Reply to Aquino-Esperanza et al.: Considerations for an Optimal Electrical Activity of the Diaphragm Threshold for Automated Detection of Ineffective Efforts

From the Authors:

We greatly appreciate the interest of Aquino-Esperanza and colleagues in our research letter (1) regarding the influence of suboptimal filtering of the electrical activity of the diaphragm (EAdi) signal on the detection of patient–ventilator asynchronies. In that letter, we raised the concern that cardiac activity–related artifacts in the EAdi signal may be mistakenly detected as ineffective efforts when the EAdi threshold is too low. Based on this work, Aquino-Esperanza and colleagues have thoughtfully reanalyzed the performance of their Better Care algorithm (2) to find an appropriate EAdi threshold for the automatic detection of ineffective efforts. They conclude that increasing the EAdi threshold from 1 μV to 2.3 μV improved the sensitivity of their algorithm and maintained adequate specificity. We appreciate this careful reanalysis and agree with the authors that a threshold >2 μV is reasonable. It should be noted that in our work (1), EAdi artifacts were mostly <4 μV, but we agree that a threshold of 4 μV would be clinically disproportionate and increase the false-negative rate.

We also agree with the authors that a personalized adaptive EAdi threshold may improve the performance of automatic detection of true ineffective efforts. We considered testing this with our dataset; however, the incidence of true ineffective efforts was too low (1). In contrast, because the processing of these EAdi artifacts is a technical issue, it might be rather impossible to distinguish between artifacts and true ineffective efforts based on a certain EAdi threshold solely. As part of our earlier work, we aimed to quantify waveform characteristics of the cardiac activity–related artifacts (e.g., slope of the inspiratory EAdi increase, timing, and amplitude) and predict the occurrence of these artifacts based on patient characteristics. For instance, we hypothesized that cardiac activity–related peaks had steeper increases (“sharp waves”, possibly consistent with fast cardiac depolarization); however, slopes of the artificial and true peaks were similar on average, and artificial peaks with both lower and higher slopes compared with true EAdi peaks were found within patients. Furthermore, factors such as the presence of ventricular hypertrophy were not related to the occurrence of these artifacts. We did not include these findings in our research letter, as this is clinically not very helpful at this time.

Importantly, the main challenge with developing a (personalized) EAdi threshold using signal characteristics is the uncertainty regarding how the ventilator algorithm processes the EAdi signal (“black box”). In addition, artifacts may look different when originating from cardiac or catheter movements (mechanical artifact) or when being secondary to inefficient filtering of the QRS complex (electrical artifact). This requires specific analysis of the raw diaphragm electromyography signal, and indeed, complex mathematical techniques might offer a solution. As the diaphragm electromyography is not available to the clinician to test this approach, we reason that using a threshold >2 μV as proposed by Aquino-Esperanza and colleagues is an appropriate practical solution for automatic detecting of ineffective efforts in large datasets. However, one should keep in mind that artifacts of larger amplitudes can be present and that careful consideration of the EAdi catheter position and signal quality is required when using EAdi for clinical decision-making and research.