A new coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (coronavirus disease 2019 [COVID-19]) was first reported in late December 2019 in Wuhan, China (1), and has progressed to become a pandemic with over 3 million cases confirmed (2) and still growing (3). No specific therapeutic agents have been identified, and its infectious nature, hospitalization rates, intensive care admissions, and mortality are very high (2,4–7). Preexisting chronic illnesses such as diabetes, hypertension, and obesity result in worst outcomes in the presence of COVID-19 infection and virus-induced respiratory dysfunction (8). In influenza-like illnesses, hyperglycemia has been reported to increase plasma glucose concentration in airway secretions. Additionally, increased viral replication in vivo and suppression of the antiviral immune response is also described (9). Increased permeability of the vasculature and subsequent collapse of alveolar epithelium have direct effects on pulmonary function (10) and may explain the higher mortality rates observed in these patients (8).
Hyperglycemia and COVID-19 Infections
The significant hyperglycemia that occurs in the acute inflammatory state of COVID-19 patients has been recognized and found to be pronounced among those with diabetes, prediabetes, and/or obesity. A bidirectional link between chronic inflammation and hyperglycemia had been already described for chronic complications of diabetes. For instance, several changes in the immune system including alterations in specific cytokines and chemokines, shifts in the number and activation state of various leukocyte populations, and increased apoptosis and tissue fibrosis are present in obesity and type 2 diabetes, suggesting that inflammation has an active role in the pathogenesis of hyperglycemia, progression to clinically overt type 2 diabetes, and chronic complications (11–14). We believe that this baseline inflammatory state could set the stage and background for further elevations in the levels of inflammatory cytokines, particularly as seen in acute infectious diseases such as COVID-19, further increasing insulin resistance, promoting proinflammatory effects of acute (stress) hyperglycemia, and ultimately leading to a poor prognosis of such patients with diabetes (15–18). Our observations are corroborated by the recent retrospective multicentered study of over 7,000 cases of COVID-19 in Hubei Province, China (17). The authors reported a significant correlation between well-controlled blood glucose and lower serum levels of inflammatory markers (interleukin-6 [IL-6], high sensitivity C-reactive protein [hsCRP], lactate dehyrogenase [LDH]) in patients with COVID-19. A recently reported study in patients with diabetes without advanced chronic complications or comorbidities at baseline supports the marked and rapidly evolving inflammatory process in the presence of SARS-CoV-2 infection (19). Despite presenting initially with mild symptoms and fever, the clinical course deteriorated very rapidly with progressive dyspnea and pneumonia with higher computerized tomography imaging severity scores. Compared with patients without diabetes, individuals with diabetes showed higher elevations in the concentrations of IL-6, ferritin, hsCRP, and d-dimer foreshadowing a raging cytokine storm and a hypercoagulable state with rapid deterioration. In addition, the insulin requirements were very high even in those patients who were insulin naive prior to admission (19).
Although the current information continues to emerge and the full impact of severely high peaks in glucose levels on disease course and mortality has only recently become available, the initial experience we collected since the peak of COVID-19 in Michigan suggests that acute and persistent surges in blood glucose levels associated with the cytokine storm herald poor prognosis. Thus, in conjunction with management of infection, inflammation, and supportive care, a rapidly instituted and tailored glucose management plan targeting hyperglycemia is critical. This can help prevent and reduce morbidity and complications leading to prolonged inpatient stay and increased resource utilization during these times when most hospitals and health care systems are overwhelmed by the COVID-19 cases (17).
Michigan is one of the states with a very high number of COVID-19 cases and high rates of complications and mortality. Here we provide a perspective on the Michigan Medicine experience and the plans implemented to effectively curb these glucose surges and expedite recovery for patients with diabetes and/or stress hyperglycemia admitted with COVID-19–related illness.
A specialized regional isolation containment unit was created for the acute care of these patients, and a stepwise general management plan that also included specified laboratory testing to monitor disease activity with a panel of inflammatory and prothrombotic markers was implemented. The overall health care delivery models have been transformed significantly, transitioning from the classical in-person consult to interprofessional consultations. This model relies on interactive chart reviews and provides recommendations to the primary teams by diabetes specialists with the goal of preserving personal protective equipment (PPE) and reducing exposure for health care providers.
Within the very first days of COVID-19 inpatient surge, a phenotype of severe hyperglycemia was noted in a large proportion of the critically ill admitted patients carrying a prior diagnosis of type 1 diabetes, type 2 diabetes, prediabetes, or severe obesity. Their glucose management was further complicated by rapid acute renal failure, tube feed initiation, vasopressor support for hypotension, steroids for acute respiratory distress syndrome, and chronic renal replacement therapy. In addition, detailed history of their diabetes management was limited, as several patients were transferred from other Michigan hospitals not connected with our electronic medical records. All these factors presented important health care challenges.
Thus, our major aim was to develop viable algorithms to provide a targeted approach to manage hyperglycemia in COVID-19–infected patients based on a personalized risk stratification that includes different levels of hyperglycemia and insulin resistance, prior diabetes control, presence of obesity, needs and type of nutritional support, renal dysfunction, vasopressor support, and disease activity.
