Machine learning-based algorithms applied to drug prescriptions and other healthcare services in the Sicilian claims database to identify acromegaly as a model for the earlier diagnosis of rare diseases.

Acromegaly is a rare disease characterized by a diagnostic delay ranging from 5 to 10 years from the symptoms' onset. The aim of this study was to develop and internally validate machine-learning algorithms to identify a combination of variables for the early diagnosis of acromegaly. This retrospective population-based study was conducted between 2011 and 2018 using data from the claims databases of Sicily Region, in Southern Italy. To identify combinations of potential predictors of acromegaly diagnosis, conditional and unconditional penalized multivariable logistic regression models and three machine learning algorithms (i.e., the Recursive Partitioning and Regression Tree, the Random Forest and the Support Vector Machine) were used, and their performance was evaluated. The random forest (RF) algorithm achieved the highest Area under the ROC Curve value of 0.83 (95% CI 0.79-0.87). The sensitivity in the test set, computed at the optimal threshold of predicted probabilities, ranged from 28% for the unconditional logistic regression model to 69% for the RF. Overall, the only diagnosis predictor selected by all five models and algorithms was the number of immunosuppressants-related pharmacy claims. The other predictors selected by at least two models were eventually combined in an unconditional logistic regression to develop a meta-score that achieved an acceptable discrimination accuracy (AUC = 0.71, 95% CI 0.66-0.75). Findings of this study showed that data-driven machine learning algorithms may play a role in supporting the early diagnosis of rare diseases such as acromegaly.

Identifying the joint signature of brain atrophy and gene variant scores in Alzheimer's Disease.

The joint modeling of genetic data and brain imaging information allows for determining the pathophysiological pathways of neurodegenerative diseases such as Alzheimer's disease (AD). This task has typically been approached using mass-univariate methods that rely on a complete set of Single Nucleotide Polymorphisms (SNPs) to assess their association with selected image-derived phenotypes (IDPs). However, such methods are prone to multiple comparisons bias and, most importantly, fail to account for potential cross-feature interactions, resulting in insufficient detection of significant associations. Ways to overcome these limitations while reducing the number of traits aim at conveying genetic information at the gene level and capturing the integrated genetic effects of a set of genetic variants, rather than looking at each SNP individually. Their associations with brain IDPs are still largely unexplored in the current literature, though they can uncover new potential genetic determinants for brain modulations in the AD continuum. In this work, we explored an explainable multivariate model to analyze the genetic basis of the grey matter modulations, relying on the AD Neuroimaging Initiative (ADNI) phase 3 dataset. Cortical thicknesses and subcortical volumes derived from T1-weighted Magnetic Resonance were considered to describe the imaging phenotypes. At the same time the genetic counterpart was represented by gene variant scores extracted by the Sequence Kernel Association Test (SKAT) filtering model. Moreover, transcriptomic analysis was carried on to assess the expression of the resulting genes in the main brain structures as a form of validation. Results highlighted meaningful genotype-phenotype interactionsas defined by three latent components showing a significant difference in the projection scores between patients and controls. Among the significant associations, the model highlighted EPHX1 and BCAS1 gene variant scores involved in neurodegenerative and myelination processes, hence relevant for AD. In particular, the first was associated with decreased subcortical volumes and the second with decreasedtemporal lobe thickness. Noteworthy, BCAS1 is particularly expressed in the dentate gyrus. Overall, the proposed approach allowed capturing genotype-phenotype interactions in a restricted study cohort that was confirmed by transcriptomic analysis, offering insights into the underlying mechanisms of neurodegeneration in AD in line with previous findings and suggesting new potential disease biomarkers.

The IHI Rochester Report 2022 on Healthcare Informatics Research: Resuming After the CoViD-19.

