Risk factors for Relative and Absolute Hypoglycemia in Patients Treated for Diabetic Ketoacidosis and Hyperosmolar Hyperglycemic Syndrome

Abstract

Introduction: Glycemic management in the intensive care unit is an evolving practice area. This evolution has included the refinement of blood glucose targets, matching glycemic management to premorbid status, and investigations into the impact of glycemic variability and relative hypoglycemia on ICU outcomes. The interplay between these phenomena and absolute hypoglycemia has yet to be investigated in hyperglycemic emergencies. Objectives: To examine the incidence of and risk factors for relative hypoglycemia and absolute hypoglycemia in patients admitted to an intensive care unit for the management of hyperglycemic emergencies. Methods: This was a retrospective, single-center, exploratory analysis of adults admitted to the medical intensive care unit for diabetic ketoacidosis or hyperosmolar hyperglycemic syndrome. The primary outcome was the incidence of relative hypoglycemia, defined as a blood glucose level 30% lower than baseline. The baseline was determined by the estimated average blood glucose calculated from hemoglobin A1c within 3 months of index admission. Secondary outcomes were ICU length of stay, glycemic variability, and incidence of absolute hypoglycemia.Results: Relative hypoglycemia was observed in 60% of patients in the cohort. Longer insulin infusion duration and higher hemoglobin A1c levels were found to statistically increase the risk of developing relative hypoglycemia. Higher glycemic variability and longer ICU length of stay were associated with the risk of developing absolute hypoglycemia. Conclusions: Relative hypoglycemia is a frequent occurrence in this patient population. Hemoglobin A1c and duration of the insulin infusion statistically influenced the risk of developing relative hypoglycemia. Higher glycemic variability and longer ICU stay were significantly associated with developing absolute hypoglycemia. While relative hypoglycemia is common in hyperglycemic emergencies, the clinical impact remains uncertain and warrants additional investigation.

Introduction

Glycemic management in the intensive care unit (ICU) is an evolving area of practice. This evolution has included the refinement of blood glucose targets, matching of glycemic management to premorbid status, and investigations into the impact of glycemic variability (GV) and relative hypoglycemia (RH) on ICU outcomes.^1-3 The interplay between GV, RH, and absolute hypoglycemia (AH) has yet to be investigated in the hyperglycemic emergencies diabetic ketoacidosis [DKA] and hyperosmolar hyperglycemic syndrome [HHS].

GV refers to the fluctuations of a patient’s blood glucose levels over some designated period. In critically ill patients, greater GV has been associated with increased length of stay (LOS) and mortality.^4,5 RH describes experiencing symptoms of hypoglycemia while serum glucose is still measured above 70 mg/dl.^5,6 In the ICU setting, symptoms are often not assessable. In this context, RH has been defined as a ≥30% decrease from estimated preadmission glycemia or any drop into the blood glucose range of 70 to 110 mg/dl for patients with preadmission hemoglobin A1c (HbA1c) ≥8.0%.^3,7,8 RH triggers the release of counterregulatory stress hormones similar to the pathophysiology of AH. ³ Initially described in outpatient diabetes management, RH’s clinical significance is becoming increasingly recognized in the ICU patient population.^3,8,9 In ICU patients, RH has been found to be a predictor of AH and has been associated with increased mortality, yet these populations have generally excluded patients with hyperglycemic emergencies.^3,8-10

Given the potential importance of minimizing GV and avoiding RH in other ICU populations, it is necessary to understand the effects of these variables during hyperglycemic emergencies. If similar associations are discovered, different treatment strategies could be developed to mitigate these negative sequelae. This study aimed to examine the incidence of and explore the association of GV, RH, and AH in patients admitted to the ICU for DKA and HHS. We hypothesized that RH would be a frequent occurrence in this population and would be associated with GV, AH, and increased LOS.

Methods

Study Design

This study was a retrospective, single-center, cohort analysis. It was approved by the University of Illinois Chicago’s Office for the Protection of Research Subjects (OPRS) Institutional Review Board (IRB; Protocol Number 2021-1209). The methods followed all institutional policies regarding the ethical treatment of research subjects and were aligned with the Helsinki Declaration of 1975. Data was extracted from a report of patients admitted with diagnosis codes for hyperglycemic emergencies from 2019 to 2020. Patients were included if they were admitted to the ICU with DKA or HHS as determined by the International Classification of Diseases (ICD)—10 codes between and had a hemoglobin A1c (HbA1c) value documented within 3 months of their admission. Patients were excluded if they were <18 years of age or underwent cardiothoracic surgery. Patients were screened by the authors until the number of patients needed to meet power were included.

