Partial least squares regression estimates of milk and blood constituents using Fourier-transform mid-infrared (FTIR) analysis have shown promise as a tool for monitoring early-lactation excessive energy deficit in dairy herds. Our objective was to analyze milk via FTIR to determine the association of early-lactation predicted milk β-hydroxybutyrate (BHB) concentrations, predicted blood nonesterified fatty acid (NEFA) concentrations, and predicted milk de novo fatty acid (FA) percentages relative to total FA concentrations, with the risk of disease or removal in early lactation (hyperketonemia, displaced abomasum, metritis, culling, or death) and average daily milk yield during the first 15 wk of lactation. We enrolled 517 multiparous Holstein cows from 2 dairy farms in New York. Composite milk samples were collected twice weekly from 3 to 18 DIM for a total of 4 timepoints (T1, T2, T3, T4) and analyzed using FTIR spectrometry for milk BHB and FA composition and predicted blood NEFA. Blood samples were collected for hyperketonemia determination (BHB ≥ 1.2 mmol/L) using a handheld meter, and farm-diagnosed occurrence of disease or removal during the first 30 DIM and average daily milk yield during the first 15 wk of lactation were collected from herd management software. The incidence of disease or removal between 3 and 18 DIM was 20.2%. Explanatory models for disease or removal were developed for each predicted constituent of interest at each timepoint using fixed-effect multivariable Poisson regression. Repeated measures ANOVA models were developed for each predicted constituent to assess differences in average daily milk yield. For all timepoints, increased risk of disease or removal was associated with higher predicted milk BHB [relative risk (RR) = 2.0; RR = 3.4; RR = 5.2; RR = 9.1], higher predicted blood NEFA (RR = 2.7; RR = 2.5; RR = 3.8; RR = 10.0), and lower predicted milk de novo FA relative percentages (RR = 2.9; RR = 3.3; RR = 5.8; RR = 7.2). Average daily milk yield was increased for cows above the cut point for predicted milk BHB (2.1 kg/d) and predicted blood NEFA (3.5 kg/d) and below the cut point for de novo FA relative percentages (2.3 kg/d). Our results suggest that FTIR-predicted milk BHB, blood NEFA, and milk de novo FA relative percentages are promising indicators of subsequent disease or removal in early lactation; their positive relationship with milk yield warrants further exploration.