137 Parturition prediction in mares using gestation length, and pH and hardness of preparturient mammary secretions

Abstract

Due to variability in gestation length, attention is placed on predictive variables of parturition. Previous studies have focused on variables of days in gestation (DIG), and pH or hardness (Hd) of preparturient mammary secretions. Research on the added benefit of using multiple variables simultaneously is limited. We hypothesized that a combination of mammary secretion pH, Hd, and DIG would improve parturition predictions within 24, 48, and 72h of birth. Our objectives were to identify the best combination of variables that provided probability estimates of parturition and to determine optimal cutoff probabilities for predicting parturition. Diagnostic criteria were compared with previously published criteria using samples collected in this study (Table 1). Mammary secretions (n = 1,524) were collected from 36 mares before 135 births from 2014 to 2021. The pH was measured with narrow range (pH 6.0–7.7, 0.3 precision) pH strips, Hd with the Predict-a-Foal Kit (0–5 squares, 0.5 precision), and DIG was calculated based on the last day bred. Backward stepwise selection with mixed logistic regression was used to select the best combination of variables to estimate probability of whether parturition would occur. Bootstrapping with 10,000 replications and data clustered by foal was utilized to improve precision of coefficients used to determine estimated probability. Decision curve analysis (DCA) was utilized to compare the net benefits of each diagnostic criteria. The best combination of variables for prediction within 24h (8 ± 4h)were pH and DIG, while prediction within 48h (32 ± 4h) and 72h (56 ± 4h) added the variable of Hd. Optimal cutoff probabilities that maximized the sensitivity (SN) and specificity (SP) of parturition diagnosis were 13.8, 10.2, and 12.7% for 24, 48, and 72h. The sum of the SN, positive predictive value (PPV), SP, and negative predictive value (NPV) illustrates how the combination is optimized over any single diagnostic criteria (Table 1). The DCA indicates greater net benefit for predictions using the combination over single diagnostic criteria within realistic threshold probabilities. Using a combination of variables to predict if parturition will occur will reduce false positives without increasing false negatives.