Study Procedures

Once included, data points were extracted and collected from electronic medical records (Cerner Powerchart 2011, CERNER© Corporation). Demographic information included patient’s age (years), sex (male or female), race (African American, Asian or Pacific Islander, Caucasian, Hispanic, or Other), weight (kilograms), diabetes history (Type 1, Type 2 or gestational), and HbA1c ([%] closest to admission date within 3 months of index admission). Admission data included glucose level (mg/dl), beta-hydroxybutyrate (mmol/l), serum pH, anion gap (AG), serum osmolality (mOsm/kg), and Acute Physiology and Chronic Health Evaluation (APACHE) II. Patient-specific glycemic details were also collected and included glucose readings throughout the duration of the insulin infusion, which included all blood glucose readings from the start through the completion of the insulin infusion. Patients experiencing DKA were transitioned from the insulin infusion when serum glucose levels fell between 180 and 239 mg/dl and ketoacidosis resolved. Patients experiencing HHS were transitioned when serum glucose fell between 240 and 299 mg/dl and HHS resolved. Both of these strategies were according to institutional protocol (Supplemental Appendix Figure 1). The number of RH events (defined as 30% below a patient’s baseline glucose), the number of AH events (defined as blood glucose levels below 70 mg/dl), and the need for dextrose administration were also collected. Baseline serum glucose was calculated from HbA1c utilizing the estimated average glucose formula from the American Diabetes Association. ¹¹ Glucose variability was calculated for each patient utilizing the coefficient of variation (CoV), which was determined by taking the standard deviation (SD) of each patient’s blood glucose readings divided by their mean serum glucose during the insulin infusion utilization and multiplying by 100 to report as a percentage. ¹²

Details regarding the patient’s hospital course were collected, including ICU LOS (hours), duration of the insulin infusion (hours), and the total amount of insulin received (units). All data were collected and stored in REDCap © (Vanderbilt University, Version 12.2.10, 2022).

Outcomes

The primary outcome was the incidences of RH in the total cohort. Patients were then separated into 2 groups: (1) those that experienced a RH event or (2) those that did not experience a RH event during their ICU admission. Outcomes of ICU LOS, incidence of AH, and GV (CoV) were then compared between these groups as secondary outcomes. Additional secondary outcomes included identifying variables independently associated with RH or AH.

Statistical Analysis

Given our hypothesis that RH would be a predictor of AH, a power calculation was conducted a priori. When comparing patients that experienced RH to those who did not, a population of 176 patients would be needed to detect a 12% absolute difference in the incidence of AH. This sample size was calculated assuming a rate of AH to be 15% and 3% in the 2 groups with 80% power and an alpha value of 0.05.

All data were analyzed and compared using descriptive statistics. Normally distributed data were presented as means with standard deviations (SD) and non-normally distributed data were presented as medians with interquartile ranges [IQR]. Comparisons of central tendency between those that experienced RH and those that did not were made with the T-test for normally distributed data and Mann Whitney U Test for non-normally distributed data. Chi-square or Fisher’s exact test were used to compare incidences between groups.

Logistic regression was utilized to determine the associations between variables and development of RH and AH. We first utilized Pearson or Spearman correlation to identify variables associated with RH and AH. Variables with P < .1 or those known to influence the outcomes of interest were included in multivariable regression modeling. A backward, conditional regression model was utilized for each analysis to determine which variables were independently associated with RH and AH. In both models, variables were assessed for collinearity using variance inflation factor (VIF) assessment, and variables were excluded if the results demonstrated collinearity. A P-value <.05 was considered statistically significant. Data was analyzed using IBM SPSS © (IBM Corp, Version 26, 2019) and Microsoft Excel © (Microsoft, Version 2205, 2021).

Results

A total of 312 patients were screened, with 136 patients excluded, leaving a total of 176 patients included in the final analysis (Figure 1). Demographic information is listed in Table 1. Patients most often presented with DKA (80.6%) and had a history of Type 2 Diabetes. The remainder of the population presented with HHS (16.5%) or mixed presentation (2.9%). The mean (SD) HbA1c was 11.33 (3.05) %, with a distribution of HbA1c depicted in the Supplemental Appendix Figure 2. At admission, the median [IQR] glucose was 447.5 [323.75-624.7] mg/dl. The median [IQR] beta-hydroxybutyrate level was elevated to 3.75 [1.1-7.2] mmol/l, and the median [IQR] pH was 7.33 [7.25-7.39].

