National-Level Outcomes of Diabetic Ketoacidosis with and without Coma: A Retrospective Analysis of 2.3 Million U.S. Hospitalizations

Introduction

Diabetic ketoacidosis (DKA) is a life-threatening diabetes complication characterized by hyperglycemia, ketosis, and metabolic acidosis [¹]. DKA remains a significant and growing global complication affecting people with both type 1 and some with type 2 diabetes, contributing to substantial morbidity and mortality [²]. The burden is particularly high in low-resource countries, for instance, in Ethiopia, 60.7% of newly diagnosed pediatric type 1 diabetes patients presented with DKA [³]. In the United States, DKA incidence among commercially insured individuals was 55.5 per 1,000 person-years, with rates increasing from 2011 but slightly declining in 2018–2019 [⁴]. Approximately one-third of U.S. youth newly diagnosed with type 1 diabetes present with DKA, which is associated with worse glycemic outcomes despite intensive insulin management [⁵]. National DKA hospitalizations have been rising due to factors including delayed diagnosis, poor medication-taking behavior, and infections, though most cases respond favorably to timely interventions [⁶]. Coma is a rare but life-threatening complication of DKA, most often resulting from severe cerebral edema or profound metabolic derangement[⁷]. It occurs in approximately 10% of cases and signifies advanced physiological decompensation, necessitating immediate intensive care[⁷]. Cerebral edema is the leading cause of coma in pediatric DKA and is often fatal or results in long-term neurological damage if not rapidly identified and treated [⁸]. Other potential causes, such as brainstem infarction, may also lead to coma and can be fatal despite aggressive ICU intervention[⁹].

Despite its clinical importance, coma in DKA has not been systematically studied in large cohort trials, resulting in a lack of standardized guidance for neurological assessment and management. [⁷]. Many studies on DKA outcomes rely on single-center or regional data with small sample sizes, limiting the generalizability of their findings [¹⁰]. National-level or population-based research on the epidemiology and outcomes of DKA-related coma remains scarce[¹¹]. Most clinical trials and observational studies on DKA do not stratify patients based on the presence or severity of coma, which obscures understanding of its specific impact on outcomes [⁸]. Although pediatric studies more often address cerebral complications, they frequently lack the long-term follow-up needed to assess neurological sequelae [¹²]. Variation in how coma is defined and diagnosed across studies further complicates data interpretation and limits consistency in reporting[⁷].

This nationwide study utilizing the National Inpatient Sample (NIS) is warranted for several reasons. The NIS represents the largest all-payer inpatient database in the United States, providing a robust platform for population-level analysis of DKA outcomes. While prior single-center or regional studies have provided valuable insights, they often lack statistical power and generalizability to assess less common presentations such as DKA with coma, which comprises only a small percentage of total DKA cases [¹¹]. The NIS dataset allows for a comprehensive assessment of clinical outcomes stratified by coma status across diverse geographical regions, hospital types, and patient demographics, overcoming the limitations of previous smaller-scale studies.

Despite the clinical significance of altered consciousness in DKA, there remains a paucity of large-scale investigations specifically examining how coma affects outcomes and resource utilization in this condition. Previous research has established that consciousness level represents a key determinant of DKA severity, yet standardized clinical approaches for neurological assessment and management of DKA remain inconsistent [⁷]. By leveraging the NIS database’s extensive demographic and clinical information, this study offers a unique opportunity to quantify the impact of coma on mortality, length of stay, and hospital charges while controlling for confounding variables.

Methods

Results

Discussion

To our knowledge, this is the most comprehensive retrospective cohort study that addressed patients with DKA in the USA over the last decade. Using multivariable regression analyses on more than 2.3 million DKA admissions from 2016 to 2022, we identified significant disparities in clinical outcomes by age, diabetes type, coma status, payer source, and hospital characteristics.

Previous studies have identified several factors associated with increased mortality in DKA, including advanced age, infections, impaired consciousness, and the presence of severe comorbidities [¹³,¹⁴,¹⁵,¹⁶]. In our study, the highest in-hospital mortality, longest hospital stays, and greatest hospital charges were observed among patients aged $\geq$65 years, those with type 2 diabetes, Asian or Pacific Islander patients, and those presenting with coma. Older adults often have multiple comorbid conditions that contribute to clinical complexity and may prolong hospitalization for reasons beyond DKA itself. Their higher acuity and increased need for medical interventions likely account for both the extended length of stay and the elevated healthcare costs.

