Research ROC analysis of prepartum feeding time can accurately predict postpartum metritis development in HF crossbred cows Tapas Kumar Patbandha a, # , Tushar Kumar Mohanty a, * , Siddhartha Shankar Layek a , Arumugam Kumaresan a , Suresh C. Kantwa a , R. Malhotra b , A.P. Ruhil c , Shiv Prasad a a Cattle Yard, Livestock Production and Management Section, National Dairy Research Institute, Karnal, Haryana, India b DESM Division, National Dairy Research Institute, Karnal, Haryana, India c ASRB, Krishi Bhawan, Pusa, ICAR, New Delhi, India article info Article history: Received 25 June 2012 Received in revised form 14 April 2013 Accepted 1 May 2013 Available online 17 June 2013 Keywords: ROC analysis feeding time metritis threshold value relative risk abstract Receiver operating characteristic (ROC) analysis is a useful statistical tool to visualize, organize, and select diagnostic variables. Accuracy of the test is represented by the area under the ROC curve (AUC), sensi- tivity (Se), and specificity (Sp). In the present study, ROC analysis of prepartum feeding time (day 11 to day 2) of 20 pluriparous Holstein-Friesian crossbred cows were recorded, and the cows were followed up for detecting postpartum metritis development if any. Prepartum daily feeding time data during whole study was grouped into 2 periods (P1 [day 11 to day 7] and P2 [day 6 to day 2]). The AUC of feeding time for the P1 and P2 periods was 0.64 (P ¼ 0.29) and 0.86 (P ¼ 0.006), respectively. Cows with daily feeding times below the optimum threshold value of 284.5 min/d during P2 (Se ¼ 75% and Sp ¼ 91.67%) had more chance of developing postpartum metritis compared with those with daily feeding time above the threshold value (positive likelihood ratio ¼ 9.0). Furthermore, relative risk (RR) analysis revealed that the risk of development of postpartum metritis increased as feeding time decreased below the optimum threshold value during P2 (RR ¼ 5.571, P ¼ 0.004). Thus, the optimum threshold value of feeding time during P2 may be used to predict risk of postpartum metritis, allowing for preventive measures to be implemented in the early postpartum period to reduce the likelihood of postpartum complications. Ó 2013 Elsevier Inc. All rights reserved. Introduction Metritis is an important postpartum health disorder, which af- fects reproductive performance and productivity of high-yielding dairy cows and buffaloes. Heavy productivity loss because of increased days open, days dry, and calving interval along with significant reduction in lactation length has been reported in crossbred cows at National Dairy Research Institute, Karnal (Balasundaram et al., 2011; Sharma et al., 2011). Earlier studies on blood metabolites and feeding behavior revealed that these pa- rameters can be effectively used for early prediction of metritis (Urton et al., 2005; Huzzey et al., 2007; Ospina et al., 2010). Pre- partum feeding time below an optimum threshold value (75 min/d) was found to be highly correlated with increased incidence of postpartum metritis (Urton et al., 2005). The optimum threshold value is defined as the threshold value that has maximum com- bined sensitivity (Se) and specificity (Sp) for predicting diseases (Urton et al., 2005; Ospina et al., 2010). The common method used to find the optimum threshold value is through receiver operating characteristic (ROC) analysis. It is a simple statistical tool used to characterize a variable in terms of area under the ROC curve (AUC) and provides optimum threshold values along with the corresponding Se, Sp, and likelihood ratio (LR). The Se measures the proportion of sick and animals that are identified by the diagnostic test as sick (true positives [TPs]), whereas Sp measures the proportion of healthy cows correctly identified by the test as healthy (true negatives [TNs]). The LR refers to the number of times a cow is more likely to develop a disease when they are below a particular threshold value relative to if they were above this threshold value. The assumption behind ROC * Address for reprint requests and correspondence: Tushar Kumar Mohanty, Cattle Yard, Livestock Production and Management Section, National Dairy Research Institute, Karnal 132001, Haryana, India; Tel: þ91-9215508002. E-mail address: mohanty.tushar@gmail.com (T.K. Mohanty). # Current address: Department of Livestock Production and Management, College of Veterinary Science and Animal Husbandry, Junagadh Agricultural University, Junagadh 362001, Gujarat, India. Contents lists available at SciVerse ScienceDirect Journal of Veterinary Behavior journal homepage: www.journalvetbehavior.com 1558-7878/$ e see front matter Ó 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jveb.2013.05.001 Journal of Veterinary Behavior 8 (2013) 362e366