Figure 1.

CONSORT diagram.

Table 1.

Demographic and Hyperglycemic Emergency Presentation Data.

Variable	N = 176
Age in years, mean (SD)	46.48 (17.1)
Weight in kg, mean (SD)	85.95 (29.7)
Male, n (%)	92 (52)
Race, n (%)
African American	87 (49.4)
Asian/Pacific Islander	4 (2.3)
Caucasian	15 (8.5)
Hispanic	26 (14.8)
Other/not listed	44 (25)
Diabetes History, n (%) a
T1DM	46 (26.1)
T2DM	129 (73.3)
Gestational	2 (1.1)
Mean HbA1c, % (SD)	11.33 (3.05)
Mean estimated average glucose in mg/dL (SD) b	278 (87.44)
Mean relative hypoglycemia cut-off (SD) c	163.12 (73.35)
Hyperglycemic Emergency Presentation
DKA, n (%)	142 (80.6)
HHS, n (%)	29 (16.5)
Mixed presentation, n (%)	5 (2.9)
Glucose at admission, mg/dl, median [IQR] d	447.5 [323.75-624.7]
APACHE II score, median [IQR]	11 [8-16]
Osmolarity in mOsm/kg, median [IQR]	301.65 [292.07-314.07]
pH, median [IQR]	7.33 [7.25-7.39]
Beta-hydroxybutyrate in mmol/l, median [IQR]	3.75 [1.1-7.2]

Note. ADA = American Diabetes Association, APACHE = acute physiology and chronic health evaluation, DKA = diabetic ketoacidosis, eAG = estimated average glucose, HHS = hyperosmolar hyperglycemic syndrome, IQR = interquartile range, kg = kilograms, mmol = millimole, mOsm = milliosmoles, n = number, pH = potential hydrogen, SD = standard deviation, T1DM = type 1 diabetes mellitus, T2DM = type 2 diabetes mellitus.

^aOne patient had both T1DM and Gestational Diabetes listed in their past medical history, so was counted twice.

^bEstimated average glucose = 28.7 × A1C − 46.7 ¹² —from A1C and eAG | ADA. Accessed May 31, 2022. https://www.diabetes.org/diabetes/a1c/a1c-and-eag

^cValue 30% below a patient’s estimated average glucose was calculated for each individual patient, the mean of these values is reported here.

^dGlucose at admission refers to the first documented blood glucose value in the patient’s chart, typically from the presentation to the emergency room.

Throughout the population, there were a total of 985 RH events (Table 2). One hundred seven (61%) patients experienced at least one RH event, and the mean (SD) number of RH events per patient was 5.6 (8.11). There were 38 individual AH readings based on glucose readings in 16 (9%) of patients in the cohort. The mean (SD) CoV was 34.08% (16.27). The mean (SD) duration of treatment with insulin was 20.93 (35.61) hours, and the mean (SD) total dose of insulin was 100.3 units (181.83). The median [IQR] LOS in the ICU was 31.88 [16.69-71.18] hours.

Table 2.

Primary and Secondary Outcomes.

Variable	N = 176
Number of relative hypoglycemic events, n	985
Number of patients who experienced at least one relative hypoglycemic event, n (%)	107 (60.8)
Average number of relative hypoglycemic events per patient (SD)	5.6 (8.1)
Average Coefficient of Variation (SD)	34.08 (16.27)
Number of absolute hypoglycemic events, n	38
Number of patients who experienced at least one absolute hypoglycemic event, n (%)	16 (9)
Average number of absolute hypoglycemic events per patient (SD)	0.31 (1.38)
Average ICU length of stay in hours (SD)	83.07 (162.4)
Average total dose of insulin in units (SD) a	100.3 (181.83)
Average duration of insulin in hours (SD)	20.93 (35.61)

Note. ICU = intensive care unit, n = number, SD = standard deviation.

^aDetermined by summation of hourly insulin rates in units/hour

Those who experienced RH were found to have greater CoV (37.11% [14.92] vs 26.75% [18], P < .0001) compared to those who did not (Table 3). LOS and number of AH events did not differ statistically between the groups (P > .05).

Table 3.

Comparison Data of Secondary Outcomes Between Patients Who Did and Did Not Develop Relative Hypoglycemia.