In 2020, the CoViD-19 pandemic spread worldwide in an unexpected way and suddenly modified many life issues, including social habits, social relationships, teaching modalities, and more. Such changes were also observable in many different healthcare and medical contexts. Moreover, the CoViD-19 pandemic acted as a stress test for many research endeavors, and revealed some limitations, especially in contexts where research results had an immediate impact on the social and healthcare habits of millions of people. As a result, the research community is called to perform a deep analysis of the steps already taken, and to re-think steps for the near and far future to capitalize on the lessons learned due to the pandemic. In this direction, on June 09th-11th, 2022, a group of twelve healthcare informatics researchers met in Rochester, MN, USA. This meeting was initiated by the Institute for Healthcare Informatics-IHI, and hosted by the Mayo Clinic. The goal of the meeting was to discuss and propose a research agenda for biomedical and health informatics for the next decade, in light of the changes and the lessons learned from the CoViD-19 pandemic. This article reports the main topics discussed and the conclusions reached. The intended readers of this paper, besides the biomedical and health informatics research community, are all those stakeholders in academia, industry, and government, who could benefit from the new research findings in biomedical and health informatics research. Indeed, research directions and social and policy implications are the main focus of the research agenda we propose, according to three levels: the care of individuals, the healthcare system view, and the population view.

Discovering predictive temporal patterns for Acute Kidney Injury from critical care data.

Acute Kidney Injury is a severe clinical condition with a high risk of multi-organs complications and mortality. For this reason, early recognition is crucial. Our proposal based on a 3-window framework discovers all hidden regularities, called Approximate Predictive Functional Dependencies, with the aim to enable early recognition of high-risk patients during hospitalization in the Intensive Care Unit (ICU). We evaluated the different severity stages according to the Kidney Disease Improving Global Outcomes (KDIGO) guidelines, building different pathological state patterns, from admission to the discharge from ICU. According to the clinical practice, for each patient, we examined various characteristics expressed as a temporal history of events that may predict a pathological state pattern. We evaluated our proposal exploiting the MIMIC-IV dataset, a collection of Electronic Medical Records from ICU. The obtained results showed promising possibilities to use this type of dependency to support clinical practice.

A manifesto on explainability for artificial intelligence in medicine.

The rapid increase of interest in, and use of, artificial intelligence (AI) in computer applications has raised a parallel concern about its ability (or lack thereof) to provide understandable, or explainable, output to users. This concern is especially legitimate in biomedical contexts, where patient safety is of paramount importance. This position paper brings together seven researchers working in the field with different roles and perspectives, to explore in depth the concept of explainable AI, or XAI, offering a functional definition and conceptual framework or model that can be used when considering XAI. This is followed by a series of desiderata for attaining explainability in AI, each of which touches upon a key domain in biomedicine.

Mining compact predictive pattern sets using classification model.

In this paper, we develop a new framework for mining predictive patterns that aims to describe compactly the condition (or class) of interest. Our framework relies on a classification model that considers and combines various predictive pattern candidates and selects only those that are important for improving the overall class prediction performance. We test our approach on data derived from MIMIC-III EHR database, focusing on patterns predictive of sepsis. We show that using our classification approach we can achieve a significant reduction in the number of extracted patterns compared to the state-of-the-art methods based on minimum predictive pattern mining approach, while preserving the overall classification accuracy of the model.

Clinical Information Systems and Artificial Intelligence: Recent Research Trends.

OBJECTIVES: This survey aims at reviewing the literature related to Clinical Information Systems (CIS), Hospital Information Systems (HIS), Electronic Health Record (EHR) systems, and how collected data can be analyzed by Artificial Intelligence (AI) techniques. METHODS: We selected the major journals (11 journals) collecting papers (more than 7,000) over the last five years from the top members of the research community, and read and analyzed the papers (more than 200) covering the topics. Then, we completed the analysis using search engines to also include papers from major conferences over the same five years. RESULTS: We defined a taxonomy of major features and research areas of CIS, HIS, EHR systems. We also defined a taxonomy for the use of Artificial Intelligence (AI) techniques on healthcare data. In the light of these taxonomies, we report on the most relevant papers from the literature. CONCLUSIONS: We highlighted some major research directions and issues which seem to be promising and to need further investigations over a medium- or long-term period.

