The impact of in-hospital initiation of a surgeon-led, anti-osteoporotic medication algorithm for patients with fragility hip fractures: a quasi-experimental study
Our latest research study was published in the Archives of Orthopaedic and Trauma Surgery, 30 June 2025.
The impact of in-hospital initiation of a surgeon-led, anti-osteoporotic medication algorithm for patients with fragility hip fractures: a quasi-experimental study
by
Ioannis I. Daskalakis, Evangelos I. Kritsotakis, Johannes D. Bastian, Ioannis V. Sperelakis & Theodoros H. Tosounidis
Access the full paper via:
Abstract
Introduction
Secondary fracture prevention is essential to current hip fracture management. However, many patients are discharged without the appropriate anti-osteoporotic medication (AOM). This study aims to evaluate the impact of the implementation of an in-hospital, surgeon-led AOM algorithm on patients with fragility hip fractures. The three outcome events of interest assessed were (a) the AOM initiation (b) the persistence to medication at 2 years of follow-up and (c) the secondary fracture incidence within 2 years of AOM initiation in the treated and control groups.
Materials and methods
This was a three-arm controlled before-after quasi-experimental study. A group of hospitalized patients with low-energy hip fractures who were prescribed AOM according to the surgeon-led AO algorithm between March 2020 and May 2022, were compared with a group of concurrent control patients who refused the treatment over the same 2 year period, and a group of historical control patients who were hospitalized for hip fragility fractures in 2 years before the algorithm was introduced (February 2018 to February 2020). AOM initiation rates, 2 year medication persistence, and risks of secondary osteoporotic fracture were assessed and compared between groups.
Results
In this study, we enrolled 598 patients (mean age 82 ± 8 years, 78% female). Post-fracture AOM initiation increased from 15% (41/281) before to 67% (213/317) after introducing the algorithm. Medication persistence after 2 years of AOM initiation was 56% (95% confidence interval [CI] 49–63%) in treated patients and 52% (95% CI 36–66%) in historical controls. Secondary osteoporotic fractures occurred in 15/213 (7%) treated patients, 8/104 (8%) concurrent control patients, and 20/281 (7%) historical-control patients over an average follow-up of 20.4 months. Multivariable Cox regression analysis did not demonstrate significantly different fracture risks in historical controls (cause-specific hazard ratio [csHR] 0.92; 95%CI 0.45–1.89) or concurrent controls (csHR 1.08; 95% CI 0.45–2.57) compared to treated patients.
Conclusion
The AO Foundation algorithm can increase AOM initiation at hospital discharge, retaining high medication persistence 2 years post-fracture. A longer follow-up period is required to evaluate the algorithm's effect on secondary fracture prevention.
Statistical methods
In our earlier study, 213 consecutive hip fracture patients were included in the intervention group for AOM, which was deemed to have adequate statistical power to detect a meaningful difference in secondary fracture risk according to medication persistence after a minimum follow-up period of 12 months (average follow-up was 17.2 ± 7.1 months). In the present study, we compared the same group of treated patients (n = 213) with concurrent (n = 104) and historical (n = 281) control patients, after extending the follow-up period to a maximum of 24 months. The sizes of the control groups were determined by the number of eligible patients as shown in Fig. 1. A power calculation showed that the available sample sizes allowed the detection of moderate effect sizes (hazard ratio > 1.5, or equivalently, < 0.67) at the 0.05 family-wise error rate (Bonferroni corrected significance level α = 0.025) with power at least 87% when comparing treated patients to concurrent controls and 99% when comparing treated patients to historical controls (Supplementary Figure S1).
Survival analysis methods were applied to assess the risks of three outcome events of interest over time: persistence to ΑΟΜ and subsequent secondary fracture or death. Persistence was expressed as time from initiation to discontinuation of ΑΟΜ. The Kaplan–Meier method was employed to estimate cumulative probabilities of medication persistence by a given time and up to 2 years from initiation. A multivariable Cox proportional hazards model was applied to estimate adjusted hazard ratios (HR) for therapy discontinuation (non-persistence) according to baseline predictors.
Cumulative incidence function (CIF) plots were used to describe secondary fracture over time, accounting for the length of follow-up and censored event times, and considering death as a competing risk that modifies the probability of observing a fracture. The CIF gives the probability of a patient experiencing a fracture (primary event) at any given time, conditioned upon not experiencing either event (primary or competing) until that time. CIFs were estimated by applying the Aalen-Johansen method and compared between treated patients and controls using a modified log-rank test. To account for variability in baseline variables, we estimated cause-specific hazard rate ratios (csHR) for each event of interest (i.e., secondary fracture and death) using the Cox proportional-hazards regression model. The csHR is the between-groups ratio of the instantaneous rates of occurrence of the event of interest in patients who have not experienced any event.
The possibility of multicollinearity was ruled out by examining variance inflation factors (supplementary Table S1). Log-linearity in hazard rates for continuous variables (age, CCI, LOS), as assumed by the Cox model, was confirmed using plots of Martingale residuals (supplementary Figures S2-S4) and by testing the possibility of non-linear fit with restricted cubic splines. The proportional hazards assumption for treatment status (i.e. constant csHR over time) was confirmed by the Grambsch-Therneau test in the multivariable models. When none of the events of interest occurred by the end of the follow-up or patients were lost to follow-up, event times were censored at the date of the last follow-up. None of the study variables had missing data. Two-sided 95% confidence intervals and p values were reported throughout. Statistical significance was considered when p < 0.05. Stata version 18.5 (Stata Corp., College Station, TX, USA) was used for all analyses.
Fig. S1. Estimated power for two-sample comparison of hazard rates between treated patients and concurrent controls (A) or historical controls (B)
Figure 1. Study flowchart of the study design and patient enrollment
Figure 3. Cumulative incidence function plots of secondary fractures and deaths following hospital discharge in relation to anti-osteoporotic treatment