Diagnosis of Rare Diseases: a scoping review of clinical decision support systems

Objectivizing issues in the diagnosis of complex rare diseases: lessons learned from testing existing diagnosis support systems on ciliopathies

BACKGROUND

There are approximately 8,000 different rare diseases that affect roughly 400 million people worldwide. Many of them suffer from delayed diagnosis. Ciliopathies are rare monogenic disorders characterized by a significant phenotypic and genetic heterogeneity that raises an important challenge for clinical diagnosis. Diagnosis support systems (DSS) applied to electronic health record (EHR) data may help identify undiagnosed patients, which is of paramount importance to improve patients' care. Our objective was to evaluate three online-accessible rare disease DSSs using phenotypes derived from EHRs for the diagnosis of ciliopathies.

METHODS

Two datasets of ciliopathy cases, either proven or suspected, and two datasets of controls were used to evaluate the DSSs. Patient phenotypes were automatically extracted from their EHRs and converted to Human Phenotype Ontology terms. We tested the ability of the DSSs to diagnose cases in contrast to controls based on Orphanet ontology.

RESULTS

A total of 79 cases and 38 controls were selected. Performances of the DSSs on ciliopathy real world data (best DSS with area under the ROC curve = 0.72) were not as good as published performances on the test set used in the DSS development phase. None of these systems obtained results which could be described as "expert-level". Patients with multisystemic symptoms were generally easier to diagnose than patients with isolated symptoms. Diseases easily confused with ciliopathy generally affected multiple organs and had overlapping phenotypes. Four challenges need to be considered to improve the performances: to make the DSSs interoperable with EHR systems, to validate the performances in real-life settings, to deal with data quality, and to leverage methods and resources for rare and complex diseases.

CONCLUSION

Our study provides insights into the complexities of diagnosing highly heterogenous rare diseases and offers lessons derived from evaluation existing DSSs in real-world settings. These insights are not only beneficial for ciliopathy diagnosis but also hold relevance for the enhancement of DSS for various complex rare disorders, by guiding the development of more clinically relevant rare disease DSSs, that could support early diagnosis and finally make more patients eligible for treatment.

Rare diseases and pyruvate kinase M2: a promising therapeutic connection

Pyruvate kinase M2 (PKM2) is a key glycolytic enzyme that regulates proliferating cell metabolism. The role of PKM2 in common diseases has been well established, but its role in rare diseases (RDs) is less understood. Over the past few years, PKM2 has emerged as a crucial player in RDs, including, neoplastic, respiratory, metabolic, and neurological disorders. Herein, we summarize recent findings and developments highlighting PKM2 as an emerging key player in RDs. We also discuss the current status of PKM2 modulation in RDs with particular emphasis on preclinical and clinical studies in addition to current challenges in the field.

Synthetic protein protease sensor platform

Introduction: Protease activity can serve as a highly specific biomarker for application in health, biotech, and beyond. The aim of this study was to develop a protease cleavable synthetic protein platform to detect protease activity in a rapid cell-free setting. Methods: The protease sensor is modular, with orthogonal peptide tags at the N and C terminal ends, which can be uncoupled via a protease responsive module located in between. The sensor design allows for several different readouts of cleavage signal. A protein 'backbone' [Green fluorescent protein (GFP)] was designed in silico to have both a C-terminal Flag-tag and N-Terminal 6x histidine tag (HIS) for antibody detection. A protease cleavage site, which can be adapted for any known protease cleavage sequence, enables the uncoupling of the peptide tags. Three different proteases-Tobacco, Etch Virus (TEV), the main protease from coronavirus SARS-COV-2 (Mpro) and Matrix Metallopeptidase 9 (MMP9)-a cancer-selective human protease-were examined. A sandwich Enzyme-Linked Immunosorbent Assay (ELISA) was developed based on antibodies against the HIS and Flag tags. As an alternative readout, a C-terminal quencher peptide separable by protease cleavage from the GFP was also included. Purified proteins were deployed in cell-free cleavage assays with their respective protease. Western blots, fluorescence assays and immunoassay were performed on samples. Results: Following the design, build and validation of protein constructs, specific protease cleavage was initially demonstrated by Western blot. The novel ELISA proved to afford highly sensitive detection of protease activity in all cases. By way of alternative readout, activation of fluorescence signal upon protease cleavage was also demonstrated but did not match the sensitivity provided by the ELISA method. Discussion: This platform, comprising a protease-responsive synthetic protein device and accompanying readout, is suitable for future deployment in a rapid, low-cost, lateral flow setting. The modular protein device can readily accommodate any desired protease-response module (target protease cleavage site). This study validates the concept with three disparate proteases and applications-human infectious disease, cancer and agricultural crop infection.