Older adults in the United States face a heightened risk of DKA-related coma due to atypical presentations that delay diagnosis., Unlike younger individuals, older adults are more likely to present with altered mental status rather than classic DKA symptoms [¹⁷]. Infections, particularly respiratory and urinary tract infections, are the primary precipitating factors for approximately 70.1% of DKA cases in this age group, rapidly accelerating metabolic decompensation[¹⁷]. Comorbidities, including renal dysfunction and hypoalbuminemia, diminish physiological reserves, increasing vulnerability to severe acidosis and coma [¹⁷]. These factors likely explain the significantly higher in-hospital mortality observed of 7.69% in older adults with DKA, compared to 3.48% and 1.34% in the age groups 35-64 and 18-34, respectively, as in our study.

Prevention strategies for older adults with diabetes in the USA must address their unique DKA risk factors. Structured diabetes self-management education (DSME) tailored for cognitive decline and complex medication regimens can significantly improve medication-taking behavior and symptom recognition [¹⁸]. Because infections are the most common DKA trigger in elderly patients, early detection and treatment are essential components of prevention [¹⁹]. Healthcare systems should also implement 24-hour telehealth services and clear sick-day protocols to prevent DKA during acute illnesses or missed insulin doses [²⁰]. Regular medication reviews and caregiver engagement can prevent unintentional insulin omissions that commonly precipitate DKA in older patients [²⁰]. Clinicians should maintain a low threshold for ketone testing in elderly individuals with hyperglycemia, even in the absence of classic DKA symptoms, given the atypical presentations common in this population (Anupama et al., 2018). Finally, prevention efforts should be personalized, incorporating socioeconomic status, mental health, and prior history of DKA into individualized risk mitigation strategies [¹⁸].

In our study, individuals with type 2 diabetes accounted for 39.9% of DKA admissions—fewer than those with type 1 diabetes, who comprised 49.1%. This distribution is consistent with previous studies from the United States and England, where the proportion of DKA admissions among patients with type 2 diabetes has ranged from 20% to 30% [²¹,²²]. Regardless of whether multivariable regression was used, prior studies consistently concluded that people with type 2 diabetes experience higher in-hospital mortality and longer hospital stays compared to those with type 1 diabetes [²³,²⁴,²⁵].

This difference is likely attributable to the older age distribution of individuals with T2DM, who often have a clustering of metabolic risk factors such as hypertension, dyslipidemia, and obesity [26]. Moreover, patients with T2DM are more likely to have established cardiovascular comorbidities and other complications that exacerbate clinical severity [26]. Greater insulin resistance and the presence of a concurrent hyperosmolar state further complicate management and contribute to prolonged hospitalization. Together, these factors also lead to higher total hospital charges among patients with DKA and type 2 diabetes.

Consciousness level is a clinical symptom that must be evaluated in patients diagnosed with DKA. Altered consciousness may result from various factors, including acidosis, elevated osmolality, the direct impact of ketone bodies, diminished cerebral blood flow, reduced glucose uptake and utilization by neuronal cells, as well as other contributing elements such as severe infections and strokes [²⁴,¹⁶]. Altered consciousness can occur from brain edema, a complication of fluid therapy in DKA [²⁶]. Altered consciousness is considered one of the parameters for evaluating the severity of DKA and is frequently linked to mortality rates. Our findings detected that comatose patients have a higher in-hospital mortality rate compared to non-comatose patients. This result was further supported by findings from Sato et al., who reported that the in-hospital mortality rate was 4.33 times higher in the alert patients’ group compared to the control group. Also, Venkatesh et al. revealed that the average GCS score was significantly higher in DKA survivors compared with non-survivors [²⁷].

