The ethics committee of each participating hospital approved this multicenter retrospective study. All patients in this study consented to participate. A total of 11 hospitals have registered orthopedic trauma surgery cases in our university database annually since 2014. The participating hospitals in the database are all affiliated with the department of orthopedic surgery at our university. Orthopedic surgeons perform the surgery at these hospitals in central Japan. We collected surgically treated cases of intertrochanteric fracture from this database.

Medication data were collected from electronic medical records at the time of admission. We counted all prescribed medications that the patient was taking regularly prior to being admitted. We did not include in the medication count over-the-counter (OTC) medications, acute medications prescribed within 1 week before admission, or herbal supplements and traditional medicines. Polypharmacy was defined as the concurrent use of five or more regularly prescribed medications [12]. The medication count was based on the number of different medications rather than the number of doses taken daily.

This study included patients who underwent surgical treatment for intertrochanteric fractures between 2016 and 2020. In total, 1864 patients were initially included. We excluded patients who were under 65 years of age, those with polytrauma, and those with a Charlson Comorbidity Index (CCI) greater than 3 points, resulting in a study population of 1608 patients. The rationale for excluding patients with a CCI greater than 3 points was to focus on patients who were likely to require fewer medications, thus limiting the influence of underlying comorbidities. This selection helped to ensure that the study primarily examined the effects of polypharmacy itself, rather than the need for multiple medications due to severe or multiple chronic conditions.

The following data were collected from the patient’s electronic medical records: age at the time of injury, sex, body mass index (BMI), residence before admission, smoking history, alcohol habit, and date of death or the date of last follow-up examination. The follow-up period extended up to 206 months, with a median duration of 12 months.

We used the CCI [13] to assess the comorbidities of patients at the time of injury. The CCI is calculated by assigning one point each to the following factors: myocardial infarction, congestive heart failure, peripheral vascular disease, dementia, chronic pulmonary disease, connective tissue disease, peptic ulcer disease, mild liver disease, and uncomplicated diabetes. Two points each are assigned to the following factors: hemiplegia, moderate-to-severe renal disease, complicated diabetes, malignancy within 5 years of the diagnosis, leukemia, or lymphoma. Three points are assigned for moderate-to-severe liver disease, and six points each are assigned to AIDS (not HIV) and metastatic solid tumors. We then divided our population into average (CCI 0), mild (CCI 1), and moderate (CCI 2) cohorts [14].

Each patient’s physical status was also assessed using the American Society of Anesthesiologists physical status (ASA-PS) score [15]. The anesthesiologist routinely sets the ASA-PS score before an anesthetic procedure on the basis of a subjective assessment, and it is available in the medical records. Then, we divided the patients into the average (ASA-PS 1), mild (ASA-PS 2), and severely ill (ASA-PS ≥ 3) groups [16].

The Parker Mobility Score, which evaluates a patient’s walking ability [17], includes three questions, each valued at 0–3 points. On the basis of the sum of the mobility assessment in three different situations (being able to get about the house, getting out of the house, and going shopping), the total score ranges from 0 to 9. For each of these three situations, mobility must be scored as follows: no difficulty (3 points), with an aid (2 points), with help from another person (1 point), or not at all (0 points). The highest overall score of 9 indicates the best possible mobility [18]. We evaluated the Parker Mobility Score at both admission and the last follow-up.

Information regarding death within 1 year postoperatively was obtained from hospital medical records and follow-up telephone interviews conducted with patients’ families or nursing home staff when necessary.

We divided fractures into two categories: stable fractures, which are typically two-part fractures, and unstable fractures, which have three or more parts [19].

We recorded complications such as delirium, deep vein thrombosis, heart failure, cerebral infarction, pneumonia, and urinary tract infections. These complications were selected on the basis of previous studies that identified them as the most common and clinically significant postoperative complications in hip fracture patients [20,21,22]. These diagnoses were made by orthopedic surgeons in consultation with hospital general physicians and were defined as conditions requiring treatment by the general physicians. Postoperative delirium was diagnosed when the following criteria were met: an acute change in mental status with a fluctuating course or inattention (reduced ability to sustain attention and follow conversations), along with disorganized thinking [23].

We conducted two separate analyses. For Analysis 1, we performed propensity score matching to reduce potential confounding factors between polypharmacy and non-polypharmacy groups. After excluding patients with insufficient data, the final study group consisted of 1346 patients. These patients were divided into two groups: the polypharmacy group taking five or more medications and the non-polypharmacy group taking fewer than five medications. The propensity scores were calculated using logistic regression with the following variables as continuous values where applicable: age, sex, BMI, CCI, residence before admission, fracture type, ASA score, and Parker Mobility Score. We employed 1:1 nearest neighbor matching without replacement with a caliper width of 0.2 standard deviations, resulting in 498 matched pairs (N = 996). The balance of covariates before and after matching was assessed using standardized mean differences (SMDs). After 1:1 nearest neighbor matching, all SMDs were reduced to < 0.1, indicating sufficient balance between groups (Supplementary Table S1). In the matched cohort, we used the Kaplan–Meier method with log-rank test to compare survival between groups. Postoperative complications were analyzed using McNemar’s test.

For analysis 2, preliminary assessment of the Cox regression model revealed potential violations of the linearity assumption for continuous variables. To address this, we categorized age (≤ 70, 71–80, 81–90, > 91 years), BMI (< 18.5, 18.5–25, ≥ 25 kg/m2 according to WHO criteria), and Parker Mobility Score (≤ 3, 4–6, ≥ 7) before analysis. We performed multiple imputation using the CART (Classification and Regression Trees) method to handle missing data, generating five imputed datasets (m = 5) with a maximum of five iterations (maxit = 5) and a random seed of 123 for reproducibility. We conducted Cox proportional hazards models using MCMC methods on the imputed dataset (N = 1608), adjusting for categorized variables, sex, CCI, residence before admission, fracture type, and ASA score. The proportional hazards assumption was verified using Schoenfeld residuals (Global p = 0.22), and no influential outliers were identified. Multicollinearity was assessed using variance inflation factors (VIF), and all analyses were performed using EZR version 1.40 [24].