Does gene therapies clinical research in rare diseases reflects the competitivity of the country: Example of France

Rare diseases are chronic, serious and generally genetic conditions affecting a small number of people, and their therapeutic management is a real challenge. They represent a considerable burden for patients, caregivers and society alike. Compared with existing symptomatic treatments, gene therapies represent a promising new approach aimed at treating these diseases by replacing a defective gene, or by abolishing or reviving a gene-derived function. France is considered one of the leading countries in the research and development of drugs for rare diseases, yet the position of French public and private stakeholders in the research and development of gene therapies for rare diseases at global and European level remains unclear. To answer this question, we used the GENOTRIAL FR database developed by OrphanDev to clarify France's involvement and competitiveness in this field. The results show that France is actively involved in gene therapy clinical trials, with a dense international collaboration network and solid expertise. However, the French medical infrastructure is mainly involved in clinical research on gene therapy candidates sponsored by several foreign countries. To a lesser extent, French public and private entities are also developing their own gene therapy candidates for various rare diseases, some of which have already reached advanced clinical phases. In conclusion, a number of technical and financial challenges need to be overcome if France is to maintain its position as a European and world leader and increase its contribution to reducing the economic and social burden of rare diseases by developing revolutionary and effective new therapies.

[Short paths to diagnosis with artificial intelligence: systematic literature review on diagnostic decision support systems]

BACKGROUND

Rare diseases are often recognized late. Their diagnosis is particularly challenging due to the diversity, complexity and heterogeneity of clinical symptoms. Computer-aided diagnostic aids, often referred to as diagnostic decision support systems (DDSS), are promising tools for shortening the time to diagnosis. Despite initial positive evaluations, DDSS are not yet widely used, partly due to a lack of integration with existing clinical or practice information systems.

OBJECTIVE

This article provides an insight into currently existing diagnostic support systems that function without access to electronic patient records and only require information that is easily obtainable.

MATERIALS AND METHODS

A systematic literature search identified eight articles on DDSS that can assist in the diagnosis of rare diseases with no need for access to electronic patient records or other information systems in practices and hospitals. The main advantages and disadvantages of the identified rare disease diagnostic support systems were extracted and summarized.

RESULTS

Symptom checkers and DDSS based on portrait photos and pain drawings already exist. The degree of maturity of these applications varies.

CONCLUSION

DDSS currently still face a number of challenges, such as concerns about data protection and accuracy, and acceptance and awareness continue to be rather low. On the other hand, there is great potential for faster diagnosis, especially for rare diseases, which are easily overlooked due to their large number and the low awareness of them. The use of DDSS should therefore be carefully considered by doctors on a case-by-case basis.

Clinical Genetics Assessment Triangle (CGAT): A simple tool to identify patients with genetic conditions

OBJECTIVE

The objective of this study was to develop a simple tool for general physicians to promptly identify and refer pediatric patients with a higher probability of having a genetic condition.

STUDY DESIGN

This retrospective, descriptive study was conducted at a tertiary pediatric hospital's Clinical Genetics Unit from June 2019 to January 2020. We included patients under 18 years of age who visited the unit, excluding those without genetic testing. Epidemiological, clinical, and genetic variables were collected from electronic medical records. The primary outcome was the diagnosis of a genetic condition based on genetic testing.

