Artificial intelligence has the potential to revolutionize modern society in all its aspects. Encouraged by the variety and vast amount of data that can be gathered from patients (e.g., medical images, text, and electronic health records), researchers have recently increased their interest in developing AI solutions for clinical care. Moreover, a diverse repertoire of methods can be chosen towards creating performant models for use in medical applications, ranging from disease prediction, diagnosis, and prognosis to opting for the most appropriate treatment for an individual patient. In this respect, the present paper aims to review the advancements reported at the convergence of AI and clinical care. Thus, this work presents AI clinical applications in a comprehensive manner, discussing the recent literature studies classified according to medical specialties. In addition, the challenges and limitations hindering AI integration in the clinical setting are further pointed out.

Background and objectives: The diagnosis of acute kidney injury (AKI) is usually based on changes in serum creatinine, which is a poor marker of early renal dysfunction. The discriminative and predictive abilities of serum and urinary cystatin C were examined for the prediction of AKI. Design, setting, participants, & measurements: In this prospective cohort study, serum and urinary cystatin C were serially measured in a heterogeneous group of patients (n = 616) presenting to a tertiary care emergency department. The primary outcome was AKI, classified according to RIFLE and AKIN criteria. The secondary outcome was an adjudication based on clinical criteria to AKI, prerenal azotemia, chronic kidney disease (CKD), and normal kidney function. Results: Patients were adjudicated to have AKI in 21.1%, prerenal azotemia in 25.8%, CKD in 2.4%, and normal kidney function in 50.7%. For the diagnosis of AKI, the discriminatory ability of urinary creatinine and cystatin C was marginal. Both serum cystatin C and serum creatinine (at presentation and 6 hours later) showed high discriminatory ability for the diagnosis of AKI. However, only serum cystatin C attained a significant early predictive power (Hosmer-Lemeshow P value > 0.05). Serum cystatin C could differentiate between AKI and prerenal azotemia, but not between AKI and CKD. Conclusions: Serum cystatin C is an early, predictive biomarker of AKI, which outperforms serum creatinine in the heterogeneous emergency department setting. However, neither biomarker discriminated between AKI and CKD. Additional biomarkers continue to be needed for improved specificity in the diagnosis of community-acquired AKI.

Background: The Manchester Triage System (MTS) is a 5-point triage scale used to triage patients presenting to the emergency department. It was introduced in the UK in 1996 and is now widespread, especially in Europe, and has been in use in our hospital since 2000 via a computerised protocol. A study was undertaken to determine whether the subgroups created by the application of MTS have different propensities for indirect triage outcomes such as death in the A&E department or being admitted to hospital. Methods: A database of 321 539 patients triaged during a 30-month period (from January 2005 to June 2007) was used. MTS codes, death outcomes, admission and admission route were used to estimate the proportions and association between MTS codes and the remaining variables by χ² univariate analysis. Results: There was a clear association between the priority group and short-term mortality as well as with the proportion of patients admitted to hospital. Conclusions: The MTS provides information that extends beyond its immediate usefulness as a prioritisation mechanism. It is a powerful tool for distinguishing between patients with high and low unadjusted risk of short-term death as well as those who will stay in hospital for at least 24 h and those who will return home. Its discriminatory power is not equal for medical and surgical specialities, which may be linked to the nature of its inbuilt discriminators.

Emergency department (ED) triage systems aim to direct the best clinical assistance to those who are in the greatest urgency and guarantee that resources are efficiently applied.The study's purpose was to determine whether the Manchester Triage System (MTS) second version is a useful instrument for determining the risk of hospital admission, intrahospital death and resource utilisation in ED and to compare it with the MTS first version.This was a prospective study of patients that attended the ED at a large hospital. It comprised a total of 25 218 cases that were triaged between 11 July and 13 October 2011. The MTS codes were grouped into two clusters: red and orange into a ‘high acuity/priority’ (HP) cluster, and yellow, green and blue into a ‘low acuity/priority’ cluster.The risk of hospital admission in the HP cluster was 4.86 times that of the LP cluster for both admission route and ages. The percentage of patient hospital admission between medical and surgical specialties, in high and low priority clusters, was similar. We found the risk of death in the HP cluster to be 5.58 times that of the risk of the low acuity/priority cluster. The MTS had an inconsistent association relative to the utilisation of x-ray, while it seemed to portray a consistent association between ECG and laboratory utilisation and MTS cluster.There were no differences between medical and surgical specialities risk of admission. This suggests that improvements were made in the second version of MTS, particularly in the discriminators of patients triaged to surgical specialties, because this was not true for the first version of MTS.

Maintaining appropriate home rehabilitation programs after stroke, with proper adherence and remote monitoring is a challenging task. Virtual reality (VR) - based serious games could be a strategy used in telerehabilitation (TR) to engage patients in an enjoyable and therapeutic approach. The aim of this review was to analyze the background and quality of clinical research on this matter to guide future research. The review was based on research material obtained from PubMed and Cochrane up to April 2020 using the PRISMA approach. The use of VR serious games has shown evidence of efficacy on upper limb TR after stroke, but the evidence strength is still low due to a limited number of randomized controlled trials (RCT), a small number of participants involved, and heterogeneous samples. Although this is a promising strategy to complement conventional rehabilitation, further investigation is needed to strengthen the evidence of effectiveness and support the dissemination of the developed solutions.

Summary: Background EHR systems have high potential to improve healthcare delivery and management. Although structured EHR data generates information in machine-readable formats, their use for decision support still poses technical challenges for researchers due to the need to preprocess and convert data into a matrix format. During our research, we observed that clinical informatics literature does not provide guidance for researchers on how to build this matrix while avoiding potential pitfalls.Objectives This article aims to provide researchers a roadmap of the main technical challenges of preprocessing structured EHR data and possible strategies to overcome them.Methods Along standard data processing stages – extracting database entries, defining features, processing data, assessing feature values and integrating data elements, within an EDPAI framework –, we identified the main challenges faced by researchers and reflect on how to address those challenges based on lessons learned from our research experience and on best practices from related literature. We highlight the main potential sources of error, present strategies to approach those challenges and discuss implications of these strategies.Results Following the EDPAI framework, researchers face five key challenges: (1) gathering and integrating data, (2) identifying and handling different feature types, (3) combining features to handle redundancy and granularity, (4) addressing data missingness, and (5) handling multiple feature values. Strategies to address these challenges include: crosschecking identifiers for robust data retrieval and integration; applying clinical knowledge in identifying feature types, in addressing redundancy and granularity, and in accommodating multiple feature values; and investigating missing patterns adequately.Conclusions This article contributes to literature by providing a roadmap to inform structured EHR data preprocessing. It may advise researchers on potential pitfalls and implications of methodological decisions in handling structured data, so as to avoid biases and help realize the benefits of the secondary use of EHR data.Citation: Ferrão JC, Oliveira MD, Janela F, Martins HMG. Preprocessing structured clinical data for predictive modeling and decision support – a roadmap to tackle the challenges.