple manner to automate and integrate the data management procedure. During the second stage, the application of the ontology was studied to monitor patients with different and multiple morbidities. After this task, the ontology design was successfully adjusted to provide useful personalized medical care. In the third and final stage, a proof-of-concept on the software required to remote monitor patients by means of the ontology-based solution was developed and evaluated.

CONCLUSIONS

Our proposed ontology provides an understandable and simple solution to address integration and personalized care challenges in home-based telemonitoring scenarios. Furthermore, our three-stage approach contributes to enhance the understanding, re-usability and transferability of our solution.

Semi-supervised detection of intracranial pressure alarms using waveform dynamics

Patient monitoring systems in intensive care units (ICU) are usually set to trigger alarms when abnormal values are detected. Alarms are generated by threshold-crossing rules that lead to high false alarm rates. This is a recognized issue that causes alarm fatigue, waste of human resources, and increased patient risks. Recently developed smart alarm models require alarms to be validated by experts during the training phase. The manual annotation process involved is time-consuming and virtually impossible to achieve for the thousands of alarms recorded in the ICU every week. To tackle this problem, we investigate in this study if the use of semi-supervised learning methods, that can naturally integrate unlabeled data samples in the model, can be used to improve the accuracy of the alarm detection. As a proof of concept, the detection system is evaluated on intracranial pressure (ICP) signal alarms. Specific morphological and trending features are extracted from the ICP signal waveform to capture the dynamic of the signal prior to alarms. This study is based on a comprehensive dataset of 4791 manually labeled alarms recorded from 108 neurosurgical patients. A comparative analysis is provided between kernel spectral regression (SR-KDA) and support vector machine (SVM) both modified for the semi-supervised setting. Results obtained during the experimental evaluations indicate that the two models can significantly reduce false alarms using unlabeled samples; especially in the presence of a restrained number of labeled examples. At a true alarm recognition rate of 99%, the false alarm reduction rates improved from 9% (supervised) to 27% (semi-supervised) for SR-KDA, and from 3% (supervised) to 16% (semi-supervised) for SVM.

Implementation of artifact detection in critical care: a methodological review

Artifact detection (AD) techniques minimize the impact of artifacts on physiologic data acquired in critical care units (CCU) by assessing quality of data prior to clinical event detection (CED) and parameter derivation (PD). This methodological review introduces unique taxonomies to synthesize over 80 AD algorithms based on these six themes: 1) CCU; 2) physiologic data source; 3) harvested data; 4) data analysis; 5) clinical evaluation; and 6) clinical implementation. Review results show that most published algorithms: a) are designed for one specific type of CCU; b) are validated on data harvested only from one OEM monitor; c) generate signal quality indicators (SQI) that are not yet formalized for useful integration in clinical workflows; d) operate either in standalone mode or coupled with CED or PD applications; e) are rarely evaluated in real-time; and f) are not implemented in clinical practice. In conclusion, it is recommended that AD algorithms conform to generic input and output interfaces with commonly defined data: 1) type; 2) frequency; 3) length; and 4) SQIs. This shall promote: a) reusability of algorithms across different CCU domains; b) evaluation on different OEM monitor data; c) fair comparison through formalized SQIs; d) meaningful integration with other AD, CED and PD algorithms; and e) real-time implementation in clinical workflows.

Smart health monitoring systems: an overview of design and modeling

Health monitoring systems have rapidly evolved during the past two decades and have the potential to change the way health care is currently delivered. Although smart health monitoring systems automate patient monitoring tasks and, thereby improve the patient workflow management, their efficiency in clinical settings is still debatable. This paper presents a review of smart health monitoring systems and an overview of their design and modeling. Furthermore, a critical analysis of the efficiency, clinical acceptability, strategies and recommendations on improving current health monitoring systems will be presented. The main aim is to review current state of the art monitoring systems and to perform extensive and an in-depth analysis of the findings in the area of smart health monitoring systems. In order to achieve this, over fifty different monitoring systems have been selected, categorized, classified and compared. Finally, major advances in the system design level have been discussed, current issues facing health care providers, as well as the potential challenges to health monitoring field will be identified and compared to other similar systems.

Reducing unnecessary lab testing in the ICU with artificial intelligence

OBJECTIVES

To reduce unnecessary lab testing by predicting when a proposed future lab test is likely to contribute information gain and thereby influence clinical management in patients with gastrointestinal bleeding. Recent studies have demonstrated that frequent laboratory testing does not necessarily relate to better outcomes.

