We used diagnosis procedure combination (DPC) data from hospitals participating in the Quality Indicator/Improvement Project (QIP) managed by Kyoto University’s Department of Healthcare Economics and Quality Management. The QIP database includes DPC data from Japanese acute care hospitals that voluntarily participated in the project [12]. DPC data provide clinical information on inpatient admissions, practices, and reimbursement, and can be utilized for research on quality improvement in clinical practice, hospital management, and healthcare systems and policies [13].

The FLS-CS states that physicians, nurses, pharmacists, physical therapists, and other professionals should cooperate to identify eligible patients, evaluate the risk of secondary fractures, and provide treatment and follow-up for secondary fracture prevention. The target patients are described as those over 50 years of age with all types of fragility fractures, with top priority given to patients with hip and clinical vertebral fractures. It also states that the evaluation of secondary fracture risk should occur as early as possible after the fracture, at least within 90 days, based on Japanese osteoporosis guidelines. This evaluation can be done using DXA or thoracolumbar spine X-rays, or risk assessment tools such as FRAX. Furthermore, evidence-based pharmacotherapy and fall prevention strategies should be initiated immediately after evaluation [14]. FLS programs were introduced on a hospital basis. The management fee is divided into categories 1 to 3 for patients who have undergone hip fracture surgery. Management fee 1 can be claimed if appropriate evaluation and treatment of osteoporosis is performed based on FLS-CS and osteoporosis guidelines during hospitalization for surgery. The evaluation may include bone mass measurement, bone metabolism marker assessment, spine X-ray, etc. Management fees 2 and 3 can be claimed in rehabilitation wards and outpatient clinics when patients for whom management fee 1 was claimed continue to be evaluated and treated for osteoporosis [11].

Surgical patients aged 50 years or older with hip fractures, as well as those with vertebral fractures, who were admitted to and discharged from QIP-participating hospitals for which data were consecutively available from June 1, 2017, to June 30, 2023, were included. Patients with high-energy trauma and multiple trauma codes (ICD-10 codes: T07.x and T79.4) as comorbidities upon admission and those with same-day admission and discharge or death as outcomes were excluded. For cases of disease that led to hospitalization, hip fractures were defined by ICD-10 codes S72.00, S72.10, and S72.20, and vertebral fractures were defined by codes S22.00, S32.00, S32.70, and T02.10. Excluded were hip fractures with terms “suspected,” “postoperative,” or “dislocation fracture” in the diagnosis and vertebral fractures with terms “suspected,” “postoperative,” “pubic,” “sciatic,” “transverse processes,” “spinous processes,” or “metastases” in the diagnosis. Surgical procedures for hip fractures were identified by claim codes, which are billing codes for surgery, including open reduction and internal fixation (femur) (K0461), intra-articular open reduction and internal fixation (hip) (K0731), hemiarthroplasty for hip (K0811), and total hip arthroplasty (K0821).

The primary outcome was the monthly proportion of patients with the implementation of secondary fracture prevention during hospitalization. This proportion was calculated as the total number of patients discharged in a certain month with the implementation of secondary fracture prevention during hospitalization divided by the total number of patients discharged in a certain month for hip and vertebral fractures, respectively. In this study, secondary fracture prevention was defined as an osteoporosis test performed and osteoporosis medication prescribed during hospitalization, according to the criteria for claiming the management fee. Osteoporosis tests were defined as bone mineral density tests or blood/urine tests measuring bone metabolism markers (1.25-dihydroxyvitamin D3, 25-hydroxyvitamin D, tartrate-resistant acid phosphatase 5b, osteocalcin, calcitonin, undercarboxylated osteocalcin, deoxypyridinoline, N-terminal telopeptide, β-C-terminal telopeptide, intact procollagen type I N-terminal propeptide, and bone-type alkaline phosphatase). The medications for osteoporosis were bisphosphonates, active vitamin D3, calcium, vitamin K2, selective estrogen receptor modulators, calcitonin, teriparatide, denosumab, estradiol, and romosozumab. Prescriptions included both dowry and discharge prescriptions. Because prescription and testing for osteoporosis are rarely performed together, we also analyzed the proportion of osteoporosis tests and the proportion of osteoporosis medications, respectively, as supplemental outcomes.

