Risk factors and mortality associated with severe spontaneous hypoglycemia in hospitalized patients: a retrospective cohort study

Abstract

Background

Severe spontaneous hypoglycemia (SSH) is a rare but high-risk condition in hospitalized patients, particularly among the critically ill. It is associated with increased mortality and may serve as a marker of disease severity.

Objective

To describe in-hospital mortality and clinical profile observed in patients with SSH.

Methods

A retrospective cohort study was conducted at Santa Izabel Hospital from November 1, 2023, to March 31, 2024. Among 5,803 hospitalized patients, 134 (2.3%) had severe hypoglycemia (≤ 40 mg/dL). After excluding insulin users and measurement errors, 47 patients (0.81%) were analyzed. Data included demographics, comorbidities, hypoglycemia recurrence, and hospital outcomes.

Results

SSH prevalence was 0.81% (n = 47) among all hospitalized patients with at least one blood glucose (BG) reading, with an overall mortality rate of 66% (n = 31). Recurrence within 24 h occurred in 46.8% (n = 22) of patients but was not significantly associated with increased mortality (82% with recurrence vs. 52% without recurrence; p = 0.063). Among total deaths (n = 31), 41.9% (n = 13) occurred within 24 h (43.8% with recurrence vs. 38.5% without recurrence) and 22.6% (n = 7) within 12 h. In the total SSH group, renal insufficiency (51%), infection (80.9%), malignancy (46.8%), and elevated lactate > 2.0 mmol/L (70.3%) were common. Patients who died had more comorbidities compared to survivors (median 4.00 vs. 2.00; [p < 0.001), elevated serum creatinine levels (1.44 vs. 0.87 mg/dL; p = 0.035), and increased serum lactate concentrations (3.60 vs. 1.15 mmol/L; p = 0.002). Fasting in the 48 h preceding the first hypoglycemic episode was significantly more common in the patients with at least one recurrence of SSH in 24 h (70.0% vs. 20.0%; p = 0.001). In the multivariable model, none of the analyzed variables remained significantly associated with mortality.

Conclusion

SSH is rare but patients who experience it have elevated in-hospital mortality. It is often present in the context of rapid clinical deterioration, with many deaths in this cohort occurring within 24 h of the first episode. SSH may be a key marker of disease severity and can be further explored as a prognostic marker to guide management.

Introduction

Spontaneous hypoglycemia in hospitalized patients, defined as a blood glucose level below 70 mg/dL without use of insulin or other glucose-lowering agents, is a serious but under-recognized phenomenon in inpatient care. Unlike iatrogenic hypoglycemia, which is typically related to overtreatment of hyperglycemia, SSH often reflects an underlying derangement in physiological homeostasis and is associated with markedly high mortality [1–5]. Patients presenting with SSH are often in the ICU, of advanced age, with renal or hepatic dysfunction, adrenal insufficiency [1, 6], multi-organ failure [7, 8], septic shock [9] and malnutrition [10–12].

Given the markedly high mortality observed among patients who develop SSH, its recognition is crucial in hospital settings. Although previous studies have explored associations between spontaneous hypoglycemia and mortality, the present study aimed to provide a detailed analysis of SSH episode recurrence and a time-to-death stratification when applicable. We hypothesize that SSH can serve as a mortality signal and prognostic indicator, marking underlying illness severity. We conducted a retrospective observational study using a clinical inpatient database to characterize in-hospital mortality and the clinical profile of patients who developed SSH, with an emphasis on critically ill patients at an academic center.

Methods

Study design

We conducted a retrospective cohort study at Santa Izabel Hospital, a tertiary teaching hospital in Salvador, Bahia, Brazil. Data were collected from patients admitted between November 1, 2023, and March 31, 2024.

Reporting standards

This study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for cohort studies, ensuring rigor and transparency in design, data collection, and analysis [13].

Participants

The study included adult patients (≥ 18 years old) admitted to intensive care, intermediate-care (step-down units), and general wards, who developed at least one episode of SSH, defined as a capillary blood glucose level ≤ 40 mg/dL in the absence of insulin or hypoglycemic agents. All patients receiving insulin or any oral hypoglycemic drug — including sulfonylureas, meglitinides, DPP-4 inhibitors, SGLT2 inhibitors, GLP-1 receptor agonists, and metformin — were excluded. Erroneous glucose measurements were also assessed and excluded from the final analysis.