Variable	RH (n = 107)	No RH (n = 69)	P-value
Coefficient of Variation, %, Mean (SD)	37.11 (14.92)	26.75 (18.0)	<.0001
Length of Stay, hours, Median [IQR]	29.03 [16.18-66.53]	34.58 [17-72.25]	.630
Absolute hypoglycemia, n (%)	13 (7.4)	3 (1.7)	.107
APACHE II Score, Mean (SD)	12.45 (0.66)	11.64 (0.65)	.38

Note. IQR = interquartile range, n = number, RH = relative hypoglycemia; SD = standard deviation.

In bivariate analyses, AH, HbA1c level, total insulin dose, insulin infusion duration, CoV, and AG were identified as variables associated with RH (P < .1). Table 4 shows the results of the multivariate logistic regression analysis of variables associated with RH. For every 1% increase in HbA1c, there was a 1.3-fold increase in the likelihood of RH (P < .001). Additionally, for each hour of insulin infusion usage, there was an increased likelihood of RH by 1.03 times (P = .037). GV was also significantly associated with the development of RH. For every 10% increase in CoV, patients’ risk of developing RH increased 10-fold (P = .01). There was not a statistical association between RH and rates of AH (OR = 2.839, 95% CI: 0.92-20, P = .063).

Table 4.

Multivariate Regression Model for Developing Relative Hypoglycemia and Absolute Hypoglycemia.

i. Regression model for developing relative hypoglycemia
Variable	Odds ratio	95% confidence interval	P-value
Absolute Hypoglycemia	2.839	0.92-20	.063
HbA1c—per 1% increase	1.2666	1.114-1.439	<.001
Total Dose of Insulin—per unit	0.999	0.994-1.004	.598
Insulin Infusion Duration—per hour	1.029	1.002-1.057	.037
Anion Gap	1.023	0.975-1.074	.345
Coefficient of Variation—per 10% increase	1.031	1.007-1.056	.01
Model information: Chi-square = 34.663, df = 6, P < .001, Nagelkerke R-squared = 0.394
ii. Regression model for developing absolute hypoglycemia
Variable	Odds ratio	95% confidence interval	P-value
HbA1c—per 1% increase)	0.844	0.683-1.045	.119
Type 1 Diabetes	0.119	0.033-0.431	.001
Length of Stay—per hour	1.003	1.00-1.006	.041
Coefficient of Variation—per 10% increase	1.036	1.002-1.071	.039

Note. Model information: Chi-square = 32.16, df = 5, P < .001, Nagelkerke R-Squared: 0.368. HbA1c = Hemoglobin A1c.

In bivariate analyses for AH, CoV, HbA1c level, Type 1 Diabetes Mellitus, ICU LOS, and AG all were correlated with the development of AH (P < .1). Table 4 shows the results of a multivariate logistic regression analysis of variables associated with AH. Hemoglobin A1c, type 1 diabetes, and AG were inversely associated with the development of AH. However, for every 10-point increase in the CoV, the risk of developing AH increased 10-fold (P = .039). Additionally, for each additional hour a patient stayed in the ICU, their risk of developing AH increased by 1.003 times (P = .041).

Discussion

In this retrospective, exploratory analysis examining the incidence of RH in adult patients being treated for DKA or HHS, RH occurred in about 60% of the population. Our population consisted mostly of patients with T2DM, which is typical for the patient population seen at our institution. The median blood glucose at presentation was greater than 200 mg/dl, driven by the inclusion of patients with hyperglycemic emergencies in the cohort. Hemoglobin-A1c, GV, and duration of the insulin infusion were risk factors associated with an increased risk of developing RH. We did not observe a statistically significant relationship between RH and AH in our cohort, which could be limited by the rate of AH and/or the number of patients included. GV and LOS were risk factors for developing AH. At the time of publication, to our knowledge, this is the first study to describe factors associated with RH and AH in patients being treated for DKA and HHS.

