Comparison of Left Ventricular Hypertrophy by Electrocardiogram and Echocardiography Using Novel Data Analytics Tool
Jacqueline Weinberg1, Puran Bhagat2,
Szymon Bieganski2, A. Shane Brown3, Melvin Kanasseril2, Tim Lavers3,
Kirk Spencer4, Craig Sable1
1Children’s National Health System, Washington, DC, USA
2Philips Healthcare, Andover, MA, USA
3Pacific Knowledge Systems, Sydney, Australia
4University of Chicago Medical Center, Chicago, IL, USA
BACKGROUND: Left ventricular hypertrophy (LVH) is a common finding on paediatric ECG leading to many referrals for echocardiography (echo). A novel data analytics client has the ability to rapidly mine data from echo and ECG databases. We hypothesized that the client would allow for comparison of ECG and echo data in a large paediatric population to evaluate ECG as a screening tool in our practice.
METHODS: A test database was created from existing echo and ECG data from September 2011 to September 2014 in a large regional practice in Washington, DC. The data analytics client was applied to this database to identify patients 0-18 years of age with LVH on ECG, with complete echo performed on the same day in an outpatient setting. ECG R wave in lead V6 (V6R) and echo left ventricle (LV) measurements, Z-scores, and LV qualitative finding codes (FC) were evaluated. V6R was correlated with Z-score of LV diameter in diastole (LVIDd) and LV mass indexed (LVMI) to body surface area.
RESULTS: Of 35,007 encounters evaluated, 2,622 cases were identified meeting inclusion criteria. Of these, 81% contained FC for normal LV structure and size. Abnormal LV FC included 10% with dilated LV, 3% with concentric LVH, and 1% with asymmetric LVH. Mean V6R was 26.8mm (2.68mV). Of 2,632 cases with LVIDd Z-score, the mean was +0.4 (±1.7), and 232 (9%) had LVID Z-score >2. Of 2,690 cases with LV mass, mean LVMI was 74.5 g/m2 (±29), and 371 (14%) had LVMI >100 g/m2. When evaluated by age, the youngest patients were most likely and the oldest patients were least likely to have normal LV FC. V6R had a weakly positive correlation with LVIDd Z-score and LVMI (Table).
CONCLUSIONS: The data analytics client was able to mine a database of ECG and echo reports, comparing LVH by ECG and LV measurements and qualitative findings by echo, identifying abnormal LV by echo in only 19% of cases. Further use of this novel tool may assist in adjusting current normal ECG values for paediatric patients.
|Total*||Normal LV FC*||LVID Z-score†||Abnormal LVIDd Z-score*||LVMI (g/m2)†||Abnormal LVMI*||V6R (mm)†||V6R vs. LVIDd Z-score‡||V6R vs. LVMI‡|
|All||2,622||2,114 (81)||+0.4 (1.6)||232 (9)||74.5 (29)||371 (14)||26.8 (9.0)||0.21 (<0.0001)||0.27 (<0.0001)|
|Male||1,675 (64)||1,364 (81)||+0.4 (1.4)||130 (8)||76.9 (26.8)||251 (15)||27.5 (8.5)||0.13 (<0.0001)||0.20 (<0.0001)|
|Female||946 (36)||750 (79)||+0.4 (2.0)||101 (11)||70.1 (32.1)||120 (13)||25.5 (9.8)||0.29 (<0.0001)||0.27 (<0.0001)|
* N (%), † mean (SD), ‡ r (p value)