Variables and outcomes

The primary outcome of the study was all cause in-hospital mortality within the SSH cohort. Variables analyzed included demographic data (age, sex), body mass index (BMI), renal insufficiency, infection (documented in medical records or inferred from antibiotic use), pre-existing liver disease, malnutrition (based on BMI or medical documentation), vasopressor use, fasting status (as documented in medical orders), and potential adrenal insufficiency (defined by concurrent hyponatremia and hyperkalemia or current corticosteroid use for > 15 days). Additional clinical variables included evidence of poor tissue perfusion (serum lactate > 2 mmol/L), hypotension in the 24 h preceding the first SSH episode, ischemic hepatitis (AST/ALT > 10x upper limit of normal), diabetes mellitus, cancer, heart failure, and recent surgery (within 48 h). We assessed SSH recurrence within 24 h and time to death following the initial SSH episode.

Data collection and sources

At Santa Izabel Hospital, blood glucose levels are measured with calibrated glucose meters, and measurements are automatically recorded in patients’ electronic charts. Data were extracted retrospectively from electronic medical records.

Sample size

We included all eligible patients during the study period, resulting in a sample size of 47 patients diagnosed with SSH. A recent literature review found similar sample sizes in retrospective cohorts exploring SSH [14].

Statistical analysis

Descriptive statistics were used to summarize patient characteristics. Continuous variables were assessed for normality using the Shapiro–Wilk test and histogram inspection. Normally distributed variables were described as means ± standard deviation (SD), and non-normally distributed variables as medians with interquartile ranges (IQR). Categorical variables were presented as absolute and relative frequencies.

Bivariate comparisons between survivors and non-survivors were conducted using the Mann–Whitney U test for continuous variables and Fisher’s exact test for categorical variables. A p-value < 0.05 was considered statistically significant.

To investigate the association between clinical variables and in-hospital mortality, univariable logistic regression models were applied. Variables with clinical relevance or with a p value < 0.10 were considered for multivariable logistic regression, limited to three covariates to avoid overfitting given the number of events (n = 31 deaths). Multicollinearity was assessed using the variance inflation factor (VIF), with a threshold of < 2 indicating acceptable collinearity. Although “hypoperfusion/hypotension” demonstrated a strong univariable association with mortality, it was excluded from the multivariable model due to its clinical collinearity with lactate and limited event distribution, which could compromise the model’s robustness. Because of the limited sample, this model is presented strictly for hypothesis generation. Statistical significance was set at two-sided p < 0.05, but effect sizes and 95% confidence intervals are emphasized over p-values owing to the exploratory nature of these analyses.

While this allowed for exploratory modeling, it imposed limitations on the number of predictors that could be safely included in the multivariable model. A post hoc power calculation indicated that to detect a difference in mortality between 50% and 79% with 80% power and a two-sided α of 0.05, a minimum of 60 patients per group would be required.

All statistical analyses were performed using Stata Statistical Software: Release 17 (StataCorp LLC, College Station, TX, USA).

Results

Using the hospital’s glycemic management system, we identified patients with blood glucose (BG) levels of 40 mg/dL or lower between November 1, 2023, and March 31, 2024. Among 5,803 patients, 134 (2.3%) met this criterion, accounting to a total of 98,059 BG readings. Following medical record review, we excluded patients on insulin therapy and instances of measurement error. Measurement errors were defined as BG readings below 40 mg/dL followed by values above 70 mg/dL within 10 min, without documented hypoglycemia symptoms or any intravenous or oral glucose intervention. Most exclusions were related to insulin-mediated hypoglycemia. This resulted in a final cohort of 47 patients (0.81% of the initial population) for subsequent analysis (Fig. 1).

Fig. 1.

A flowchart showing the selection of study participants

The median age was 73 years, with a slight male predominance (55.3%). Hypoglycemia recurrence within 24 h of the index event was common, affecting 22 patients (46.8%). The number of hypoglycemic episodes among those with recurrence varied, with each patient presenting a distinct profile — some with multiple events, including complex recurrence patterns – but most presented only one subsequent episode. All patients had at least one comorbidity, with over half presenting with four or more. (Table 1)

Table 1.

