Predictive models for cochlear implant outcomes: Performance, generalizability, and the impact of cohort size.

While cochlear implants have helped hundreds of thousands of individuals, it remains difficult to predict the extent to which an individual's hearing will benefit from implantation. Several publications indicate that machine learning may improve predictive accuracy of cochlear implant outcomes compared to classical statistical methods. However, existing studies are limited in terms of model validation and evaluating factors like sample size on predictive performance. We conduct a thorough examination of machine learning approaches to predict word recognition scores (WRS) measured approximately 12 months after implantation in adults with post-lingual hearing loss. This is the largest retrospective study of cochlear implant outcomes to date, evaluating 2,489 cochlear implant recipients from three clinics. We demonstrate that while machine learning models significantly outperform linear models in prediction of WRS, their overall accuracy remains limited (mean absolute error: 17.9-21.8). The models are robust across clinical cohorts, with predictive error increasing by at most 16% when evaluated on a clinic excluded from the training set. We show that predictive improvement is unlikely to be improved by increasing sample size alone, with doubling of sample size estimated to only increasing performance by 3% on the combined dataset. Finally, we demonstrate how the current models could support clinical decision making, highlighting that subsets of individuals can be identified that have a 94% chance of improving WRS by at least 10% points after implantation, which is likely to be clinically meaningful. We discuss several implications of this analysis, focusing on the need to improve and standardize data collection.

A MultiCenter Analysis of Factors Associated with Hearing Outcome for 2,735 Adults with Cochlear Implants.

While the majority of cochlear implant recipients benefit from the device, it remains difficult to estimate the degree of benefit for a specific patient prior to implantation. Using data from 2,735 cochlear-implant recipients from across three clinics, the largest retrospective study of cochlear-implant outcomes to date, we investigate the association between 21 preoperative factors and speech recognition approximately one year after implantation and explore the consistency of their effects across the three constituent datasets. We provide evidence of 17 statistically significant associations, in either univariate or multivariate analysis, including confirmation of associations for several predictive factors, which have only been examined in prior smaller studies. Despite the large sample size, a multivariate analysis shows that the variance explained by our models remains modest across the datasets (R2=0.12-0.21). Finally, we report a novel statistical interaction indicating that the duration of deafness in the implanted ear has a stronger impact on hearing outcome when considered relative to a candidate's age. Our multicenter study highlights several real-world complexities that impact the clinical translation of predictive factors for cochlear implantation outcome. We suggest several directions to overcome these challenges and further improve our ability to model patient outcomes with increased accuracy.

The distribution of the time taken for an epidemic to spread between two communities.

Assessing the risk of disease spread between communities is important in our highly connected modern world. However, the impact of disease- and population-specific factors on the time taken for an epidemic to spread between communities, as well as the impact of stochastic disease dynamics on this spreading time, are not well understood. In this study, we model the spread of an acute infection between two communities ('patches') using a susceptible-infectious-removed (SIR) metapopulation model. We develop approximations to efficiently evaluate the probability of a major outbreak in a second patch given disease introduction in a source patch, and the distribution of the time taken for this to occur. We use these approximations to assess how interventions, which either control disease spread within a patch or decrease the travel rate between patches, change the spreading probability and median spreading time. We find that decreasing the basic reproduction number in the source patch is the most effective way of both decreasing the spreading probability, and delaying epidemic spread to the second patch should this occur. Moreover, we show that the qualitative effects of interventions are the same regardless of the approximations used to evaluate the spreading time distribution, but for some regions in parameter space, quantitative findings depend upon the approximations used. Importantly, if we neglect the possibility that an intervention prevents a large outbreak in the source patch altogether, then intervention effectiveness is not estimated accurately.

A computational model of the topographic distribution of ventilation in healthy human lungs.

The topographic distribution of ventilation in the lungs is determined by the interaction of several factors, including lung shape, airway tree geometry, posture, and tissue deformation. Inter-species differences in lung structure-function and technical difficulty in obtaining high resolution imaging of the upright human lung means that it is not straightforward to experimentally determine the contribution of each of these factors to ventilation distribution. We present a mathematical model for predicting the topological distribution of inhaled air in the upright healthy human lung, based on anatomically structured model geometries and biophysical equations for model function. Gravitational deformation of the lung tissue is predicted using a continuum model. Airflow is simulated in anatomically based conducting airways coupled to geometrically simplified terminal acinar units with varying volume-dependent compliances. The predicted ventilation distribution is hence governed by local tissue density and elastic recoil pressure, airway resistance and acinar compliance. Results suggest that there is significant spatial variation in intrinsic tissue properties in the lungs. The model confirms experimental evidence that in the healthy lungs tissue compliance has a far greater effect than airway resistance on the spatial distribution of ventilation, and hence a realistic description of tissue deformation is essential in models of ventilation.

Evidence for minimal oxygen heterogeneity in the healthy human pulmonary acinus.

It has been suggested that the human pulmonary acinus operates at submaximal efficiency at rest due to substantial spatial heterogeneity in the oxygen partial pressure (Po(2)) in alveolar air within the acinus. Indirect measurements of alveolar air Po(2) could theoretically mask significant heterogeneity if intra-acinar perfusion is well matched to Po(2). To investigate the extent of intra-acinar heterogeneity, we developed a computational model with anatomically based structure and biophysically based equations for gas exchange. This model yields a quantitative prediction of the intra-acinar O(2) distribution that cannot be measured directly. Temporal and spatial variations in Po(2) in the intra-acinar air and blood are predicted with the model. The model, representative of a single average acinus, has an asymmetric multibranching respiratory airways geometry coupled to a symmetric branching conducting airways geometry. Advective and diffusive O(2) transport through the airways and gas exchange into the capillary blood are incorporated. The gas exchange component of the model includes diffusion across the alveolar air-blood membrane and O(2)-hemoglobin binding. Contrary to previous modeling studies, simulations show that the acinus functions extremely effectively at rest, with only a small degree of intra-acinar Po(2) heterogeneity. All regions of the model acinus, including the peripheral generations, maintain a Po(2) >100 mmHg. Heterogeneity increases slightly when the acinus is stressed by exercise. However, even during exercise the acinus retains a reasonably homogeneous gas phase.

Pulmonary gas exchange in anatomically-based models of the lung.

Pulmonary gas exchange can be investigated at different scales of interest. Our approach is to couple models of gas exchange to anatomically-detailed models of the airway and pulmonary vascular trees. We are linking a hierarchy of models from the capillary segment up to the whole lung, so that a change in the detailed small-scale behaviour has a flow-on effect to function at a larger scale. The anatomically-based models will be used to understand how regional perturbations to the structure or function of the airway and vascular trees and the state of health of the functional tissue affect gas exchange. We are interested in the degree to which the system can be perturbed before it is detected by standard laboratory measures.