Database and setting
Yokohama City is the most populous city in Japan and is governed by the local government for the Greater Tokyo Area, which includes Tokyo metropolis. Yokohama City has a population of approximately 3.75 million (January 2021), and the population distribution is as follows: individuals aged < 15 years, 11.9%; 15–64 years, 62.8%; and ≥ 65 years, 25.2% (January 2021) [2]. The Yokohama Original Medical Database (YoMDB) [8] was built by Yokohama City after approval from the Information Disclosure and Personal Information Protection Review Board based on the Yokohama City Ordinance. This large medical invoice database can only be accessed by administrative officers from departments in charge of medical care and includes data on all residents of Yokohama City who have the following three government-funded insurance types: National Health Insurance, Long Life Medical Care System, and Public Assistance. The National Health Insurance is a system for farmers, self-employed people, and other such individuals. The Long Life Medical Care System is the system for all individuals aged ≥ 75 years. The personal numbers of insurance and Public Assistance are hashed to protect the identity of individuals, and information that could reveal an individual’s identity, such as their name and treatment details, is deleted to make the database secure.
This database includes 68.3% of Yokohama City residents aged 65–69 years; 84.1% of residents aged 70–74 years, with 99.7% of them aged 75–79 years; and 98.6% of residents aged > 80 years. Hence, it has a strong representation of the elderly and is an especially reliable database for those older than 75 years.
Case selection
The data from 2014 and 2015 included those of 2,486,834 people and 29,411,895 medical invoices. Target patients were selected based on three criteria: age ≥ 65 years, death caused by a malignant neoplasm based on International Classification of Diseases (ICD)-10 codes (C00-C97), and additional HEC charges. The additional charge could have been applied by an insurance-participating medical facility when planned/emergent home medical care and nursing visits were provided more than twice for a total of 15 days (within 14 days before the day of death).
Statistical analyses and definition of variables
Patients’ demographics, including age, sex, insurance type, cancer type, and institution type, were collected from the database. In this study, we divided the patients into the < 80-year and ≥ 80-year groups according to their age. The groups were compared according to each criterion for treatment planning in actual clinical practice. Cancer type was classified as follows: oral and oropharyngeal (International Classification of Diseases-10th revision [ICD-10] code C00-C14, C32), esophageal and gastric (C15, C16), colon (C18-C20), hepatobiliary and pancreatic (C22-C25), lung (C33-C34), skin (C43-C44), breast (C50), gynecologic (C53-C56), prostate (C61), urinary tract (C64-C68), brain (C70-C72), thyroid (C73), and blood (C81-C85, C88-C96) cancers.
HEC was classified based on the three types of home visits as follows: p-HV, e-HV, and i-HV. Furthermore, i-HVs were divided according to the timing of the visit as follows: daytime i-HV (visit during clinic hours), midnight i-HV (visit from 22:00 to 06:00), and night/holiday i-HV (visit at a time other than clinic hours and midnight). These classifications were consistent with the medical fee point system in Japan.
Medical fee point data related to home medical care (coded as C000 to C171-2) were obtained based on the invoice of medical treatment for the target patients during the 2-year period and were analyzed. Admission was defined as the claim of the basic admission fee (coded as A100-109) or special admission fee (coded as A300-317), including admission to the critical care unit, intensive care unit, high care unit, palliative care unit, and community-based integrated care unit.
Emergent admission could not be directly confirmed due to the payment system. Cases with data on planned/emergent home medical care fee and hospitalization fee in the following month were referred to as cases of emergent hospitalization. Survival time at home was defined as the period from the first home medical care event to the month in which the additional fee for end-of-life care was charged.
Comparisons were made between the groups of patients aged < 80 years and ≥ 80 years. Statistical analyses were performed using Student’s t-test (to compare the time of e-HV and i-HV and the number of home visits per month), χ2 test (to compare baseline characteristics, emergent admission, central venous nutrition, oxygen use, opioid use, p-HVs by doctors or nurses three times a week or more, and death at home), and multiple regression (to compare survival time after HEC introduction). Sex, insurance type, and type of institution were added as covariates in the multiple regression analysis. P values < 0.05 were considered statistically significant. All statistical analyses were performed using Statistical Package for the Social Sciences, version 27 (International Business Machines Corporation, Armonk, NY, USA).