RESULTS

Among 445 patients, 304 were included; 163 (53.6%) were male, and mean age was 7.4 years (SD 5.1 years). A genetic condition was diagnosed in 139 patients (45.7%). Using a multiple logistic regression model, five variables significantly contributed to reaching a diagnosis: suspected diagnosis at referral (OR 3.45, P < 0.001), short stature (OR 3.11, P < 0.001), global developmental delay/intellectual disability (OR 2.65, P < 0.001), dysmorphic craniofacial features (OR 1.99, P = 0.035), and multiple congenital anomalies (OR 2.54, P = 0.033). The association strength (OR) increased when these variables were paired with each other. The study's findings are presented in the form of a triangle, known as the Clinical Genetics Assessment Triangle (CGAT), which summarizes the results. A decision tree model is applied to guide clinical department referrals based on the affected sides of the triangle.

CONCLUSIONS

The CGAT has the potential to enable general physicians to promptly identify pediatric patients with an increased probability of having a genetic condition.

Requirements analysis for an AI-based clinical decision support system for general practitioners: a user-centered design process

BACKGROUND

As the first point of contact for patients with health issues, general practitioners (GPs) are frequently confronted with patients presenting with non-specific symptoms of unclear origin. This can result in delayed, prolonged or false diagnoses. To accelerate and improve the diagnosis of diseases, clinical decision support systems would appear to be an appropriate tool. The objective of the project 'Smart physician portal for patients with unclear disease' (SATURN) is to employ a user-centered design process based on the requirements analysis presented in this paper to develop an artificial Intelligence (AI)-based diagnosis support system that specifically addresses the needs of German GPs.

METHODS

Requirements analysis for a GP-specific diagnosis support system was conducted in an iterative process with five GPs. First, interviews were conducted to analyze current workflows and the use of digital applications in cases of diagnostic uncertainty (as-is situation). Second, we focused on collecting and prioritizing tasks to be performed by an ideal smart physician portal (to-be situation) in a workshop. We then developed a task model with corresponding user requirements.

RESULTS

Numerous GP-specific user requirements were identified concerning the tasks and subtasks: performing data entry (open system, enter patient data), reviewing results (receiving and evaluating results), discussing results (with patients and colleagues), scheduling further diagnostic procedures, referring to specialists (select, contact, make appointments), and case closure. Suggested features particularly concerned the process of screening and assessing results: e.g., the system should focus more on atypical patterns of common diseases than on rare diseases only, display probabilities of differential diagnoses, ensure sources and results are transparent, and mark diagnoses that have already been ruled out. Moreover, establishing a means of using the platform to communicate with colleagues and transferring patient data directly from electronic patient records to the system was strongly recommended.

CONCLUSIONS

Essential user requirements to be considered in the development and design of a diagnosis system for primary care could be derived from the analysis. They form the basis for mockup-development and system engineering.

A diagnostic support system based on pain drawings: binary and k-disease classification of EDS, GBS, FSHD, PROMM, and a control group with Pain2D

BACKGROUND AND OBJECTIVE

The diagnosis of rare diseases (RDs) is often challenging due to their rarity, variability and the high number of individual RDs, resulting in a delay in diagnosis with adverse effects for patients and healthcare systems. The development of computer assisted diagnostic decision support systems could help to improve these problems by supporting differential diagnosis and by prompting physicians to initiate the right diagnostic tests. Towards this end, we developed, trained and tested a machine learning model implemented as part of the software called Pain2D to classify four rare diseases (EDS, GBS, FSHD and PROMM), as well as a control group of unspecific chronic pain, from pen-and-paper pain drawings filled in by patients.

METHODS

Pain drawings (PDs) were collected from patients suffering from one of the four RDs, or from unspecific chronic pain. The latter PDs were used as an outgroup in order to test how Pain2D handles more common pain causes. A total of 262 (59 EDS, 29 GBS, 35 FSHD, 89 PROMM, 50 unspecific chronic pain) PDs were collected and used to generate disease specific pain profiles. PDs were then classified by Pain2D in a leave-one-out-cross-validation approach.

RESULTS

Pain2D was able to classify the four rare diseases with an accuracy of 61-77% with its binary classifier. EDS, GBS and FSHD were classified correctly by the Pain2D k-disease classifier with sensitivities between 63 and 86% and specificities between 81 and 89%. For PROMM, the k-disease classifier achieved a sensitivity of 51% and specificity of 90%.

CONCLUSIONS

Pain2D is a scalable, open-source tool that could potentially be trained for all diseases presenting with pain.

