Performance Status (KPS), Glasgow Prognostic Score (GPS), and PN re- quirements were recorded. Univariate and multivariate analyses were performed including Kaplan-Meier curves, Cox Regression and correlation analyses. In total, 107 HPN patients (68 women, 39 men, mean age 57 years) with advanced cancer were identified. The main indications for HPN were bowel obstruction (74.3%) and high output ostomies (14.3%). Cancer cachexia was present in 87.1% of patients. The hazard ratio (HR) for upper gastrointestinal and ‘other’ cancers vs gynaecological malignancy was 1.75 (p ¼ 0.077) and 2.11 (p ¼ 0.05), respectively. KPS score, GPS, PN volume and PN potassium levels significantly predicted survival (HRKPS50 vs <50 ¼ 0.47; HRGPS ¼ 2 vs GPS ¼ 0 ¼ 3.19). In multivariate Cox regression analyses after adjustment for covariates, KPS and GPS remained significant predictors (p<0.05), whilst PN volume reached borderline significance (p¼0.094) (Table 1). In general, people who score over 150 in the 3-month and over 100 in the 12-month survival prediction nomograms, respec- tively, had less than 20% survival probability. Internal validation is also discussed. Performance status, prognostic scoring and PN requirements may predict survival in patients with advanced cancer receiving HPN. PN volume and potassium content might assist in decision making as predictors of survival. Further research and education of healthcare professionals is needed to identify which patients would most benefit from HPN and ensure timely referral and access to HPN. LEAN BODY MASS IN COMPUTED TOMOGRAPHY AS A MARKER OF NUTRITIONAL STATUS IN ENTEROCUTANEOUS FISTULAE: A CROSS SECTIONAL STUDY Konstantinos C. Fragkos a , Kenneth Cheung b , Debbie Thong a , Niamh Keane a , Shameer Mehta a , Farooq Rahman a , Andrew Plumb b , Simona Di Caro a . a Nutrition and Intestinal Failure Service, UK; b Imaging, University College London Hospitals NHS Foundation Trust, NW1 2PG, UK We aimed to investigate the correlation between body weight, body mass index, albumin levels and radiological indices of lean body mass in patients undergoing enterocutaneous fistula (ECF) repair with surgical outcomes. Biochemistry parameters and anthropometric characteristics at the time of ECF surgery were collected for a set of patients undergoing ECF repair. Skeletal muscle and visceral and subcutaneous adiposity was measured at the level of the L3 region (slice and volumetric analysis, Image J). Statistical analysis was performed with percentages, means, Spearman’s rho and dendrograms based on hierarchical clustering. Twenty seven patients (14 females, 55 ± 3.0 years) were identified in one year. Aetiology of ECF was: surgical complications in 13 patients, extensive bowel disease in the remaining (Crohn’s disease, diverticulitis, radiation enteropathy). 36% of patients had a BMI less than 18.5 kg/m 2 , 50% had albumin less than 30 g/L. None of the L3 measurements were statistically different when compared between patients with albumin less or over 30 g/ L. Parenteral nutrition was given peri-operatively in 48.1% of patients. Pre-operative albumin had a strong positive correlation with all radio- logical indices (rho 0.402e0.522, p<0.05) when sepsis was cleared. Weight and BMI were strongly positively correlated with radiological indices (rho 0.527e0.918). Dendrograms identified two clusters that correlated with imaging indices: heamatological parameters (largely negative correla- tions) and nutritional parameters (weight, albumin and others) (mainly positive correlations) None of the L3 measurements were different with respect to ECF recurrence and PN administration (p > 0.05). L3 measurements strongly correlate with nutritional parameters at the time of surgery when sepsis is cleared. Surgical outcomes might be improved with early identification of patients requiring nutritional optimisation at initial radiology scan. INTRODUCING A DEMENTIA e SPECIFIC NUTRITION AND MEALTIME ASSESSMENT TOOL M. Hannon. Raheny Community Nursing Unit (under the management of Beaumont Hospital), Dublin 5, Ireland It is estimated that there are 55,000 people with dementia in Ireland (1). It is estimated that 90% of residents in Raheny Community Nursing Unit have cognitive impairment. The positive psychological and social aspects of eating are important pleasures of life, which can persist into old age (2). For many people with dementia food and mealtimes may add a sense of meaning, order, and structure to the day. It can provide opportunities for a person to make choices regarding their care and promote a sense of in- dependence (3). Complex nutritional problems arise in dementia over the course of the disease with the progressive decline in cognitive, behavioural and physical functions. Behaviours such as wandering, pacing, refusal, spitting out food or indifference to food can also have a significant impact. Cognitive issues affecting intake include the visual perception of food on a plate or the inability to recognise food and what to do with it. In addition to dementia-related issues, people with dementia may also be affected by other co-morbidities (4). There are multiple tools available to assess behaviour at mealtimes which affect dietary intake. Food charts are an established method of recording nutritional intake (5). However, there appears to be no tools available which combine factors affecting nutritional intake and a person’s actual dietary intake. For residents with dementia who are at a high risk of malnutrition, they may not be able to participate fully or at all in dietetic consultation to jointly agree a care plan, express nutritional preferences or desires. Therefore as carers we must use our expertise to interpret a resident’s behaviour to advocate for and encourage patient-specific care regarding food choices. The aim of this quality improvement initiative was to create and implement a tool to record the actual food intake of long-term care residents with dementia and the physiological and social factors which may affect their intake at mealtimes. The purpose is to create patient- specific care plans to improve nutritional outcomes and improve Quality of Life. Using Quality Improvement methodologies a nutrition and mealtime assessment tool was created and implemented on 25 bed unit in Raheny Community Nursing Unit. This change initiative resulted in a more detailed record of a resident’s dietary intake There was also information recorded on the level of assistance required at mealtimes, any factors affecting nutritional intake, texture modifications to diet and fluids, location of meals and who assisted or accompanied the resident. The newly introduced nutrition assessment tool has been well-received overall by nursing and healthcare staff due to its ease of use. It allows for greater flexibility in recording residents’ intake, particularly when they don’t eat at standard mealtimes. It also provides information at a glance about the textures and consistencies of meals and fluids and therapeutic diets a resident may be taking. This promotes the handover of accurate information. This has a safety implication when there is a high turnover of staff or the use of agency staff. Primarily it has allowed more specific patient centred nutritional care planning for a resident by the Dietitian as there is more qualitative information available. It has also highlighted factors affecting intake which could be addressed by the Dietitian or escalated to the medical team. The author has identified the need to provide education to staff on the use of the tool. Table 1 Multivariate cox regression. HR 95% CI p-value KPS Score <50 Ref. 50 0.49 0.30 0.80 0.004 GPS 0 Ref. 1 1.91 0.77 4.75 0.162 2 3.60 1.52 8.55 0.004 N/A 2.46 1.06 5.71 0.036 PN Volume (L/day) <2 Ref. 2 1.66 0.93 2.98 0.087 PN Potassium (mmol/day) <60 Ref. 60 1.41 0.90 2.22 0.138 Abstracts / Clinical Nutrition ESPEN 28 (2018) 245e274 259