General characteristics of participants (n = 47)

Character	Total
Median age (variacao)	73 (22–91) years
Mean+/-SD	67.3 +/−18
Male sex	55.3%
Female sex	44.7%
Number of comorbidities
1 comorbidity	5 (10.6%)
2 comorbidities	10 (21.3%)
3 comorbidities	7 (14.9%)
4 comorbidities	11 (23.4%)
5 comorbidities	6 (12.8%)
6 comorbidities	5 (10.6%)
7 comorbidities	3 (6.4%)
≥ 4 comorbidities	25 (53.2%)
Hypoglycemia Recurrence
With recurrence	22 (46.8%)
Frequency of recurrent episodes:
1 episode	10 (21.3%)
2 episodes	5 (10.6%)
3 episodes	3 (6.4%)
5 episodes	4 (8.5%)
7 episodes	1 (2.1%)
9 episodes	1 (2.1%)

We assessed the treatment of choice for hypoglycemia among all 47 patients in the cohort, but 8 (17.0%) had missing documentation regarding the method of correction. Among the 39 patients with available data, the majority—32 (82.1%)—received 25% IV glucose (G25%) as the initial therapy, which is the suggested prescription by protocol in our hospital’s electronic system. Among the 43 patients with available data on glucose recovery, the median time to achieve blood glucose levels above 70 mg/dL following the initial treatment was 52.0 min (IQR: 59.5). The shortest recorded recovery time was 7 min, while the longest reached 722 min (12 h and 2 min). The mean recovery time was 101.7 min. Four patients (8.5%) had missing data for this variable and were excluded from the analysis.

Among the 47 patients included in the cohort, 36 (76.6%) were receiving antibiotic therapy at the time of the hypoglycemic episode, and 33 of 47 patients (70.2%) had a documented infection. Considering the presence of either condition, 38 of 47 patients (80.9%) had either a documented infection or were receiving antibiotics. Renal insufficiency, either acute or chronic, was observed in 23 of 45 patients (51.1%). Regarding hemodynamic status, 23 of 36 patients (63.9%) with available data experienced hypotension, hypoperfusion, or required vasopressor support during the 24 h preceding the initial episode of SSH. Specifically, 19 (40.4%) required vasopressors within this period.

Serum lactate levels were available for 37 patients; among them, 26 (70.3%) exhibited elevated levels (> 2.0 mmol/L). Fasting within the 48 h prior to the first SSH episode was reported in 19 patients (40.4%). Notably, malignancy was the most frequently reported comorbidity (48.9%). Indicators suggestive of adrenal insufficiency or ischemic hepatitis were observed in a minority of cases, but no confirmed diagnoses were recorded. Table 2 summarizes the most common findings.

Table 2.

Comorbidities in patients with spontaneous severe hypoglycemia

Comorbidity	n/Total
Antibiotic therapy or documented infection	38/47
Antibiotic therapy	36/47
Elevated lactate (> 2.0 mmol/L) (no data n = 10)	26/37
Documented infection	33/47
Renal failure (no data n = 2)	23/45
Hypotension within 24 h before SSH	23/47
Confirmed cancer diagnosis	23/47
Fasting within 48 h before first SSH episode	19/47
Need for vasopressors	19/47
Suspected adrenal insufficiency	11/47
Diabetic	10/47
Heart failure	7/47
Signs of ischemic hepatitis (no data n = 21)	3/26
Corticosteroid use for > 15 days	4/47
Previous liver disease	2/47

We analyzed the distribution of clinical conditions across the entire cohort of patients with SSH and compared their frequency between those with and without recurrent hypoglycemia within the first 24 h of the index episode. Figure 2 illustrates the prevalence of clinical conditions in the total cohort, and stratified by recurrence status within 24 h.

Fig. 2.

