Study protocol
To perform the study, we used the SimulResp versions 2015.10.11.01 (2015 version) and 2012.06.09.01 (2012 version). The two versions were considered equivalent as the formula within the software remained unchanged, only the design of the interface changed. All tests were performed by OF or DB. OF ran the SimulResp 2012 version, under windows 7 (Microsoft, Albuquerque, NM), on a 2011 computer SONY VAIO (Sony, Minato, Japan) equipped with a Intel Core i3 processor (Intel, Santa Clara, CA) and DB ran the 2015 version, under window 8.1, on a 2016 computer ASUS X541UV (Asus, New Taipei, Taiwan) equipped with a 2.50 GHz dual core Intel Core i7-6500U processor. Blood gas values: pH, arterial partial pressure in CO2(PaCO2), arterial partial pressure in O2(PaO2) and arterial oxygen saturation in O2 (SaO2) were simulated at different time points for several types of subjects with different characteristics. For each combination, the following subjects’ characteristics were specified: age, weight, height, type of patient, and ventilation mode (spontaneous (SV) or mechanical (MV) ventilation). The characteristics were systematically entered in the same order: age, weight, height, type of patient, gender, ventilation mode. “Personalized” type of patient was selected when mechanical ventilation was chosen, in order to enter the ventilation parameters: inspired oxygen fraction (FiO2), positive end expiratory pressure (PEEP), respiratory rate (RR) and tidal volume (Vt). By default, the gender was set to male and switched to female as needed. When available hemoglobin value was entered or considered as normal when missing. Each simulation was repeated three times to address potential input error and in case of a mismatch, a fourth simulation was performed to verify or eventually replace the outlier.
A preliminary study was conducted to assess for several assumptions of the SimulResp configuration: the range of age within which SimulResp was supposed to be accurate and its ability to remain stable even when the simulation speed was modified, from 2 to 4000 times. Blood gas values (pH, PaCO2, PaO2 and SaO2) were simulated for several fictive healthy subjects with different characteristics: gender (M, F) and age (1, 2 ,4, 6, 8, 10, 12, 14, 16, 18 years old), with a 50th percentile weight and height at different simulation speeds (from 1 to 4000) and were collected after a virtual patient clinical evolution (VPCE) of 30 minutes.
First phase of the study
The first phase intends to assess the accuracy, the robustness, the repeatability and the reproducibility of SimulResp when simulating blood gas values of healthy fictive subjects. The first phase of the study consisted in assessing SimulResp’s predictions with simulated healthy subjects. Based on the results of the preliminary study, the tests were restricted to subjects from 8 to 18 years old (8, 10, 12, 14, 16, 18 years), with different characteristics; gender (M, F) and weight (10th, 50th and 90th percentile). Blood gas values were collected 3 times for each patient at a VPCE of 30 minutes, 3 and 24 hours, with a simulation speed of respectively 64, 258 and 1048. This study was conducted for both spontaneously breathing and mechanically ventilated subjects. For mechanically ventilated subjects, the following ventilation parameters were set: FiO2 21%, PEEP 3 cmH2O, Vt 7.5mL/kg, normal RR for age (9)