This study used an administrative claims database collected from hospitals marketed by Medical Data Vision Co., Ltd. [15]. The database has over 40 million patient records and covers approximately 26% of all hospitals in Japan that have adopted a diagnosis procedure combination system. From this data, it is possible to obtain a variety of information related to patient treatments, such as diagnoses, prescription drugs, surgeries, and procedures. The data used were anonymized so that individual patients could not be identified and informed consent was not obtained.

The study population consisted of patients administered TEIC between 2010 and 2019. Patients aged 15 years or older who were administered TEIC for at least three consecutive days were identified from this population, and the date of the first TEIC administration was defined as the index date. We restricted our sample to patients for whom some data existed up to the month preceding the index date. Then, patients who had received ursodeoxycholic acid (UDCA) and/or glycyrrhizin (GL) as of the month preceding the index date were excluded to ensure that patients with suspected pre-existing liver injuries were excluded. Because TDM was defined as a claim for treatment and management fees for specific drugs, it excluded patients who had been treated with a TDM drug, except for TEIC, within seven days of the index date (Additional file 1: Table S1). Finally, patients with missing data were excluded, and those that met the inclusion criteria for the study were selected.

Validation studies on positive predictive value of liver injury definition using the International Classification of Diseases, 10th edition (ICD-10) codes have not been conducted in Japan. Therefore, in this study, liver injury was defined as a diagnosis indicated by ICD-10 codes (K711/K719 [toxic drug-induced liver injury], K720 [acute and subacute liver injury], K769 [unspecified liver injury]) and treatment indicated by the administration of UDCA and/or GL recorded in the same month. Additionally, all suspected diagnoses were excluded. The treatment was defined as another treatment if there was an interval of 7 days between teicoplanin doses, and the treatment duration was defined as the period from the index date to the final administration date of the first treatment.

Teicoplanin was described in the package insert as 400 mg or 800 mg in two divided doses on the first day and 200 mg or 400 mg once daily thereafter [9]. In addition, TDM guidelines recommend doses higher than the package insert [10]. Therefore, the loading dose of TEIC was calculated for a total dose of 3 days from the index date. Loading doses were classified into the following three categories: 800–1,600 mg (the package insert dose),  > 1,600 mg (guideline dose), and  < 800 mg.

TDM was defined as reimbursement claims of treatment and management fees for specific drugs, according to previous reports [16, 17]. Similarly, pharmacist management for individual patients, placement of pharmacists on hospital wards, and establishment of infection control teams in healthcare facilities were defined by drug management and guidance fees, inpatient pharmaceutical service premiums, and infection prevention and control premiums, respectively [17].

Furthermore, no reports or guidelines clearly define drugs with potential risk of liver injury. Therefore, in this study, drugs that have already been reported to cause liver injury were classified according to their effects based on the serious disease manual, which is regularly updated by the Ministry of Health, Labor and Welfare (MHLW) [18]. Among these, drugs with a high frequency of risk occurrence for liver injury were classified into the following six categories: anti-infection drugs, antipyretic analgesics and anti-inflammatory drugs, anticancer drugs, gastrointestinal drugs, psychiatric or neurological drugs, and metabolic disease drugs. These drugs were defined as concomitant medications if their administration overlapped with the treatment period of TEIC.

Data on concomitant drugs at risk of liver injury were extracted from among oral and injectable drugs containing the relevant ingredients [18]. In addition, as in previous reports [19], the site of infection was defined using the Japanese disease code that is uniquely assigned to each disease. Details are provided in the supplementary files (Additional file 2: Table S2, Additional file 3: Table S3).

Diagnostic information for the month of admission was used to calculate Charlson Comorbidity Index (CCI). Scores were calculated using a program that can be introduced into the Stata software version 17.0 (Stata Corp., College Station, TX, USA).

The implementation of loading doses during TEIC administration was surveyed over time and compared after classifying the implementation of TDM. In the target population, propensity score matching was performed based on loading doses according to the guideline doses (guideline group) or below the package insert dose (non-guideline group). The nearest neighbor within caliper without replacement was used, and patient background was adjusted. The propensity score was calculated using logistic regression, and 1:1 matching was performed using a caliper of 0.2. Clinical variables included information that might influence the loading dose and that could be collected clinically before the start of TEIC treatment. In this study, bed size, age, CCI, weight, reimbursement, site of infection, and clinical department were selected. The balance between the two groups was evaluated using the Mann–Whitney U test for continuous variables and the chi-square test for categorical variables in the before propensity score matching population, and standardized mean difference (SMD) in the matched population. SMD was considered balanced if it was  < 0.1. In addition, risk factors for 30-day mortality and liver injury in patients administered TEIC were evaluated using univariable and multivariable conditional logistic regression analysis.

We used Stata software and EZR software (Saitama Medical Center, Jichi Medical University, Saitama, Japan) [20] for statistical analysis, with a significance level set at p < 0.05.