Inpatient Hypoglycemia Linked to Mortality in Diabetes Study

Key Takeaways

• At an Australian hospital, documented inpatient hypoglycemia occurred in adults with and without diabetes, predominantly among those with diabetes.
• Among adults with diabetes, basal–bolus insulin, premixed insulin, sulfonylureas, and malnutrition were each associated with recorded hypoglycemia after adjustment.
• Adults with diabetes and documented hypoglycemia had higher inpatient and 12-month mortality and longer hospital stays than diabetes controls without hypoglycemia.

Low capillary glucose during an adult hospital admission can arise amid diabetes treatment or in someone without diabetes. The relationship between treatment exposure, nutritional status, and subsequent outcomes is less clear when the event occurs in hospital. Investigators assessed those relationships across adult admissions at an Australian tertiary hospital.

A retrospective case–control study of inpatient hypoglycaemia was conducted at Nepean Hospital in New South Wales, Australia, from July 2018 through June 2019. Eligible adults were at least 18 years old.

Documented hypoglycemia meant a finger-prick capillary glucose reading below 4 mmol/L (72 mg/dL). Coders identified potential cases, which researchers confirmed in electronic glucose records, retaining the first episode per admission or change in care. Adults with diabetes and hypoglycemia were compared with randomly sampled diabetes admissions without hypoglycemia. Affected adults without diabetes were assessed separately, without an unaffected non-diabetes comparison group.

Among 64,699 adult admissions, 7602 were admissions with diabetes (11.7%); these were admission denominators, not participant counts. Researchers extracted medication exposure, dietitian-assessed nutritional status, mortality, and length of stay. Univariate analyses followed by standard multivariable logistic regression examined hypoglycemia associations within the diabetes comparison.

Investigators recorded 251 hypoglycemia episodes in 234 people: 202 episodes in 187 people with diabetes and 49 episodes in 47 people without diabetes. Episode counts and affected-person counts describe different units.

The authors estimated documented hypoglycemia in 2.7% of admissions with diabetes, based on coder-identified cases rather than independent patient-level incidence ascertainment. Within the diabetes case–control comparison, basal–bolus insulin was associated with hypoglycemia (adjusted odds ratio [aOR], 26.6; 95% confidence interval [CI], 7.9–106.6), as was premixed insulin (aOR, 5.9; 95% CI, 1.6–25.9). These estimates describe associations with recorded events, not treatment effects.

Sulfonylurea exposure was also associated with hypoglycemia among adults with diabetes (aOR, 3.2; 95% CI, 1.1–9.6). Malnutrition had a separate adjusted association (aOR, 4.2; 95% CI, 1.2–18.0). Both findings came from the diabetes case–control analysis.

In unadjusted comparisons of diabetes cases with diabetes controls, inpatient mortality was 10.2% versus 3%, and total 12-month mortality was 20.9% versus 5.5%. Median hospital stay was 12 versus 4 days. Total 12-month mortality included deaths during admission and was determined from hospital records.

Among affected adults without diabetes, total 12-month mortality was 36.2%. That figure describes the affected group alone, not a comparison with unaffected adults without diabetes.

Retrospective data from one Australian center cannot establish that hypoglycemia caused deaths or prolonged admissions; illness severity and episode timing complicate interpretation. The authors considered hypoglycemia a possible marker of intercurrent illness severity. Coder-based identification and infrequent overnight checks could miss events, while hospital-record follow-up could miss deaths elsewhere. Administered insulin dose and timing were not collected. Too few recurrent admissions supported a reliable mixed-effects model, so investigators used standard multivariable logistic regression. Without an unaffected non-diabetes group, outcome risk could not be compared within adults without diabetes. These Australian data do not establish inpatient hypoglycemia rates or outcome associations for U.S. hospitals. No intervention was tested.

Investigators reported that documented hypoglycemia was associated with higher mortality and longer stays in the diabetes comparison, while insulin regimens, sulfonylureas, and malnutrition were associated with its occurrence. Mortality among affected adults without diabetes remained a descriptive observation.

Clinician Questions

Were episodes of inpatient hypoglycemia always symptomatic in adults with diabetes at Nepean Hospital?

Only a minority of hospitalized adults with diabetes and documented inpatient hypoglycemia at Nepean Hospital had recorded hypoglycemic symptoms. Symptom documentation was often missing, so recorded symptoms do not establish the true frequency of symptomatic episodes.

Were adults with gestational diabetes included in the Nepean Hospital inpatient hypoglycemia comparison?

Adults with gestational diabetes were included in the diabetes group in Nepean Hospital’s inpatient hypoglycemia analysis to reflect its admissions. Investigators noted that pregnancy-specific glucose targets differ from those for other diabetes types; they did not present a gestational-diabetes-specific risk estimate.