We conducted a multi-center, quasi-experimental study using electronic medical record databases from Himeji Primary Emergency Medical Center for Nights and Holidays (Site A), Kobe Children's Primary Emergency Medical Center (Site B), and Hanshin-Kita Children’s First-Aid Center (Site C). The period from April 2016 to September 2018 was designated the "pre-intervention" period, and the period from October 2018 to December 2019 was designated the "post-intervention" period. No specific ASPs had been conducted at these facilities. We obtained opt-out consent from the patients’ parents, and this study was approved by the institutional ethics committees of Kobe Children's Hospital (Approval Number 2020-61), Japanese Red Cross Society Himeji Hospital (Approval Number 2020-39), and Kobe Children's Primary Emergency Medical Center (Approval Number 1).
Patients and Setting
Hyogo prefecture is one of the most populous regions in Japan, covering an area of 8,400 km2. The prefecture’s population in 2019 was approximately 5.5 million. Himeji city, Kobe city, and Hanshin-Kita area are all included in Hyogo prefecture with areas of 534 km2, 554 km2, and 481 km2, respectively. These cities’ population in 2019 were 540,000 (including 77,000 children aged 15 years or younger), 1.5 million (including 0.2 million children aged 15 years or younger), and 720,000 (including 0.1 million children aged 15 years or younger), respectively. Sites A, B, and C are the regional pediatric PECs providing after-hour care on weekdays, weekends, and holidays for patients aged 15 years or younger. These centers treat approximately 20,000, 30,000, and 25,000 children per year and accept patients with various medical conditions (including patients arriving by ambulance), except for those requiring surgical interventions and those with severe burns. In these centers, blood tests (complete blood counts and C-reactive protein levels), urinalysis, rapid tests for respiratory diseases, and diagnostic imaging are available. More than 50 physicians are employed as part-time employees in each facility. Each of these part-time employees usually works in their clinic or hospital during the day. Because of the variety of specialties, they do not necessarily have education in primary care. In each PEC, the number of prescription days is limited to 1 to 2 days to encourage follow-up with the patient's regular clinic.
In October 2018, Sites A and B initiated monitoring all antimicrobial prescribing trends at their facilities (particularly including 3GCs and penicillin). In addition, we performed specific ASPs at each of the facilities as described below.
Himeji Primary Emergency Medical Center for Nights and Holidays (Site A)
We published a facility-specific guideline for children in October 2018, referring to national guidelines: “Manual of Antimicrobial Stewardship. The 1st Edition ”. The facility-specific guideline recommended the choice (except 3GCs), dosage, and duration of antibiotics for each pediatric infection. Similarly, this guideline stated that antibiotics were unnecessary for common cold, upper respiratory tract infection, bronchitis, bronchiolitis, and gastroenteritis because they were mostly caused by viruses, and that the watch-and-see strategy can be adopted for mild acute otitis media without prescribing antibiotics: this was posted in the consultation room so that all physicians could check them at any time. We collected prescription data once every six months and shared the information of these data in regular meetings twice a year. Since this direct presentation was provided only to some physicians attending the meeting, the main intervention at Site A was the publish of facility-specific guideline.
Kobe Children's Primary Emergency Medical Center (Site B) 
We investigated all antimicrobial prescribing trends and the appropriateness of prescribing 3GCs every month. The results of the monitoring and investigation were provided as a monthly newsletter that summarized recent antimicrobial prescribing patterns and the facility’s prescribing targets. Similarly, the newsletters provided recommendations for specific infections and introduced national guidelines. They were posted in the consultation rooms and staff lounges so that physicians viewed it at any time.
Hanshin-Kita Children’s First-Aid Center (Site C)
This facility was a negative control where we did not perform the specific ASP during the period. We retrospectively surveyed the facility’s antimicrobial prescribing status in December 2019.
Data were collected using a search tool for electronic medical records at each pediatric PEC. The collected data included the number of patients, patient age, diagnoses, the total number of antimicrobial prescriptions, and the number of specific antimicrobial prescriptions (3GCs and amoxicillin). The patient diagnoses were based on medical receipt data.
Endpoint and Statistical Analysis
The primary outcome was the trend in the 3GC prescription rate. The secondary outcomes were the trend in the amoxicillin prescription rate and all antibiotic prescription rate. The antimicrobial prescription rate (%) was defined as “the number of specific antimicrobial prescriptions/number of patients×100”.
Patient age and diagnosis at each facility were compared pre- and post-intervention. Age was categorized into four categories (<1, 1–5, 6–10, and 11–15 years), and the diagnosis was classified into 12 categories (upper respiratory tract infections, pharyngitis/tonsillitis, bronchitis, gastroenteritis, sinusitis, streptococcal pharyngitis, pneumonia, acute otitis media, urinary tract infections, skin and soft tissue infections, and others). “Others” mainly included non-infectious diseases, such as allergic urticaria, Kawasaki disease, constipation, and foreign bodies.
Monthly obtained data were divided into pre- and post-intervention and compared using Poisson regression analysis for the primary and secondary outcomes at three centers. The difference-in-difference method was used to assess the effect of the intervention, the duration, and the difference in facilities on the outcomes. All statistical analyses were conducted using R version 4.0.5 (R Core Team, 2019, Vienna, Austria), and Bonferroni correction for multiple comparisons was applied for P values. Two-tailed P values below .0167 were considered statistically significant.