Impact of Hospital Dysglycaemia on Patient Outcomes

Sugar & Survival

Does Dysglycaemia During Hospitalisation Drive Adverse Outcomes?

A Research Catch-Up | April 2026

Setting the Scene

The hospital environment is a metabolic battlefield. Blood sugar instability — both high and low — is common among inpatients, yet its true impact on patient outcomes is often overlooked.

🔴

Hyperglycaemia

Blood glucose persistently elevated above normal range

🔵

Hypoglycaemia

Blood glucose dropping below safe thresholds

⚡

Dysglycaemia

The umbrella term: any abnormal glycaemic excursion during admission

"Not just a diabetic problem — dysglycaemia affects 1 in 3 hospital patients"

The Evidence Gap

Why Does It Matter?

Despite widespread glucose monitoring in hospitals, the systematic relationship between dysglycaemia and specific adverse outcomes — mortality, ICU escalation, infection, and prolonged stay — remains incompletely characterised.

This study asks:

how much does blood sugar instability actually cost patients?

Inconsistent thresholds across studies

Mixed patient populations

Lack of continuous outcome data

Limited infection outcome focus

We set out to close this gap.

The Research Question

Primary Research Objectives

Does dysglycaemia during hospitalisation independently predict adverse outcomes?

💀

Mortality

(In-hospital death)

🏥

ICU Admission

🦠

Hospital-Acquired Infection

📅

Length of Stay

How We Studied It

Methods at a Glance

Study Overview

Patient Cohort

Hospitalised adult inpatients with ≥2 blood glucose measurements

Glycaemic Exposure

Dysglycaemia defined using time-above/below threshold + mean glucose

4 Outcomes

Death | ICU | Infection | Length of Stay

Statistical Models

Logistic regression (binary outcomes), Linear regression (LOS)

Adjustments

Age, sex, diabetes status, admission diagnosis, comorbidities

Who Was in the Study?

Study Population

N = 2,847

Total patients included

64.3 yrs

Median age

51% Male

Sex distribution

38%

Had known diabetes

Adult inpatients ≥18 yrs

Admitted for ≥24 hours

≥2 BGL measurements

Acute medical/surgical wards

ICU at admission

Palliative care on arrival

Incomplete glucose data

Retrospective cohort study | Single tertiary centre | 2020–2023

Defining the Exposure

What counts as Dysglycaemia?

Hyperglycaemia

Any BGL >10 mmol/L

Time above threshold (TAR) >20% of readings

Hypoglycaemia

Any BGL <4.0 mmol/L

Time below range (TBR) >5% of readings

Glycaemic Variability

Coefficient of variation >36%

SD >3 mmol/L

Composite dysglycaemia = any one or more of the above criteria met

The Results: At a Glance

Dysglycaemia was independently associated with all 4 adverse outcomes

In-Hospital Mortality

💀

OR 2.4

(95% CI 1.7–3.4)

p<0.001

2.4× higher odds of death

ICU Admission

🏥

OR 1.9

(95% CI 1.4–2.6)

p<0.001

1.9× higher odds of escalation

Hospital-Acquired Infection

🦠

OR 1.7

(95% CI 1.2–2.4)

p=0.003

1.7× higher odds of infection

Length of Stay

📅

β +3.2 days

(95% CI +2.1–+4.3)

p<0.001

+3.2 extra days in hospital

Outcome 1: In-Hospital Mortality

Patients who experienced dysglycaemia had 2.4 times higher odds of dying in hospital compared to normoglycaemic patients, after adjusting for age, sex, diabetes status, and severity of illness.

OR 2.4

(95% CI 1.7 – 3.4)

p < 0.001

Hyperglycaemia alone

OR 2.1

Hypoglycaemia alone

OR 3.1

Both (combined)

OR 4.2

No dysglycaemia

3.1%

Dysglycaemia present

10.8%

Outcome 2: ICU Admission

Dysglycaemia significantly predicted unplanned ICU escalation. Patients with blood glucose instability were nearly twice as likely to require ICU-level care, suggesting glycaemic control may be a modifiable target for preventing deterioration.

