The importance of patient characteristics, operators, and image quality for accurate diagnoses of heart failure by general practitioners using hand-held ultrasound

Abstract

Abstract Background Echocardiography is a cornerstone in diagnosing heart failure (HF). Hand-held ultrasound devices (HUDs) are established diagnostic tools. After a period of training, inexperienced users can utilize HUDs in HF diagnostics. Developments have allowed for decision-support software such as automatic analyses of ejection fraction (autoEF), mitral annular plane systolic excursion (autoMAPSE) and telemedicine. It is not well established how patient characteristics, operator qualifications and image quality influence the clinical usefulness of HUDs. Furthermore, it is not known whether associations of patient-, operator- and image quality characteristics with diagnostic accuracy is consistent across different decision-support software. Purpose To evaluate associations of patient-, operator- and image quality characteristics with the accuracy of HF diagnostics by general practitioners (GPs) using HUDs with and without decision-support software. Method After a period of training, five GPs examined outpatients with suspected HF. They examined the patients clinically and subsequently by adding HUD examinations, autoEF and autoMAPSE. Recorded images were interpreted by external cardiologists using a telemedical software program, giving immediate feedback to the GPs. After each step the GPs considered whether the patients had HF. Diagnostic accuracy was evaluated by comparing to reference echocardiograms by cardiologists. Image quality was evaluated by external cardiologists scoring each of the categories; LV view, LV alignment, apical misposition, mitral annular assessment and visible segments. The associations of possible predictors with the diagnostic accuracy were tested by univariate regression. Results Of 166 patients, 28 were diagnosed with HF after reference examinations. The GPs correctly diagnosed 72% of the patients after HUD examinations, 55% after autoMAPSE, 57% after autoEF and 76% after telemedical support. Atrial fibrillation (AF) was present in 40 patients, 27 had chronic obstructive pulmonary disease and 121 had body mass index >25 kg/m2. Ongoing AF was significantly associated with reduced accuracy using HUD examination alone and in combination with autoEF (p<0.05). No other patient characteristics were associated with GPs' diagnostic accuracy. There were significant differences between different operators when autoMAPSE was used (p<0.05). Image quality indices were not significantly associated with diagnostic accuracy (p≥0.2). Conclusion Atrial fibrillation and operator characteristics were associated with reduced diagnostic accuracy when basing diagnosis on automatic decision support software. No other association of neither patient-, operator- nor image quality characteristics with the accuracy of GP diagnostics was found. Thus, it is important that the clinical society is aware of the limited to modest accuracy of HF diagnostics by inexperienced users of hand-held ultrasound including novel decision-support software. Funding Acknowledgement Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Research council of NorwayNorwegian University of Science and Technology