The University of Michigan Hospital provided care to ∼500 COVID-19 patients since the COVID-19 crisis hit southeast Michigan. Approximately 160 out of 500 patients had known diabetes and were referred to us for management. However, some patients who came in critical condition and had no prior known diabetes also developed hyperglycemia. We also observed that among patients we followed, ∼43% were African American, which is high but in line with other observations showing a disproportionately high burden of severe disease among African Americans (20). The algorithm was rapidly developed, and although it continued to be refined daily in the first couple of weeks based on the emerging observations by our team, it is fair to state that it was used in a more or less refined form in up to 200 patients. In addition, given the very large number of patients we had to follow on a daily basis, we have activated several hyperglycemia management teams to cover all these patients.
Preliminary in-house experience has confirmed that the severity of hyperglycemia and marked insulin resistance were also associated with a characteristic inflammatory biomarker signature that includes rapid elevations and changes in the levels of hsCRP, procalcitonin, triglycerides, IL-6, and d-dimers; thus, these were also included in the risk stratified approach. Several prior studies have reported that procalcitonin levels may be important predictors for a more severe form of disease (17,21). See examples in two randomly selected patients admitted with COVID-19–related pneumonia, acute respiratory distress syndrome, and important surges in inflammatory biomarkers who developed severe hyperglycemia followed by our team (Fig. 1A and B).
In the next steps, we created protocols for insulin delivery for nurses entering individual patient rooms. We prioritized reducing the number of glucose checks as much as safely possible in order to minimize health care providers’ exposures while also conserving PPE. The U.S. Food and Drug Administration recently approved the use of continuous glucose monitors for inpatient glucose measurements (22). While this definitely helps in reducing exposure and conserving PPE, in our experience, their accuracy is not validated in the most critical patients due to superimposed hypotension, use of vasopressors, and possibly high-dose acetaminophen, which can falsely elevate glucose levels. Additionally, given the complex care of these patients, the extra burden of teaching the use of a new tool on nurses was not sustainable. Thus, for convenience, we included arterial and venous blood glucose values, which are frequently drawn in ventilated patients and in patients receiving high dose of intravenous vitamin C infusion, to replace point-of-care glucose checks, and we reduced the number of glucose checks to every 6 h in the majority of cases. To further reduce burden on primary teams, and for timely insulin dose adjustments to parallel changes in inflammatory and the rest of disease progress markers, our team was performing insulin dose adjustments multiple times a day and was in charge of writing all insulin orders for inpatient hyperglycemia management. This was a critical component for success, given the very fluid clinical status of the severely ill COVID-19 patients, necessitating a very close watch and constant changes in insulin regimens for successful titrations.
The tailored protocols developed are described in Tables 1 and 2 with overall targeted blood glucose goals of 150–180 mg/dL. However, blood glucose levels <200 mg/dL were also targeted in some patients with very labile and critical forms of disease, particularly since most were also on continuous tube feeding and thus in a constant postprandial state.
For critically ill patients with severe hyperglycemia (blood glucose >450 or 500 mg/dL), an insulin infusion was initiated with titration often requiring very high rates—up to 12–20 units/h and occasionally up to 40 units/h. Once glucose ranges were within 200–300 mg/dL at lower hourly insulin drip rates, we would transition to subcutaneous insulin as soon as possible given the extenuating health care considerations described above.
Scheduled regular insulin, a sliding scale, and basal insulin adequately timed with other nursing interventions, especially arterial blood gas checks for ventilator settings, helped successfully lower glucose levels into goal range without increasing nurse contact, thus decreasing overall burden and PPE use. Our algorithms to predict labile glucose values with significant hyper- and hypoglycemia were improved by monitoring the changes in inflammatory biomarkers levels checked by the intensive care unit (ICU) teams, thus allowing us to prompt up or down titrations of insulin doses more confidently to prevent either further glucose surges or hypoglycemia. Given that insulin resistance reduces dramatically as a patient’s clinical condition improves, we proactively reduced insulin doses as soon as reductions in inflammatory biomarkers trends were documented. This flexible approach following trends in frequently monitored inflammatory markers to help us guide insulin titrations was a critical part of our evaluation. Our observations and developed algorithms were in fact in concordance with the recent publication by Hamdy and Gabbay (23) outlining a similar experience and administration of regular insulin every 6 h in the management of diabetes in COVID-19 patients in ICU at the Joslin Diabetes Center. In addition, similar to our Perspective, the recent review by Al-Jaghbeer and Lansang (24) provides broad guidance in management of hyperglycemia in COVID-19 patients in ICU.
Our Perspective has several limitations. First, the data described and the algorithm we have developed are not a result of a randomized clinical trial or a research study but instead are based on our direct observations in the patients with severe COVID-19 disease we followed. Thus, we do not have a control population to compare differences in outcomes. Second, we acknowledge that given the very fluid status associated with this pandemic and the rapid rise in the number of severe cases admitted, there are many confounding factors that we are unable to account for at this time. In addition, we are unable to provide at this time more specific data on the direct effectiveness of this algorithm on several important outcomes such as mortality, time to recovery, length of intensive care or overall hospital stay, or rate of severe complications associated with the algorithm. Lastly, these initial observations may only be applicable to patients with phenotypes and socioeconomic status similar to those who were admitted to the University of Michigan Hospitals. Thus, these need to be confirmed in larger and controlled studies that are including the new evidences on disease course, risk factors, management, and prognosis of COVID-19 infection that are emerging globally.
Strengths of the algorithm are the fact that the continuous management of the insulin orders by our diabetes team allowed us to proactively and effectively react to surges in glucose levels driven by disease activity and significantly decrease the burden on the primary teams.