Determinants of implementing artificial intelligence-based clinical decision support tools in healthcare: a scoping review protocol

INTRODUCTION

Artificial intelligence (AI), the simulation of human intelligence processes by machines, is being increasingly leveraged to facilitate clinical decision-making. AI-based clinical decision support (CDS) tools can improve the quality of care and appropriate use of healthcare resources, and decrease healthcare provider burnout. Understanding the determinants of implementing AI-based CDS tools in healthcare delivery is vital to reap the benefits of these tools. The objective of this scoping review is to map and synthesise determinants (barriers and facilitators) to implementing AI-based CDS tools in healthcare.

METHODS AND ANALYSIS

This scoping review will follow the Joanna Briggs Institute methodology and the Preferred Reporting Items for Systematic reviews and Meta-Analysis extension for Scoping Reviews checklist. The search terms will be tailored to each database, which includes MEDLINE, Embase, CINAHL, APA PsycINFO and the Cochrane Library. Grey literature and references of included studies will also be searched. The search will include studies published from database inception until 10 May 2022. We will not limit searches by study design or language. Studies that either report determinants or describe the implementation of AI-based CDS tools in clinical practice or/and healthcare settings will be included. The identified determinants (barriers and facilitators) will be described by synthesising the themes using the Theoretical Domains Framework. The outcome variables measured will be mapped and the measures of effectiveness will be summarised using descriptive statistics.

ETHICS AND DISSEMINATION

Ethics approval is not required because all data for this study have been previously published. The findings of this review will be published in a peer-reviewed journal and presented at academic conferences. Importantly, the findings of this scoping review will be widely presented to decision-makers, health system administrators, healthcare providers, and patients and family/caregivers as part of an implementation study of an AI-based CDS for the treatment of coronary artery disease.

The Experience of a Tertiary Reference Hospital in the Study of Rare Neurological Diseases

Background and Objectives: Rare diseases (RDs) are life-threatening or chronically impairing conditions that affect about 6% of the world's population. RDs are often called 'orphan' diseases, since people suffering from them attract little support from national health systems. Aim: The aim of this study is to describe the clinical characteristics of, and the available laboratory examinations for, patients who were hospitalized in a tertiary referral center and finally received a diagnosis associated with a Rare Neurological Disease (RND). Materials and Methods: Patients that were hospitalized in our clinic from 1 January 2014 to 31 March 2022 and were finally diagnosed with an RND were consecutively included. The RND classification was performed according to the ORPHAcode system. Results: A total of 342 out of 11.850 (2.9%) adult patients admitted to our department during this period received a diagnosis associated with an RND. The most common diagnosis (N = 80, 23%) involved an RND presenting with dementia, followed by a motor neuron disease spectrum disorder (N = 64, 18.7%). Family history indicative of an RND was present in only 21 patients (6.1%). Fifty-five (16%) people had previously been misdiagnosed with another neurological condition. The mean time delay between disease onset and diagnosis was 4.24 ± 0.41 years. Conclusions: Our data indicate that a broad spectrum of RNDs may reach a tertiary Neurological Center after a significant delay. Moreover, our data underline the need for a network of reference centers, both at a national and international level, expected to support research on the diagnosis and treatment of RND.

The Power of Clinical Diagnosis for Deciphering Complex Genetic Mechanisms in Rare Diseases

Complex genetic disease mechanisms, such as structural or non-coding variants, currently pose a substantial difficulty in frontline diagnostic tests. They thus may account for most unsolved rare disease patients regardless of the clinical phenotype. However, the clinical diagnosis can narrow the genetic focus to just a couple of genes for patients with well-established syndromes defined by prominent physical and/or unique biochemical phenotypes, allowing deeper analyses to consider complex genetic origin. Then, clinical-diagnosis-driven genome sequencing strategies may expedite the development of testing and analytical methods to account for complex disease mechanisms as well as to advance functional assays for the confirmation of complex variants, clinical management, and the development of new therapies.

