Reducing unnecessary lab testing in the ICU with artificial intelligence

Evaluation of the impact of diagnostic blood loss and red blood cell transfusion in very-low-birth-weight anaemic neonates during hospitalization: A multi-centre retrospective clinical study

BACKGROUND AND OBJECTIVES

Diagnostic blood loss is a significant factor in the development of anaemia in neonates with very low birth weight. This study aimed to assess the clinical efficacy of intervention approaches involving varying diagnostic blood loss and red blood cell transfusion volumes in neonates with very low birth weights experiencing anaemia during hospitalization.

MATERIALS AND METHODS

A total of 785 newborns with anaemia weighing less than 1500 g were enrolled from 32 hospitals in China. The study involved monitoring diagnostic blood loss and red blood cell transfusion and evaluating relevant interventions such as red blood cell transfusion and clinical outcomes. Three intervention approaches were established based on the difference between blood loss and transfusion (Intervention Approaches 0, 1 and 2). The primary outcomes measured were unsatisfactory weight gain during hospitalization and neonatal mortality. The secondary outcomes included related complications.

RESULTS

In the non-hospital-acquired anaemia group, Intervention Approach 2 had the highest incidence of below-normal weight gain (odds ratio [OR]: 3.019, 95% confidence interval [CI]: 1.081-8.431, p = 0.035). Multivariate analysis revealed that Intervention Approach 1 had a protective effect on weight gain. In the hospital-acquired anaemia group, Intervention Approach 2 had the highest incidence of below-normal weight gain (OR: 3.335, 95% CI: 1.785-6.234, p = 0.000) and mortality (OR: 5.341, 95% CI: 2.449-11.645, p = 0.000), while Intervention Approach 1 had the lowest incidence of intraventricular haemorrhage. Intervention Approach 1 demonstrated favourable outcomes in both anaemia groups.

CONCLUSION

Intervention Approach 1 improved weight gain and reduced mortality and complications in both the non-hospital-acquired and hospital-acquired anaemia groups.

Interventions to improve appropriateness of laboratory testing in the intensive care unit: a narrative review

Healthcare expenses are increasing, as is the utilization of laboratory resources. Despite this, between 20% and 40% of requested tests are deemed inappropriate. Improper use of laboratory resources leads to unwanted consequences such as hospital-acquired anemia, infections, increased costs, staff workload and patient stress and discomfort. The most unfavorable consequences result from unnecessary follow-up tests and treatments (overuse) and missed or delayed diagnoses (underuse). In this context, several interventions have been carried out to improve the appropriateness of laboratory testing. To date, there have been few published assessments of interventions specific to the intensive care unit. We reviewed the literature for interventions implemented in the ICU to improve the appropriateness of laboratory testing. We searched literature from 2008 to 2023 in PubMed, Embase, Scopus, and Google Scholar databases between April and June 2023. Five intervention categories were identified: education and guidance (E&G), audit and feedback, gatekeeping, computerized physician order entry (including reshaping of ordering panels), and multifaceted interventions (MFI). We included a sixth category exploring the potential role of artificial intelligence and machine learning (AI/ML)-based assisting tools in such interventions. E&G-based interventions and MFI are the most frequently used approaches. MFI is the most effective type of intervention, and shows the strongest persistence of effect over time. AI/ML-based tools may offer valuable assistance to the improvement of appropriate laboratory testing in the near future. Patient safety outcomes are not impaired by interventions to reduce inappropriate testing. The literature focuses mainly on reducing overuse of laboratory tests, with only one intervention mentioning underuse. We highlight an overall poor quality of methodological design and reporting and argue for standardization of intervention methods. Collaboration between clinicians and laboratory staff is key to improve appropriate laboratory utilization. This article offers practical guidance for optimizing the effectiveness of an intervention protocol designed to limit inappropriate use of laboratory resources.

The Evolution and Future of Intensive Care Management in the Era of Telecritical Care and Artificial Intelligence

Critical care practice has been embodied in the healthcare system since the institutionalization of intensive care units (ICUs) in the late '50s. Over time, this sector has experienced many changes and improvements in providing immediate and dedicated healthcare as patients requiring intensive care are often frail and critically ill with high mortality and morbidity rates. These changes were aided by innovations in diagnostic, therapeutic, and monitoring technologies, as well as the implementation of evidence-based guidelines and organizational structures within the ICU. In this review, we examine these changes in intensive care management over the past 40 years and their impact on the quality of care available to patients. Moreover, the current state of intensive care management is characterized by a multidisciplinary approach and the use of innovative technologies and research databases. Advancements such as telecritical care and artificial intelligence are being increasingly explored, especially since the COVID-19 pandemic, to reduce the length of hospitalization and ICU mortality. With these advancements in intensive care and ever-changing patient needs, critical care experts, hospital managers, and policymakers must also explore appropriate organizational structures and future enhancements within the ICU.

Barriers and facilitators to utilizing digital health technologies by healthcare professionals

Digital technologies change the healthcare environment, with several studies suggesting barriers and facilitators to using digital interventions by healthcare professionals (HPs). We consolidated the evidence from existing systematic reviews mentioning barriers and facilitators for the use of digital health technologies by HP. Electronic searches were performed in five databases (Cochrane Database of Systematic Reviews, Embase®, Epistemonikos, MEDLINE®, and Scopus) from inception to March 2023. We included reviews that reported barriers or facilitators factors to use technology solutions among HP. We performed data abstraction, methodological assessment, and certainty of the evidence appraisal by at least two authors. Overall, we included 108 reviews involving physicians, pharmacists, and nurses were included. High-quality evidence suggested that infrastructure and technical barriers (Relative Frequency Occurrence [RFO] 6.4% [95% CI 2.9-14.1]), psychological and personal issues (RFO 5.3% [95% CI 2.2-12.7]), and concerns of increasing working hours or workload (RFO 3.9% [95% CI 1.5-10.1]) were common concerns reported by HPs. Likewise, high-quality evidence supports that training/educational programs, multisector incentives, and the perception of technology effectiveness facilitate the adoption of digital technologies by HPs (RFO 3.8% [95% CI 1.8-7.9]). Our findings showed that infrastructure and technical issues, psychological barriers, and workload-related concerns are relevant barriers to comprehensively and holistically adopting digital health technologies by HPs. Conversely, deploying training, evaluating HP's perception of usefulness and willingness to use, and multi-stakeholders incentives are vital enablers to enhance the HP adoption of digital interventions.

Routine Versus On-Demand Blood Sampling in Critically Ill Patients: A Systematic Review

OBJECTIVES

We aimed to provide an overview of the current evidence on routine versus on-demand blood sampling in critical care. We assessed the reported proportion of patients exposed to daily routine blood sampling, the tests performed, characteristics associated with more frequent blood sampling, and the reported benefits and harms of routine blood sampling compared with on-demand sampling.

DATA SOURCES

We systematically searched the Cochrane Library, the Excerpta Medica Database, and the Medical Literature Analysis and Retrieval System Online for studies assessing routine versus on-demand blood testing in critically ill patients from inception to September 2022.

STUDY SELECTION

Abstracts and full texts were assessed independently and in duplicate by two reviewers.

STUDY EXTRACTION

Data were extracted independently and in duplicate by two reviewers using predefined extraction forms.

