Table 1

Unbiased multivariate logistic regression analysis of 42 regressors (variables) against the primary outcome of death/ICU admission within 30 days of COVID-19 diagnosis in patients with diabetes mellitus (n=268 patients)

RegressorEstimateSEP valueMarginal effect (%)
Male gender0.8270.4530.06817.9
Ethnicity: South Asian−1.3520.7360.066−29.2
Ethnicity: black0.2160.5780.7094.7
Ethnicity: white−0.0910.5080.858−2.0
Type 1 diabetes0.9931.1350.38221.5
Active foot disease0.0750.8390.9291.6
Ischemic heart disease1.5470.5690.00733.4
Heart failure−0.9560.6130.119−20.7
Chronic obstructive pulmonary disease−0.4990.6610.450−10.8
Active cancer−0.3830.7270.598−8.3
GLP-1 receptor agonist1.0401.7550.55322.5
Dipeptidyl peptidase-IV inhibitor−0.8250.9350.378−17.8
Total number of medications for diabetes0.1370.8000.8643.0
ACE inhibitor0.5360.5130.29711.6
Angiotensin II receptor blocker0.2000.5290.7054.3
Antiplatelet drug−0.7350.4860.130−15.9
White cell count−0.1120.1460.442−2.4
Platelet count−0.0090.0030.0010.2
Serum sodium0.0770.0390.0501.7
eGFR on diagnosis−0.0120.0100.224−0.3
C reactive protein0.0020.0020.5360.0
Capillary blood glucose on diagnosis0.0340.0380.3690.7
Respiratory rate on diagnosis0.0470.0410.2521.0
Heart rate on diagnosis−0.0040.0140.774−0.1
Systolic blood pressure0.0030.0100.7930.1
Diastolic blood pressure0.0150.0170.3910.3
NEWS on diagnosis−0.0010.1100.9900.0
Inspired oxygen delivered on diagnosis0.0000.0110.9740.0
Oxygen saturations on diagnosis0.0200.0360.5750.4
Maximum inspired oxygen required0.0720.012<0.00011.6
  • This is an unselected multivariate logistic (logit) analysis of all variables that were collected for patients admitted with swab-positive COVID-19 who had diabetes mellitus, as applied to the primary outcome of death or ICU admission within 30 days. 268 patients are included with 42 variables, with the only exclusions being those patients/variables for which ≥5% data points were unknown. For this reason, HbA1c is not included as a regressor as it would have reduced the number of patients included to 168, although of note in that regression HbA1c did not survive multiple correction (and neither did body weight). For categorical variables, a positive ‘estimate’ indicates an increased risk of the primary outcome (death or ICU admission) with that variable present, and a negative estimate indicates a reduced risk of the primary outcome if that variable is present. The p value is a measure of the confidence of that variable being an independent predictor of the primary outcome corrected for all of the other regressors listed. For continuous variables, a positive ‘estimate’ indicates an increasing risk of the primary outcome as the variable increases. Since in logistic regressions estimated coefficients cannot be interpreted as a measure of the contribution of the effect, we have also calculated marginal effects along with their SEs. A positive marginal effect indicates that an increase in that variable is associated with a fully adjusted increased risk of the primary outcome. The converse applies for negative marginal effects. For categorical variables, the marginal effect indicates the percentage increased risk of the primary outcome, if that variable exists. So, for example, patients with diabetes (all other things being equal) have a 33% increased risk of death/ICU if they have IHD.

  • Statistically significant P values are shown in bold.

  • eGFR, estimated glomerular filtration rate; GLP-1, glucagon-like peptide-1; HbA1c, glycated hemoglobin; ICU, intensive care unit; IHD, ischemic heart disease; NEWS, National Early Warning Score.