Study design and setting
A prospective cohort study was conducted to investigate the clinical characteristics of hospitalised pneumonia patients beginning in October 2010 in Kurashiki Central Hospital, a 1,166-bed tertiary hospital in Okayama Prefecture, Japan (UMIN000004353). In particular, CAP patients who were admitted between 2013 and 2018 were analysed. CAP was defined as recommended by the Infectious Diseases Society of America/American Thoracic Society guidelines . These guidelines required any of the following: presence of infiltrate on chest imaging; some clinical symptoms (cough, sputum production, fever, dyspnoea, and pleuritic chest pain); and physical findings of coarse crackles on auscultation and elevated inflammatory biomarkers, including C-reactive protein and white blood cell count on laboratory testing. The exclusion criteria were age < 15 years and nursing and healthcare-associated pneumonia (NHCAP). NHCAP patients were classified according to the Japanese Respiratory Society guidelines . Patient information (age, sex, underlying medical problems, antibiotic therapy, pneumonia severity, mortality rate) and causative pathogens were evaluated. Pneumonia severity was evaluated based on the CURB-65  and the Pneumonia Severity Index (PSI) . Patients’ clinical characteristics and causative pathogens before and after the west Japan heavy rainfall event were compared using prospective cohort data.
This study was performed as part of a clinical study of pneumonia (trial registration: UMIN000004353), and it was approved by the institutional review board of Kurashiki Central Hospital (approval number 3379). All patients gave their informed consent to participate in this study. The research was conducted according to the principles of the World Medical Association Declaration of Helsinki.
Details of the heavy rainfall event in western Japan
Severe flooding occurred during the so-called west Japan heavy rainfall event between June 28 and July 8, 2018. Due to the influence of the seasonal rain front and typhoons, this storm produced record-breaking heavy rainfall nationwide, particularly in western Japan. In the Chugoku and Shikoku regions, including Okayama Prefecture, there was approximately two to three times the normal rainfall for the whole of July based on averages from 1982 to 2018 . A total of 123 observation points in Japan recorded the maximum 72-hour precipitation amount, and Kurashiki city was included among them . The rainfall amount was 270.5 mm just over the period of July 5 to 7, whereas rainfall for all of July was 323.5 mm in Kurashiki city. This is 2.8 times the average rainfall for this period of 115 mm. Subsequent river flooding and landslides occurred in many areas of Japan. A total of 223 persons died, and 8 were declared missing nationwide; in Kurashiki city alone, there were 52 victims .
In addition, the heavy rainfall caused the Oda River, which is a tributary of the Takahashi River that runs through Kurashiki city, to flood and subsequently breach its banks. As a result, Mabi town in Kurashiki city, which is located 10 km from our hospital, was wholly under water (Fig. 1). Patients affected by the flood visited our hospital because it is the largest hospital covering the affected area.
After the heavy rain, many people came to assist with reconstruction work from inside and outside the affected area. There were 17,287 volunteers in July, and 4,640 to 24,958 volunteers worked from August to November. They removed mud and debris from houses still standing and helped clear the significant wreckage. The removal of major debris in the Mabi town area to a temporary depot was completed on August 25, 2018, but work at the temporary depot has been on-going.
To investigate the impact of the disaster on hospitalised CAP patients, the study period covered the period from July 1 to November 30 in 2018 for the reasons described below, and the data were compared for the same period from 2013 to 2017.
The first reason for the chosen study period was the effect of influenza virus as a causative pathogen in CAP. Klein et al. reported that bacterial coinfection with influenza is common in admitted patients, and that the most common causative bacteria is Streptococcus pneumoniae, followed by Staphylococcus aureus . In Japan, influenza virus activity increases from December to March. Since it was necessary to exclude the effect of coinfection with influenza virus, the period spanning week 47 to week 15 was omitted .
The second reason for the chosen study period was that it was necessary to evaluate patients who had engaged in reconstruction work, primarily that involving the removal of rubble, earth, and sand after several months. After the west Japan heavy rainfall event, the Kurashiki municipal office announced the completion of primary debris removal on August 25, although work at the temporary debris storage facilities continued after that date. An estimated 5,000 to 25,000 volunteers participated in the clean-up from July to November.
To identify causative pathogens, sputum and blood samples were used for culture and serology, as well as a urinary antigen test to detect S. pneumoniae and Legionella pneumophila serogroup 1. Other Legionella species were identified using culture on Wadowsky-Yee-Okuda-α medium. Causative pathogens were defined based on a previous report  if the following criteria were met: (1) positive sputum culture of more than 1 + on a qualitative test or 105 on a quantitative test, with reference to the sputum Gram stain; (2) positive blood culture (excluding contaminating normal skin flora); (3) positive pleural fluid culture; (4) positive urinary antigen test for S. pneumoniae and L. pneumophila; (5) seroconversion or 4-fold increase in antibodies against Mycoplasma pneumoniae and Chlamydophila pneumoniae; and (6) ≥ 1:320 on a single antibody test for M. pneumoniae PA antibody (FUJIREBIO, Tokyo, Japan) or a cut-off index of ≥ 2.0 on a C. pneumoniae IgM antibody test using the Hitazyme assay (Hitachi Chemical, Tokyo, Japan).
Continuous variables are expressed as medians and interquartile range, whereas categorical variables are presented as percentages. Continuous variables were analysed using the non-parametric Mann-Whitney U-test, and categorical variables were compared using Fisher’s exact test. Regarding the comparison of hospitalised pneumonia patients from 2013 to 2017 and 2018, the data were evaluated using Poisson regression analysis. An interrupted time series analysis and segmented Poisson regression with offset terms to detect changes in levels were performed. The rate of causative pathogens per month was also evaluated if the causative pathogens showed significant differences. Autocorrelation was tested using the autocorrelation function and partial autocorrelation function, with a threshold of 0.2. Previous studies reported that CAP patients with S. pneumoniae are more numerous in winter and that those with L. pneumophila are the most numerous in summer . In addition, increased precipitation is associated with an increased risk for legionellosis [19–21]. Therefore, seasonality was adjusted based on Fourier terms, and the model was adjusted by monthly rainfall. As mentioned above, the duration of the intervention was determined to be between July and November 2018. A P value < 0.05 was considered significant. Analyses were performed using R software (version 3.0.3, Vienna, Austria).