Critically ill patients with COVID-19 are at increased risk of complications such as cardiovascular events, coagulation disorders, acute respiratory distress syndrome (ARDS), and nosocomial infections that could contribute to poorer clinical outcomes. It has been reported that the rate of nosocomial infections in critically ill COVID-19 patients ranges from 30 to 50%.1–3 A variety of factors have been suggested as contributors to this increased risk of developing nosocomial infections: an impaired immune system triggered by SARS-CoV-2 infection, encompassing phenomena such as immune hyper-response and inflammation; the increased risk of ARDS; the need for invasive ventilation; the use of empirical antibiotic therapy; and long ICU stays. Interestingly, recent studies found hyperglycemia to be a risk factor for adverse outcomes and mortality in severe COVID-19,4,5 but there is still a scarcity of data on the impact of hyperglycemia on the risk of nosocomial infections in critically ill COVID-19 patients. Thus, we aimed to investigate the association between hyperglycemia at ICU admission and the risk of nosocomial bacterial pneumonia in COVID-19 in the ICU setting, along with the risk of in-hospital mortality.

The CIBERESUCICOVID is a prospective, multicenter, and observational study that included consecutive adult patients admitted to 55 Spanish ICUs for severe COVID-19 between 5 February and 21 December 2021. Data was collected as previously described.6,7 Descriptive statistics were used for the basic features of the study data. Categorical variables were compared using the chi-squared test or Fisher's exact test, whereas continuous variables were compared using the non-parametric Mann–Whitney U test. We analyzed the association between hyperglycemia (serum glucose>126mg/dL)8 at ICU admission and nosocomial bacterial pneumonia9 by means of a mixed-effects multivariable model,10 defined by a binomial probability distribution and a logit link function, with centers as a random effect. To evaluate the effect of hyperglycemia on in-hospital mortality, a Fine-Gray competing risks model11 stratified on the centre variable was used, considering discharge from hospital as competing risk for mortality. The study received approval from the Institution's Internal Review Board (Comité Ètic d’Investigació Clínica, registry number HCB/2020/0370) and we obtained informed consent from either patients or their relatives.

During the study period, 6225 patients were admitted to ICU in the 55 Spanish hospitals due to COVID-19. We included 5763 patients in this analysis; of these, 3704 (64%) presented hyperglycemia at ICU admission and 2059 (36%) did not. Baseline characteristics, clinical features, complications, and outcomes of the cohort according to the presence of hyperglycemia at ICU admission are shown in Table 1. Median age was 65 (56–72) years and 71% of patients were male. Hypertension (55%), obesity (43%), diabetes mellitus (33%), and chronic lung disease (16%) were the most frequent comorbidities. Patients were admitted to the ICU after a median of 74–9 days from the presentation of symptoms and of 1 (0–3) days from hospital admission. Patients with hyperglycemia were more frequently older and had more comorbidities (33% were diabetic, for example, compared to 9% of those not presenting hyperglycemia), and, overall, they presented more severe COVID-19 at ICU admission, as measured by the APACHE II and SOFA scores. Moreover, patients with hyperglycemia received corticosteroids more frequently and presented significantly more complications, e.g., a need for invasive mechanical ventilation, ARDS, acute renal failure, nosocomial pneumonia, and acute cardiac injury. The lengths of hospital and ICU stays were also significantly longer, and in-hospital, ICU, and 90-day mortality rates were significantly higher in patients with hyperglycemia. As regards the primary outcomes – namely rates of nosocomial bacterial pneumonia and in-hospital mortality, both were significantly higher in patients with hyperglycemia than in those without it (31% vs. 24% and 33% vs. 25%, respectively, both p<0.001). Several variables assessed at ICU admission were associated with an increased risk of developing nosocomial bacterial pneumonia in the multivariable analysis (Table 2, Panel A), notably male gender, older age, administration of antibiotics prior to ICU admission, mechanical ventilation, and corticosteroids, but hyperglycemia was not one of them. However, the factors associated with in-hospital mortality (Table 2, Panel B) did include hyperglycemia at ICU admission.

