TY - JOUR
T1 - A novel method of calculating stroke volume using point-of-care echocardiography
AU - Aligholizadeh, Ehson
AU - Teeter, William
AU - Patel, Rajan
AU - Hu, Peter
AU - Fatima, Syeda
AU - Yang, Shiming
AU - Ramani, Gautam
AU - Safadi, Sami
AU - Olivieri, Peter
AU - Scalea, Thomas
AU - Murthi, Sarah
N1 - Publisher Copyright:
© 2020 The Author(s).
PY - 2020/8/20
Y1 - 2020/8/20
N2 - Background: Point-of-care transthoracic echocardiography (POC-TTE) is essential in shock management, allowing for stroke volume (SV) and cardiac output (CO) estimation using left ventricular outflow tract diameter (LVOTD) and left ventricular velocity time integral (VTI). Since LVOTD is difficult to obtain and error-prone, the body surface area (BSA) or a modified BSA (mBSA) is sometimes used as a surrogate (LVOTDBSA, LVOTDmBSA). Currently, no models of LVOTD based on patient characteristics exist nor have BSA-based alternatives been validated. Methods: Focused rapid echocardiographic evaluations (FREEs) performed in intensive care unit patients over a 3-year period were reviewed. The age, sex, height, and weight were recorded. Human expert measurement of LVOTD (LVOTDHEM) was performed. An epsilon-support vector regression was used to derive a computer model of the predicted LVOTD (LVOTDCM). Training, testing, and validation were completed. Pearson coefficient and Bland-Altman were used to assess correlation and agreement. Results: Two hundred eighty-seven TTEs with ideal images of the LVOT were identified. LVOTDCM was the best method of SV measurement, with a correlation of 0.87. LVOTDmBSA and LVOTDBSA had correlations of 0.71 and 0.49 respectively. Root mean square error for LVOTDCM, LVOTDmBSA, and LVOTDBSA respectively were 13.3, 37.0, and 26.4. Bland-Altman for LVOTDCM demonstrated a bias of 5.2. LVOTDCM model was used in a separate validation set of 116 ideal images yielding a linear correlation of 0.83 between SVHEM and SVCM. Bland Altman analysis for SVCM had a bias of 2.3 with limits of agreement (LOAs) of-24 and 29, a percent error (PE) of 34% and a root mean square error (RMSE) of 13.9. Conclusions: A computer model may allow for SV and CO measurement when the LVOTD cannot be assessed. Further study is needed to assess the accuracy of the model in various patient populations and in comparison to the gold standard pulmonary artery catheter. The LVOTDCM is more accurate with less error compared to BSA-based methods, however there is still a percentage error of 33%. BSA should not be used as a surrogate measure of LVOTD. Once validated and improved this model may improve feasibility and allow hemodynamic monitoring via POC-TTE once it is validated.
AB - Background: Point-of-care transthoracic echocardiography (POC-TTE) is essential in shock management, allowing for stroke volume (SV) and cardiac output (CO) estimation using left ventricular outflow tract diameter (LVOTD) and left ventricular velocity time integral (VTI). Since LVOTD is difficult to obtain and error-prone, the body surface area (BSA) or a modified BSA (mBSA) is sometimes used as a surrogate (LVOTDBSA, LVOTDmBSA). Currently, no models of LVOTD based on patient characteristics exist nor have BSA-based alternatives been validated. Methods: Focused rapid echocardiographic evaluations (FREEs) performed in intensive care unit patients over a 3-year period were reviewed. The age, sex, height, and weight were recorded. Human expert measurement of LVOTD (LVOTDHEM) was performed. An epsilon-support vector regression was used to derive a computer model of the predicted LVOTD (LVOTDCM). Training, testing, and validation were completed. Pearson coefficient and Bland-Altman were used to assess correlation and agreement. Results: Two hundred eighty-seven TTEs with ideal images of the LVOT were identified. LVOTDCM was the best method of SV measurement, with a correlation of 0.87. LVOTDmBSA and LVOTDBSA had correlations of 0.71 and 0.49 respectively. Root mean square error for LVOTDCM, LVOTDmBSA, and LVOTDBSA respectively were 13.3, 37.0, and 26.4. Bland-Altman for LVOTDCM demonstrated a bias of 5.2. LVOTDCM model was used in a separate validation set of 116 ideal images yielding a linear correlation of 0.83 between SVHEM and SVCM. Bland Altman analysis for SVCM had a bias of 2.3 with limits of agreement (LOAs) of-24 and 29, a percent error (PE) of 34% and a root mean square error (RMSE) of 13.9. Conclusions: A computer model may allow for SV and CO measurement when the LVOTD cannot be assessed. Further study is needed to assess the accuracy of the model in various patient populations and in comparison to the gold standard pulmonary artery catheter. The LVOTDCM is more accurate with less error compared to BSA-based methods, however there is still a percentage error of 33%. BSA should not be used as a surrogate measure of LVOTD. Once validated and improved this model may improve feasibility and allow hemodynamic monitoring via POC-TTE once it is validated.
KW - Cardiac output
KW - Echocardiography
KW - Fluid resuscitation
KW - Hemodynamic monitoring
KW - POCUS
UR - http://www.scopus.com/inward/record.url?scp=85089769469&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85089769469&partnerID=8YFLogxK
U2 - 10.1186/s12947-020-00219-w
DO - 10.1186/s12947-020-00219-w
M3 - Article
C2 - 32819371
AN - SCOPUS:85089769469
SN - 1476-7120
VL - 18
JO - Cardiovascular Ultrasound
JF - Cardiovascular Ultrasound
IS - 1
M1 - 37
ER -