Background Smartphones have made it possible for patients to digitally report symptoms before physical primary care visits. Using machine learning (ML), these data offer an opportunity to support decisions about the appropriate level of care (triage). Objective The purpose of this study was to explore the interrater reliability between human physicians and an automated ML-based triage method. Methods After testing several models, a naïve Bayes triage model was created using data from digital medical histories, capable of classifying digital medical history reports as either in need of urgent physical examination or not in need of urgent physical examination. The model was tested on 300 digital medical history reports and classification was compared with the majority vote of an expert panel of 5 primary care physicians (PCPs). Reliability between raters was measured using both Cohen κ (adjusted for chance agreement) and percentage agreement (not adjusted for chance agreement). Results Interrater reliability as measured by Cohen κ was 0.17 when comparing the majority vote of the reference group with the model. Agreement was 74% (138/186) for cases judged not in need of urgent physical examination and 42% (38/90) for cases judged to be in need of urgent physical examination. No specific features linked to the model’s triage decision could be identified. Between physicians within the panel, Cohen κ was 0.2. Intrarater reliability when 1 physician retriaged 50 reports resulted in Cohen κ of 0.55. Conclusions Low interrater and intrarater agreement in triage decisions among PCPs limits the possibility to use human decisions as a reference for ML to automate triage in primary care.

Non-parametric data-driven analysis techniques can be used to study datasets with few assumptions about the data and underlying experiment. Variations of independent component analysis (ICA) have been the methods mostly used on fMRI data, e.g., in finding resting-state networks thought to reflect the connectivity of the brain. Here we present a novel data analysis technique and demonstrate it on resting-state fMRI data. It is a generic method with few underlying assumptions about the data. The results are built from the statistical relations between all input voxels, resulting in a whole-brain analysis on a voxel level. It has good scalability properties and the parallel implementation is capable of handling large datasets and databases. From the mutual information between the activities of the voxels over time, a distance matrix is created for all voxels in the input space. Multidimensional scaling is used to put the voxels in a lower-dimensional space reflecting the dependency relations based on the distance matrix. By performing clustering in this space we can find the strong statistical regularities in the data, which for the resting-state data turns out to be the resting-state networks. The decomposition is performed in the last step of the algorithm and is computationally simple. This opens up for rapid analysis and visualization of the data on different spatial levels, as well as automatically finding a suitable number of decomposition components.

A metabolite corrected arterial input function is a prerequisite for quantification of positron emission tomography (PET) data by compartmental analysis. This quantitative approach is also necessary for radioligands without suitable reference regions in brain. The measurement is laborious and requires cannulation of a peripheral artery, a procedure that can be associated with patient discomfort and potential adverse events. A non invasive procedure for obtaining the arterial input function is thus preferable. In this study, we present a novel method to obtain image-derived input functions (IDIFs). The method is based on calculation of the Pearson correlation coefficient between the time-activity curves of voxel pairs in the PET image to localize voxels displaying blood-like behavior. The method was evaluated using data obtained in human studies with the radioligands [ 11 C]flumazenil and [ 11 C]AZ10419369, and its performance was compared with three previously published methods. The distribution volumes (V T ) obtained using IDIFs were compared with those obtained using traditional arterial measurements. Overall, the agreement in V T was good (B3% difference) for input functions obtained using the pairwise correlation approach. This approach performed similarly or even better than the other methods, and could be considered in applied clinical studies. Applications to other radioligands are needed for further verification.

Although we know that asking questions is an essential aspect of online tutoring, there is limited research on this topic. The aim of this paper was to identify commonly used direct question types and explore the effects of using these question types on conversation intensity, approach to tutoring, perceived satisfaction and perceived learning. The research setting was individual online synchronous tutoring in mathematics. The empirical data was based on 13,317 logged conversations and a questionnaire. The tutors used a mix of open, more student-centred questions, and closed, more teacher-centred questions. In contrast to previous research, this study provides a more positive account indicating that it is indeed possible to train tutors to focus on asking questions, rather than delivering content. Frequent use of many of the question types contributed to increased conversation intensity. However, there were few question types that were associated with statistically significant effects on perceived satisfaction or learning. There are no silver bullet question types that by themselves led to positive effects on perceived satisfaction and learning. The question types could be used by teachers and teacher students when reflecting on what types of questions they are asking, and what kind of questions they could be asking.

No abstract