OR 1.9

(95% CI 1.4 – 2.6)

p < 0.001

Persisted after adjustment for admission severity

Effect strongest in non-diabetic patients (OR 2.3)

Timing matters: early dysglycaemia (Day 1–2) highest risk

No dysglycaemia

6.2%

Dysglycaemia present

14.1%

Outcome 3: Hospital-Acquired Infection

The immune-suppressive effects of hyperglycaemia are well-known in vitro, but this study confirms the clinical signal: dysglycaemic patients had 1.7× higher odds of developing a hospital-acquired infection — including urinary tract infections, pneumonia, and surgical site infections.

OR 1.7

(95% CI 1.2 – 2.4)

p = 0.003

UTI

OR 1.6

Pneumonia/LRTI

OR 2.0

Surgical site

OR 1.8

Bloodstream

OR 1.5

No dysglycaemia

8.4%

38%

Dysglycaemia present

17.6%

80%

After adjusting for immunosuppression, antibiotic use, and invasive device exposure.

Outcome 4: Length of Stay

Using linear regression, dysglycaemia was independently associated with a 3.2-day increase in hospital stay. Even after adjusting for comorbidities, surgical procedures, and infection events, the glycaemic effect on LOS remained statistically robust.

β = +3.2 days

(95% CI +2.1 – +4.3)

p < 0.001

Median LOS no dysglycaemia:

4.8 days

Median LOS with dysglycaemia:

8.1 days

Hyperglycaemia only:

+2.6 days

Hypoglycaemia only:

+4.1 days

Both:

+6.3 days

Normoglycaemic

4.8 days

Dysglycaemic

8.1 days

Putting It All Together

Dysglycaemia: A Consistent Independent Predictor

Across all four outcomes, the direction and magnitude of effect were consistent —

dysglycaemia is not a bystander.

What Does This Mean?

Implications, Limitations & What's Next

Clinical Implications

<ul style="margin: 0; padding-left: 20px; display: flex; flex-direction: column; gap: 20px;"> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%2300e5ff%22/></svg>') no-repeat left 10px; list-style-type: none;">Routine glycaemic monitoring should be standard across all inpatients, not just diabetics</li> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%2300e5ff%22/></svg>') no-repeat left 10px; list-style-type: none;">Glucose targets need to be actively managed — dysglycaemia is modifiable</li> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%2300e5ff%22/></svg>') no-repeat left 10px; list-style-type: none;">Early identification of glycaemic instability may trigger timely intervention</li> </ul>

Study Limitations

<ul style="margin: 0; padding-left: 20px; display: flex; flex-direction: column; gap: 20px;"> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%23ffb300%22/></svg>') no-repeat left 10px; list-style-type: none;">Single-centre retrospective design</li> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%23ffb300%22/></svg>') no-repeat left 10px; list-style-type: none;">Causality cannot be inferred</li> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%23ffb300%22/></svg>') no-repeat left 10px; list-style-type: none;">Glucose measurement frequency varied by clinical need</li> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%23ffb300%22/></svg>') no-repeat left 10px; list-style-type: none;">Incomplete insulin/steroid prescription data</li> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%23ffb300%22/></svg>') no-repeat left 10px; list-style-type: none;">Selection bias in who received monitoring</li> </ul>

Future Directions

<ul style="margin: 0; padding-left: 20px; display: flex; flex-direction: column; gap: 20px;"> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%23b088f9%22/></svg>') no-repeat left 10px; list-style-type: none;">Multicentre prospective replication</li> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%23b088f9%22/></svg>') no-repeat left 10px; list-style-type: none;">Continuous glucose monitoring (CGM) in ward setting</li> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%23b088f9%22/></svg>') no-repeat left 10px; list-style-type: none;">Randomised trial of glycaemic intervention on LOS</li> <li style="padding-left: 10px; margin-bottom: 5px; background: url('data:image/svg+xml;utf8,<svg xmlns=%22http://www.w3.org/2000/svg%22 width=%2210%22 height=%2210%22 viewBox=%220 0 10 10%22><circle cx=%225%22 cy=%225%22 r=%224%22 fill=%22%23b088f9%22/></svg>') no-repeat left 10px; list-style-type: none;">Machine learning prediction model</li> </ul>

The Take-Home

Dysglycaemia in hospital is

common, measurable, and dangerous.

It independently predicts

death, ICU escalation, infection, and longer stays.

It is also

modifiable

— and that makes it a

target worth pursuing.

Research summary prepared April 2026

Thank you | Questions welcome