Prevalence of clinical conditions in the total cohort and in patients with and without hypoglycemia recurrence within 24 h

Infection or recent antibiotic use was the most frequently associated condition (38/47), particularly among patients with recurrent hypoglycemia (20/22). Fasting in the 48 h preceding the first hypoglycemia episode was significantly associated with recurrence, being present in 70.0% of the recurrent group (14/20) and 20.0% of the non-recurrent group (5/25) (p = 0.001). Other conditions such as hemodynamic compromise, renal insufficiency, and malignancy were common but did not reach statistical significance. Less frequent findings, including heart failure, prolonged corticosteroid use, and ischemic hepatitis, showed no notable differences between groups. Full data are detailed in Table 2.

Bar graph comparing the prevalence of selected clinical conditions in the total cohort (black), among patients with recurrent hypoglycemia within 24 h (gray), and non-recurrent cases (white). Percentages are shown above each bar. Variables were selected based on clinical relevance and completeness of data.

Note

Fasting was significantly more frequent among patients with hypoglycemia recurrence (p = 0.001). Statistical comparisons were performed using Fisher’s exact test or the Chi-square test,depending on cell counts.

Association between clinical variables and in-hospital mortality

The overall hospital mortality rate was 66.0% (31 of 47 patients). Mortality was significantly associated with a higher number of comorbidities (median 4.00 [IQR: 3.00–5.00] vs. 2.00 [1.00–3.00]; p < 0.001), elevated serum creatinine levels (1.44 [0.94–1.72] vs. 0.87 [0.68–1.27] mg/dL; p = 0.035), and increased serum lactate concentrations (3.60 [2.30–7.90] vs. 1.15 [1.02–2.03] mmol/L; p = 0.002). Patients who died also experienced a higher number of recurrent severe hypoglycemia (SSH) episodes within 24 h (1.00 [0.00–2.00] vs. 0.00 [0.00–0.25]; p = 0.036). A comparison of clinical variables between patients who died and those who were discharged is presented in Table 3.

Table 3.

Comparison of creatinine levels, serum lactate, number of comorbidities, and number of hypoglycemia recurrences within 24 h (median and IQR) between patients who died vs. those who were discharged from the hospital

	Died	Discharged	p-value
Creatinine (mg/Ll)	1.44 (0.94–1.72)	0.87 (0.68–1.27)	0.035
Lactate (mmol/L)	3.6 (2.30–7.90)	1.15 (1.02–2.03)	0.002
Recurrences in 24 h	1.0 (0–2.0)	0.00 (0-0.25)	0.036
Number of comorbidities	4.00 (3.0–5.0)	2.00 (1.0–3.0)	< 0.001

Among categorical variables, hospital mortality was significantly associated with the presence of hypoperfusion, hypotension, or vasopressor use (p = 0.012), use of antibiotic therapy (p = 0.026), and the presence of documented infection at the time of the hypoglycemic episode (p = 0.019), considering valid denominators after exclusion of missing data. No statistically significant associations with mortality were observed for the other variables analyzed.

Univariable logistic regression showed a higher number of comorbidities was associated with a 2.5-fold increase in the odds of in-hospital mortality (OR = 2.50; 95% CI: 1.41–4.41; p = 0.002). Elevated serum lactate levels were also significantly associated with higher mortality risk (OR = 2.19; 95% CI: 1.06–4.54; p = 0.035). Although higher serum creatinine levels showed a trend toward increased mortality (OR = 3.44), this did not reach statistical significance (95% CI: 0.94–12.52; p = 0.061). Full results are presented in Table 4.

Table 4.

Univariable logistic Regression – Hospital mortality

Variable	OR	95% CI	p-value
Number of comorbidities	2.5	1.41–4.41	0.002
Creatinine	3.44	0.94–12.52	0.061
Lactate	2.19	1.06–4.54	0.035
Hypoperfusion/Hypotension	35.2	3.55–349.15	0.002
Infection/Antibiotic use	3.07	0.69–13.61	0.14
Recurrent hypoglycemia	2.94	0.73–12.62	0.128
Neoplasia	1.14	0.35–3.81	0.828
Suspected adrenal insufficiency	0.97	0.25–3.80	0.961
Fasting	0.95	0.26–3.37	0.94
Heart failure	0.46	0.08–2.47	0.363
Ischemic hepatitis	0.65	0.13–3.29	0.601
Corticosteroid ≥ 15d	0.99	0.15–6.51	0.99

Note: P-values in this table originate from univariable logistic regression models assessing the association between each independent variable and the odds of in-hospital death. These values may differ from those reported in bivariate analyses (e.g., Mann–Whitney U test, Fisher’s exact test), which evaluate unadjusted group-level differences without estimating effect sizes or accounting for probabilistic modeling.