Patients were categorized and described according to the month of discharge for each fracture across three distinct time periods: from June 2017 to May 2019, before FLS-CS was published (before intervention); from June 2019, when FLS-CS was published, to March 2022, before the management fee was introduced (after the publication of FLS-CS); and after April 2022, after the management fee was introduced (after the introduction of the management fee). Categorical variables are presented as frequencies (percentages). The continuous variable length of hospital stay is presented as the median (interquartile range), and age is presented as the mean (standard deviation). One-way analysis of variance was used to compare the means of the three groups. A non-parametric Kruskal–Wallis test was used to compare the medians. A chi-squared test was used to compare categorical variables. We presented the proportion of hospitals that implemented secondary fracture prevention by duration for each fracture, along with the proportion of hospitals that have claimed the management fee since April 2022 and those that have not claimed it.

Interrupted time series analysis (ITSA) was used to estimate the effect of the publication of FLS-CS and the introduction of the management fee on the implementation of secondary fracture prevention. ITSA is considered a robust quasi-experimental approach to evaluating the effect of interventions at a population level at a well-defined point in time. It can statistically evaluate the magnitude and timing of intervention-induced changes in outcomes, whether the change was immediate, delayed, transient, or long-term, and whether factors other than the intervention could explain these changes [15, 16].

ITSA was conducted as follows: A generalized linear model assuming a Poisson distribution was used with a logarithmic link function. The objective variable was the number of secondary fracture prevention performed for each month with the proportion calculated using the number of patients per month of discharge as the offset term. The first intervention point was June 2019, coinciding with the publication of FLS-CS, whereas the second intervention point was April 2022, following the introduction of the management fee. The measurement interval was set at 1 month, yielding a total of 73 time points. Overdispersion, seasonality, and autocorrelation were considered in the analysis. If there was overdispersion, a pseudo-Poisson distribution was used. Seasonality was considered by including a harmonic term (sines and cosines) with 12-month periods. The validity of the Poisson regression model was evaluated using correlograms (autocorrelation and partial autocorrelation functions) and residuals. Immediate and monthly rates of change resulting from the two interventions were estimated. The rate of change was reported as the logarithm of the coefficient estimated from the model (IRR: incidence rate ratio) along with a 95% confidence interval (CI). The following models were used in this study.

$$\mathit\;y=\mathit\;n+_+_\times time+_\times \;intervention1+_\times \;time \;after \;intervention1+_\times \;intervention2+_\times\; time \;after \;intervention2$$

where y is the number of cases in which secondary fracture prevention was performed, n is the total number of cases per month, time is the number of months since June 2017 (counted from 0), intervention1 is a dummy variable representing the publication of FLS-CS, set at 0 until May 2019 and 1 after June 2019, time after intervention1 represents the number of months since FLS-CS was published (June 2019 and after are counted from 0), intervention2 represents the introduction of the management fee, a dummy variable set at 0 until March 2022 and 1 after April 2022, and time after intervention2 represents the number of months since the management fee was introduced (April 2022 or later was counted from 0). β0 is the baseline intercept, β1 is the baseline trend, β2 represents the immediate proportion change after the publication of FLS-CS, β3 represents the level change after the publication of FLS-CS, β4 represents the level change after the introduction of the management fee, and β5 represents the slope change after the introduction of the management fee. The same method was used to analyze the proportion of osteoporosis tests and the proportion of osteoporosis medications.

As a sensitivity analysis, the impact of the intervention was analyzed with a time lag of 3 or 6 months from the publication of FLS-CS, considering the potential delay between publication and its actual effects. Furthermore, due to potential variations in the definition of secondary fracture prevention across hospitals regarding the criteria for claiming the management fee, the analysis was conducted with a broader definition of secondary fracture prevention as osteoporosis tests or medications prescribed during hospitalization.

All statistical tests were two-sided, with P values less than 0.05 considered statistically significant. All statistical analyses were performed using R ver. 4.2.1 (Package tsModel Ver0.6–1).