RH was initially recognized as an outpatient phenomenon but is becoming increasingly recognized in ICU patients. Kwan et al conducted a similar retrospective cohort study examining the incidence of RH in over 1500 ICU patients. ³ The authors reported a 52.2% incidence of RH in their study population. Our study, though smaller, found a similar incidence in RH in the DKA and HHS population, which was excluded from many of the studies examining RH.^3,8,12 Krinsley et al found a large amount of RH events in a study of 3500 ICU patients while observing that greater time with RH was associated with increased mortality. ⁸ More recently, Okazaki et al examined relative dysglycemia’s impact on mortality by calculating an A1c derived average glucose and glycemic ratios based on minimum and maximum glucose readings in 1700 ICU patients. ¹⁰ These authors concluded that relative dysglycemia, including hypoglycemia, hyperglycemia, and the 2 combined, was associated with an increase in-hospital mortality after adjustments of covariates. ¹⁰ We did not evaluate mortality in our analysis given the traditionally low mortality for these hyperglycemic emergencies at our institution, but given the relationship between RH and mortality, future larger analyses including a variety of institutions should explore this outcome. What was very clear from our analysis was that RH is a frequent occurrence meaning that many of the patients being treated for hyperglycemic emergencies are potentially experiencing a counterregulatory stress hormones like the pathophysiology of AH. ³ The sequelae of this potential stress state remain unclear.

 RH has been found to be a predictor of AH in ICU patients. Kwan et al demonstrated in their review that over 65% of patients who experienced an AH event had at least one RH event preceding it. ³ However, the authors only included patients with at least one glucose reading and a HbA1c measurement during their admission or within the previous 3 months but did not comment on including DKA and HHS patients in their analysis, making direct comparison of our results challenging. ³ Our investigation did not observe a statistical association between RH and AH. This could be secondary to our cohort’s low overall incidence of AH and/or the relatively small number of patients in our analysis. The low numbers of AH may be due to the protocolized nature of managing DKA or HHS at our institution or the degree of change in glucose that would necessitate reaching levels of AH. To prevent AH during insulin infusion therapy, the American Diabetes Association suggests protocols include the initiation of dextrose infusions if glucose readings fall below 300 mg/dl and continue until resolution of DKA and HHS. ¹³ Many of our patients in this study experienced blood glucose readings that fell below 300 mg/dl and may have had appropriate preventative measures taken to prevent them from developing AH despite frequently experiencing RH. Larger investigations exploring this relationship would be warranted to confirm or refute these findings.

GV has also been found to be an important phenomenon in ICU patients. Egi and colleagues described this concept in a cohort of over 7000 ICU patients and found that the SD of glucose concentration was an independent predictor of ICU mortality. ¹⁴ Todi and Battacharya examined the impact of GV on mortality in over 2000 critically ill patients across multiple ICU settings. ¹⁵ Their study found that GV, defined by SD and GV index, was associated with increased mortality in their cohort of mixed ICU patients. Krinsley et al also examined GV as part of their study. ⁴ Utilizing CoV to represent GV, the authors found that GV was associated with increased mortality across their entire population. Though our study did not examine mortality, it did examine the impact of GV on other outcomes in DKA and HHS patients. We found that GV was a predictor of both RH and AH. These results highlight the importance of trends in serum glucose ranges, in addition to singular abnormal values when designing a glycemic management regimen for ICU patients. At our institution, our DKA/HHS treatment protocol only incorporates the measured glucose value to determine the insulin rate. Perhaps, more complex algorithms that capture the relative change in glucose through GV in additional to measured glucose values would be advantageous in DKA/HHS management to prevent the association of GV with RH and AH.

It is important to note the differences in the use of continuous infusion insulin therapy for managing hyperglycemic emergencies versus essential hyperglycemia in critically ill patients. Generally speaking, the management of DKA and HHS focuses on the correction of fluid status, electrolyte abnormalities, and hyperglycemia.^13,16 Utilization of insulin will correct hyperglycemia relatively quickly, but the underlying acidemia and/or electrolyte abnormalities typically persist for a longer period of time hence the addition of dextrose once blood glucose values fall below a certain threshold (less than 300 mg/dl at our institution).^13,16 The addition of dextrose helps prevent the development of AH which is demonstrated by the low incidence of absolute hypoglycemia in our cohort. This contrasts the management of hyperglycemia with continuous infusion insulin, where rates are adjusted per protocol to a specific target without the consideration of insulin for other purposes. Additionally, the glycemic goals during treatment of DKA/HHS utilizing protocols are static numeric glucose targets. They are not patient specific. This leads to 2 observances in relation to our study. First, patients with higher HbA1c (and therefore, higher estimated average glucose) will by nature of the existing protocols experience higher rates of RH. We observed this association. Second, is that there is a great opportunity to tailor therapy according to the pre-morbid glycemic control of our patients. The complexity of this becomes challenging, but potentially through technology, we could create patient specific algorithms incorporating baseline HbA1c and GV to avoid both RH and AH.