DATA SYNTHESIS

Of 12,212 records screened, 298 full-text articles were assessed for eligibility. We included 70 studies; 50 nonrandomized interventional studies and 20 observational studies. Exposure to routine blood testing was 52-100% (very low certainty of evidence). Blood testing seemed to occur more frequently in medical intensive care settings with a median of 18 blood tests per patient day (interquartile range, 10-33) (very low certainty of evidence). Mixed biochemistry seemed to be the most frequently performed blood tests across all settings (five tests per patient day; interquartile range, 2-10) (very low certainty of evidence). Reductions in routine blood testing seemed to be associated with reduced transfusion rates and costs without apparent adverse patient outcomes (low certainty of evidence).

CONCLUSIONS

In this systematic review, routine blood testing in critically ill patients was common and varied considerably. A reduction in routine blood testing appeared to be associated with reduced transfusion rates and costs without adverse effects, but the evidence was very uncertain.

Confidence-based laboratory test reduction recommendation algorithm

BACKGROUND

We propose a new deep learning model to identify unnecessary hemoglobin (Hgb) tests for patients admitted to the hospital, which can help reduce health risks and healthcare costs.

METHODS

We collected internal patient data from a teaching hospital in Houston and external patient data from the MIMIC III database. The study used a conservative definition of unnecessary laboratory tests, which was defined as stable (i.e., stability) and below the lower normal bound (i.e., normality). Considering that machine learning models may yield less reliable results when trained on noisy inputs containing low-quality information, we estimated prediction confidence to assess the reliability of predicted outcomes. We adopted a "select and predict" design philosophy to maximize prediction performance by selectively considering samples with high prediction confidence for recommendations. Our model accommodated irregularly sampled observational data to make full use of variable correlations (i.e., with other laboratory test values) and temporal dependencies (i.e., previous laboratory tests performed within the same encounter) in selecting candidates for training and prediction.

RESULTS

The proposed model demonstrated remarkable Hgb prediction performance, achieving a normality AUC of 95.89% and a Hgb stability AUC of 95.94%, while recommending a reduction of 9.91% of Hgb tests that were deemed unnecessary. Additionally, the model could generalize well to external patients admitted to another hospital.

CONCLUSIONS

This study introduces a novel deep learning model with the potential to significantly reduce healthcare costs and improve patient outcomes by identifying unnecessary laboratory tests for hospitalized patients.

Prediction of ICU Patients' Deterioration Using Machine Learning Techniques

INTRODUCTION

Assessing vital sign measurements within hospital settings presents a valuable opportunity for data analysis and knowledge extraction. By generating adaptable, personalized prediction models of patient vital signs, these models can yield clinically relevant insights not achievable through population-based models. This study aims to compare several statistical forecasting models to determine their real-life applicability.

OBJECTIVES

The primary objectives of this paper are to evaluate whether the following measurements: blood pressure, oxygen saturation, temperature and heart rate can predict deterioration in Intensive Care Unit (ICU) patients. Additionally, we aim to identify which of these measurements contributes most significantly to our prediction. Lastly, we seek to determine the most accurate data mining technique for real-life data applications.

METHODS

This retrospective chart review study utilized data from patients admitted to the ICU at a tertiary hospital between January and December, 2019. Data mining techniques for prediction included logistic regression, support vector machine classifier, k-nearest neighbors (KNN), gradient boosting classifier, and Naive Bayes classifier. A comprehensive comparison of these techniques was performed, focusing on accuracy, precision, recall, and F-measure.

RESULTS

To achieve the research objectives, the SelectKBest class was applied to extract the most contributory features for prediction. Blood pressure ranked first with a score of 9.98, followed by respiratory rate, temperature, and heart rate. Analysis of 653 patient records indicated that 129 patients expired, while 542 patients were discharged either to their homes or other facilities. Among the five training models, two demonstrated the highest accuracy in predicting patient deterioration or survival at 88.83% and 84.72%, respectively. The gradient boosting classifier accurately predicted 115 out of 129 expired patients, while the KNN correctly predicted 109 out of 129 expired patients.

CONCLUSION

Machine learning has the potential to enhance clinical deterioration prediction compared to traditional methods. This allows healthcare professionals to implement preventative measures and improve patients' quality of life, ultimately increasing average life expectancy. Although our research focused exclusively on ICU patients, data mining techniques can be applied in various contexts both within and outside the hospital setting.

Artificial Intelligence and Machine Learning in Patient Blood Management: A Scoping Review

Machine learning (ML) and artificial intelligence (AI) are widely used in many different fields of modern medicine. This narrative review gives, in the first part, a brief overview of the methods of ML and AI used in patient blood management (PBM) and, in the second part, aims at describing which fields have been analyzed using these methods so far. A total of 442 articles were identified by a literature search, and 47 of them were judged as qualified articles that applied ML and AI techniques in PBM. We assembled the eligible articles to provide insights into the areas of application, quality measures of these studies, and treatment outcomes that can pave the way for further adoption of this promising technology and its possible use in routine clinical decision making. The topics that have been investigated most often were the prediction of transfusion (30%), bleeding (28%), and laboratory studies (15%). Although in the last 3 years a constantly increasing number of questions of ML in PBM have been investigated, there is a vast scientific potential for further application of ML and AI in other fields of PBM.

External Validation of a Laboratory Prediction Algorithm for the Reduction of Unnecessary Labs in the Critical Care Setting

BACKGROUND

Unnecessary laboratory tests contribute to iatrogenic harm and are a major source of waste in the health care system. We previously developed a machine learning algorithm to help clinicians identify unnecessary laboratory tests, but it has not been externally validated. In this study, we externally validate our machine learning algorithm.

METHODS

To externally validate the machine learning algorithm that was originally trained on the Medical Information Mart for Intensive Care (MIMIC) III database, we tested the algorithm in a separate institution. We identified and abstracted data for all patients older than 18 years admitted to the intensive care unit at Memorial Hermann Hospital in Houston, Texas (MHH) from January 1, 2020 to November 13, 2020. Using the transfer learning style, we performed external validation of the machine learning algorithm.

RESULTS

A total of 651 MHH patients were included. The model performed well in predicting abnormality (area under the curve [AUC] 0.98 for MIMIC III and 0.89 for MHH). The model performed similarly in predicting transitions from normal laboratory range to abnormal (AUC 0.71 for MIMIC III and 0.70 for MHH). The performance of the model in predicting the actual laboratory value was also similar in the MIMIC III (accuracy 0.41) and MHH data (0.45).

CONCLUSIONS

We externally validated the machine learning model and showed that the model performed similarly, supporting the generalizability to other settings. While this model demonstrated good performance for predicting abnormal labs and transitions, it does not perform well enough for prediction of laboratory values in most clinical applications.

State of the Art of Machine Learning-Enabled Clinical Decision Support in Intensive Care Units: Literature Review

Background

Modern clinical care in intensive care units is full of rich data, and machine learning has great potential to support clinical decision-making. The development of intelligent machine learning–based clinical decision support systems is facing great opportunities and challenges. Clinical decision support systems may directly help clinicians accurately diagnose, predict outcomes, identify risk events, or decide treatments at the point of care.

Objective

We aimed to review the research and application of machine learning–enabled clinical decision support studies in intensive care units to help clinicians, researchers, developers, and policy makers better understand the advantages and limitations of machine learning–supported diagnosis, outcome prediction, risk event identification, and intensive care unit point-of-care recommendations.

Methods

We searched papers published in the PubMed database between January 1980 and October 2020. We defined selection criteria to identify papers that focused on machine learning–enabled clinical decision support studies in intensive care units and reviewed the following aspects: research topics, study cohorts, machine learning models, analysis variables, and evaluation metrics.

