Citations

Prediction models aim to use available data to predict a health state or outcome that has not yet been observed. Prediction is primarily relevant to clinical practice, but is also used in research, and administration. While prediction modeling involves estimating the relationship between patient factors and outcomes, it is distinct from casual inference. Prediction modeling thus requires unique considerations for development, validation, and updating. This document represents an effort from editors at 31 respiratory, sleep, and critical care medicine journals to consolidate contemporary best practices and recommendations related to prediction study design, conduct, and reporting. Herein, we address issues commonly encountered in submissions to our various journals. Key topics include considerations for selecting predictor variables, operationalizing variables, dealing with missing data, the importance of appropriate validation, model performance measures and their interpretation, and good reporting practices. Supplemental discussion covers emerging topics such as model fairness, competing risks, pitfalls of “modifiable risk factors”, measurement error, and risk for bias. This guidance is not meant to be overly prescriptive; we acknowledge that every study is different, and no set of rules will fit all cases. Additional best practices can be found in the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines, to which we refer readers for further details.

In 12 anesthetized paralyzed dogs, pulmonary gas exchange and intrapulmonary inspired gas distribution were compared between continuous-flow ventilation (CFV) and conventional mechanical ventilation (CMV). Nine dogs were studied while they were lying supine, and three dogs were studied while they were lying prone. A single-lumen catheter for tracheal insufflation and a double-lumen catheter for bilateral endobronchial insufflation [inspired O2 fraction = 0.4; inspired minute ventilation = 1.7 +/- 0.3 (SD) 1.kg-1.min-1] were evaluated. Intrapulmonary gas distribution was assessed from regional 133Xe clearances. In dogs lying supine, CO2 elimination was more efficient with endobronchial insufflation than with tracheal insufflation, but the alveolar-arterial O2 partial pressure difference was larger during CFV than during CMV, regardless of the type of insufflation. By contrast, endobronchial insufflation maintained both arterial PCO2 and alveolar-arterial O2 partial pressure difference at significantly lower levels in dogs lying prone than in dogs lying supine. In dogs lying supine, the dependent lung was preferentially ventilated during CMV but not during CFV. In dogs lying prone, gas distribution was uniform with both modes of ventilation. The alveolar-arterial O2 partial pressure difference during CFV in dogs lying supine was negatively correlated with the reduced ventilation of the dependent lung, which suggests that increased ventilation-perfusion mismatching was responsible for the increase in alveolar-arterial O2 partial pressure difference. The more efficient oxygenation during CFV in dogs lying prone suggests a more efficient matching of ventilation to perfusion, presumably because the distribution of blood flow is also nearly uniform.

To determine the sensitivity of pulmonary resistance (RL) to changes in breathing frequency and tidal volume, we measured RL in intact anesthetized dogs over a range of breathing frequencies and tidal volumes centering around those encountered during quiet breathing. To investigate mechanisms responsible for changes in RL, the relative contribution of airway resistance (Raw) and tissue resistance (Rti) to RL at similar breathing frequencies and tidal volumes was studied in six excised, exsanguinated canine left lungs. Lung volume was sinusoidally varied, with tidal volumes of 10, 20, and 40% of vital capacity. Pressures were measured at three alveolar sites (PA) with alveolar capsules and at the airway opening (Pao). Measurements were made during oscillation at five frequencies between 5 and 45 min-1 at each tidal volume. Resistances were calculated by assuming a linear equation of motion and submitting lung volume, flow, Pao, and PA to a multiple linear regression. RL decreased with increasing frequency and decreased with increasing tidal volume in both isolated and intact lungs. In isolated lungs, Rti decreased with increasing frequency but was independent of tidal volume. Raw was independent of frequency but decreased with tidal volume. The contribution of Rti to RL ranged from 93 +/- 4% (SD) with low frequency and large tidal volume to 41 +/- 24% at high frequency and small tidal volume. We conclude that the RL is highly dependent on breathing frequency and less dependent on tidal volume during conditions similar to quiet breathing and that these findings are explained by changes in the relative contributions of Raw and Rti to RL.

Obesity is associated with important decrements in lung volumes. Despite this, ventilation remains normally or near normally distributed at least for moderate decrements in functional residual capacity (FRC). We tested the hypothesis that this is because maximum flow increases presumably as a result of an increased lung elastic recoil. Forced expiratory flows corrected for thoracic gas compression volume, lung volumes, and forced oscillation technique at 5–11–19 Hz were measured in 133 healthy subjects with a body mass index (BMI) ranging from 18 to 50 kg/m2. Short-term temporal variability of ventilation heterogeneity was estimated from the interquartile range of the frequency distribution of the difference in inspiratory resistance between 5 and 19 Hz (R5–19_IQR). FRC % predicted negatively correlated with BMI ( r = −0.72, P < 0.001) and with an increase in slope of either maximal ( r = −0.34, P < 0.01) or partial flow-volume curves ( r = −0.30, P < 0.01). Together with a slight decrease in residual volume, this suggests an increased lung elastic recoil. Regression analysis of R5–19_IQRagainst FRC % predicted and expiratory reserve volume (ERV) yielded significantly higher correlation coefficients by nonlinear than linear fitting models ( r2= 0.40 vs. 0.30 for FRC % predicted and r2= 0.28 vs. 0.19 for ERV). In conclusion, temporal variability of ventilation heterogeneities increases in obesity only when FRC falls approximately below 65% of predicted or ERV below 0.6 liters. Above these thresholds distribution is quite well preserved presumably as a result of an increase in lung recoil.