Something not highlighted in our study was the role a clinical pharmacist plays in glycemic management in the ICU. At our institution, a dedicated clinical pharmacy specialist is part of the interdisciplinary team and rounds daily with the treatment team, providing verbal recommendations on rounds. At the time of this study, there were no official pharmacy-driven glycemic management protocols at our institution. This role has been explored in numerous studies. Ngyuen and colleagues examined the impact of their pharmacists’ role in managing glucose levels in surgical ICU patients and found a statistically significant decrease in hyperglycemic events in the units with a dedicated clinical pharmacist. ¹⁷ In patients with diabetes, patients that were at high risk for severe hypoglycemic events (blood glucose less than 70 mg/dl), Cook and colleagues demonstrated that the implementation of a pharmacist-based glycemic control protocol significantly decreased the number of severe hypoglycemic events in their institution. ¹⁸ Additionally, pharmacists play numerous roles in managing DKA and HHS, participating in institutional insulin infusion protocol development and adherence, monitoring of these patients, and actively participating in transitions of care. ¹⁹ Given our findings, perhaps pharmacists with the use of technology, could create the patient-specific tailored treatment algorithms that may be necessary incorporating general treatment goals, patients baseline glucose control, and aspects of treatment response such as GV. Further studies to examine the role which a clinical pharmacist in this capacity can impact relative hypoglycemia and glycemic variability should be explored.

The main strength of this study is how it highlights the rates of RH in an ICU population of patients being treated for hyperglycemic emergencies. This study found that RH was a common occurrence in hyperglycemic emergency patients, but the impact of these findings requires further investigation. As our study was retrospective and we determined RH as a calculated value based on HbA1c, it is unclear if patients were symptomatic or if counterregulatory stress hormones were being triggered, in theory leading to the negative sequalae of RH. Based on what we know about RH in the ICU, ⁶ it is likely that both were present, but future prospective observational and interventional studies may be warranted. As our study only investigated the rates of RH and AH during the insulin infusion, future studies should evaluate RH, AH, and GV after the completion of the insulin infusion during the subcutaneous insulin administration period to fully capture the entire course of DKA and HHS. Given that GV predicted RH and AH during DKA and HHS treatment in this study, GV as a predictor these variables after the insulin infusion stopped could be another area of further investigation.

Limitations

Our study was not without limitations, many of which are secondary to the nature of retrospective chart review. External validity is potentially limited due to the single center nature of the study and may not reflect the hyperglycemic emergency population at every institution. Being a retrospective chart review, collection of the data was reliant on accurate documentation and was at risk of both recall and miscalculation bias. To circumvent this limitation, consistent data collection through trained data collectors was utilized. In addition, all data was collected from our institution’s historic electronic medical record limiting the functionality of the software. Despite limited functionality, data was not missing for any of the outcomes related to the primary and secondary outcomes. Additionally, one of the limitations of this study was the small amount of AH observed in the cohort. This was likely due to our institution’s specific protocol for management of hyperglycemic emergencies as described above. While we met power with the number of total patients, given the small number of AH cases and rates of AH less than we anticipated, we were unable to fully assess the influence of RH on AH in this patient population. Despite these limitations, this study remains one of the first to investigate the concepts of RH, AH, and GV in DKA and HHS. Additionally, it should be noted that the association of HbA1c and RH was across a range of HbA1c values of 5.4% to 19.2%. If patients present with HbA1c outside of this range, the association may not be present. It is also imperative to mention that the current derived protocols for treating DKA and HHS very often causes RH for those patients with extreme elevations in their baseline HbA1c. Perhaps, patient specific DKA and HHS protocols should be considered. As RH was not statistically associated with AH with the current study sample and power, it is unclear if the RH seen in our investigation has impact on other outcomes. Much larger studies would be warranted to better explore these parameters. This study intended to explore this phenomenon in a patient population that up until this point were not included in the literature. As previously mentioned, future studies into the clinical impacts of RH, AH, and GV in this patient population are warranted.

Conclusion

Glycemic management in the ICU is an evolving area of critical care medicine, with current standards reflecting the “one size fits all” approach. The concept of RH is becoming increasingly recognized in ICU patients as individualized glycemic goals are considered. RH was found to be a frequent event in patients treated for DKA or HHS. Both the duration of the insulin infusion and higher HbA1c levels were found to statistically increase the risk of developing RH. Higher GV was found to be related to RH and AH and could serve as an additional target for treatment. The clinical impact of developing RH in hyperglycemic emergencies remains unknown, but it appears that a patient-specific tailored approach may be warranted. Further investigation is needed.