Assessment of Subclinical Ketosis Prevalence, Milk Quality Alterations, and Predictive Modeling in Dairy Herds

Subclinical ketosis (SCK) is a metabolic disorder affecting dairy cows during early lactation, significantly impacting health and productivity. Data on its prevalence in Northern Algeria remain scarce. This study aimed to assess the prevalence and risk factors of SCK in dairy cows, investigate alterations in milk composition, and develop a predictive model adapted to local farming conditions. A cross-sectional survey was conducted involving 164 cows from 15 herds. Milk β-hydroxybutyrate (BHB) levels were measured using test strips (PortaBHB®), with a threshold of ≥100 µmol/L defining SCK. Milk components (fat, protein, lactose, and solids-not-fat [SNF]) were analyzed via Lactoscan. Physiological and nutritional parameters—including body condition score (BCS), lactation rank, and forage-to-concentrate ratio (F/C ratio)—were recorded and statistically analyzed. The results showed that SCK was very common, with 61.6% of cases occurring between 22 and 80 days in milk. Logistic regression identified low BCS (≤3.0) and moderate F/C ratio (1.5–2.5) as significant risk factors, with odds ratios of 6.65 (p=0.001) and 3.40 (p=0.010). Ketotic cows had significantly reduced milk protein (3.11% vs. 3.20%, p=0.036), lactose (4.22% vs. 4.38%, p=0.017), and SNF (8.04% vs. 8.26%, p=0.026), indicative of metabolic disruption. A Random Forest model combining physiological and milk data predicted SCK with good accuracy (68%) and discrimination (AUC=0.77). These findings highlight a high SCK prevalence and metabolic milk alterations in Algerian dairy herds, supporting the model’s potential as a tool for early detection and improved herd management.