Results

A total of 643 papers were collected, and using our selection criteria, 97 studies were found. Studies were categorized into 4 topics—monitoring, detection, and diagnosis (13/97, 13.4%), early identification of clinical events (32/97, 33.0%), outcome prediction and prognosis assessment (46/97, 47.6%), and treatment decision (6/97, 6.2%). Of the 97 papers, 82 (84.5%) studies used data from adult patients, 9 (9.3%) studies used data from pediatric patients, and 6 (6.2%) studies used data from neonates. We found that 65 (67.0%) studies used data from a single center, and 32 (33.0%) studies used a multicenter data set; 88 (90.7%) studies used supervised learning, 3 (3.1%) studies used unsupervised learning, and 6 (6.2%) studies used reinforcement learning. Clinical variable categories, starting with the most frequently used, were demographic (n=74), laboratory values (n=59), vital signs (n=55), scores (n=48), ventilation parameters (n=43), comorbidities (n=27), medications (n=18), outcome (n=14), fluid balance (n=13), nonmedicine therapy (n=10), symptoms (n=7), and medical history (n=4). The most frequently adopted evaluation metrics for clinical data modeling studies included area under the receiver operating characteristic curve (n=61), sensitivity (n=51), specificity (n=41), accuracy (n=29), and positive predictive value (n=23).

Conclusions

Early identification of clinical and outcome prediction and prognosis assessment contributed to approximately 80% of studies included in this review. Using new algorithms to solve intensive care unit clinical problems by developing reinforcement learning, active learning, and time-series analysis methods for clinical decision support will be greater development prospects in the future.

Predicting abnormal laboratory blood test results in the intensive care unit using novel features based on information theory and historical conditional probability: Observational Study (Preprint)

Background

Redundancy in laboratory blood tests is common in intensive care units (ICUs), affecting patients’ health and increasing health care expenses. Medical communities have made recommendations to order laboratory tests more judiciously. Wise selection can rely on modern data-driven approaches that have been shown to help identify low-yield laboratory blood tests in ICUs. However, although conditional entropy and conditional probability distribution have shown the potential to measure the uncertainty of yielding an abnormal test, no previous studies have adapted these techniques to include them in machine learning models for predicting abnormal laboratory test results.

Objective

This study aimed to address the limitations of previous reports by adapting conditional entropy and conditional probability to extract features for predicting abnormal laboratory blood test results.

Methods

We used an ICU data set collected across Alberta, Canada, which included 55,689 ICU admissions from 48,672 patients. We investigated the features of conditional entropy and conditional probability by comparing the performances of 2 machine learning approaches for predicting normal and abnormal results for 18 blood laboratory tests. Approach 1 used patients’ vitals, age, sex, and admission diagnosis as features. Approach 2 used the same features plus the new conditional entropy–based and conditional probability–based features. Both approaches used 4 different machine learning models (fuzzy model, logistic regression, random forest, and gradient boosting trees) and 10 metrics (sensitivity, specificity, accuracy, precision, negative predictive value [NPV], F₁ score, area under the curve [AUC], precision-recall AUC, mean G, and index balanced accuracy) to assess the performance of the approaches.

Results

Approach 1 achieved an average AUC of 0.86 for all 18 laboratory tests across the 4 models (sensitivity 78%, specificity 84%, precision 82%, NPV 75%, F₁ score 79%, and mean G 81%), whereas approach 2 achieved an average AUC of 0.89 (sensitivity 84%, specificity 84%, precision 83%, NPV 81%, F₁ score 83%, and mean G 84%). We found that the inclusion of the new features resulted in significant differences for most of the metrics in favor of approach 2. Sensitivity significantly improved for 8 and 15 laboratory tests across the different classifiers (minimum P<.001 and maximum P=.04). Mean G and index balanced accuracy, which are balanced performance metrics, also improved significantly across the classifiers for 6 to 10 and 6 to 11 laboratory tests. The most relevant feature was the pretest probability feature, which is the probability that a test result was normal when a certain number of consecutive prior tests was already normal.

Conclusions

The findings suggest that conditional entropy–based features and pretest probability improve the capacity to discriminate between normal and abnormal laboratory test results. Detecting the next laboratory test result is an intermediate step toward developing guidelines for reducing overtesting in the ICU.

How well do critical care audit and feedback interventions adhere to best practice? Development and application of the REFLECT-52 evaluation tool

BACKGROUND

Healthcare Audit and Feedback (A&F) interventions have been shown to be an effective means of changing healthcare professional behavior, but work is required to optimize them, as evidence suggests that A&F interventions are not improving over time. Recent published guidance has suggested an initial set of best practices that may help to increase intervention effectiveness, which focus on the "Nature of the desired action," "Nature of the data available for feedback," "Feedback display," and "Delivering the feedback intervention." We aimed to develop a generalizable evaluation tool that can be used to assess whether A&F interventions conform to these suggestions for best practice and conducted initial testing of the tool through application to a sample of critical care A&F interventions.

METHODS

We used a consensus-based approach to develop an evaluation tool from published guidance and subsequently applied the tool to conduct a secondary analysis of A&F interventions. To start, the 15 suggestions for improved feedback interventions published by Brehaut et al. were deconstructed into rateable items. Items were developed through iterative consensus meetings among researchers. These items were then piloted on 12 A&F studies (two reviewers met for consensus each time after independently applying the tool to four A&F intervention studies). After each consensus meeting, items were modified to improve clarity and specificity, and to help increase the reliability between coders. We then assessed the conformity to best practices of 17 critical care A&F interventions, sourced from a systematic review of A&F interventions on provider ordering of laboratory tests and transfusions in the critical care setting. Data for each criteria item was extracted by one coder and confirmed by a second; results were then aggregated and presented graphically or in a table and described narratively.

RESULTS

In total, 52 criteria items were developed (38 ratable items and 14 descriptive items). Eight studies targeted lab test ordering behaviors, and 10 studies targeted blood transfusion ordering. Items focused on specifying the "Nature of the Desired Action" were adhered to most commonly-feedback was often presented in the context of an external priority (13/17), showed or described a discrepancy in performance (14/17), and in all cases it was reasonable for the recipients to be responsible for the change in behavior (17/17). Items focused on the "Nature of the Data Available for Feedback" were adhered to less often-only some interventions provided individual (5/17) or patient-level data (5/17), and few included aspirational comparators (2/17), or justifications for specificity of feedback (4/17), choice of comparator (0/9) or the interval between reports (3/13). Items focused on the "Nature of the Feedback Display" were reported poorly-just under half of interventions reported providing feedback in more than one way (8/17) and interventions rarely included pilot-testing of the feedback (1/17 unclear) or presentation of a visual display and summary message in close proximity of each other (1/13). Items focused on "Delivering the Feedback Intervention" were also poorly reported-feedback rarely reported use of barrier/enabler assessments (0/17), involved target members in the development of the feedback (0/17), or involved explicit design to be received and discussed in a social context (3/17); however, most interventions clearly indicated who was providing the feedback (11/17), involved a facilitator (8/12) or involved engaging in self-assessment around the target behavior prior to receipt of feedback (12/17).

CONCLUSIONS

Many of the theory-informed best practice items were not consistently applied in critical care and can suggest clear ways to improve interventions. Standardized reporting of detailed intervention descriptions and feedback templates may also help to further advance research in this field. The 52-item tool can serve as a basis for reliably assessing concordance with best practice guidance in existing A&F interventions trialed in other healthcare settings, and could be used to inform future A&F intervention development.