DESIGN

Data preprocessing, feature selection, and classification were performed and an artificial intelligence tool, fuzzy modeling, was used to identify lab tests that do not contribute an information gain. There were 11 input variables in total. Ten of these were derived from bedside monitor trends heart rate, oxygen saturation, respiratory rate, temperature, blood pressure, and urine collections, as well as infusion products and transfusions. The final input variable was a previous value from one of the eight lab tests being predicted: calcium, PTT, hematocrit, fibrinogen, lactate, platelets, INR and hemoglobin. The outcome for each test was a binary framework defining whether a test result contributed information gain or not.

PATIENTS

Predictive modeling was applied to recognize unnecessary lab tests in a real world ICU database extract comprising 746 patients with gastrointestinal bleeding.

MAIN RESULTS

Classification accuracy of necessary and unnecessary lab tests of greater than 80% was achieved for all eight lab tests. Sensitivity and specificity were satisfactory for all the outcomes. An average reduction of 50% of the lab tests was obtained. This is an improvement from previously reported similar studies with average performance 37% by [1-3].

CONCLUSIONS

Reducing frequent lab testing and the potential clinical and financial implications are an important issue in intensive care. In this work we present an artificial intelligence method to predict the benefit of proposed future laboratory tests. Using ICU data from 746 patients with gastrointestinal bleeding, and eleven measurements, we demonstrate high accuracy in predicting the likely information to be gained from proposed future lab testing for eight common GI related lab tests. Future work will explore applications of this approach to a range of underlying medical conditions and laboratory tests.

An automatic critical care urine meter

Nowadays patients admitted to critical care units have most of their physiological parameters measured automatically by sophisticated commercial monitoring devices. More often than not, these devices supervise whether the values of the parameters they measure lie within a pre-established range, and issue warning of deviations from this range by triggering alarms. The automation of measuring and supervising tasks not only discharges the healthcare staff of a considerable workload but also avoids human errors in these repetitive and monotonous tasks. Arguably, the most relevant physiological parameter that is still measured and supervised manually by critical care unit staff is urine output (UO). In this paper we present a patent-pending device that provides continuous and accurate measurements of patient's UO. The device uses capacitive sensors to take continuous measurements of the height of the column of liquid accumulated in two chambers that make up a plastic container. The first chamber, where the urine inputs, has a small volume. Once it has been filled it overflows into a second bigger chamber. The first chamber provides accurate UO measures of patients whose UO has to be closely supervised, while the second one avoids the need for frequent interventions by the nursing staff to empty the container.

Anaesthesia monitor alarms: a theory-driven approach

The development of physiologic monitors has contributed to the decline in morbidity and mortality in patients undergoing anaesthesia. Diverse factors (physiologic, technical, historical and medico-legal) create challenges for monitor alarm designers. Indeed, a growing body of literature suggests that alarms function sub-optimally in supporting the human operator. Despite existing technology that could allow more appropriate design, most anaesthesia alarms still operate on simple, pre-set thresholds. Arguing that more alarms do not necessarily make for safer alarms is difficult in a litigious medico-legal environment and a competitive marketplace. The resultant commitment to the status quo exposes the risks that a lack of an evidence-based theoretical framework for anaesthesia alarm design presents. In this review, two specific theoretical foundations with relevance to anaesthesia alarms are summarised. The potential significance that signal detection theory and cognitive systems engineering could have in improving anaesthesia alarm design is outlined and future research directions are suggested.

PRACTITIONER SUMMARY

The development of physiologic monitors has increased safety for patients undergoing anaesthesia. Evidence suggests that the full potential of the alarms embedded within those monitors is not being realised. In this review article, the authors propose a theoretical framework that could lead to the development of more ergonomic anaesthesia alarms.

Signal quality and data fusion for false alarm reduction in the intensive care unit

Due to a lack of integration between different sensors, false alarms (FA) in the intensive care unit (ICU) are frequent and can lead to reduced standard of care. We present a novel framework for FA reduction using a machine learning approach to combine up to 114 signal quality and physiological features extracted from the electrocardiogram, photoplethysmograph, and optionally the arterial blood pressure waveform. A machine learning algorithm was trained and evaluated on a database of 4107 expert-labeled life-threatening arrhythmias, from 182 separate ICU visits. On the independent test data, FA suppression results with no true alarm (TA) suppression were 86.4% for asystole, 100% for extreme bradycardia and 27.8% for extreme tachycardia. For the ventricular tachycardia alarms, the best FA suppression performance was 30.5% with a TA suppression rate below 1%. To reduce the TA suppression rate to zero, a reduction in FA suppression performance to 19.7% was required.

