Group-Level Ranking-Based Hubness Analysis of Human Brain Connectome Reveals Significant Interhemispheric Asymmetry and Intraparcel Heterogeneities.

Objective: Graph theory applications are commonly used in connectomics research to better understand connectivity architecture and characterize its role in cognition, behavior and disease conditions. One of the numerous open questions in the field is how to represent inter-individual differences with graph theoretical methods to make inferences for the population. Here, we proposed and tested a simple intuitive method that is based on finding the correlation between the rank-ordering of nodes within each connectome with respect to a given metric to quantify the differences/similarities between different connectomes. Methods: We used the diffusion imaging data of the entire HCP-1065 dataset of the Human Connectome Project (HCP) (n = 1,065 subjects). A customized cortical subparcellation of HCP-MMP atlas (360 parcels) (yielding a total of 1,598 ROIs) was used to generate connectivity matrices. Six graph measures including degree, strength, coreness, betweenness, closeness, and an overall "hubness" measure combining all five were studied. Group-level ranking-based aggregation method ("measure-then-aggregate") was used to investigate network properties on population level. Results: Measure-then-aggregate technique was shown to represent population better than commonly used aggregate-then-measure technique (overall rs: 0.7 vs 0.5). Hubness measure was shown to highly correlate with all five graph measures (rs: 0.88-0.99). Minimum sample size required for optimal representation of population was found to be 50 to 100 subjects. Network analysis revealed a widely distributed set of cortical hubs on both hemispheres. Although highly-connected hub clusters had similar distribution between two hemispheres, average ranking values of homologous parcels of two hemispheres were significantly different in 71% of all cortical parcels on group-level. Conclusion: In this study, we provided experimental evidence for the robustness, limits and applicability of a novel group-level ranking-based hubness analysis technique. Graph-based analysis of large HCP dataset using this new technique revealed striking hemispheric asymmetry and intraparcel heterogeneities in the structural connectivity of the human brain.

Development and Validation of a Machine Learning Model to Aid Discharge Processes for Inpatient Surgical Care.

Importance: Inpatient overcrowding is associated with delays in care, including the deferral of surgical care until beds are available to accommodate postoperative patients. Timely patient discharge is critical to address inpatient overcrowding and requires coordination among surgeons, nurses, case managers, and others. This is difficult to achieve without early identification and systemwide transparency of discharge candidates and their respective barriers to discharge. Objective: To validate the performance of a clinically interpretable feedforward neural network model that could improve the discharge process by predicting which patients would be discharged within 24 hours and their clinical and nonclinical barriers. Design, Setting, and Participants: This prognostic study included adult patients discharged from inpatient surgical care from May 1, 2016, to August 31, 2017, at a quaternary care teaching hospital. Model performance was assessed with standard cross-validation techniques. The model's performance was compared with a baseline model using historical procedure median length of stay to predict discharges. In prospective cohort analysis, the feedforward neural network model was used to make predictions on general surgical care floors with 63 beds. If patients were not discharged when predicted, the causes of delay were recorded. Main Outcomes and Measures: The primary outcome was the out-of-sample area under the receiver operating characteristic curve of the model. Secondary outcomes included the causes of discharge delay and the number of avoidable bed-days. Results: The model was trained on 15 201 patients (median [interquartile range] age, 60 [46-70] years; 7623 [50.1%] men) discharged from inpatient surgical care. The estimated out-of-sample area under the receiver operating characteristic curve of the model was 0.840 (SD, 0.008; 95% CI, 0.839-0.844). Compared with the baseline model, the neural network model had higher sensitivity (52.5% vs 56.6%) and specificity (51.7% vs 82.6%). The neural network model identified 65 barriers to discharge. In the prospective study of 605 patients, causes of delays included clinical barriers (41 patients [30.1%]), variation in clinical practice (30 patients [22.1%]), and nonclinical reasons (65 patients [47.8%]). Summing patients who were not discharged owing to variation in clinical practice and nonclinical reasons, 128 bed-days, or 1.2 beds per day, were classified as avoidable. Conclusions and Relevance: This cohort study found that a neural network model could predict daily inpatient surgical care discharges and their barriers. The model identified systemic causes of discharge delays. Such models should be studied for their ability to increase the timeliness of discharges.

A robust estimation model for surgery durations with temporal, operational, and surgery team effects.

For effective operating room (OR) planning, surgery duration estimation is critical. Overestimation leads to underutilization of expensive hospital resources (e.g., OR time) whereas underestimation leads to overtime and high waiting times for the patients. In this paper, we consider a particular estimation method currently in use and using additional temporal, operational, and staff-related factors provide a statistical model to adjust these estimates for higher accuracy.The results show that our method increases the accuracy of the estimates, in particular by reducing large errors. For the 8093 cases we have in our data, our model decreases the mean absolute deviation of the currently used scheduled duration (42.65 ± 0.59 minutes) by 1.98 ± 0.28 minutes. For the cases with large negative errors, however, the decrease in the mean absolute deviation is 20.35 ± 0.74 minutes (with a respective increase of 0.89 ± 0.66 minutes in large positive errors). We find that not only operational and temporal factors, but also medical staff and team experience related factors (such as number of nurses and the frequency of the medical team working together) could be used to improve the currently used estimates. Finally, we conclude that one could further improve these predictions by combining our model with other good prediction models proposed in the literature. Specifically, one could decrease the mean absolute deviation of 39.98 ± 0.58 minutes obtained via the method of Dexter et al (Anesth Analg 117(1):204-209, 2013) by 1.02 ± 0.21 minutes by combining our method with theirs.