Termviewer - A Web Application for Streamlined Human Phenotype Ontology (HPO) Tagging and Document Annotation

Clinical notes from electronic health records (EHRs) contain a large amount of clinical phenotype data on patients that can provide insights into the phenotypic presentation of various diseases. A number of Natural Language Processing (NLP) algorithms have been utilized in the past few years to annotate medical concepts, such as Human Phenotype Ontology (HPO) terms, from clinical notes. However, efficient use of NLP algorithms requires the use of high-quality clinical notes with phenotype descriptions, and erroneous annotations often exist in results from these NLP algorithms. Manual review by human experts is often needed to compile the correct phenotype information on individual patients. Here we develop TermViewer, a web application that allows multi-party collaborative annotation and quality assessment of clinical notes that have already been processed and tagged by NLP algorithms. TermViewer allows users to view clinical notes with HPO terms highlighted, and to easily classify high-quality notes and revise incorrect tagging of HPO terms. Currently, TermViewer combines MetaMap and cTAKES, two of the most widely used NLP tools for tagging medical terms, and identifies where these two tools agree and disagree, allowing users to perform collaborative manual reviews of computationally generated HPO annotations. TermViewer can be a stand-alone tool for analyzing notes or become part of a machine-learning pipeline where tagged HPO terms can be used as additional input data. TermViewer is available at https://github.com/WGLab/TermViewer.

Diagnostic delay in rare diseases: data from the Spanish rare diseases patient registry

BACKGROUND

According to the International Rare Diseases Research Consortium (IRDiRC), a known rare disease (RD) should be diagnosable within a year. This study sought: firstly, to ascertain how long it takes to obtain the diagnosis of a RD in Spain, along with its associated time trend; and secondly, to identify and measure diagnostic delay (defined by the IRDiRC as any period exceeding a year) by reference to the characteristics of RDs and the persons affected by them.

METHODS

Using data sourced from the Spanish Rare Diseases Patient Registry, we performed a descriptive analysis of the time elapsed between symptom onset and diagnosis of each RD, by sex, age and date of symptom onset, and type of RD. We analysed the time trend across the period 1960-2021 and possible change points, using a Joinpoint regression model and assuming a Poisson distribution. The multivariate analysis was completed with backward stepwise logistic regression.

RESULTS

Detailed information was obtained on 3304 persons with RDs: 56.4% had experienced delay in diagnosis of their RDs, with the mean time taken being 6.18 years (median = 2; IQR 0.2-7.5). Both the percentage of patients with diagnostic delay and the average time to diagnosis underwent a significant reduction across the study period (p < 0.001). There was a higher percentage of diagnostic delays: in women (OR 1.25; 95% CI 1.07-1.45); in cases with symptom onset at age 30-44 years (OR 1.48; 95% CI 1.19-1.84): and when analysed by type of RD, in mental and behavioural disorders (OR 4.21; 95% CI 2.26-7.85), followed by RDs of the nervous system (OR 1.39; 95% CI 1.02-1.88).

CONCLUSIONS

This is the first study to quantify time to diagnosis of RDs in Spain, based on data from a national registry open to any RD. Since over half of all persons affected by RDs experience delay in diagnosis, new studies are needed to ascertain the factors associated with this delay and the implications this has on the lives of patients and their families.

[What is confirmed in the diagnostics of autoinflammatory fever diseases?]

Periodic fever syndromes (PFS) are a group of rare autoinflammatory diseases, which are characterized by disorders of the innate immune reaction and life-long recurrent episodes of inflammatory symptoms. This article describes the diagnostic approach. In addition to the patient medical history, physical examination and laboratory determinations, gene tests are becoming increasingly more important. The panel diagnostics using high throughput sequencing or next generation sequencing (NGS) is the method of choice for the detection of a genetic cause of PFS. This article discusses the diagnostic decision support systems (DDSS) that can play a future role in the diagnosis of rare diseases, especially those with complex patterns of symptoms.

[Interdisciplinary care path and potential IT support for people with rare diseases in Germany]

INTRODUCTION

Due to the high variability and, at the same time, rare occurrence of rare diseases, the diagnosis of these patients (approx. 4 million people in Germany) can turn into an odyssey. The large time interval between the appearance of symptoms and the final diagnosis of the rare disease leads to a delay in the appropriate treatment. The often long period of uncertainty about the cause of symptoms as well as non-specific or even wrong therapies can have negative effects on both the course of disease and the patients' quality of life. For a better understanding of the current care situation and IT landscape, the interdisciplinary care pathway for people with rare diseases will be modelled and the possible uses of IT applications identified that have the potential to improve diagnosis, treatment and therapy of rare diseases.

