Before collecting data, each model ran for a 2-minute period to allow for piston stabilization. Data were then collected for a 1-minute period. This procedure was repeated three times for each model. Although the ASL 5000 provided several data output values, the variables of interest for this validation study were VT and TI. Data from the ASL 5000 were reviewed in the postrun analysis section of the software. Using the utilities option, data were exported to Excel (Microsoft, Redmond, CA) for management and SAS (Version 9.3; SAS Institute Inc, Cary, NC) for analysis.
The primary outcomes of interest (VT, TI) obtained from the ASL 5000 were compared with data abstracted from the literature by a one-sample t test. Follow-up testing was performed using Wilcoxon signed rank test to account for departures from normality, with similar findings noted. For repeated measures of primary variables of interest, the Friedman test with post hoc multiple pairwise comparisons was used to assess for differences between each of the experimental iterations, for each model. Statistical significance was established at a P value of less than 0.05, unless otherwise noted.
Post hoc assessment of clinical relevance, as it relates to statistical significance, was completed once all results were available.
One-sample testing revealed evidence of statistically significant differences between published characteristics or the clinical standard used in this study and values produced by the ASL 5000 (VT and TI) for each iteration and model (P < 0.01 for all). The maximum differences of means (experimental iteration mean − clinical standard mean) are the following: term infant without lung disease (TI = 0.09 s, VT = 0.29 mL), infant with severe BPD (TI = 0.08 s, VT = 0.17 mL), child without lung disease (TI = 0.10 s, VT = 0.17 mL), and child with neuromuscular disease (TI = 0.09 s, VT = 0.57 mL). The maximum difference of means in VT noted for all models was less than 1.6% of the scripted values. The maximum difference of means in TI ranges was 18% for all models. Table 5 provides the summary statistics by model and iteration.
The Friedman test results provided strong evidence of an overall difference in VT and TI for the infant with severe BPD model (P < 0.01 for both) and evidence of an overall difference in TI for the child with neuromuscular disease model (P = 0.03). There were no statistically significant differences in the term infant without lung disease or child without lung disease models for TI and VT (P = 0.3, 0.1, 0.9, and 0.7, respectively). Table 6 provides results of Friedman test along with Tukey multiple pairwise comparisons for each of the models. Post hoc analysis of Friedman test with multiple pairwise comparisons using Tukey adjusted α demonstrated statistically significant differences for VT between the first and second iterations (difference of medians = 0.01 mL), as well as the first and third iterations (difference of medians = 0.02 mL) in the infant with severe BPD model. Inspiratory time was significantly different, in the infant with severe BPD model, between the first and third iterations (difference of medians = 0.01 s) as well as between the second and third iterations (difference in medians < 0.01 s). In the child with neuromuscular disease model, TI was significantly different between the first and third iterations as well as between the second and third iterations (difference in medians < 0.01 s for both).
Simulation studies are often used to examine device performance using models of the respiratory system19 because they carry no risk of harm and, compared with using humans or animals, are less cumbersome to work with. However, if the simulated model is unable to consistently replicate the model characteristics, it is difficult to distinguish whether the differences in the variables of interest were due to inconsistencies in the model or differences in the operational characteristics of the devices evaluated. This study used the ASL 5000 lung simulator. The precision with which this simulator can produce consistent results is dependent on the lack of variation with which the piston moves within the casing. This variation is most likely to occur between experimental iterations.20 Therefore, time or the number of repeated iterations of each model was an important factor to consider when validating a simulated model. Repeating each script more than once enabled the researchers to determine whether the ASL 5000 was able to produce consistent breath characteristics over repeated iterations.
Once the model was scripted, simulator calibration and a stabilization period were the essential steps performed before collecting our data. Although the ASL 5000 was calibrated before use and a stabilization period was provided, there exists the potential for variation in piston movement, which can affect results. A 2-minute stabilization period was used to prevent any initial disruption in piston motion, which may occur as the script was initializing, from contaminating data collection. The 2-minute period was chosen from our previous experience modeling disease states and using pediatric models to evaluate respiratory equipment.19–22
The ASL 500 has a 500-Hz sampling rate. Based on this sampling rate and our previous experience with device evaluation using simulated models, a 1-minute data collection period was considered sufficient to obtain respiratory breath data.19–22 With respect to this study, the inconsistencies or variation in VT and TI between iterations were manifested in small changes (10ths–100ths of mL and seconds, respectively). Variation was minimal enough between experimental iterations to prevent analysis by traditional repeated measures analysis of variance, necessitating the Friedman test approach. This minimal variation provides evidence of simulator precision over time.
Statistically significant differences were found between ASL 5000 values and published or target values for VT and TI. In evaluating those differences, it was essential to consider their relevance to the outcome of our work. The greatest magnitude of differences between published values and the values obtained from the ASL 5000 were negligible. Maximum difference in VT was less than 1.6%, which equates to 0.6 mL. Similarly, the maximum difference in TI was 18% or less than 0.1 second. During normal breathing, variations in VT and TI do occur, and because the variations we observed between published values were so small, the researchers assessed these differences as not clinically relevant.
Although statistically significant, the magnitude of differences and standard deviations found in VT and TI between the iterations for each model were very small, relative to the scripted values obtained from the literature. Our post hoc determination of clinically irrelevant differences was supported in the literature reviews used to construct the models we sought to validate.