The presence of hypoperfusion or hypotension was strongly associated with mortality (OR = 35.20; 95% CI: 3.55–349.15; p = 0.002), while infection or antibiotic use (OR = 3.07; 95% CI: 0.69–13.61; p = 0.140) and recurrent hypoglycemia (OR = 2.94; 95% CI: 0.73–12.62; p = 0.128) did not reach statistical significance. (Table 4)

Multivariable logistic regression – hospital mortality

To investigate independent predictors of in-hospital mortality, a multivariable logistic regression model was constructed. Variable selection was guided by clinical relevance and statistical significance in univariable analysis (p < 0.10). The final model included three covariates: number of comorbidities, serum creatinine, and serum lactate concentrations. The inclusion was limited to three variables to prevent model overfitting, following the rule of one variable per ten outcome events, considering that 31 patients in the cohort died. Although “hypoperfusion/hypotension” demonstrated a strong univariable association with mortality, it was excluded from the multivariable model due to its clinical collinearity with lactate and limited event distribution, which could compromise the model’s robustness. None of the variables remained statistically significant in the adjusted model; full results are presented in Table 5.

Table 5.

Multivariable logistic regression – hospital mortality

Variable	Adjusted OR	95% CI	p-value
Comorbidities	1.21	0.53–2.75	0.655
Creatinine	2.90	0.53–15.95	0.221
Lactate	1.82	0.76–4.39	0.182

Note: Variables included in the multivariable model were selected based on a p-value < 0.10 in univariable logistic regression and clinical plausibility. Inclusion was restricted to three covariates to comply with the rule of one variable per ten outcome events (n=31 deaths).

Time-to-death analysis for severe spontaneous hypoglycemia

Of the 31 patients who died, 13 (41.9%) died within 24 h of the initial hypoglycemic episode, and 7 (22.6%) died within 12 h. The remaining 18 patients (58.1%) died after more than 24 h. Overall, time-to-death was documented in hours and minutes for all 13 early deaths. The mean time to death was 10.1 h (SD: 6.9), with a median of 9.0 h (IQR: 6.5–12.6), ranging from 43 min to 23 h and 43 min. Although early mortality was more frequent among patients with recurrent hypoglycemia, the observed differences were not statistically significant in this cohort (≤ 24 h: p = 0.328; ≤12 h: p = 0.690; Fisher’s exact test). Full details are presented in Fig. 3.

Fig. 3.

– Distribution of hospital and early mortality by hypoglycemia recurrence group

Note

Hospital mortality was 31/47 (66.0%). Among the 31 patients who died, 13 (41.9%) died ≤ 24 h, 7 (22.6%) died ≤ 12 h, and 18 (58.1%) died > 24 h. Among the 22 patients with recurrent hypoglycemia, 18 died (81.8%). Of these, 16 had complete time-to-death data: 7/16 (43.8%) died ≤ 24 h, 4/16 (25.0%) died ≤ 12 h, and 5/16 (31.25%) died > 24 h. Among the 25 non-recurrent patients, 13 died (52.0%). All 13 had complete time-to-death data: 5/13 (38.5%) died ≤ 24 h, 3/13 (23.1%) died ≤ 12 h, and 5/13 (38.5%) died > 24 h. Fisher’s exact test was used to compare proportions across groups. All percentages reflect valid denominators with missing data excluded.

Discussion

This single-center retrospective observational study examined 47 hospitalized patients experiencing severe spontaneous hypoglycemia (SSH), defined as blood glucose ≤ 40 mg/dL without exogenous glucose-lowering agents. Our findings align with previous research suggesting that SSH primarily functions as prognostic marker rather than a pathophysiological mechanism per se leading to mortality.

SSH as a marker of disease severity rather than a direct cause of death and SSH pathophysiology.