Our study revealed that Medicare beneficiaries experienced the poorest outcomes among all payer groups, characterized by the longest hospital stays (6.58 days), the highest in-hospital mortality (6.35%), and the highest charges ($74,860). Prior research analysis reveals Medicare beneficiaries face a 3.3-fold higher adjusted risk of recurrent DKA hospitalizations compared to privately insured individuals, even after controlling for socioeconomic variables [²⁸]. This heightened vulnerability is partly attributable to demographic characteristics, as Medicare patients typically present with advanced age and multiple comorbidities, including cardiovascular and renal disease, significantly increasing DKA episode severity and mortality rates [²³]. Systemic healthcare delivery issues further compound these challenges, including fragmented care, insufficient diabetes education, and limited access to specialist endocrinology services [²⁸]. Notably, younger individuals qualifying for Medicare through disability represent a particularly high-risk group, as they often face both socioeconomic disadvantages and complex medical needs that exacerbate DKA vulnerability [²⁸]. Our findings further support these disparities, as Medicare patients required more intensive care, longer hospitalizations, and incurred higher total costs, consistent with more complex clinical presentations and greater resource needs. After adjustment for sex, multivariate regression analysis showed that females were associated with significantly shorter hospital stays, lower in-hospital mortality rates, and fewer hospital charges. However, the difference in length of stay, while statistically significant, was minimal and likely not clinically meaningful. Although the effect sizes for mortality and hospital charges were statistically significant, they were relatively small; however, the differences in hospital charges may still be meaningful from a health system and cost-management perspective. Previous studies evaluated that females with DKA and females with decompensated diabetes had lower in-hospital mortality and lower mortality, respectively [¹³,²⁹]. Conversely, another study concluded no significant differences in in-hospital mortality between males and females hospitalized with DKA [³⁰]. Our study was aligned with the analysis of the large-scale Japanese dataset; both concluded that there was a difference among sexes as a factor that affects the mortality rate, but it was still less clinically important when compared to other factors like coma, age, or DM type [²⁵].

For adjustment for hospital teaching status and hospital bed size, we found that urban teaching hospitals and hospitals with large hospital bed sizes had the highest mortality rates, hospital stays, and total charges. More complicated cases come to teaching hospitals; thus, they carried the highest mortality rate as they were admitted to teaching hospitals due to more supported resources and more frequent intensive care [³¹]. Thus, patients who were hospitalized in teaching hospitals had a longer hospital stay and total charges because of the complexity of the admitted cases and ICU utilization. Either hospital stay may be prolonged due to the nature of the educational process.

For adjustment for hospital regions, our findings suggest that Midwest hospitals carried a lower risk for developing mortality, but these results also had less clinical importance, which concluded that this is a less clinically important factor when compared to other factors. But total charges for Midwest hospitals carried the best results compared to other regions, as they carried the least total charges. The hospital stay results were fewer for Midwest and West hospitals compared to South and Northeast hospitals.

Additionally, our results revealed that self-payers carried the shortest hospital stay period and the fewest total charges compared to other Primary Expected Payer groups, despite still carrying no remarkable difference according to mortality. Finally, hospital location and income quartile appear to be the least important factors affecting outcomes in this study.

Strengths and limitations: This retrospective cohort study contains a huge number of patients, which exceeded 2 million in the last decade; thus, the data can be validated through a wider portion of patients than previous studies in the USA. Also, this retrospective cohort study is favorable because it contains three outcomes: in-hospital mortality, hospital stay, and total charges. Also, it contains a multivariate regression model for each outcome, which will give an insight into the hospital stay and total charges that are more important outcomes for the decision-making process among hospitals, patients, and physicians. However, our study had some limitations. We were unable to address the severity of DKA in this study, as the database didn’t contain laboratory data (e.g., levels of plasma glucose, glycated hemoglobin, insulin, ketone bodies, electrolytes, anti-glutamic acid decarboxylase antibodies, and c-peptide). As we addressed comatose patients as severe cases, we couldn’t address the severity of DKA among non-comatose patients. There are other confounders we could access through our analysis (e.g., BMI, infection status, Charlson Comorbidity Index, or scores for daily activity). HHS may not have been fully excluded from our study population, as there is a clear distinction between DKA and HHS in clinical settings.

Conclusion

We found that age (>65 years), consciousness level (comatose patients), race (Asian or Pacific Islander), and hospital teaching status (urban teaching) are the strongest independent factors associated with increased in-hospital mortality, hospital stay length, and total charges using multivariate regression models through a large dataset of hospitals in the USA. All these factors should be considered for treatment, controlling hospital resources, and the decision-making process.