QUADRES PREUS Nº 2

properly identifying injections following a different medication, not including saline flushes (as this was assessed previously in Section 4.2.1.2), over 10 repetitions of each medication combination of the 4 medications (propofol, epinephrine, esmolol, and succinylcholine) and over the same injection volumes as before (1.0, 2.0, 5.0, 8.0mL). For example, 10 repetitions were performed for 2.0mL propofol injection following 2.0mL esmolol injection, and visa-versa, for all combinations of medications at all injection volumes. Similar to previous experiments, injection volumes were not permuted during this experiment; i.e. injection volumes were blocked so that all 2.0mL volumes were performed together, and 2.0mL injection volumes were not performed with 5.0mL injection volumes. To make this point clearer, 2.0mL propofol was never injected after 5.0mL esmolol—2.0mL injections of medication_A were always paired with 2.0mL injections of medication_B. The full set of all 12 injection pairings is shown in Table 15.

Table 15. Injection pairings for different medication injections without saline flush between injections ID Lower Salt Concentration followed by

Higher Salt Concentration ID

Higher Salt Concentration followed by Lower Salt Concentration A Propofol followed by Epinephrine G Epinephrine followed by Propofol B Propofol followed by Esmolol H Esmolol followed by Propofol C Propofol followed by Succinylcholine I Succinylcholine followed by Propofol D Epinephrine followed by Esmolol J Esmolol followed by Epinephrine E Epinephrine followed by Succinylcholine K Succinylcholine followed by Epinephrine F Esmolol followed by Succinylcholine L Succinylcholine followed by Esmolol *IDs are given to each injection pairing to condense future tables and for future in-text

referencing.

For each of the 12 injection pairings, 10 repetitions of injections were performed at each of the 4 levels of volume, resulting in a total of 240 injections for each of the 4 individual medications. Thus, there were a total of 960 trials of injections performed.

Results

The results from the paired injection trials are summarized in Table 16. All trials successfully identified the paired injections properly except for paired injections of 1.0mL for injection IDs (I) and (K), which corresponded to propofol (0.3 g/L) and epinephrine (1.0 g/L) being injected after succinylcholine (3.0 g/L), respectively. Out of 960 injections, only 11 were improperly identified by the system—the system was over 98% accurate over all injections during this experiment.

Table 16. Frequency of properly detecting the injection of paired injections for each of the 12 medication

pairings across all volumes

Paired Injection Volume of Medication to Administer (mL) ID Previous Salt Concentration (g/L) Salt Concentration to Detect (g/L) 1.0 2.0 5.0 8.0 A 0.3 1.0 10 10 10 10 B 0.3 2.0 10 10 10 10 C 0.3 3.0 10 10 10 10 D 1.0 2.0 10 10 10 10 E 1.0 3.0 10 10 10 10 F 2.0 3.0 10 10 10 10 G 1.0 0.3 10 10 10 10 H 2.0 0.3 10 10 10 10 I 3.0 0.3 2 10 10 10 J 2.0 1.0 10 10 10 10 K 3.0 1.0 7 10 10 10 L 3.0 2.0 10 10 10 10 Discussion

The results from the paired injection experiments are not surprising. In the previous experiments in Section 4.2.1.2 where medications were always injected after a saline flush, the medications were properly identified in 160 out of 160 injections. In those cases, a medication with a higher concentration of salt solution was injected after a medication with a lower concentration of salt solution. Similarly, in this experiment for cases (A) through (F), a medication with a higher concentration of salt solution was injected after a medication with a lower concentration of salt solution. In these cases, the medications were properly identified in 240 out of 240 injections.

In the previous experiments in Section 4.2.1.3 where saline flushes were performed after each medication injection, saline flushes (0.0g/L) of 1.0mL were not properly identified by the system in 10 out of 10 injections when following injections of succinylcholine (3.0g/L). Additionally, from previous experiments, saline flushes of 1.0mL were not properly identified by

the system in 9 out of 10 injections when following injections of esmolol (2.0g/L). In this experiment, it was observed that propofol (0.3 g/L) injections of 1.0mL were not properly identified by the system in 8 out of 10 injections when following injections of succinylcholine, and epinephrine (1.0 g/L) injections were also not properly identified by the system in 3 out of 10 injections when following injections of succinylcholine. In all of these cases, the system did not properly identify the injected medication when the injected medication had a lower salt concentration by at least 2.0g/L than the previously injected medication. These findings suggest that the system’s current detection method is flawed for low-volume injections when a medication with -2.0g/L difference or more is injected after a medication with a higher salt concentration (i.e. a 0.5g/L medication injected after a 3.0g/L medication).

From the results of all injection experiments performed in this dissertation, when a medication with a higher concentration of salt solution is injected after a medication with a lower concentration of salt solution, there is high probability that it will be properly identified, especially with larger injection volumes. Conversely, when a medication with a lower concentration of salt solution is injected after a medication with a higher concentration of salt solution, there is a low probability that it will be properly identified with smaller injection volumes, especially when the difference in salt concentrations between the medications is high. This behavior is not amenable for detecting user errors during training scenarios, as the user may inject any volume of the medications in any order.

One possible solution is to limit the medications to be mapped across a tighter range of salt concentrations, but that would require low variability in conductivity measurements across all combinations of medication injections in order to achieve high accuracy when identifying

medication injections. Future testing could explore the limits to the range of salt concentrations while maximizing the amount of medications that are possible to properly identify by the system. Another possible solution, as mentioned during previous discussion, is to develop a mathematical model that instantaneously predicts what simulated medication is being injected based upon the initial injection rate, rate of change in conductivity, and knowing what was previously injected. Future work should explore development of such a model and perform similar experiments on accuracy to compare with the current detection methods.