TABLE 1

Model-based estimates for 30-day mortality in reference to categories of admission glucose levels

Non-fasting AGL mmol·L⁻¹	Patients		Ref. value	Bivariate analysis (Model A)	Adjusted for PSI score (Model B)	Adjusted for CURB-65 score (Model C)
Non-fasting AGL mmol·L⁻¹	At risk n	Died n (%)	Ref. value	Bivariate analysis (Model A)	Adjusted for PSI score (Model B)	Adjusted for CURB-65 score (Model C)
4–<6	317	22 (6.9)	5	2.15 (1.42–3.24)	1.94 (1.25–3.02)	2.07 (1.34–3.19)
6–<8	733	38 (5.2)	7	Ref.
8–<10	277	10 (3.6)	9	1.29 (1.09–1.54)	1.08 (0.91–1.27)	1.10 (0.94–1.29)
10–<12	98	8 (8.2)	11	1.84 (1.33–2.54)	1.33 (0.97–1.81)	1.40 (1.04–1.90)
12–<14	52	5 (9.6)	13	2.49 (1.60–3.88)	1.66 (1.06–2.59)	1.81 (1.18–2.78)
≥14	72	12 (16.7)	15	3.19 (1.86–5.45)	2.03 (1.15–3.56)	2.28 (1.32–3.93)

Data in the first two columns are from a single imputation dataset. Model-based estimates of ORs (95% CIs) for 30-day mortality for several admission glucose level (AGL) categories. ORs were calculated for reference (Ref.) values in the centre of categories. Model A is for a second-degree polynomial fit of AGL only. Models B and C are the adjusted models with pneumonia severity index (PSI) and CURB-65 scores (confusion, urea >7 mmol·L⁻¹, respiratory rate ≥30 breaths·min⁻¹, blood pressure <90 mmHg (systolic) or ≤60 mmHg (diastolic), age ≥65 years), respectively. Formulas for the final logistic regression models, in which x denotes admission glucose level, β denotes model betas for CURB-65 categories, and z denotes PSI score, were as follows. Model A: . Model B: . Model C: . Model C betas were 0.92(CURB 65 score 1), 1.63(CURB 65 score 2), 2.42(CURB 65 score 3), 1.88(CURB 65 score 4) and 3.17(CURB 65 score 5).