TRIAL REGISTRATION

Not applicable.

Prediction of blood lactate values in critically ill patients: a retrospective multi-center cohort study

Elevations in initially obtained serum lactate levels are strong predictors of mortality in critically ill patients. Identifying patients whose serum lactate levels are more likely to increase can alert physicians to intensify care and guide them in the frequency of tending the blood test. We investigate whether machine learning models can predict subsequent serum lactate changes. We investigated serum lactate change prediction using the MIMIC-III and eICU-CRD datasets in internal as well as external validation of the eICU cohort on the MIMIC-III cohort. Three subgroups were defined based on the initial lactate levels: (i) normal group (< 2 mmol/L), (ii) mild group (2-4 mmol/L), and (iii) severe group (> 4 mmol/L). Outcomes were defined based on increase or decrease of serum lactate levels between the groups. We also performed sensitivity analysis by defining the outcome as lactate change of > 10% and furthermore investigated the influence of the time interval between subsequent lactate measurements on predictive performance. The LSTM models were able to predict deterioration of serum lactate values of MIMIC-III patients with an AUC of 0.77 (95% CI 0.762-0.771) for the normal group, 0.77 (95% CI 0.768-0.772) for the mild group, and 0.85 (95% CI 0.840-0.851) for the severe group, with only a slightly lower performance in the external validation. The LSTM demonstrated good discrimination of patients who had deterioration in serum lactate levels. Clinical studies are needed to evaluate whether utilization of a clinical decision support tool based on these results could positively impact decision-making and patient outcomes.

Clinical decision-making framework against over-testing based on modeling implicit evaluation criteria

Different statistical methods include various subjective criteria that can prevent over-testing. However, no unified framework that defines generalized objective criteria for various diseases is available to determine the appropriateness of diagnostic tests recommended by doctors. We present the clinical decision-making framework against over-testing based on modeling the implicit evaluation criteria (CDFO-MIEC). The CDFO-MIEC quantifies the subjective evaluation process using statistics-based methods to identify over-testing. Furthermore, it determines the test's appropriateness with extracted entities obtained via named entity recognition and entity alignment. More specifically, implicit evaluation criteria are defined-namely, the correlation among the diagnostic tests, symptoms, and diseases, confirmation function, and exclusion function. Additionally, four evaluation strategies are implemented by applying statistical methods, including the multi-label k-nearest neighbor and the conditional probability algorithms, to model the implicit evaluation criteria. Finally, they are combined into a classification and regression tree to make the final decision. The CDFO-MIEC also provides interpretability by decision conditions for supporting each clinical decision of over-testing. We tested the CDFO-MIEC on 2,860 clinical texts obtained from a single respiratory medicine department in China with the appropriate confirmation by physicians. The dataset was supplemented with random inappropriate tests. The proposed framework excelled against the best competing text classification methods with a Mean_F₁ of 0.9167. This determined whether the appropriate and inappropriate tests were properly classified. The four evaluation strategies captured the features effectively, and they were imperative. Therefore, the proposed CDFO-MIEC is feasible because it exhibits high performance and can prevent over-testing.

Iatrogenic anaemia and transfusion thresholds in ICU patients with COVID-19 disease at a tertiary care hospital

BACKGROUND

Patients with severe COVID-19 disease frequently develop anaemia as the result of multiple mechanisms and often receive transfusions. The aims of this study were to assess the impact of repeated blood samplings on patients' anaemic state using standard-volume tubes, in comparison with the hypothetical use of low-volume tubes and to evaluate the transfusion policy adopted.

STUDY DESIGN AND METHODS

Transfusion data of mechanically ventilated non-bleeding patients with COVID-19 disease hospitalized in ICU for a minimum of 20 days were recorded. The total volume of blood drawn for samplings with standard-volume tubes and the corresponding red blood cell mass (RBCM) removed during hospitalization for each patient were calculated and compared with the hypothetical use of low-volume tubes.

RESULTS

Twenty-four patients fulfilled the inclusion criteria. Ten patients were anaemic at ICU admission (41.7 %). Overall, 6658 sampling tubes were employed, for a total of 16,786 mL of blood. The median RBCM subtracted by blood samplings per patient accounted for about one third of the total patients' RBCM decrease until discharge. The use of low-volume tubes would have led to a median saving of about one third of the drawn RBCM. Eleven patients were transfused (45.8 %) at a mean Hb value of 7.7 (± 0.5) g/dL.

CONCLUSION

The amount of blood drawn for sampling has a significant role in the development of anaemia and the use of low-volume tubes could minimize the problem. Large high-powered studies are warranted to assess the more appropriate transfusion thresholds in non-bleeding critically ill patients with COVID-19 disease.

Artificial intelligence-based prediction of transfusion in the intensive care unit in patients with gastrointestinal bleeding

Objective

Gastrointestinal (GI) bleeding commonly requires intensive care unit (ICU) in cases of potentialhaemodynamiccompromise or likely urgent intervention. However, manypatientsadmitted to the ICU stop bleeding and do not require further intervention, including blood transfusion. The present work proposes an artificial intelligence (AI) solution for the prediction of rebleeding in patients with GI bleeding admitted to ICU.

Methods

A machine learning algorithm was trained and tested using two publicly available ICU databases, the Medical Information Mart for Intensive Care V.1.4 database and eICU Collaborative Research Database using freedom from transfusion as a proxy for patients who potentially did not require ICU-level care. Multiple initial observation time frames were explored using readily available data including labs, demographics and clinical parameters for a total of 20 covariates.

Results

The optimal model used a 5-hour observation period to achieve an area under the curve of the receiving operating curve (ROC-AUC) of greater than 0.80. The model was robust when tested against both ICU databases with a similar ROC-AUC for all.

Conclusions

The potential disruptive impact of AI in healthcare innovation is acknowledge, but awareness of AI-related risk on healthcare applications and current limitations should be considered before implementation and deployment. The proposed algorithm is not meant to replace but to inform clinical decision making. Prospective clinical trial validation as a triage tool is warranted.

The Value of Artificial Intelligence in Laboratory Medicine

OBJECTIVES

As laboratory medicine continues to undergo digitalization and automation, clinical laboratorians will likely be confronted with the challenges associated with artificial intelligence (AI). Understanding what AI is good for, how to evaluate it, what are its limitations, and how it can be implemented are not well understood. With a survey, we aimed to evaluate the thoughts of stakeholders in laboratory medicine on the value of AI in the diagnostics space and identify anticipated challenges and solutions to introducing AI.

METHODS

We conducted a web-based survey on the use of AI with participants from Roche's Strategic Advisory Network that included key stakeholders in laboratory medicine.

RESULTS

In total, 128 of 302 stakeholders responded to the survey. Most of the participants were medical practitioners (26%) or laboratory managers (22%). AI is currently used in the organizations of 15.6%, while 66.4% felt they might use it in the future. Most had an unsure attitude on what they would need to adopt AI in the diagnostics space. High investment costs, lack of proven clinical benefits, number of decision makers, and privacy concerns were identified as barriers to adoption. Education in the value of AI, streamlined implementation and integration into existing workflows, and research to prove clinical utility were identified as solutions needed to mainstream AI in laboratory medicine.

CONCLUSIONS

This survey demonstrates that specific knowledge of AI in the medical community is poor and that AI education is much needed. One strategy could be to implement new AI tools alongside existing tools.