METHODS

In order to achieve these goals, an initial care pathway was modelled on the basis of process descriptions which are commonly used in the literature, discussed in detail, and agreed upon in a first workshop with six experts from outpatient and inpatient care as well as employees of Centers for Rare Diseases. In a second workshop, ten experts analyzed the resulting care pathway with regard to the possible use of IT applications, and the identification was agreed upon. The experts included those involved in the process, in particular physicians, patients / patient representatives, health care researchers, and experts in hospital IT, IT security, and data protection.

RESULTS

The two workshops resulted in process models including the specification of possible uses for IT applications. The most important steps in the care pathway for people with rare diseases in Germany include: neonatal screening, seeking medical advice, outpatient care by general practitioners, outpatient care by specialists, care by specialist outpatient clinic, care by clinic, care by a Center for Rare Diseases: case review and case conference and treatment and therapy. The discussion of the possible uses of IT applications resulted in a focus on registers (e. g. with regard to experts, treatment and therapy options) as well as on digital tools, such as "digital findings and findings platform" and "digital referral with referral tracking".

DISCUSSION

Our results show that the care pathway is very heterogeneous and complex. Thus, the sub-processes show different variants with many branches and repetitions. They also illustrate that the care for people with rare diseases requires a high level of interdisciplinary collaboration; diagnosis as well as treatment and therapy often take place across sectors and in cooperation between different medical health care institutions and professions. When analyzing the current IT landscape, it becomes clear that IT applications can be used at many process steps in the care for people with rare diseases and have a high potential. Therefore, they must be used to inform decisions about the adequate diagnosis and treatment as well as communication about the clinical pictures and the patient's case between practitioners and medical care sectors.

CONCLUSION

The interdisciplinary collaboration highlights the need for cooperation between the various parties involved in the process, which requires the identification and implementation of interfaces between the stakeholders and their systems. However, it is not enough to include the view of the processes; the data perspective is also required. Creating interoperability also enables the use of IT applications. The basis for this is the results obtained.

Establishing a second-generation artificial intelligence-based system for improving diagnosis, treatment, and monitoring of patients with rare diseases

Patients with rare diseases are a major challenge for healthcare systems. These patients face three major obstacles: late diagnosis and misdiagnosis, lack of proper response to therapies, and absence of valid monitoring tools. We reviewed the relevant literature on first-generation artificial intelligence (AI) algorithms which were designed to improve the management of chronic diseases. The shortage of big data resources and the inability to provide patients with clinical value limit the use of these AI platforms by patients and physicians. In the present study, we reviewed the relevant literature on the obstacles encountered in the management of patients with rare diseases. Examples of currently available AI platforms are presented. The use of second-generation AI-based systems that are patient-tailored is presented. The system provides a means for early diagnosis and a method for improving the response to therapies based on clinically meaningful outcome parameters. The system may offer a patient-tailored monitoring tool that is based on parameters that are relevant to patients and caregivers and provides a clinically meaningful tool for follow-up. The system can provide an inclusive solution for patients with rare diseases and ensures adherence based on clinical responses. It has the potential advantage of not being dependent on large datasets and is a dynamic system that adapts to ongoing changes in patients' disease and response to therapy.

CADA: phenotype-driven gene prioritization based on a case-enriched knowledge graph

Many rare syndromes can be well described and delineated from other disorders by a combination of characteristic symptoms. These phenotypic features are best documented with terms of the Human Phenotype Ontology (HPO), which are increasingly used in electronic health records (EHRs), too. Many algorithms that perform HPO-based gene prioritization have also been developed; however, the performance of many such tools suffers from an over-representation of atypical cases in the medical literature. This is certainly the case if the algorithm cannot handle features that occur with reduced frequency in a disorder. With Cada, we built a knowledge graph based on both case annotations and disorder annotations. Using network representation learning, we achieve gene prioritization by link prediction. Our results suggest that Cada exhibits superior performance particularly for patients that present with the pathognomonic findings of a disease. Additionally, information about the frequency of occurrence of a feature can readily be incorporated, when available. Crucial in the design of our approach is the use of the growing amount of phenotype–genotype information that diagnostic labs deposit in databases such as ClinVar. By this means, Cada is an ideal reference tool for differential diagnostics in rare disorders that can also be updated regularly.

