Surprising and novel multivariate sequential patterns using odds ratio for temporal evolution in healthcare.

BACKGROUND: Pattern mining techniques are helpful tools when extracting new knowledge in real practice, but the overwhelming number of patterns is still a limiting factor in the health-care domain. Current efforts concerning the definition of measures of interest for patterns are focused on reducing the number of patterns and quantifying their relevance (utility/usefulness). However, although the temporal dimension plays a key role in medical records, few efforts have been made to extract temporal knowledge about the patient's evolution from multivariate sequential patterns. METHODS: In this paper, we propose a method to extract a new type of patterns in the clinical domain called Jumping Diagnostic Odds Ratio Sequential Patterns (JDORSP). The aim of this method is to employ the odds ratio to identify a concise set of sequential patterns that represent a patient's state with a statistically significant protection factor (i.e., a pattern associated with patients that survive) and those extensions whose evolution suddenly changes the patient's clinical state, thus making the sequential patterns a statistically significant risk factor (i.e., a pattern associated with patients that do not survive), or vice versa. RESULTS: The results of our experiments highlight that our method reduces the number of sequential patterns obtained with state-of-the-art pattern reduction methods by over 95%. Only by achieving this drastic reduction can medical experts carry out a comprehensive clinical evaluation of the patterns that might be considered medical knowledge regarding the temporal evolution of the patients. We have evaluated the surprisingness and relevance of the sequential patterns with clinicians, and the most interesting fact is the high surprisingness of the extensions of the patterns that become a protection factor, that is, the patients that recover after several days of being at high risk of dying. CONCLUSIONS: Our proposed method with which to extract JDORSP generates a set of interpretable multivariate sequential patterns with new knowledge regarding the temporal evolution of the patients. The number of patterns is greatly reduced when compared to those generated by other methods and measures of interest. An additional advantage of this method is that it does not require any parameters or thresholds, and that the reduced number of patterns allows a manual evaluation.

Spatiotemporal Data Modelling for Epidemiological Research in Hospitals.

Nosocomial infections are a great source of concern for healthcare organizations. The spatial layout of hospitals and the movements of patients play significant roles in the spread of outbreaks. However, the existing models are ad-hoc for a specific hospital and research topic. This work shows the design of a data model to study the spread of infections among hospital patients. Its spatial dimension describes the hospital layout with several levels of detail, and the temporal dimension describes everything that happens to the patients in the form of events, which can relate to the spatial dimension. The model is meant to be sufficiently general to fit any hospital layout and to be used for different epidemiological research topics. We proved the model's suitability by defining six queries based on patients' movements and contacts that could assist in several epidemiological research tasks, such as discovering potential transmission routes. The model was implemented as an RDF* knowledge graph, and the queries were in SPARQL*. Finally, we designed two experiments in which two outbreaks of Clostridium difficile were analyzed using several queries (four in the first experiment and two in the second) on a knowledge graph (105,000 nodes, 185,000 edges) with synthetic data.

Spatial-temporal simulation for hospital infection spread and outbreaks of Clostridioides difficile.

Validated and curated datasets are essential for studying the spread and control of infectious diseases in hospital settings, requiring clinical information on patients' evolution and their location. The literature shows that approaches based on Artificial Intelligence (AI) in the development of clinical-support systems have benefits that are increasingly recognized. However, there is a lack of available high-volume data, necessary for trusting such AI models. One effective method in this situation involves the simulation of realistic data. Existing simulators primarily focus on implementing compartmental epidemiological models and contact networks to validate epidemiological hypotheses. Nevertheless, other practical aspects such as the hospital building distribution, shifts or safety policies on infections has received minimal attention. In this paper, we propose a novel approach for a simulator of nosocomial infection spread, combining agent-based patient description, spatial-temporal constraints of the hospital settings, and microorganism behavior driven by epidemiological models. The predictive validity of the model was analyzed considering micro and macro-face validation, parameter calibration based on literature review, model alignment, and sensitive analysis with an expert. This simulation model is useful in monitoring infections and in the decision-making process in a hospital, by helping to detect spatial-temporal patterns and predict statistical data about the disease.