A deep learning solution to recommend laboratory reduction strategies in ICU

OBJECTIVE

To build a machine-learning model that predicts laboratory test results and provides a promising lab test reduction strategy, using spatial-temporal correlations.

MATERIALS AND METHODS

We developed a global prediction model to treat laboratory testing as a series of decisions by considering contextual information over time and across modalities. We validated our method using a critical care database (MIMIC III), which includes 4,570,709 observations of 12 standard laboratory tests, among 38,773 critical care patients. Our deep-learning model made real-time laboratory reduction recommendations and predicted the properties of lab tests, including values, normal/abnormal (whether labs were within the normal range) and transition (normal to abnormal or abnormal to normal from the latest lab test). We reported area under the receiver operating characteristic curve (AUC) for predicting normal/abnormal, evaluated accuracy and absolute bias on prediction vs. observation against lab test reduction proportion. We compared our model against baseline models and analyzed the impact of variations on the recommended reduction strategy.

RESULTS

Our best model offered a 20.26 % reduction in the number of laboratory tests. By applying the recommended reduction policy on the hold-out dataset (7755 patients), our model predicted normality/abnormality of laboratory tests with a 98.27 % accuracy (AUC, 0.9885; sensitivity, 97.84 %; specificity, 98.80 %; PPV, 99.01 %; NPV, 97.39 %) on 20.26 % reduced lab tests, and recommended 98.10 % of transitions to be checked. Our model performed better than the greedy models, and the recommended reduction strategy was robust.

DISCUSSION

Strong spatial and temporal correlations between laboratory tests can be used to optimize policies for reducing laboratory tests throughout the hospital course. Our method allows for iterative predictions and provides a superior solution for the dynamic decision-making laboratory reduction problem.

CONCLUSION

This work demonstrates a machine-learning model that assists physicians in determining which laboratory tests may be omitted.

Automatic generation of minimum dataset and quality indicators from data collected routinely by the clinical information system in an intensive care unit

BACKGROUND

Quality indicators (QIs) are being increasingly used in medicine to compare and improve the quality of care delivered. The feasibility of data collection is an important prerequisite for QIs. Information technology can improve efforts to measure processes and outcomes. In intensive care units (ICU), QIs can be automatically measured by exploiting data from clinical information systems (CIS).

OBJECTIVE

To describe the development and application of a tool to automatically generate a minimum dataset (MDS) and a set of ICU quality metrics from CIS data.

METHODS

We used the definitions for MDS and QIs proposed by the Spanish Society of Critical Care Medicine and Coronary Units. Our tool uses an extraction, transform, and load process implemented with Python to extract data stored in various tables in the CIS database and create a new associative database. This new database is uploaded to Qlik Sense, which constructs the MDS and calculates the QIs by applying the required metrics. The tool was tested using data from patients attended in a 30-bed polyvalent ICU during a six-year period.

RESULTS

We describe the definitions and metrics, and we report the MDS and QI measurements obtained through the analysis of 4546 admissions. The results show that our ICU's performance on the QIs analyzed meets the standards proposed by our national scientific society.

CONCLUSIONS

This is the first step toward using a tool to automatically obtain a set of actionable QIs to monitor and improve the quality of care in ICUs, eliminating the need for professionals to enter data manually, thus saving time and ensuring data quality.

A systematic literature review on unnecessary diagnostic testing: The role of ICT use

BACKGROUND

The negative impact of unnecessary diagnostic tests on healthcare systems and patients has been widely recognized. Medical researchers in various countries have been devoting effort to reduce unnecessary diagnostic tests by using different types of interventions, including information and communications technology-based (ICT-based) intervention, educational intervention, audit and feedback, the introduction of guidelines or protocols, and the reward and punishment of staff. We conducted a review of ICT based interventions and a comparative analysis of their relative effectiveness in reducing unnecessary tests.

METHOD

A systematic Boolean search in PubMed, EMBase and EBSCOhost research databases was performed. Keyword search and citation analysis were also conducted. Empirical studies reporting ICT based interventions, and their implications on relative effectiveness in reducing unnecessary diagnostic tests (pathology tests or medical imaging) were evaluated independently by two reviewers based on a rigorously developed coding protocol.

RESULTS

92 research articles from peer-reviewed journals were identified as eligible. 47 studies involved a single-method intervention and 45 involved multi-method interventions. Regardless of the number of interventions involved in the studies, ICT-based interventions were utilized by 71 studies and 59 of them were shown to be effective in reducing unnecessary testing. A clinical decision support (CDS) tool appeared to be the most adopted ICT approach, with 46 out of 71 studies using CDS tools. The CDS tool showed effectiveness in reducing test volume in 38 studies and reducing cost in 24 studies.

CONCLUSIONS

This review investigated five frequently utilized intervention methods, ICT-based, education, introduction of guidelines or protocols, audit and feedback, and reward and punishment. It provides in-depth analysis of the efficacy of different types of interventions and sheds insights about the benefits of ICT based interventions, especially those utilising CDS tools, to reduce unnecessary diagnostic testing. The replicability of the studies is limited due to the heterogeneity of the studies in terms of context, study design, and targeted types of tests.

[Laboratory testing in intensive care medicine]

Despite the tremendous technological developments in medicine, careful history-taking and clinical examination remain the cornerstones of diagnostics. Numerous laboratory tests are ordered in intensive care and emergency medicine. The rate of overutilization of these tests during initial patient admission is almost 50%. Patient history may be frequently insufficient for conducting targeted laboratory testing, and concern about not overlooking a pathology also contributes to laboratory test overutilization. On the other hand, laboratory test profiles are frequently defined a priori to simplify the management process. However, these profiles are commonly based on symptoms rather than on a suspected diagnosis. Several laboratory variables are outside the normal range in critically ill patients. However, normal ranges are defined on the basis of data from healthy subjects, and these do not allow for a clear distinction between stress adaptation and clinically relevant changes that require correction. Pathophysiological changes due to the acute injury in critically ill patients and the reaction of the organism to the injury or even to the treatment itself can lead to changes in laboratory values. Untargeted laboratory tests contribute to iatrogenic anemia and increased costs. The results of such tests are either hardly noticed or, in the worst case, lead to further unnecessary diagnostic steps and unjustified therapeutic measures. Point-of-care laboratory tests, including blood gas analysis, blood count, serum electrolytes, and lactate, to assess the patient's homeostatic state and laboratory data for the relevant critical care scores are uniformly required. Beyond that, every laboratory test should be chosen wisely based on a concrete clinical question.

Audit and feedback to improve laboratory test and transfusion ordering in critical care: a systematic review

BACKGROUND

Laboratory tests and transfusions are sometimes ordered inappropriately, particularly in the critical care setting, which sees frequent use of both. Audit and Feedback (A&F) is a potentially useful intervention for modifying healthcare provider behaviors, but its application to the complex, team-based environment of critical care is not well understood. We conducted a systematic review of the literature on A&F interventions for improving test or transfusion ordering in the critical care setting.

METHODS

Five databases, two registries, and the bibliographies of relevant articles were searched. We included critical care studies that assessed the use of A&F targeting healthcare provider behaviors, alone or in combination with other interventions to improve test and transfusion ordering, as compared to historical practice, no intervention, or another healthcare behaviour change intervention. Studies were included only if they reported laboratory test or transfusion orders, or the appropriateness of orders, as outcomes. There were no restrictions based on study design, date of publication, or follow-up time. Intervention characteristics and absolute differences in outcomes were summarized. The quality of individual studies was assessed using a modified version of the Effective Practice and Organisation of Care Cochrane Review Group's criteria.