Normalizing Spontaneous Reports Into MedDRA: Some Experiments With MagiCoder.

Text normalization into medical dictionaries is useful to support clinical tasks. A typical setting is pharmacovigilance (PV). The manual detection of suspected adverse drug reactions (ADRs) in narrative reports is time consuming and natural language processing (NLP) provides a concrete help to PV experts. In this paper, we carry out experiments for testing performances of MagiCoder, an NLP application designed to extract MedDRA terms from narrative clinical text. Given a narrative description, MagiCoder proposes an automatic encoding. The pharmacologist reviews, (possibly) corrects, and then, validates the solution. This drastically reduces the time needed for the validation of reports with respect to a completely manual encoding. In previous work, we mainly tested MagiCoder performances on Italian written spontaneous reports. In this paper, we include some new features, change the experiment design, and carry on more tests about MagiCoder. Moreover, we do a change of language, moving to English documents. In particular, we tested MagiCoder on the CADEC dataset, a corpus of manually annotated posts about ADRs collected from the social media.

From narrative descriptions to MedDRA: automagically encoding adverse drug reactions.

CONTEXT: The collection of narrative spontaneous reports is an irreplaceable source for the prompt detection of suspected adverse drug reactions (ADRs). In such task qualified domain experts manually revise a huge amount of narrative descriptions and then encode texts according to MedDRA standard terminology. The manual annotation of narrative documents with medical terminology is a subtle and expensive task, since the number of reports is growing up day-by-day. OBJECTIVES: Natural Language Processing (NLP) applications can support the work of people responsible for pharmacovigilance. Our objective is to develop NLP algorithms and tools for the detection of ADR clinical terminology. Efficient applications can concretely improve the quality of the experts' revisions. NLP software can quickly analyze narrative texts and offer an encoding (i.e., a list of MedDRA terms) that the expert has to revise and validate. METHODS: MagiCoder, an NLP algorithm, is proposed for the automatic encoding of free-text descriptions into MedDRA terms. MagiCoder procedure is efficient in terms of computational complexity. We tested MagiCoder through several experiments. In the first one, we tested it on a large dataset of about 4500 manually revised reports, by performing an automated comparison between human and MagiCoder encoding. Moreover, we tested MagiCoder on a set of about 1800 reports, manually revised ex novo by some experts of the domain, who also compared automatic solutions with the gold reference standard. We also provide two initial experiments with reports written in English, giving a first evidence of the robustness of MagiCoder w.r.t. the change of the language. RESULTS: For the current base version of MagiCoder, we measured an average recall and precision of 86.9% and 91.8%, respectively. CONCLUSIONS: From a practical point of view, MagiCoder reduces the time required for encoding ADR reports. Pharmacologists have only to review and validate the MedDRA terms proposed by the application, instead of choosing the right terms among the 70 K low level terms of MedDRA. Such improvement in the efficiency of pharmacologists' work has a relevant impact also on the quality of the subsequent data analysis. We developed MagiCoder for the Italian pharmacovigilance language. However, our proposal is based on a general approach, not depending on the considered language nor the term dictionary.

A Methodological Framework for the Integrated Design of Decision-Intensive Care Pathways-an Application to the Management of COPD Patients.