When we separately analyzed diabetic and non-diabetic patients according to the presence of hyperglycemia at ICU admission, we found that the prevalence of hyperglycemia amongst the former was 86.7%, as against 56.9% in the latter. Furthermore, hyperglycemia did not impact outcomes in patients with diabetes, who, overall, presented poorer outcomes than non-diabetic patients (Supplementary Tables 1 and 2). Moreover, when multivariable analyses of risk factors for bacterial pneumonia and in-hospital mortality were run for both groups separately, hyperglycemia was not found to predict either outcome in diabetic or non-diabetic patients (Supplementary Tables 3 and 4). A receiving operating curve showed a cut-off of 150mg/dL for hyperglycemia, allowing for a better discrimination of outcomes than 126mg/dL in the overall cohort (Supplementary Figure 1), although it did not predict bacterial pneumonia in the multivariable analysis (Supplementary Table 5).

The prevalence of hyperglycemia at ICU admission in our cohort was found to be strikingly high, even though only a quarter of the patients were diabetic. Moreover, although a higher percentage of patients in the hyperglycemia group received systemic corticosteroids, the vast majority of patients only started receiving corticosteroids once they had been admitted to the ICU – i.e., after hyperglycemia was detected. The most likely explanation is the systemic inflammation induced by COVID-19 itself, which has been widely demonstrated in several previous studies.12 However, we did not assess the percentage of non-diabetic patients who developed, during admission for COVID-19, de novo diabetes that persisted after discharge or until death. This omission prevented us from comparing our results with those of Cromer and colleagues,13 who recently found that diabetes diagnosed at the presentation of COVID-19 is associated with lower glucose but higher inflammatory markers and a greater likelihood of ICU admission, suggesting that stress hyperglycemia is a significant physiological mechanism. Interestingly, these authors found that 31.2% of their patients were diabetic – in line with our results –, whereas only 13.2% of their patients fulfilled the criteria for newly diagnosed diabetes at some point. Approximately half of these individuals experienced a regression of DM.13 Remarkably, however, although hyperglycemia was found to be a risk factor for in-hospital mortality in the overall cohort, only non-diabetic patients developed significantly poorer outcomes when presenting hyperglycemia at admission. This finding is crucial to the interpretation of our findings and the subsequent extraction of clinical conclusions, and it also presents the ability to predict mortality more accurately, with a higher cut-off than that of the traditional definition of hyperglycemia (i.e., 150mg/dL than 126mg/dL).

Interestingly, while the prevalence of bacterial pneumonia in our cohort was high overall, presentation of hyperglycemia did not predict bacterial pneumonia in the whole cohort or in patients with diabetes, whereas non-diabetic patients with hyperglycemia did have a significantly higher prevalence of bacterial pneumonia than those not presenting hyperglycemia at admission. The higher incidence of bacterial nosocomial pneumonia found in COVID-19 patients compared to other critically ill has mostly been linked to the long duration of invasive mechanical ventilation, the high incidence of ARDS, and immune-suppressive treatment.14 The non-diabetic patients in our cohort with hyperglycemia had higher rates of invasive mechanical ventilation and immune-suppressive treatment, but not ARDS. While the quality of evidence on the risk factors for ventilator-associated hospital-acquired pneumonia unrelated to COVID-19 is moderate-good, this is not the case for pneumonia not associated with the use of ventilators.15 Furthermore, systematic reviews and meta-analyses have consistently found corticosteroids to be a risk factor, along with invasive mechanical ventilation and diabetes, whereas this is not true of hyperglycemia.14 Further studies are needed to investigate the risk factors for developing nosocomial pneumonia, particularly in non-ventilated critically ill COVID-19 patients.

Our study is limited by the lack of data on antidiabetic therapy, disease baseline control in diabetics (e.g., glycosylated hemoglobin), and the rate of new diagnoses of diabetes during the index episode.

In summary, we found that, overall, patients presenting with hyperglycemia at ICU admission had more aggressive and severe COVID-19 (i.e., less time between hospital and ICU admissions), as well as poorer outcomes, including in-hospital mortality. Hyperglycemia appears to be a better predictor of poor outcomes in non-diabetic than in diabetic patients. Early detection and management of hyperglycemia is required in hospitalized COVID-19 patients.