The association between SSH and illness severity in our cohort was demonstrated by several key findings. Patients who died following SSH episodes had significantly higher serum creatinine levels (1.44 [0.94–1.72] vs. 0.87 [0.68–1.27] mg/dL; p = 0.035), and serum lactate concentrations (3.60 [2.30–7.90] vs. 1.15 [1.02–2.03] mmol/L; p = 0.002). Compared to survivor’s renal dysfunction and tissue hypoperfusion were more prevalent, although it did not correlate significantly with mortality in our multivariate analysis. Furthermore, a higher burden of comorbidities was associated with in-hospital mortality (median 4.00 [IQR: 3.00–5.00] vs. 2.00 [1.00–3.00]; p < 0.001)). This relationship was confirmed in univariable logistic regression (OR = 2.50; 95% CI: 1.41–4.41; p = 0.002), though also not remaining statistically significant in the multivariate model. As previously stated, the pathophysiology of SSH is multifactorial. Most of our cohort was comprised of ICU patients. This patient profile is consistent with high prevalence of comorbidities that can contribute to SSH. In sepsis, metabolic dysregulation is driven by an inflammatory cascade marked by elevated cytokines, which disrupt normal glucose homeostasis. This inflammatory state impairs gluconeogenesis from key substrates, such as alanine, lactate, and pyruvate, though it may remain partially preserved from glycerol. Sepsis also leads to increased glycolysis and insulin resistance, often coupled with glycogen depletion, increasing the susceptibility to hypoglycemia. Hypotension and decreased tissue perfusion increase peripheral utilization of glucose because the shift to anaerobic metabolism requires 18 times more glucose to produce the same amount of energy (ATP) as aerobic metabolism [15–19].

Kidney dysfunction may also play an important role in the onset of hypoglycemia in hospitalized patients. The kidneys are responsible for glucose production through gluconeogenesis, accounting for approximately 40% of this process. In renal failure, this capacity is reduced, which may predispose to hypoglycemia. The kidneys also contribute to the degradation of insulin. When renal function is impaired, insulin may persist in the circulation for longer, leading to excessive glucose utilization and, consequently, hypoglycemia [6]. Patients with renal failure often experience anorexia, leading to decreased food intake and reduced glucose availability. Furthermore, hemodialysis can lead to reductions in plasma glucose as blood is filtered. Even with low-permeability membranes, glucose can be removed due to its low molecular weight. Autonomic neuropathy, common in diabetic patients with kidney disease, can impair the normal counter-regulatory response to hypoglycemia [19]– [20].

Advanced age is also a known risk factor, as it reduces physiological reserves and increases prevalence of key comorbidities that collectively impair glucose production and clearance such as hepatic and kidney insufficiency [18] Liver dysfunction can also contribute to SSH development as glycogenolysis and gluconeogenesis are compromised. Impaired liver function leads to diminished glucose output and reduced response to counter-regulatory hormones [16]. Hypoalbuminemia, frequently observed in malnourished and cirrhotic patients, has been associated with elevated risk of severe hypoglycemia [11, 19].

Mortality, recurrence, and clinical deterioration

Our findings align with previous studies that have linked SSH to high in-hospital mortality, particularly among critically ill patients with multiple comorbidities [8, 9]. The observed mortality rate of 66.0% in our cohort exceeds those reported in some retrospective studies, such as Saliba et al. (26.7%) and Akirov et al. (up to 41.8%). This disparity could be a reflex of different patient profiles or be attributed to our smaller sample size. Importantly, while we identified a non-significant trend toward higher mortality among patients with recurrent episodes (81.8% vs. 52.0%), Boucai et al. reported a dose-response relationship between the number of spontaneous hypoglycemia episodes and death, which weakened after adjusting for comorbidities—supporting the hypothesis that SSH may be more a marker of critical illness than a direct cause of death. This interpretation is reinforced by studies such as Kosiborod et al., who found elevated mortality in spontaneous but not insulin-induced hypoglycemia, suggesting that the etiology of hypoglycemia affects prognostic implications [2, 5, 21–23].

Mortality was non-significantly higher among patients with recurrent SSH (≥ 1 additional episode within 24 h of the index event) compared to those with a single episode (81.8% [18/22] vs. 52.0% [13/25]; p = 0.063). A higher number of SSH episodes within the first 24 h was also significantly associated with death (1.00 [0.00–2.00] vs. 0.00 [0.00–0.25]; p = 0.036).