These differences or inconsistencies were smaller than those that naturally occur with breathing in human subjects. Small variations between iterations were less than variations in VT breaths that occur in human subjects without lung disease.6
Limitations of the Simulated Models
The propensity for great variations in VT in infants and children with lung disease and respiratory muscle weakness depends on the severity of lung disease and the amount of muscle wasting and respiratory muscle dysfunction, respectively.23 We did not account for this variability in our child with neuromuscular disease model because of the specific intent with which we will use the model. This is, however, a limitation of our model and perhaps its use. There are situations where it would be essential to account for and script the previously mentioned variations in a simulated model. Those variations in respiratory muscle function would be important to incorporate in a model if the model were used to evaluate a specific interaction between a simulated subject and a device. An example of this would be constructing a simulated model (n of 1) to determine ventilator settings, which would enhance synchrony for a neuromuscular patient transitioning from a ventilator in the intensive care unit to a portable ventilator for home use. A simulated model of the patient, used to determine the settings for ventilator parameters to enhance patient-ventilator synchrony, would replace the need for a “trial-and-error” method of adjusting these settings while the patient is receiving mechanical ventilation. In an n of 1 study, where the clinician may want to model the characteristics of a patient to determine ventilator settings that would improve synchrony or evaluate the effect certain ventilator changes will have on lung distension, it is crucial to use a double-compartment lung model. In our clinical practice, we routinely use the ASL at the bedside to evaluate sensitivity and positive end expiratory pressure (PEEP) settings before transitioning a technology dependent patient a ventilator used in an intensive care unit to a portable home ventilator. For this use, it is standard to use a double-compartment lung model.
Because a single-lung compartment was used for each of our models, their utility may be limited. A single-compartment lung model does not account for differences in heterogeneity in respiratory mechanics, which occurs with disease states. Therefore, our models may not accurately reflect the heterogeneity in respiratory mechanics, which may be an important consideration, such as in the previously mentioned n of 1 construct.
Because inconsistencies can occur with simulated models, it is essential to evaluate magnitude of differences between the desired and measured variable(s) of interest before use. In our study, the magnitude of variation within and between iterations for each model was minimal with clinically irrelevant significant effects noted in the disease state models. The ASL 5000 lung simulator was demonstrated to be an effective way of validating these neonatal and pediatric respiratory models.
1. Blakeman TC, Branson RD. Evaluation of 4 new generation portable ventilators. Respir Care
2. Blakeman TC, Rodriquez D Jr, Hanseman D, et al. Bench evaluation of 7 home-care ventilators. Respir Care
3. Boussen S, Gainnier M, Michelet P. Evaluation of ventilators used during transport of critically ill patients: a bench study. Respir Care
4. Gonzales JF, Russian CJ, Gregg Marshall S, et al. Comparing the effects of rise time and inspiratory cycling criteria on 6 different mechanical ventilators. Respir Care
5. Hsia D, DiBlasi RM, Richardson P, et al. The effects of flexible bronchoscopy on mechanical ventilation in a pediatric
lung model. Chest
6. Gerhardt T, Hehre D, Feller R, et al. Pulmonary mechanics
in normal infants and young children during first 5 years of life. Pediatr Pulmonol
8. Gappa M, Pillow JJ, Allen J, et al. Lung function tests in neonates and infants with chronic lung disease: lung and chest-wall mechanics. Pediatr Pulmonol
9. Gerhardt T, Hehre D, Feller R, et al. Serial determination of pulmonary function in infants with chronic lung disease. J Pediatr
10. Papastamelos C, Panitch HB, Allen JL. Chest wall compliance in infants and children with neuromuscular disease
. Am J Respir Crit Care Med
1996;154(4 Pt 1):1045–1048.
11. Wohl ME, Stigol LC, Mead J. Resistance of the total respiratory system in healthy infants and infants with bronchiolitis. Pediatrics
12. Hazinski M. Children are different. In: Hazinski M. Manual of Pediatric
Critical Care. 1st ed. St. Louis, Mosby-Year Book, 1999:1–18.
13. Hjalmarson O, Sandberg KL. Lung function at term reflects severity of bronchopulmonary dysplasia
. J Pediatr
14. Brostrom EB, Thunqvist P, Adenfelt G, et al. Obstructive lung disease in children with mild to severe BPD. Respir Med
15. May C, Kennedy C, Milner AD, et al. Lung function abnormalities in infants developing bronchopulmonary dysplasia
. Arch Dis Child
16. Greenspan JS, Wolfson MR, Locke RG, et al. Increased respiratory drive and limited adaptation to loaded breathing in bronchopulmonary dysplasia
. Pediatr Res
17. Rochester DF, Esau SA. Assessment of ventilatory function in patients with neuromuscular disease
. Clin Chest Med
18. Mulreany LT, Weiner DJ, McDonough JM, et al. Noninvasive measurement of the tension-time index in children with neuromuscular disease
. J Appl Physiol
19. Volsko TA, Hoffman J, Conger A, et al. The effect of targeting scheme on tidal volume delivery during volume control mechanical ventilation. Respir Care
20. Van Fleet H, Dunn DK, McNinch NL, et al. Evaluation of functional characteristics of 4 oscillatory positive pressure devices in a simulated cystic fibrosis model. Respir Care
21. Volsko TA, Fedor K, Amadei J, et al. High flow through a nasal cannula and CPAP effect in a simulated infant
model. Respir Care
22. Bailes SA, Firestone KS, Dunn DK, et al. Evaluating the effect of flow and interface type on pressures delivered with bubble CPAP in a simulated model. Respir Care
23. LoMauro A, Romei M, Priori R, et al. Alterations of thoraco-abdominal volumes and asynchronies in patients with spinal muscle atrophy type III. Respir Physiol Neurobiol
Keywords:© 2018 Society for Simulation in Healthcare
Pediatric; infant; pulmonary mechanics; bronchopulmonary dysplasia; neuromuscular disease; simulation; physiologic modeling