Background: Surgery remains the first choice of curative treatment, for patients with non-small lung cancer, the proportion of patients undergoing surgery has risen in recent years. Post-operative compli- cations are well recognised following curative lung cancer surgery but there is limited data on readmission rates and causes . The UK Thoracic Surgery Group (TSG), a subgroup of the National Lung Cancer Forum (NLCNF) conducted a multicentre audit to assess readmission potential causes and patient experience. Method: The audit involved 6 UK thoracic surgical centres with prospective data collection over 3 months from primary lung cancer resection patients. Patients were contacted 1 month post discharge by telephone. Data collection included demographics, socioeconomic, smoking status, comorbidities, surgery, postoperative recovery, discharge satisfaction and readmission details. Result: 268 patients underwent thoracic surgery, the overall readmission rate was 11% (30), with variable readmission rate across the centres (range 3-24%), most readmission occurred within 7 days of discharge 47% (14) with patients being readmitted to a hospital that did not performed the procedure 43%(17). The most common cause of readmission was mainly pulmonary related with chest infections being largest cause, pain, wound infection and pneumothorax were also common. Length of stay following readmission was longer than initial surgical stay median 8 (range 0-94) vs 5 (range 2-27).Type of surgical approach had no impact on readmission. However readmission was associated with smoking, post-operative complications, discharge with drain, length of stay post-surgery and the patient’s readiness for discharge (see table 1). Conclusion: This audit provides a broad overview of the pattern and trend of readmissions rates within 30 days post discharge following lung cancer resection. Whilst not every readmission can be avoided, there is opportunity to identify and pre- vent patient readmission. Listening to patient’s assessment of their readiness for discharge is crucial to facilitating patient compliance with discharge and confidence in community carers. Keywords: Re-admis- sion, Thoracic Surgery, NSCLC MA18.01 Non-Small Cell Lung Cancer Risk Assessment with Artificial Neural Networks T. Chaunzwa, 1 Y. Xu, 2 D. Christiani, 3 A. Shafer, 4 N. Diao, 4 M. Lanuti, 5 R. Mak, 6 H. Aerts 2 1 Howard Hughes Medical Institute, Chevy Chase, MD/ US, 2 Radiation Oncology, Harvard Medical School, Boston, MA/US, 3 Environmental Health, Epidemiology, Harvard School of Public Health, Boston, MA/US, 4 Harvard T.H.Chan School of Public Health, Boston, MA/ US, 5 Thoracic Surgery, Massachusetts General Hospital, Boston, MA/US, 6 Radiation Oncology, Brigham and Womens Hospital, Boston ,MA/US Background: Lung cancer is a heterogeneous disease with many clinically important subtypes. Given the complexity of classification, there is room for innovative risk assessment tools to help ascertain prognosis and management. In this work we tested an Artificial Neural Network (ANN) to stratify patients into clinically significant low and high risk categories. Method: CT imaging, survival, and cancer staging data was extracted for a sample of 311 patients with Stage-I (n ¼ 186) and Stage-II (n ¼ 125) non-small cell lung cancer (NSCLC) from the comprehensive Boston Lung Cancer Survival (BLCS) cohort. Median follow-up from time of diagnosis was 3.5 years, with 86% 2-year sur- vival. A deep convolutional neural network pretrained on ImageNet was used, with fine-tuning of the last convolutional layers, dense layers, and softmax for stratification. Inputs of this model were 50 x 50 mm2 image patches. Training was performed on 182 labeled CT scans (112 Stage-I and 70 Stage-II). 46 cases were used for initial cross-validation, with an independent test set of 83 cases. The median prediction probability from the ANN was used as a cutoff to divide patients into low and high risk groups. Result: The model was able to perform classification of cancer stage on the heterogeneous test set (AUC ¼ 0.73, p< 0.0005). The test set was split evenly into low risk (n ¼ 42) and high risk (n¼ 41) groups based on model predictions. There was statistically significant separation in the Kaplan Meier-estimates for survivorship in the two stratified groups (p < 0.02). Conclusion: ANNs can be effective tools for quantitative risk stratification in NSCLC. In addition to the potential for real-time clinical decision support, ANNs may also help create new paradigms in lung cancer risk assessment. The models have the capacity to perform suprahuman computations, which can help meet future demands of clinical practice, given expanding digital-imaging volumes. Keywords: Neural network, Sur- vival, Staging MA18.02 The Impact of Treatment Evolution in NSCLC (iTEN) Model: Development and Validation D. Moldaver, 1 M. Hurry, 2 D. Tran, 1 W. Evans, 3 P. Cheema, 4 R. Sangha, 5 R. Burkes, 6 B. Melosky, 7 E. Orava, 2 D. Grima 1 1 Cornerstone Research Group, Burlington/CA, 2 Astrazeneca Canada, Toronto, ON/CA, 3 Department of Oncology, McMaster University, Hamilton, ON/CA, 4 William Osler Health System, University of Toronto, Toronto/CA, 5 Cross Cancer Institute, Edmonton, AB/CA, 6 Department of Medicine, Division of Medical Oncology, Mount Sinai Hospital, University of Toronto, Toronto, ON/CA, 7 Medical Oncology, BC Cancer Agency, Vancouver, BC/CA Background: The iTEN model was developed to estimate the survival impact of new treatments for advanced NSCLC (aNSCLC) patients. The structure and key assumptions of the iTEN model and outputs vali- dated against published real-world survival data are presented. S418 Journal of Thoracic Oncology Vol. 13 No. 10S