Table 4

Most significant predictors from the logistic regression model

PredictorsBiochemical hypoglycemia (BG <4 mmol/L)Clinically significant hypoglycemia (BG <3 mmol/L)
CoefficientP valuez scoreCoefficientP valuez score
PrevLowGlucose3 (+)3.842<0.00128.424.021<0.00120.39
Albumin level (−)−0.078<0.001−27.22−0.074<0.001−19.51
Intravenous insulin (+)0.639<0.00115.430.501<0.0019.82
Procedure indication (+)0.485<0.00114.870.339<0.0016.81
Sulfonylurea (+)0.572<0.00114.240.311<0.0015.35
Type 2 diabetes (−)−0.820<0.001−13.68−0.656<0.001−7.88
Weight (−)−0.010<0.001−7.42−0.012<0.001−6.38
Oxygen saturation (+)0.059<0.0016.310.067<0.0015.30
Metformin (−)−0.212<0.001−6.02−0.258<0.001−5.02
Long-acting human insulin (+)0.011<0.004.770.010<0.0015.35
Rapid-acting human insulin (+)0.023<0.0014.11NS
Mixed insulin analog (+)0.007<0.0013.67NS
  • Factors with a positive coefficient value increase the risk of hypoglycemia, and factors with a negative coefficient value decrease the risk of hypoglycemia. The factors are listed in order of effect size on the logistic regression model, e.g., an increase in albumin value reduces the risk of hypoglycemia, people with type 2 diabetes have an decreased risk of hypoglycemia. A “+” or “−” sign is given to each of the factors to indicate the effect direction. NS, not significant.