Close this search box.

Impact of COVID-19 versus other pneumonia on in-hospital mortality and functional decline among Japanese dialysis patients: a retrospective cohort study – Scientific Reports

Source of data

The study participants were identified from the Diagnosis Procedure Combination (DPC) inpatient database, which is an administrative claims database in Japan. This database includes more than 1000 hospitals, including all 82 university hospitals, and covers more than half of all admissions in the country16. The DPC database provides information on various aspects such as diagnosis and comorbidities at hospital admission and cause of death coded according to the International Classification of Disease and Related Health Problems, 10th Revision (ICD-10)17. It also includes patient information such as age, gender, Body Mass Index (BMI), admission and discharge status, ADL at admission and discharge, and a comorbidity score known as the Charlson Comorbidity Index18, which is updated for risk adjustment19,20. This study was performed in accordance with the ethical principles laid down in the 1964 Declaration of Helsinki, and was approved by the ethics committee of Tokyo Medical and Dental University (No. M2000-788). The requirement for informed consent was waived by the ethics committee of Tokyo Medical and Dental University due to the anonymous nature of the data.

In 2020, there were 338,256 pneumonia-related cases as any of a primary diagnosis during hospitalization, reason for admission, or disease that required the highest medical care cost in the DPC database. The inclusion criteria for this study were patients who were at least 18 years old, had a hospital stay of at least 24 h, and had either COVID-19 or pneumonia as the primary diagnosis for hospitalization (Fig. S1). COVID-19 and non-COVID-19 pneumonia were recognized with the ICD-10 codes. Several exclusion criteria were applied, including second or subsequent admissions, death within 24 h of admission, a BMI less than 15 or greater than 50, incomplete information on BMI, ADL, and admission type (emergent or non-emergent), patients with aspiration pneumonia, and patients who initiated hemodialysis or peritoneal dialysis during hospitalization. A total of 123,378 patients were included before matching, comprising 66,692 non-COVID-19/ND patients, 54,132 COVID-19/ND patients, 1894 non-COVID-19/D patients, and 660 COVID-19/D patients. After a propensity score matching (PSM), 2136 patients were divided into four subgroups, each consisting of 534 patients (Fig. S1).

Patients who received maintenance hemodialysis and peritoneal dialysis were recognized with the coding of patient care procedures as follows: chronic maintenance hemodialysis with < 4 h per session, ≥ 4 h and < 5 h per session, ≥ 5 h per session, or chronic maintenance hemodiafiltration or continuous peritoneal dialysis. One COVID-19/D patient was dependent of both hemodialysis and peritoneal dialysis and the remaining dialysis patients were on maintenance hemodialysis. There were 13 patients post transplantation, including 7 post kidney transplant, 5 post stem cell transplant, and 1 post liver transplant patients.

Patient characteristics

The Barthel Index scores at admission and discharge were calculated based on 10 functional abilities: feeding, bathing, dressing, grooming, toileting, bowel control, bladder control, chair transfer, ambulation, and climbing stairs21. The Barthel Index ranges from 0 to 100 points, with the highest scores indicating greater independence in physical functions and lower scores indicating a more bedridden status. The Barthel Index scores were used to classify the level of dependence into four groups: total dependence (0–20 points), severe dependence (21–60 points), moderate dependence (61–90 points), and mild dependence or complete independence (91–100 points)22,23. In addition, other clinical data on inpatients were collected, including age, gender, BMI, dialysis dependency, the updated Charlson Comorbidity Index excluding renal disease20,23, comorbidities, and admission type (emergency or non-emergency). The age groups used in the analysis were 18–49, 50–59, 60–69, 70–79, and greater than 80 years24.


The primary outcome of this study was the occurrence of in-hospital deaths from any cause. The secondary outcome was a decline in physical function, defined as a decrease of at least 20% in the Barthel Index score at discharge compared to that at admission. We also evaluated a risk of death directly from pneumonia or death due to other diseases. The database identified deaths from COVID-19 or non-COVID-19 pneumonia as a primary diagnosis during hospitalization or those from other reasons. Other outcomes included hospital length of stay and medical care cost. The long-term hospitalization was defined as a stay of 30 days or longer. The high medical cost was defined as the highest quartile of participants. Patients were followed until discharge, transfer, or in-hospital death20.

Data analyses

Baseline characteristics were presented as numerical values (%) or medians (interquartile ranges). The Wald confidence interval for proportions was examined. We estimated the propensity score using a logistic regression model. To minimize potential confounding effects, differences in age (18–49, 50–59, 60–69, 70–79, and greater than 80 years) and sex were adjusted. Based on the propensity score using a 1:1 scheme, Non-COVID-19/D patients (prematching, N = 1894; postmatching, N = 534) were first matched with COVID-19/D patients (prematching, N = 660; postmatching, N = 534). Subsequently, COVID-19/ND (prematching, N = 54,132; postmatching, N = 534) and Non-COVID-19/ND patients (prematching, N = 66,692; postmatching, N = 534) were adjusted with each COVID-19/D patient. The caliper width used for matching was set at 0.25 of the standard deviation of the propensity score. The cumulative hazard after hospitalization for pneumonia was assessed using the Nelson–Aalen estimator among the four groups. The log-rank test was used for a statistical comparison. Risks of mortality or functional decline were estimated using logistic regression analyses among postmatching patients, adjusting for age, gender, BMI, Barthel Index score at admission, and Charlson Comorbidity Index score. Statistical analyses were performed using JMP Pro 12.0 software (SAS Institute Inc., Cary, USA). p-values of less than 0.05 were considered statistically significant.