Healthcare processes are by nature complex, mostly due to their multidisciplinary character that requires continuous coordination between care providers. They encompass both organizational and clinical tasks, the latter ones driven by medical knowledge, which is inherently incomplete and distributed among people having different expertise and roles. Care pathways refer to planning and coordination of care processes related to specific groups of patients in a given setting. The goal in defining and following care pathways is to improve the quality of care in terms of patient satisfaction, costs reduction, and medical outcome. Thus, care pathways are a promising methodological tool for standardizing care and decision-making. Business process management techniques can successfully be used for representing organizational aspects of care pathways in a standard, readable, and accessible way, while supporting process development, analysis, and re-engineering. In this paper, we introduce a methodological framework that fosters the integrated design, implementation, and enactment of care processes and related decisions, while considering proper representation and management of organizational and clinical information. We focus here and discuss in detail the design phase, which encompasses the simulation of care pathways. We show how business process model and notation (BPMN) and decision model and notation (DMN) can be combined for supporting intertwined aspects of decision-intensive care pathways. As a proof-of-concept, the proposed methodology has been applied to design care pathways related to chronic obstructive pulmonary disease (COPD) in the region of Veneto, in Italy.

Telemedicine for Developing Countries. A Survey and Some Design Issues.

BACKGROUND: Developing countries need telemedicine applications that help in many situations, when physicians are a small number with respect to the population, when specialized physicians are not available, when patients and physicians in rural villages need assistance in the delivery of health care. Moreover, the requirements of telemedicine applications for developing countries are somewhat more demanding than for developed countries. Indeed, further social, organizational, and technical aspects need to be considered for successful telemedicine applications in developing countries. OBJECTIVE: We consider all the major projects in telemedicine, devoted to developing countries, as described by the proper scientific literature. On the basis of such literature, we want to define a specific taxonomy that allows a proper classification and a fast overview of telemedicine projects in developing countries. Moreover, by considering both the literature and some recent direct experiences, we want to complete such overview by discussing some design issues to be taken into consideration when developing telemedicine software systems. METHODS: We considered and reviewed the major conferences and journals in depth, and looked for reports on the telemedicine projects. RESULTS: We provide the reader with a survey of the main projects and systems, from which we derived a taxonomy of features of telemedicine systems for developing countries. We also propose and discuss some classification criteria for design issues, based on the lessons learned in this research area. CONCLUSIONS: We highlight some challenges and recommendations to be considered when designing a telemedicine system for developing countries.

Thirty years of artificial intelligence in medicine (AIME) conferences: A review of research themes.

BACKGROUND: Over the past 30 years, the international conference on Artificial Intelligence in MEdicine (AIME) has been organized at different venues across Europe every 2 years, establishing a forum for scientific exchange and creating an active research community. The Artificial Intelligence in Medicine journal has published theme issues with extended versions of selected AIME papers since 1998. OBJECTIVES: To review the history of AIME conferences, investigate its impact on the wider research field, and identify challenges for its future. METHODS: We analyzed a total of 122 session titles to create a taxonomy of research themes and topics. We classified all 734 AIME conference papers published between 1985 and 2013 with this taxonomy. We also analyzed the citations to these conference papers and to 55 special issue papers. RESULTS: We identified 30 research topics across 12 themes. AIME was dominated by knowledge engineering research in its first decade, while machine learning and data mining prevailed thereafter. Together these two themes have contributed about 51% of all papers. There have been eight AIME papers that were cited at least 10 times per year since their publication. CONCLUSIONS: There has been a major shift from knowledge-based to data-driven methods while the interest for other research themes such as uncertainty management, image and signal processing, and natural language processing has been stable since the early 1990s. AIME papers relating to guidelines and protocols are among the most highly cited.

Spatiotemporal data visualisation for homecare monitoring of elderly people.