Monitor alarm fatigue: an integrative review

Alarm fatigue is a national problem and the number one medical device technology hazard in 2012. The problem of alarm desensitization is multifaceted and related to a high false alarm rate, poor positive predictive value, lack of alarm standardization, and the number of alarming medical devices in hospitals today. This integrative review synthesizes research and non-research findings published between 1/1/2000 and 10/1/2011 using The Johns Hopkins Nursing Evidence-Based Practice model. Seventy-two articles were included. Research evidence was organized into five main themes: excessive alarms and effects on staff; nurse's response to alarms; alarm sounds and audibility; technology to reduce false alarms; and alarm notification systems. Non-research evidence was divided into two main themes: strategies to reduce alarm desensitization, and alarm priority and notification systems. Evidence-based practice recommendations and gaps in research are summarized.

Anaesthesia monitoring using fuzzy logic

OBJECTIVE

Humans have a limited ability to accurately and continuously analyse large amount of data. In recent times, there has been a rapid growth in patient monitoring and medical data analysis using smart monitoring systems. Fuzzy logic-based expert systems, which can mimic human thought processes in complex circumstances, have indicated potential to improve clinicians' performance and accurately execute repetitive tasks to which humans are ill-suited. The main goal of this study is to develop a clinically useful diagnostic alarm system based on fuzzy logic for detecting critical events during anaesthesia administration.

METHOD

The proposed diagnostic alarm system called fuzzy logic monitoring system (FLMS) is presented. New diagnostic rules and membership functions (MFs) are developed. In addition, fuzzy inference system (FIS), adaptive neuro fuzzy inference system (ANFIS), and clustering techniques are explored for developing the FLMS' diagnostic modules. The performance of FLMS which is based on fuzzy logic expert diagnostic systems is validated through a series of off-line tests. The training and testing data set are selected randomly from 30 sets of patients' data.

RESULTS

The accuracy of diagnoses generated by the FLMS was validated by comparing the diagnostic information with the one provided by an anaesthetist for each patient. Kappa-analysis was used for measuring the level of agreement between the anaesthetist's and FLMS's diagnoses. When detecting hypovolaemia, a substantial level of agreement was observed between FLMS and the human expert (the anaesthetist) during surgical procedures.

CONCLUSION

The diagnostic alarm system FLMS demonstrated that evidence-based expert diagnostic systems can diagnose hypovolaemia, with a substantial degree of accuracy, in anaesthetized patients and could be useful in delivering decision support to anaesthetists.

A low cost device for monitoring the urine output of critical care patients

In critical care units most of the patients' physiological parameters are sensed by commercial monitoring devices. These devices can also supervise whether the values of the parameters lie within a pre-established range set by the clinician. The automation of the sensing and supervision tasks has discharged the healthcare staff of a considerable workload and avoids human errors, which are common in repetitive and monotonous tasks. Urine output is very likely the most relevant physiological parameter that has yet to be sensed or supervised automatically. This paper presents a low cost patent-pending device capable of sensing and supervising urine output. The device uses reed switches activated by a magnetic float in order to measure the amount of urine collected in two containers which are arranged in cascade. When either of the containers fills, it is emptied automatically using a siphon mechanism and urine begins to collect again. An electronic unit sends the state of the reed switches via Bluetooth to a PC that calculates the urine output from this information and supervises the achievement of therapeutic goals.

A device for automatically measuring and supervising the critical care patient's urine output

Critical care units are equipped with commercial monitoring devices capable of sensing patients’ physiological parameters and supervising the achievement of the established therapeutic goals. This avoids human errors in this task and considerably decreases the workload of the healthcare staff. However, at present there still is a very relevant physiological parameter that is measured and supervised manually by the critical care units’ healthcare staff: urine output. This paper presents a patent-pending device capable of automatically recording and supervising the urine output of a critical care patient. A high precision scale is used to measure the weight of a commercial urine meter. On the scale’s pan there is a support frame made up of Bosch profiles that isolates the scale from force transmission from the patient’s bed, and guarantees that the urine flows properly through the urine meter input tube. The scale’s readings are sent to a PC via Bluetooth where an application supervises the achievement of the therapeutic goals. The device is currently undergoing tests at a research unit associated with the University Hospital of Getafe in Spain.