As hypothesized, SSH often preceded rapid clinical decline. A substantial proportion of deaths (41.9%; 13/31) occurred within 24 h of the initial SSH episode. The elevated overall mortality in the recurrent group suggests that repeated episodes of SSH may signal high risk of impending clinical deterioration, warranting clinical attention and assessment of underlying causes. These findings raise the question of whether SSH could be integrated into early warning systems or serve as a trigger for rapid response team activation. Given its association with poor outcomes, even a single episode of SSH—particularly in non-diabetic, older patients—should alert for heightened surveillance and consideration of care escalation. In some contexts, SSH recurrence might support the initiation of discussions regarding goals of care, palliative transition, or limitation of invasive interventions. For instance, a critically ill elderly patient with multiple comorbidities who experiences recurrent episodes of SSH despite optimal supportive measures may be better served by a shift in focus toward comfort-oriented care. It remains to be explored if timely interventions can improve patient outcomes.

Implications for clinical practice

A particular strength of our study was the time-to-death analysis, which was not previously explored in the literature. This demonstrated the rapid mortality following SSH onset. A substantial portion of the cohort – 27.7% (13/47) – died within 24 h of the index SSH event, with 14.89% (7/47) dying in the first 12 h. At the bedside, awareness of SSH as a prognostic marker could influence decision-making by clinicians. While this observational study cannot establish causality between SSH itself and death, recognizing the occurrence of SSH, particularly when in the appropriate scenario, may serve as a critical short-term marker of acute physiological decompensation. Incorporating SSH into clinical risk stratification tools could improve response times in hospital settings, offering intensivists, palliative-care doctors and physicians alike a valuable prognostic signal. Future work should explore the utility of including SSH in institutional rapid response algorithms and examine if timely interventions can improve patient outcomes.

Limitations

This study has several limitations. First, its retrospective, cross-sectional design inherently limits our ability to establish causality relationships between observed factors and SSH or mortality; only associations can be inferred. Second, the modest sample size of 47 patients, reflecting the rarity of SSH, limited the study`s statistical power and increases the potential for type II errors and widens confidence intervals—particularly in the exploratory multivariable model. Although that model was constrained to three covariates and vetted for collinearity, the resulting estimates remain imprecise; therefore, all multivariable findings should be interpreted as hypothesis-generating rather than definitive. Larger, prospective cohorts with an appropriate non-SSH comparison group are required to corroborate these preliminary observation. Additionally, the single-center design and the limited observational period may restrict generalizability to broader populations. Finally, detailed biochemical analyses, such as hormone profiles or inflammatory markers, were not available, limiting deeper mechanistic insights into SSH pathophysiology.

Conclusion

The present study reinforces that patients who present with severe spontaneous hypoglycemia (SSH) have notably elevated in-hospital mortality. Our findings suggest SSH primarily serves as a marker of critical illness severity, linked to underlying conditions such as infection, renal dysfunction, and hemodynamic instability, showing potential as an indicator of impeding clinical deterioration.

Author contributions

A.F: Contributed to the conceptualization of the study, including project ideation, hypothesis development, and initial study framework. Actively participated in the study design and analytical strategy development and was responsible for the statistical analysis of the data. Additionally, she was involved in the manuscript drafting, performed the critical revision and clarity improvements, and coordinated project timelines and management.D.P: Responsible for data collection and contributed to the manuscript drafting. Performed critical text revision and clarity enhancements and developed graphs, tables, and figures to illustrate the results.

Funding

This work was the result of the authors’ initiative. There was no support of research or publication.

funds.

Data availability

Data is provided within the manuscript.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval

This study was approved by the Ethics Committee of Santa Izabel Hospital. Given its retrospective nature and the exclusive use of anonymized electronic medical records, the requirement for informed consent was formally waived. All procedures were conducted in accordance with the Declaration of Helsinki and institutional ethical standards for research involving human data.

Human ethics and consent to participate declarations

not applicable.

Clinical trial number

This study is a retrospective observational cohort and does not constitute a clinical trial. Therefore, no clinical trial registration is required.

Conflict of interest

The authors have no conflicts of interest to declare that are relevant to the content of this article.