OBJECTIVE: Elderly people who live alone can be assisted by home monitoring systems that identify risk scenarios such as falls, fatigue symptoms or burglary. Given that these systems have to manage spatiotemporal data, human intervention is required to validate automatic alarms due to the high number of false positives and the need for context interpretation. The goal of this work was to provide tools to support human action, to identify such potential risk scenarios based on spatiotemporal data visualisation. METHODS AND MATERIALS: We propose the MTA (multiple temporal axes) model, a visual representation of temporal information of the activity of a single person at different locations. The main goal of this model is to visualize the behaviour of a person in their home, facilitating the identification of health-risk scenarios and repetitive patterns. We evaluate the model's insight capacity compared with other models using a standard evaluation protocol. We also test its practical suitability of the MTA graphical model in a commercial home monitoring system. In particular, we implemented 8VISU, a visualization tool based on MTA. RESULTS: MTA proved to be more than 90% accurate in identify non-risk scenarios, independently of the length of the record visualised. When the spatial complexity was increased (e.g. number of rooms) the model provided good accuracy form up to 5 rooms. Therefore, user preferences and user performance seem to be balanced. Moreover, it also gave high sensitivity levels (over 90%) for 5-8 rooms. Fall is the most recurrent incident for elderly people. The MTA model outperformed the other models considered in identifying fall scenarios (66% of correctness) and was the second best for burglary and fatigue scenarios (36% of correctness). Our experiments also confirm the hypothesis that cyclic models are the most suitable for fatigue scenarios, the Spiral and MTA models obtaining most positive identifications. CONCLUSIONS: In home monitoring systems, spatiotemporal visualization is a useful tool for identifying risk and preventing home accidents in elderly people living alone. The MTA model helps the visualisation in different stages of the temporal data analysis process. In particular, its explicit representation of space and movement is useful for identifying potential scenarios of risk, while the spiral structure can be used for the identification of recurrent patterns. The results of the experiments and the experience using the visualization tool 8VISU proof the potential of the MTA graphical model to mine temporal data and to support caregivers using home monitoring infrastructures.

Mining approximate temporal functional dependencies with pure temporal grouping in clinical databases.

Functional dependencies (FDs) typically represent associations over facts stored by a database, such as "patients with the same symptom get the same therapy." In more recent years, some extensions have been introduced to represent both temporal constraints (temporal functional dependencies - TFDs), as "for any given month, patients with the same symptom must have the same therapy, but their therapy may change from one month to the next one," and approximate properties (approximate functional dependencies - AFDs), as "patients with the same symptomgenerallyhave the same therapy." An AFD holds most of the facts stored by the database, enabling some data to deviate from the defined property: the percentage of data which violate the given property is user-defined. According to this scenario, in this paper we introduce approximate temporal functional dependencies (ATFDs) and use them to mine clinical data. Specifically, we considered the need for deriving new knowledge from psychiatric and pharmacovigilance data. ATFDs may be defined and measured either on temporal granules (e.g.grouping data by day, week, month, year) or on sliding windows (e.g.a fixed-length time interval which moves over the time axis): in this regard, we propose and discuss some specific and efficient data mining techniques for ATFDs. We also developed two running prototypes and showed the feasibility of our proposal by mining two real-world clinical data sets. The clinical interest of the dependencies derived considering the psychiatry and pharmacovigilance domains confirms the soundness and the usefulness of the proposed techniques.

Visually defining and querying consistent multi-granular clinical temporal abstractions.