RESULTS

We identified 16 studies, including 13 uncontrolled before-after studies, one randomized controlled trial, one controlled before-after study, and one controlled clinical trial (quasi-experimental). These studies described 17 interventions, mostly (88%) multifaceted interventions with an A&F component. Feedback was most often provided in a written format only (41%), more than once (53%), and most often only provided data aggregated to the group-level (41%). Most studies saw a change in the hypothesized direction, but not all studies provided statistical analyses to formally test improvement. Overall study quality was low, with studies often lacking a concurrent control group.

CONCLUSIONS

Our review summarizes characteristics of A&F interventions implemented in the critical care context, points to some mechanisms by which A&F might be made more effective in this setting, and provides an overview of how the appropriateness of orders was reported. Our findings suggest that A&F can be effective in the context of critical care; however, further research is required to characterize approaches that optimize the effectiveness in this setting alongside more rigorous evaluation methods.

TRIAL REGISTRATION

PROSPERO CRD42016051941.

A Narrative Review on Hospital-Acquired Anemia: Keeping Blood where It Belongs

Hospital-acquired anemia (HAA) is a prevalent condition that is independently associated with worse clinical outcomes including prolongation of hospital stay and increased morbidity and mortality. While multifactorial in general, iatrogenic blood loss has been long recognized as one of the key contributing factors to development and worsening of HAA during hospital stay. Patients can be losing over 50 mL of blood per day to diagnostic blood draws. Strategies such as elimination of unnecessary laboratory tests that are not likely to alter the course of management, use of pediatric-size or small-volume tubes for blood collection to reduce phlebotomy volumes and avoid blood wastage, use of closed blood sampling devices, and substituting invasive tests with point-of-care testing alone or bundled together have generally been shown to be effective in reducing the volume of iatrogenic blood loss, hemoglobin decline, and blood transfusions, with no negative impact on the availability of test results for the clinical team. These strategies are important components of Patient Blood Management programs and their adoption can lead to improved clinical outcomes for patients.

Machine learning can accurately predict pre-admission baseline hemoglobin and creatinine in intensive care patients

Patients admitted to the intensive care unit frequently have anemia and impaired renal function, but often lack historical blood results to contextualize the acuteness of these findings. Using data available within two hours of ICU admission, we developed machine learning models that accurately (AUC 0.86-0.89) classify an individual patient's baseline hemoglobin and creatinine levels. Compared to assuming the baseline to be the same as the admission lab value, machine learning performed significantly better at classifying acute kidney injury regardless of initial creatinine value, and significantly better at predicting baseline hemoglobin value in patients with admission hemoglobin of <10 g/dl.

Routine vs on-demand blood sampling in critically ill patients-Protocol for a scoping review

BACKGROUND

In intensive care units, blood sampling is done commonly as part of daily routine. It remains unknown whether this practice is associated with harms or benefits, as not all routine blood tests may be clinically indicated, and blood sampling done without specific indications may be problematic. Accordingly, we aim to assess the body of evidence regarding the usage of routine vs on-demand blood sampling in critically ill patients in a scoping review.

METHODS

We will conduct a scoping review in accordance with the Preferred Reporting Items for Systematic and Meta-Analysis (PRISMA) statement as well as the PRISMA Extension for Scoping Reviews (PRISMA-ScR). Using a PICO-based search strategy, we will systematically search the Cochrane Library, Embase and Medline for relevant studies regardless of design. Two authors will independently screen studies for inclusion and extract data. We will provide a descriptive analysis of the data and asses the quality of evidence in accordance with the Grading of Recommended Assessment, Development and Evaluation approach.

DISCUSSION

The outlined scoping review will provide an important overview on the current body of evidence regarding the use of daily routine vs on-demand blood sampling in critical care settings. The findings of this scoping review will guide further research.

Predicting lab values for gastrointestinal bleeding patients in the intensive care unit: A comparative study on the impact of comorbidities and medications

Since a significant number of frequent laboratory blood tests are unnecessary and these tests may have complications, developing a system that could identify unnecessary tests is essential. In this paper, a value prediction approach is presented to predict the values of Calcium and Hematocrit laboratory blood tests for upper gastrointestinal bleeding patients and patients with unspecified hemorrhage in their gastrointestinal tract. The data have been extracted from the MIMIC-II database. By considering the issues of MIMIC-II in the process of data extraction and using expert knowledge, comprehensive preprocessing has been performed to validate the data. The first prediction system is developed using zero order Takagi-Sugeno fuzzy modeling and the sequential forward selection method. The results of this prediction system for target laboratory tests are promising. In the second proposed prediction system, patients are clustered using their comorbidity information before the final prediction phase. For each cluster, a medication feature is constructed and added to the data for the final feature selection. Although it was expected that clustering patients based on comorbidity data could improve the results of value prediction, the results were not improved in average. The reason for this could be the small number of abnormal laboratory test samples and their dispersion in clusters. These abnormal values would be more dispersed in the model with clustering phase, when they are scattered over different clusters.

A New Insight Into Missing Data in Intensive Care Unit Patient Profiles: Observational Study

Background

The data missing from patient profiles in intensive care units (ICUs) are substantial and unavoidable. However, this incompleteness is not always random or because of imperfections in the data collection process.

Objective

This study aimed to investigate the potential hidden information in data missing from electronic health records (EHRs) in an ICU and examine whether the presence or missingness of a variable itself can convey information about the patient health status.

Methods

Daily retrieval of laboratory test (LT) measurements from the Medical Information Mart for Intensive Care III database was set as our reference for defining complete patient profiles. Missingness indicators were introduced as a way of representing presence or absence of the LTs in a patient profile. Thereafter, various feature selection methods (filter and embedded feature selection methods) were used to examine the predictive power of missingness indicators. Finally, a set of well-known prediction models (logistic regression [LR], decision tree, and random forest) were used to evaluate whether the absence status itself of a variable recording can provide predictive power. We also examined the utility of missingness indicators in improving predictive performance when used with observed laboratory measurements as model input. The outcome of interest was in-hospital mortality and mortality at 30 days after ICU discharge.

Results

Regardless of mortality type or ICU day, more than 40% of the predictors selected by feature selection methods were missingness indicators. Notably, employing missingness indicators as the only predictors achieved reasonable mortality prediction on all days and for all mortality types (for instance, in 30-day mortality prediction with LR, we achieved area under the curve of the receiver operating characteristic [AUROC] of 0.6836±0.012). Including indicators with observed measurements in the prediction models also improved the AUROC; the maximum improvement was 0.0426. Indicators also improved the AUROC for Simplified Acute Physiology Score II model—a well-known ICU severity of illness score—confirming the additive information of the indicators (AUROC of 0.8045±0.0109 for 30-day mortality prediction for LR).

Conclusions

Our study demonstrated that the presence or absence of LT measurements is informative and can be considered a potential predictor of in-hospital and 30-day mortality. The comparative analysis of prediction models also showed statistically significant prediction improvement when indicators were included. Moreover, missing data might reflect the opinions of examining clinicians. Therefore, the absence of measurements can be informative in ICUs and has predictive power beyond the measured data themselves. This initial case study shows promise for more in-depth analysis of missing data and its informativeness in ICUs. Future studies are needed to generalize these results.