The use of networks in spatial and temporal computational models for outbreak spread in epidemiology: A systematic review.

OBJECTIVES: To examine recent literature in order to present a comprehensive overview of the current trends as regards the computational models used to represent the propagation of an infectious outbreak in a population, paying particular attention to those that represent network-based transmission. METHODS: a systematic review was conducted following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. Papers published in English between 2010 and September 2021 were sought in the ACM Digital Library, IEEE Xplore, PubMed and Scopus databases. RESULTS: Upon considering their titles and abstracts, 832 papers were obtained, of which 192 were selected for a full content-body check. Of these, 112 studies were eventually deemed suitable for quantitative and qualitative analysis. Emphasis was placed on the spatial and temporal scales studied, the use of networks or graphs, and the granularity of the data used to evaluate the models. The models principally used to represent the spreading of outbreaks have been stochastic (55.36%), while the type of networks most frequently used are relationship networks (32.14%). The most common spatial dimension used is a region (19.64%) and the most used unit of time is a day (28.57%). Synthetic data as opposed to an external source were used in 51.79% of the papers. With regard to the granularity of the data sources, aggregated data such as censuses or transportation surveys are the most common. CONCLUSION: We identified a growing interest in the use of networks to represent disease transmission. We detected that research is focused on only certain combinations of the computational model, type of network (in both the expressive and the structural sense) and spatial scale, while the search for other interesting combinations has been left for the future.

Meaningful time-related aspects of alerts in Clinical Decision Support Systems. A unified framework.

Alerts are a common functionality of clinical decision support systems (CDSSs). Although they have proven to be useful in clinical practice, the alert burden can lead to alert fatigue and significantly reduce their usability and acceptance. Based on a literature review, we propose a unified framework consisting of a set of meaningful timestamps that allows the use of state-of-the-art measures for alert burden, such as alert dwell time, alert think time, and response time. In addition, it can be used to investigate other measures that could be relevant as regards dealing with this problem. Furthermore, we provide a case study concerning three different types of alerts to which the framework was successfully applied. We consider that our framework can easily be adapted to other CDSSs and that it could be useful for dealing with alert burden measurement thus contributing to its appropriate management.

A methodology based on multiple criteria decision analysis for combining antibiotics in empirical therapy.

BACKGROUND: The current situation of critical progression in resistance to more effective antibiotics has forced the reuse of old highly toxic antibiotics and, for several reasons, the extension of the indications of combined antibiotic therapy as alternative options to broad spectrum empirical mono-therapy. A key aspect for selecting an appropriate and adequate antimicrobial therapy is that prescription must be based on local epidemiology and knowledge since many aspects, such as prevalence of microorganisms and effectiveness of antimicrobials, change from hospitals, or even areas and services within a single hospital. Therefore, the selection of combinations of antibiotics requires the application of a methodology that provides objectivity, completeness and reproducibility to the analysis of the detailed microbiological, epidemiological, pharmacological information on which to base a rational and reasoned choice. METHODS: We proposed a methodology for decision making that uses a multiple criteria decision analysis (MCDA) to support the clinician in the selection of an efficient combined empiric therapy. The MCDA includes a multi-objective constrained optimization model whose criteria are the maximum efficacy of therapy, maximum activity, the minimum activity overlapping, the minimum use of restricted antibiotics, the minimum toxicity of antibiotics and the activity against the most prevalent and virulent bacteria. The decision process can be defined in 4 steps: (1) selection of clinical situation of interest, (2) definition of local optimization criteria, (3) definition of constraints for reducing combinations, (4) manual sorting of solutions according to patient's clinical conditions, and (5) selection of a combination. EXPERIMENTS AND RESULTS: In order to show the application of the methodology to a clinical case, we carried out experiments with antibiotic susceptibility tests in blood samples taken during a five years period at a university hospital. The validation of the results consists of a manual review of the combinations and experiments carried out by an expert physician that has explained the most relevant solutions proposed according to current clinical knowledge and their use. CONCLUSION: We show that with the decision process proposed, the physician is able to select the best combined therapy according to different criteria such as maximum efficacy, activity and minimum toxicity. A method for the recommendation of combined antibiotic therapy developed on the basis of a multi-objective optimization model may assist the physicians in the search for alternatives to the use of broad-spectrum antibiotics or restricted antibiotics for empirical therapy. The decision proposed can be easily reproduced for any local epidemiology and any different clinical settings.