OBJECTIVE: The main goal of this work is to propose a framework for the visual specification and query of consistent multi-granular clinical temporal abstractions. We focus on the issue of querying patient clinical information by visually defining and composing temporal abstractions, i.e., high level patterns derived from several time-stamped raw data. In particular, we focus on the visual specification of consistent temporal abstractions with different granularities and on the visual composition of different temporal abstractions for querying clinical databases. BACKGROUND: Temporal abstractions on clinical data provide a concise and high-level description of temporal raw data, and a suitable way to support decision making. Granularities define partitions on the time line and allow one to represent time and, thus, temporal clinical information at different levels of detail, according to the requirements coming from the represented clinical domain. The visual representation of temporal information has been considered since several years in clinical domains. Proposed visualization techniques must be easy and quick to understand, and could benefit from visual metaphors that do not lead to ambiguous interpretations. Recently, physical metaphors such as strips, springs, weights, and wires have been proposed and evaluated on clinical users for the specification of temporal clinical abstractions. Visual approaches to boolean queries have been considered in the last years and confirmed that the visual support to the specification of complex boolean queries is both an important and difficult research topic. METHODOLOGY: We propose and describe a visual language for the definition of temporal abstractions based on a set of intuitive metaphors (striped wall, plastered wall, brick wall), allowing the clinician to use different granularities. A new algorithm, underlying the visual language, allows the physician to specify only consistent abstractions, i.e., abstractions not containing contradictory conditions on the component abstractions. Moreover, we propose a visual query language where different temporal abstractions can be composed to build complex queries: temporal abstractions are visually connected through the usual logical connectives AND, OR, and NOT. RESULTS: The proposed visual language allows one to simply define temporal abstractions by using intuitive metaphors, and to specify temporal intervals related to abstractions by using different temporal granularities. The physician can interact with the designed and implemented tool by point-and-click selections, and can visually compose queries involving several temporal abstractions. The evaluation of the proposed granularity-related metaphors consisted in two parts: (i) solving 30 interpretation exercises by choosing the correct interpretation of a given screenshot representing a possible scenario, and (ii) solving a complex exercise, by visually specifying through the interface a scenario described only in natural language. The exercises were done by 13 subjects. The percentage of correct answers to the interpretation exercises were slightly different with respect to the considered metaphors (54.4--striped wall, 73.3--plastered wall, 61--brick wall, and 61--no wall), but post hoc statistical analysis on means confirmed that differences were not statistically significant. The result of the user's satisfaction questionnaire related to the evaluation of the proposed granularity-related metaphors ratified that there are no preferences for one of them. The evaluation of the proposed logical notation consisted in two parts: (i) solving five interpretation exercises provided by a screenshot representing a possible scenario and by three different possible interpretations, of which only one was correct, and (ii) solving five exercises, by visually defining through the interface a scenario described only in natural language. Exercises had an increasing difficulty. The evaluation involved a total of 31 subjects. Results related to this evaluation phase confirmed us about the soundness of the proposed solution even in comparison with a well known proposal based on a tabular query form (the only significant difference is that our proposal requires more time for the training phase: 21 min versus 14 min). DISCUSSION AND CONCLUSIONS: In this work we have considered the issue of visually composing and querying temporal clinical patient data. In this context we have proposed a visual framework for the specification of consistent temporal abstractions with different granularities and for the visual composition of different temporal abstractions to build (possibly) complex queries on clinical databases. A new algorithm has been proposed to check the consistency of the specified granular abstraction. From the evaluation of the proposed metaphors and interfaces and from the comparison of the visual query language with a well known visual method for boolean queries, the soundness of the overall system has been confirmed; moreover, pros and cons and possible improvements emerged from the comparison of different visual metaphors and solutions.

Querying temporal clinical databases on granular trends.

This paper focuses on the identification of temporal trends involving different granularities in clinical databases, where data are temporal in nature: for example, while follow-up visit data are usually stored at the granularity of working days, queries on these data could require to consider trends either at the granularity of months ("find patients who had an increase of systolic blood pressure within a single month") or at the granularity of weeks ("find patients who had steady states of diastolic blood pressure for more than 3 weeks"). Representing and reasoning properly on temporal clinical data at different granularities are important both to guarantee the efficacy and the quality of care processes and to detect emergency situations. Temporal sequences of data acquired during a care process provide a significant source of information not only to search for a particular value or an event at a specific time, but also to detect some clinically-relevant patterns for temporal data. We propose a general framework for the description and management of temporal trends by considering specific temporal features with respect to the chosen time granularity. Temporal aspects of data are considered within temporal relational databases, first formally by using a temporal extension of the relational calculus, and then by showing how to map these relational expressions to plain SQL queries. Throughout the paper we consider the clinical domain of hemodialysis, where several parameters are periodically sampled during every session.