An Optimal Policy for Patient Laboratory Tests in Intensive Care Units

Laboratory testing is an integral tool in the management of patient care in hospitals, particularly in intensive care units (ICUs). There exists an inherent trade-off in the selection and timing of lab tests between considerations of the expected utility in clinical decision-making of a given test at a specific time, and the associated cost or risk it poses to the patient. In this work, we introduce a framework that learns policies for ordering lab tests which optimizes for this trade-off. Our approach uses batch off-policy reinforcement learning with a composite reward function based on clinical imperatives, applied to data that include examples of clinicians ordering labs for patients. To this end, we develop and extend principles of Pareto optimality to improve the selection of actions based on multiple reward function components while respecting typical procedural considerations and prioritization of clinical goals in the ICU. Our experiments show that we can estimate a policy that reduces the frequency of lab tests and optimizes timing to minimize information redundancy. We also find that the estimated policies typically suggest ordering lab tests well ahead of critical onsets-such as mechanical ventilation or dialysis-that depend on the lab results. We evaluate our approach by quantifying how these policies may initiate earlier onset of treatment.

Big data and targeted machine learning in action to assist medical decision in the ICU

Historically, personalised medicine has been synonymous with pharmacogenomics and oncology. We argue for a new framework for personalised medicine analytics that capitalises on more detailed patient-level data and leverages recent advances in causal inference and machine learning tailored towards decision support applicable to critically ill patients. We discuss how advances in data technology and statistics are providing new opportunities for asking more targeted questions regarding patient treatment, and how this can be applied in the intensive care unit to better predict patient-centred outcomes, help in the discovery of new treatment regimens associated with improved outcomes, and ultimately how these rules can be learned in real-time for the patient.

Predictive modeling in urgent care: a comparative study of machine learning approaches

OBJECTIVE

The growing availability of rich clinical data such as patients' electronic health records provide great opportunities to address a broad range of real-world questions in medicine. At the same time, artificial intelligence and machine learning (ML)-based approaches have shown great premise on extracting insights from those data and helping with various clinical problems. The goal of this study is to conduct a systematic comparative study of different ML algorithms for several predictive modeling problems in urgent care.

DESIGN

We assess the performance of 4 benchmark prediction tasks (eg mortality and prediction, differential diagnostics, and disease marker discovery) using medical histories, physiological time-series, and demographics data from the Medical Information Mart for Intensive Care (MIMIC-III) database.

MEASUREMENTS

For each given task, performance was estimated using standard measures including the area under the receiver operating characteristic (AUC) curve, F-1 score, sensitivity, and specificity. Microaveraged AUC was used for multiclass classification models.

RESULTS AND DISCUSSION

Our results suggest that recurrent neural networks show the most promise in mortality prediction where temporal patterns in physiologic features alone can capture in-hospital mortality risk (AUC > 0.90). Temporal models did not provide additional benefit compared to deep models in differential diagnostics. When comparing the training-testing behaviors of readmission and mortality models, we illustrate that readmission risk may be independent of patient stability at discharge. We also introduce a multiclass prediction scheme for length of stay which preserves sensitivity and AUC with outliers of increasing duration despite decrease in sample size.

Strategies to Reduce Inappropriate Laboratory Blood Test Orders in Intensive Care Are Effective and Safe: A Before-And-After Quality Improvement Study

Unnecessary pathology tests performed in intensive care units (ICU) might lead to increased costs of care and potential patient harm due to unnecessary phlebotomy. We hypothesised that a multimodal intervention program could result in a safe and effective reduction in the pathology tests ordered in our ICU. We conducted a single-centre pre- and post-study using multimodal interventions to address commonly ordered routine tests. The study was performed during the same six month period (August to February) over three years: 2012 to 2013 (pre-intervention), 2013 to 2014 (intervention) and 2014 to 2015 (post-intervention). Interventions consisted of staff education, designing new pathology forms, consultant-led pathology test ordering and intensive monitoring for a six-month period. The results of the study showed that there was a net savings of over A$213,000 in the intervention period and A$175,000 in the post-intervention period compared to the pre-intervention period. There was a 28% reduction in the tests performed in the intervention period (P <0.0001 compared to pre-intervention period) and 26% in the post-intervention period (P <0.0001 compared to pre-intervention period). There were no ICU or hospital mortality differences between the groups. There were no significant haemoglobin differences between the groups. A multimodal intervention safely reduced pathology test ordering in the ICU, resulting in substantial cost savings.

Laboratory Tests and X-ray Imaging in a Surgical Intensive Care Unit: Checking the Checklist

Context

Patients in the surgical intensive care unit (ICU) frequently undergo laboratory and imaging testing. These tests can lead to iatrogenic anemia and radiation exposure. Many of these tests may be unnecessary for the management of a patient's illness in the surgical ICU, and their ordering may be a reflex rather than in response to a clinical question. Checklists have been used in critical care to identify and address patient care strategies.

Objective

To examine whether adding a “diagnostic testing” section to a daily checklist used for patient rounds in a surgical ICU would decrease the amount of laboratory tests and chest x-ray imaging ordered.

Methods

An additional section was added to an established ICU daily checklist, which included the following 2 questions: “Is a [chest x-ray] needed for clinical management tomorrow?” and “What laboratory tests are medically necessary for tomorrow?” Comparison was made between 3-month preintervention (control group) and intervention (intervention group) periods. Medical records of hospitalized patients during the preintervention and intervention periods were compared to determine differences in the number of tests ordered per day during each period.

Results

A total of 307 adult patients at a single institution were included in the analysis: 155 in the control group and 152 in the intervention group. The patients in each group were similar in terms of sex, age, Sequential Organ Failure scores, Charlson Comorbidity Index scores, elective admission status, surgical procedures, number of days of mechanical ventilation, ICU length of stay, and in-hospital death. No statistical reductions in laboratory tests or chest x-ray imaging ordered per day from the preintervention to intervention period were found.

Conclusion

The addition of the diagnostic testing section to the daily checklist did not result in a reduction of the amount of tests ordered per day. Further research on test appropriateness and the possible addition of a clinician decision-making tool could be studied in the future to assist with reduction of tests ordered in the surgical ICU.

Clinical decision support tool for Co-management signalling

INTRODUCTION

Co-management between internists and surgeons of selected patients is becoming one of the pillars of modern clinical management in large hospitals. Defining the patients to be co-managed is essential. The aim of this study is to create a decision tool using real-world patient data collected in the preoperative period, to support the decision on which patients should have the co-management service offered.

METHODS

Data was collected from the electronic clinical health records of patients who had an International Classification of Diseases, 9th edition (ICD-9) code of colorectal surgery during the period between January 2012 and October 2014 in a 200 bed private teaching hospital in Lisbon. ICD-9 codes of colorectal surgery [48.5 and 48.6 (anterior rectal resection and abdominoperineal resection), 45.7 (partial colectomy), 45.8 (Total Colectomy), and 45.9 (Bowel Anastomosis)] were used. Only patients above 18 years old were considered. Patients with more than one procedure were excluded from the study. From these data the authors investigated the construction of predictive models using logistic regression and Takagi-Sugeno fuzzy modelling.

RESULTS

Data contains information obtained from the clinical records of a cohort of 344 adult patients. Data from 398 emergent and elective surgeries were collected, from which 54 were excluded because they were second procedures for the same patients. Four preoperative variables were identified as being the most predictive of co-management, in multivariable regression analysis. The final model performed well after being internally validated (0.81 AUC, 77% accuracy, 74% sensitivity, 78% specificity, 93% negative predictive value). The results indicate that the decision process can be more objective and potentially automated.

