Diabetes Was an Independent Risk Factor for Severe COVID-19: Inpatient Data from 38 Hospitals in Chiba, Japan

Written by Hirotsugu Suwanai, MD, PhD

Journal Club of the Suwanai · Diabetes & Infection

Are people with diabetes more likely to become seriously ill with COVID-19? This study sought the answer in real inpatient care data.

In March 2025, PLoS One published our multicenter retrospective study using inpatient claims data (DPC data) from 38 hospitals in Chiba Prefecture, Japan. I was the first and corresponding author. Among about 10,000 inpatients, we compared in-hospital death, mechanical ventilation and dialysis between people with and without diabetes. Read the original paper ↗

Key point: how to read these results ↓

What data did we analyze?

We studied patients hospitalized with COVID-19 between February 2020 and November 2021. The Diagnosis Procedure Combination (DPC) is Japan’s case-mix payment system for acute-care hospitals and includes diagnoses, comorbidities, drugs and procedures such as mechanical ventilation, ICU care and dialysis.

Of 11,601 patients, we excluded those under 18 (402), pregnant women (114) and patients with a diabetes diagnosis but no diabetes medication (309), leaving 10,776 patients: 7,679 without diabetes (controls) and 3,097 with diabetes. Nearly all had type 2 diabetes; 15 had type 1.

In the diabetes group, 88.4% received insulin and 44.2% oral agents, most often DPP-4 inhibitors. Patients with diabetes were older (67.1 vs 55.7 years), more often male (68.2% vs 54.9%), had higher BMI (25.6 vs 23.6) and arrived by ambulance more often (40.7% vs 22.2%). Original paper, Table 1

In-hospital mortality was about four times higher

In-hospital mortality. Source: Suwanai et al., PLoS One (2025), Table 2.

In-hospital mortality was 12.9% with diabetes versus 3.5% without. Mean hospital stay was 17.8 vs 10.2 days. ICU admission occurred in 17.0% (vs 2.8%) and steroids were used in 77.8% (vs 28.3%), showing that more patients with diabetes required intensive treatment.

The risk persisted after adjustment

Because the groups differed in age, sex and other factors, we performed logistic regression adjusted for age, sex, BMI and ambulance use. Diabetes remained a significant risk factor for every outcome.

Adjusted for age, sex, BMI and ambulance use. Source: Suwanai et al., PLoS One (2025), Fig. 4.

The especially high odds ratios for mechanical ventilation and CHDF suggest that COVID-19 affects multiple organs, including the kidneys, and that respiratory and renal failure may coincide more often in people with diabetes. Original paper, results and discussion

The effect of diabetes stood out in people in their 50s

Diabetes prevalence rose with age, peaking in the 70s, but the odds ratio of death associated with diabetes was high from age 40 and peaked at 12.8 in people in their 50s (95% CI 3.71–44.1). Because older patients die more often even without diabetes, the additional effect of diabetes may be most visible in middle age. Stratifying by age reduces numbers and widens confidence intervals, however.

Changes over the course of the pandemic

Monthly mortality in the diabetes group was high in March–June 2020 and again in October–December 2020 as admissions rose, then declined in 2021 as vaccination spread. The data predate the Omicron variant, which was associated with lower mortality.

Journal Club view: how to read these results

First, the definition of the diabetes group: patients who received diabetes medication during admission. This is more reliable than diagnosis codes alone, but it could include people who started insulin for in-hospital hyperglycemia, for example due to steroid therapy. Sicker patients are more likely to become hyperglycemic and be treated, which could concentrate severe cases in the diabetes group.

Second, adjustment was limited. DPC data lack laboratory values, vital signs, imaging, COVID-19 severity and vaccination history, and comorbidities such as hypertension and kidney disease were not in the model. The odds ratios should be read as strengths of association within these limits, not as how many times more dangerous diabetes is by itself. Outcomes that are rare in controls, such as mechanical ventilation, also tend to yield large odds ratios.

Still, showing consistently higher risks — not only for death but also for ventilation, ICU care and dialysis — in about 10,000 patients, based on treatments actually delivered, is a strength. Real-world data reveal what happened in everyday practice in ways clinical trials often cannot.

Preparing for the next epidemic

The vulnerability of people with diabetes to severe infection may recur with future emerging infections. These results reinforce basic measures: glucose and weight management, vaccination and early medical attention when unwell.

The next step is to identify which people with diabetes are at highest risk using data that include glucose, HbA1c, severity and vaccination history.

All Journal Club of the Suwanai articles →