Current Challenges and Future Opportunities for XAI in Machine Learning-Based Clinical Decision Support Systems: A Systematic Review

Machine Learning and Artificial Intelligence (AI) more broadly have great immediate and future potential for transforming almost all aspects of medicine. However, in many applications, even outside medicine, a lack of transparency in AI applications has become increasingly problematic. This is particularly pronounced where users need to interpret the output of AI systems. Explainable AI (XAI) provides a rationale that allows users to understand why a system has produced a given output. The output can then be interpreted within a given context. One area that is in great need of XAI is that of Clinical Decision Support Systems (CDSSs). These systems support medical practitioners in their clinic decision-making and in the absence of explainability may lead to issues of under or over-reliance. Providing explanations for how recommendations are arrived at will allow practitioners to make more nuanced, and in some cases, life-saving decisions. The need for XAI in CDSS, and the medical field in general, is amplified by the need for ethical and fair decision-making and the fact that AI trained with historical data can be a reinforcement agent of historical actions and biases that should be uncovered. We performed a systematic literature review of work to-date in the application of XAI in CDSS. Tabular data processing XAI-enabled systems are the most common, while XAI-enabled CDSS for text analysis are the least common in literature. There is more interest in developers for the provision of local explanations, while there was almost a balance between post-hoc and ante-hoc explanations, as well as between model-specific and model-agnostic techniques. Studies reported benefits of the use of XAI such as the fact that it could enhance decision confidence for clinicians, or generate the hypothesis about causality, which ultimately leads to increased trustworthiness and acceptability of the system and potential for its incorporation in the clinical workflow. However, we found an overall distinct lack of application of XAI in the context of CDSS and, in particular, a lack of user studies exploring the needs of clinicians. We propose some guidelines for the implementation of XAI in CDSS and explore some opportunities, challenges, and future research needs.

Evaluation of a clinical decision support system for rare diseases: a qualitative study

BACKGROUND

Rare Diseases (RDs) are difficult to diagnose. Clinical Decision Support Systems (CDSS) could support the diagnosis for RDs. The Medical Informatics in Research and Medicine (MIRACUM) consortium developed a CDSS for RDs based on distributed clinical data from eight German university hospitals. To support the diagnosis for difficult patient cases, the CDSS uses data from the different hospitals to perform a patient similarity analysis to obtain an indication of a diagnosis. To optimize our CDSS, we conducted a qualitative study to investigate usability and functionality of our designed CDSS.

METHODS

We performed a Thinking Aloud Test (TA-Test) with RDs experts working in Rare Diseases Centers (RDCs) at MIRACUM locations which are specialized in diagnosis and treatment of RDs. An instruction sheet with tasks was prepared that the participants should perform with the CDSS during the study. The TA-Test was recorded on audio and video, whereas the resulting transcripts were analysed with a qualitative content analysis, as a ruled-guided fixed procedure to analyse text-based data. Furthermore, a questionnaire was handed out at the end of the study including the System Usability Scale (SUS).

RESULTS

A total of eight experts from eight MIRACUM locations with an established RDC were included in the study. Results indicate that more detailed information about patients, such as descriptive attributes or findings, can help the system perform better. The system was rated positively in terms of functionality, such as functions that enable the user to obtain an overview of similar patients or medical history of a patient. However, there is a lack of transparency in the results of the CDSS patient similarity analysis. The study participants often stated that the system should present the user with an overview of exact symptoms, diagnosis, and other characteristics that define two patients as similar. In the usability section, the CDSS received a score of 73.21 points, which is ranked as good usability.

CONCLUSIONS

This qualitative study investigated the usability and functionality of a CDSS of RDs. Despite positive feedback about functionality of system, the CDSS still requires some revisions and improvement in transparency of the patient similarity analysis.