CONCLUSIONS

The authors developed a prediction model based on preoperative characteristics, in order to support the decision for the co-management of surgical patients in the postoperative ward setting. The model is a simple bedside decision tool that uses only four numerical variables.

Systematic Review of Non-ASCP Choosing Wisely Recommendations Relevant to Pathology and Laboratory Medicine

OBJECTIVES

To determine non-American Society for Clinical Pathology pathology- and laboratory-related Choosing Wisely recommendations that drive effective test utilization in the laboratory.

METHODS

Data were collected via a two-part web-based survey distributed to a broad sample of pathologists and laboratory professionals from a variety of institutions.

RESULTS

Pathologists' most relevant recommendation: "Do not transfuse more units of blood than absolutely necessary"; highest priority: "Do not transfuse more than the minimum number of RBC units necessary to relieve symptoms of anemia or to return a patient to a safe hemoglobin range (7-8 g/dL in stable, noncardiac inpatients)." Laboratory professionals' most relevant recommendation: "Avoid testing for a Clostridium difficile infection in the absence of diarrhea"; highest priority: "Do not routinely transfuse stable, asymptomatic hospitalized patients with a hemoglobin level greater than 7 to 8 g/dL."

CONCLUSIONS

Most of the highest priority, most relevant recommendations among those surveyed concerned utilization of blood products and transfusion management.

Rational Clinical Pathology Assessment in the Intensive Care Unit

Blood tests are ordered on a daily basis in intensive care units (ICU). There are no widely accepted guidelines for testing requirements. This study investigated the impact on ICU laboratory test costs of a multi-strategy change in practice involving routine blood testing. A single centre, prospective, interventional study using historical controls was undertaken to investigate the impact of ICU specialist authorisation of high-volume routine tests on ICU laboratory test costs. Prior to commencement of the study, ICU nursing and junior ICU doctors were able to order tests. During the six-month intervention period, the ICU specialists authorised routine blood tests. Adverse events related to not performing blood tests were also recorded. Overall ICU laboratory test costs decreased by 12.3% over the six months (P=0.0022 versus historical control) with a mean compliance of 51% with the test authorisation protocol. The costs of frequently ordered tests (classified as high-volume) decreased by 20% (P=0.0022 versus historical control). These accounted for an average of 54 ± 3% of the overall ICU blood test costs (blood gas analyses 17%, simple chemistry tests consisting of electrolytes, liver function, calcium, phosphate, magnesium 14%, coagulation 12% and full blood count 11%). Two protocol-related adverse events were recorded and judged as minor and were resolved by ordering tests during the day. No adverse patient outcomes resulted from these two events. Blood testing authorisation by an ICU specialist was associated with significant cost savings in ICU and no adverse patient outcomes.

Personalized Identification of Differentially Expressed Modules in Osteosarcoma

BACKGROUND Osteosarcoma (OS), an aggressive malignant neoplasm, is the most common primary bone cancer mainly in adolescents and young adults. Differentially expressed modules tend to distinguish differences integrally. Identifying modules individually has been crucial for understanding OS mechanisms and applications of custom therapeutic decisions in the future. MATERIAL AND METHODS Samples came from individuals were used from control group (n=15) and OS group (n=84). Based on clique-merging, module-identification algorithm was used to identify modules from OS PPI networks. A novel approach - the individualized module aberrance score (iMAS) was performed to distinguish differences, making special use of accumulated normal samples (ANS). We performed biological process ontology to classify functionally modules. Then Support Vector Machine (SVM) was used to test distribution results of normal and OS group with screened modules. RESULTS We identified 83 modules containing 2084 genes from PPI network in which 61 modules were significantly different. Cluster analysis of OS using the iMAS method identified 5 modules clusters. Specificity=1.00 and Sensitivity=1.00 proved the distribution outcomes of screened modules were mainly consistent with that of total data, which suggested the efficiency of 61 modules. CONCLUSIONS We conclude that a novel pipeline that identified the dysregulated modules in individuals of OS. The constructed process is expected to aid in personalized health care, which may present fruitful strategies for medical therapy.

Controlling testing volume for respiratory viruses using machine learning and text mining

Viral testing for pediatric inpatients with respiratory symptoms is common, with considerable associated charges. In an attempt to reduce testing volumes, we studied whether data available at the time of admission could aid in identifying children with low likelihood of having a particular viral origin of their symptoms, and thus safely forgo broad viral testing. We collected clinical data for 1,685 pediatric inpatients receiving respiratory virus testing from 2010-2012. Machine-learning on the data allowed us to construct pre-test models predicting whether a patient would test positive for a particular virus. Text mining improved the predictions for one viral test. Cost-sensitive models optimized for test sensitivity showed reasonable test specificities and an ability to reduce test volume by up to 46% for single viral tests. We conclude that diverse forms of data in the electronic medical record can be used productively to build models that help physicians reduce testing volumes.

Laboratory Tests Ordering Pattern by Medical Residents From a Brazilian University Hospital

OBJECTIVES

The adequacy of laboratory test orders by medical residents is a longstanding issue. The aim of this study is to analyze the number, types, and pattern of repetition of tests ordered by medical residents.

METHODS

We studied all tests ordered over a 1-year period for inpatients of an internal medicine ward in a university hospital. Types, results, and repetition pattern of tests were analyzed in relation to patients' diagnoses.

RESULTS

We evaluated 117,666 tests, requested for 1,024 inpatients. The mean number of tests was 9.5 per day. The test repetition pattern was similar, regardless of patients' diagnoses, previous test results, or duration of stay. The probability of an abnormal result after a sequence of three normal tests was lower than 25%, regardless of the diagnosis.

CONCLUSIONS

Number of tests and repetition were both high, imposing costs, discomfort, and risks to patients, thus warranting further investigation.

Low Diagnostic Utility of Rechecking Hemoglobins Within 24 Hours in Hospitalized Patients

BACKGROUND

Clinicians often repeat hemoglobin tests within a 24 hour period to detect or monitor anemia. We sought to determine the percentage of hemoglobin tests repeated within a single hospital day that were at least 1.0 g/dL lower than the first test.

METHODS

We performed a retrospective cross-sectional analysis of hospitalized adults on medical or surgical services over 1 year at a single academic hospital. Using patient and laboratory data in the electronic health record, we analyzed the proportion of repeated hemoglobin test results that were at least 1 g/dL less than the initial hemoglobin value of that day, excluding days when transfusions were administered.

RESULTS

A total of 88,722 hemoglobin tests were obtained from 12,877 unique patients, who contributed a total of 86,859 hospitalization days. In 12,230 (14.1%) of those days, 2 or more hemoglobin tests were obtained within a single day. In the 6969 days with 2 hemoglobin tests obtained and no transfusions given, 949 (13.5%) were ≥1 g/dL lower than the initial hemoglobin value of that day, and 260 (3.7%) were ≥2 g/dL lower. Repeated tests did not often reach transfusion thresholds: 482 (6.9%) of repeat hemoglobin values were <8 g/dL, and 64 (0.9%) were <7 g/dL.

CONCLUSIONS

Hemoglobin tests were repeated in 14% of hospital days. For patients who had 2 hemoglobin tests obtained on the same day, 13.5% demonstrated a clinically significant drop. This information may be helpful to clinicians when considering whether repeat testing is appropriate.