A review of diagnostic accuracy in the smartphone era

Abstract

Background Rapid and sophisticated development in mobile and wireless technologies has contributed to the transformation of healthcare service delivery across the globe. While over one billion smartphones were sold to end users in 2014, there are on average 107 mobile phone subscriptions per 100 persons worldwide. The diverse functionality and utility of these devices contributed to their board application to the general public as well as those working in a clinical setting. The use of smartphone plays a fundamental role in patient management as it may improve access to care, outcomes in diseases that are high in incidence and prevalence and where time-to-treatment is a key predictor of outcome as well as reduce costs. Though there are many benefits that smartphones may bring to the clinical setting, it is important to understand and assess the quality of support they offer. The aim of this review is to assess smartphone’s performance and reliability as an aid in clinical diagnosis, in particular the level of accuracy it achieves and it’s potential to be fully integrated into mainstream clinical practice.

Methods A literature search using keywords was performed in PubMed. Following the defined inclusion and exclusion criteria, 11 relevant papers were identified and included in this review. Their compliance with the Standards for the Reporting of Diagnostic Accuracy (STARD) was critically assessed to evaluate the quality of their methods and reporting.

Findings All 11 articles produced findings to demonstrate that smartphone, equipped with the right specifications and functions, is a reliable aid to enable the formation of accurate diagnosis. Studies that reported statistical findings on accuracy, sensitivity and specificity (required by STARD), as well as positive and negative predictive values suggested highly sensitive and specific results with largely good level of accuracy achieved. In determining whether the diagnostic agreement between the index tests and the reference standards were better than chance alone, majority of the studies revealed ‘moderate’ to ‘almost perfect’ level of concordance with k-value ranging from 0.60 to 1. Respective internal and external validity and the findings’ implications to healthcare were also discussed.

Conclusion The evidence gathered in this review deemed not sufficient to conclude the reliability and accuracy of smartphone as a tool to be integrated into clinical practice to aid the formation of diagnosis. However, as technology continues to advance, quality of smartphones can be expected to improve further for boarder adoption in mainstream clinical practice. Rigorous evaluation and validation, establishment of appropriate standards, as well as study of best practice in the integration of these devices in clinical practice are essential to ensure quality and safety.