Hip Fracture Surgery Timing and Blood Transfusion Risk in Patients on DOACs

Overview

This study looks at patients with hip fractures who are taking direct oral anticoagulants (DOACs), a type of blood thinner. In many hospitals, surgery for these patients is delayed because of concerns about bleeding, but waiting longer can also increase risks such as complications and longer hospital stays. The purpose of this study is to find out whether operating within 24 hours is as safe as delaying surgery beyond 24 hours. Specifically, the investigators want to know if early surgery does not lead to a higher need for blood transfusions compared to delayed surgery.

Detailed Description Hip fractures are one of the most common reasons for hospital admission among older adults, and surgery is usually performed as quickly as possible because delays are linked to higher risks of complications, longer hospital stays, and even increased mortality. However, in patients who are taking direct oral anticoagulants (DOACs), surgeons often postpone surgery due to concerns about bleeding. While waiting allows time for the anticoagulant effect to diminish, it also exposes patients to the risks of surgical delay. At present, there is no clear international consensus on the safest timing for surgery in this group of patients. Some hospitals routinely wait 24-48 hours, while other hospitals have a protocol that allows to operate within 24 hours based on expert-opinion. This variation in clinical practice across hospitals in the Netherlands creates a unique opportunity to study the question as a natural experiment. This multicenter cohort study will compare outcomes between patients who undergo hip fracture surgery within 24 hours of emergency department presentation ("early surgery") and those who undergo surgery after 24 hours ("delayed surgery"). The primary focus is whether early surgery leads to a higher risk of requiring a blood transfusion after the operation. Secondary analyses will evaluate perioperative hemoglobin changes, postoperative complications (including infections, delirium, and thromboembolic events), hospital length of stay, mortality rates, and functional recovery at three months. Data Collection and Quality This study relies on clinical data routinely documented in the electronic health record (EHR). No additional procedures or interventions are carried out. Before the start of the study, a standardized preparation phase was implemented across all participating hospitals to ensure that every variable required for analysis would be registered consistently and completely in the EHR. To achieve this, variables that are frequently underreported were identified, including the timing and indication of the last DOAC dose, nutritional and frailty assessments, and reasons for surgical delay, and actively addressed them in advance. Posters and concise checklists were distributed to all relevant clinical teams (emergency medicine, anesthesiology, geriatrics, surgery, and nursing) to raise awareness and provide clear reminders. By embedding these prompts into daily clinical practice, a shared understanding among staff was established that these data points are critical for the study and must be documented accurately in the patient record. With this groundwork in place, all eligible patients are flagged locally by the principal investigator at each site. The central research team then retrospectively extracts the required information directly from the EHR. Data Management and Quality Assurance All study data are obtained directly from the electronic health record (EHR) and subsequently transferred into a secure study database managed by the coordinating center. Prior to study initiation, a study-specific data dictionary was developed, clearly defining each variable, its source in the EHR, coding, and units of measurement. Because all variables are documented as part of routine care, the main safeguard for data quality is the preparation phase in which clinical teams were instructed to record these variables in a standardized way. To support this, range and consistency checks (e.g., implausible hemoglobin values or mismatch between transfusion status and transfusion counts) are applied during data extraction and cleaning. Data quality is monitored centrally through multiple checks. Completeness of the dataset is reviewed on a recurring basis, and any ambiguities are resolved in consultation with the local principal investigator. If missing or inconsistent data are detected, the hospital concerned will be informed and instructed to improve its documentation procedures to prevent further gaps. In addition, a random sample of records will undergo source data verification against the EHR to confirm accuracy. All data management procedures follow Good Clinical Practice (GCP) principles and comply with institutional and national privacy regulations; direct identifiers remain within the local hospital, and only coded datasets are used for central analysis. Sample Size and Rationale for Non-Inferiority This study is designed as a non-inferiority prospective cohort in the form of a natural experiment. Prior data suggest transfusion rates of \~15% after delayed surgery and \~10% after early surgery in hip fracture patients using DOACs. In consultation with clinical experts, a non-inferiority margin of 5% was set as clinically acceptable. Non-inferiority will be concluded if the upper bound of the one-sided 95% CI for the risk difference does not exceed this margin. With an alpha of 0.05 (one-sided), 90% power, and the above event rates, 374 patients (187 per group) are required. As the calculation is based on the per-protocol population, inclusion will continue until both groups each contain 187 patients treated according to their hospital's surgical timing protocol (\<24h in early-surgery centers, \>24h in delayed-surgery centers). Because intention-to-treat (ITT) analyses include all eligible patients by hospital policy regardless of actual timing, the ITT population is expected to exceed 374. No loss to follow-up is anticipated, as the primary outcome is assessed during hospitalization. Statistical Analysis The primary outcome is the requirement for postoperative blood transfusion (yes/no). Non-inferiority will be tested by calculating the absolute risk difference between groups with a one-sided 95% CI. A multivariable logistic regression will also be performed to estimate the odds ratio (OR) for transfusion, adjusting for confounders including age, sex, DOAC type, and surgical procedure. With an expected 56 events, inclusion of up to five covariates is feasible without overfitting. The primary analysis will follow the per-protocol principle, excluding patients whose actual surgical timing deviated from their hospital's standard policy. An additional ITT analysis will be conducted to preserve the quasi-randomization of the natural experiment and test robustness. Inter-hospital variability will be assessed using the intraclass correlation coefficient; if clustering is present, multilevel logistic regression will be applied. Secondary Outcomes Secondary outcomes will be analyzed according to their measurement scale and distribution. Continuous variables, such as hemoglobin drop, bleeding index, and hospital length of stay, will first be tested for normality and analyzed with t-tests if normally distributed or Mann-Whitney U tests if not. Ordinal or categorical variables with more than two categories, such as the number of transfusion units, will be compared using Chi-square tests. Binary outcomes, including postoperative complications, thromboembolic events, and hemoglobin drop \>2 mmol/L, will be analyzed using Chi-square tests or Fisher's exact test if cell counts are small. Handling of Missing Data The investigators do not anticipate substantial missing data, given the extensive preparatory work undertaken to ensure consistent documentation across all participating hospitals. However, if missing values do occur, they will first be examined for patterns and mechanisms (missing completely at random, at random, or not at random). Where appropriate, multiple imputation by chained equations (MICE) will be applied to account for missing values in key variables. Extreme or implausible values will be flagged and verified against the EHR; if they cannot be corrected, they will be treated as missing. Expected Impact If early surgery proves to be as safe as delayed surgery, the results of this study could support more uniform and evidence-based guidelines. This may reduce unnecessary delays in surgery, improve outcomes, shorten hospital stays, and enhance recovery for thousands of older patients worldwide who sustain a hip fracture while using DOACs.