Can Arterial Blood Gas, Electrolyte and Acid-Base Analysis at Admission be Used to Predict Survival to Hospital Discharge for Different Causes of Colic?

BACKGROUND

Predicting outcome in horses presenting with colic remains challenging.

OBJECTIVES

To test whether arterial blood samples in horses admitted for colic predict outcome to hospital discharge for different colic types.

METHODS

Arterial blood samples collected on admission from 358 horses undergoing medical or surgical management of colic were evaluated for pH, PaO2, PaCO2, Na+, K+, iCa2+, Cl-, HCO3 - (P), HCO3 - (P, set), Base (B), Base (ecf) and anion gap. Categories were small intestinal non-strangulating (SINS) or strangulating (SIS) lesions, large colon non-volvulus (LCNV) or volvulus (LCV), small colon non-strangulating (SCNS) or strangulating (SCS) lesions, viscus rupture or other. Multivariable logistic regression models were developed based on survival, or not, to hospital discharge. Odd ratios (ORs) with 95% confidence intervals (95%CI) and area under the curve receiver operator characteristics (AUROC), sensitivity, specificity and positive and negative predictive values were calculated at a cut-off value of p = 0.5.

RESULTS

A total of 295/358 (82.4%) horses survived to hospital discharge. Variables retained as significantly associated with survival were PaO2 for SINS (OR 1.15, 95%CI 1.04-1.27), PaO2 (OR 1.06, 95% CI 1.01-1.11) and Na+ (OR 1.24, 95% CI 1.02-1.52) for SIS, Ca2+ (OR 175.1, 95% CI 2.20-13958) and HCO3 - (P) (OR 1.18, 95% CI 1.01-1.37) for LCNV and PaCO2 (OR 1.47, 95% CI 1.05-2.06) for LCV. AUROCs showed acceptable-excellent discrimination (range: 0.7-0.9), excellent sensitivity (range: 91%-100%) but poor-fair specificity (range: 8%-50%).

CONCLUSIONS

Arterial blood is good at predicting survival based on colic type but less accurate at predicting those horses which do not survive to hospital discharge.