mHealth for image-based diagnostics of acute burns in resource-poor settings : studies on the role of experts and the accuracy of their assessments

Sammanfattning: Background The global burden of injuries could be reduced by the provision of adequate emergency care but this is impeded by barriers like poor access to specialised care. Accurate and timely diagnostics supported by mobile technology could enable more effective use of scarce resources. Acute burns are one type of injury amenable to image-based diagnostic consultation. Aim To increase the knowledge about how remote diagnostic assistance for burn injuries influences the role and work of clinicians. Focus is placed on medical experts and the context of resource-poor settings. Methods The thesis comprises one qualitative and three quantitative studies. In study I (I), on burn injury epidemiology, case reports of burns patients (n=1 915) treated in eight emergency centres during one year (baseline) were reviewed. Gender differences were examined using incidence rate ratios and a two-sample test of proportions. In study II (II), image quality on handheld devices was investigated via an online questionnaire including 18 clinical and non-clinical images viewed on a smartphone, tablet and computer by 27 medical experts. Perception of quality was rated on each image. Linear regression clustered by participant was applied, comparing the ratings for the tablet and smartphone with the ratings for the computer. In study III (III), diagnostic accuracy of assessments of acute burns made via smartphones and tablets was assessed using an online questionnaire with 51 burns images. Each image was assessed by 26 medical experts using their own device. Diagnostic accuracy and intra-rater reliability were assessed with a two-way mixed effect intraclass correlation coefficient (ICC) with 95% confidence interval. In study IV (IV), perceptions of medical experts of remote diagnostic assistance were explored using semi-structured interviews with 15 medical experts. Data were analysed with thematic analysis and positioning theory. Results Burns were most often caused by hot liquids (65.2%) and of minor and moderate severity (80.4%) with the majority of patients treated as outpatients (73.9%). Young children had the highest incidence rates (75.4 per 10 000) and gender differences were mainly observed in adults. A higher proportion of women were treated as outpatients and a higher proportion of men were transferred to higher levels of care (I). Clinicians rated the quality of images viewed on smartphones and tablets higher than when viewed on computer (II). Assessments of burn size by clinicians were accurate overall (ICC=0.82) while the accuracy of burn depth assessments was low (ICC=0.53) except for depth assessments made by South African burns specialists of child burns (III). Four positions of experts were described in current remote consultations: clinical specialist, gatekeeper, mentor and educator. Experts were already engaged in image-based consultation and were confident that images could improve the accuracy of consultation. Verbal communication was described as crucial for critical situations. The app is expected to improve security, information quality, access to experts, and learning (IV). Conclusions Altogether, the findings contribute important information for the implementation of an app for remote diagnostic consultation in resource-poor settings. The acute burns compare in many ways as those seen in other levels of care with only a few potentially challenging characteristics. Clinicians from the field are satisfied with the quality of images seen on smartphones and tablets and can accurately assess burn size by looking at images on their smartphones. Burn depth assessments may be improved by more additional information which can be requested in real clinical practice. The four positions described by the experts were all expected to change with the use of the new app. The current use of WhatsApp for remote burn consultation highlight the changing field in which experts work and can both facilitate and hinder implementation of the app.