Impact of expert knowledge on the detection of patients at risk of antimicrobial therapy failure by clinical decision support systems.

Antimicrobial Susceptibility Tests (ASTs) are performed in hospitals to detect whether an infectious agent is resistant or susceptible to a set of antimicrobials. When AST results are available, the evaluation of the patient's antimicrobial therapy is a critical task to ensure its effectiveness against the found microorganism. Since not all the available antimicrobials can be tested in ASTs, clinicians rely on their expert knowledge to complement AST results and prescribe the most appropriate antimicrobials for each infection. Our goal is to help physicians in this task by improving the detection of antimicrobial therapies at risk of failure by Clinical Decision Support Systems (CDSSs). With this aim, we have incorporated the EUCAST expert rules in antimicrobial susceptibility testing into a CDSS to improve the results of ASTs. In order to achieve this, we have combined both ontologies and production rules. Furthermore, we have evaluated the impact of EUCAST expert rules on the detection of antimicrobial therapies at risk of failure. We performed a retrospective study with one year of clinical data, obtaining a total of 148 alerts from which 62 (41.9%) were based on the additional expert knowledge. Furthermore, the evaluation of the clinical relevance of 27 alerts resulted in 8 of them (29.7%) being clinically relevant. Of these, 6 were based on expert knowledge. Finally, an alarm fatigue study suggests that waiting between 48 and 72 h from the reception of the AST results can significantly reduce the number of alerts that are unnecessary in our CDSS because they are already being addressed in the hospital's daily workflow. In conclusion, we demonstrate that the incorporation of expert knowledge improves the capabilities of CDSSs as regards detecting the risk of antimicrobial therapy failure, which may improve the institutional outcomes in antimicrobial stewardship.

A decision support system for antibiotic prescription based on local cumulative antibiograms.

BACKGROUND: Local cumulative antibiograms are useful tools with which to select appropriate empiric or directed therapies when treating infectious diseases at a hospital. However, data represented in traditional antibiograms are static, incomplete and not well adapted to decision-making. METHODS: We propose a decision support method for empiric antibiotic therapy based on the Number Needed to Fail (NNF) measure. NNF indicates the number of patients that would need to be treated with a specific antibiotic for one to be inadequately treated. We define two new measures, Accumulated Efficacy and Weighted Accumulated Efficacy in order to determine the efficacy of an antibiotic. We carried out two experiments: the first during which there was a suspicion of infection and the patient had empiric therapy, and the second by considering patients with confirmed infection and directed therapy. The study was performed with 15,799 cultures with 356,404 susceptibility tests carried out over a four-year period. RESULTS: The most efficient empiric antibiotics are Linezolid and Vancomycin for blood samples and Imipenem and Meropenem for urine samples. In both experiments, the efficacies of recommended antibiotics are all significantly greater than the efficacies of the antibiotics actually administered (P < 0.001). The highest efficacy is obtained when considering 2 years of antibiogram data and 80% of the cumulated prevalence of microorganisms. CONCLUSION: This extensive study on real empiric therapies shows that the proposed method is a valuable alternative to traditional antibiograms as regards developing clinical decision support systems for antimicrobial stewardship.