A predictive model integrating three-dimensional displacement for osteonecrosis of femoral head following femoral neck fractures: a retrospective cohort study Authors: Kai Zhao, Jian Wang, Haifeng Li Published: 2026 ABSTRACT BackgroundFracture displacement is a recognized risk factor for osteonecrosis of the femoral head (ONFH) after femoral neck fractures (FNFs), yet the relationship between ONFH and three-dimensional (3D) displacement has not been clearly quantified. This study aims to develop a computed tomography-based method to measure 3D displacement at the FNF site and to build a predictive model that integrates these parameters.MethodsWe retrospectively analyzed 302 FNF patients treated with closed reduction and internal fixation within 3 years. Univariate analysis and multivariable logistic regression were used to assess the association between ONFH and multiple variables, including 3D displacement. The prediction model for ONFH was evaluated using receiver operating characteristic curve, a nomogram, calibration curve, and decision curve analysis (DCA).ResultsDuring the 3-year follow-up period, ONFH was identified in 68 cases (incidence rate 22.5%). Univariate analysis showed that Garden classification, age, time to surgery, osteoporosis, comminution, fixation removal, and displacement parameters (vertical, separation, and comprehensive) were significantly associated with ONFH (P<0.05). Multivariable logistic analysis confirmed that vertical displacement, separation displacement, age, and time to surgery were independent predictors [P<0.05; odds ratio (95% confidence interval): 1.20 (1.01–1.42), 1.19 (1.08–1.30), 0.92 (0.88–0.95), 1.87 (1.41–2.49), respectively]. The model showed an area under the curve >0.91, sensitivity >81%, specificity >89%, and a Youden index of 0.769. Calibration curves showed good agreement between predicted and actual probabilities (Hosmer-Lemeshow test, P>0.05), while DCA indicated clinical usefulness within a risk threshold of 0.10–0.90. In the nomogram, separation displacement contributed the highest score.ConclusionsONFH is related to factors such as Garden classification, age, time to surgery, osteoporosis, comminuted status, fixation removal, and vertical, separation, and comprehensive displacement. Among these, age, time to surgery, and vertical and separation displacement are independent risk factors. The predictive model that incorporates 3D displacement can quantify the risk of postoperative ONFH and shows good predictive ability in FNF patients. However, further external validation is still required. Introduction Femoral neck fractures (FNFs) account for about 54% of hip fractures and 3.58% of all fractures, and their incidence has been increasing over time (1,2). The optimal treatment strategy is still debated in clinical practice. Although previous studies report a 35% revision surgery rate after closed reduction and internal fixation (CRIF), this approach is still commonly used for patients under 70 years old and for most non-displaced fractures (3-6). Due to the special anatomical structure and biomechanical features of the femoral neck, osteonecrosis of the femoral head (ONFH) after internal fixation has long been a key concern in clinical research. Its incidence ranges from 20% to 37.9%, and about 95% of cases occur within 3 years (7-11). This condition may lead to hip dysfunction, pain, disability, and psychological burden, and it is associated with high disability and mortality rates. As a result, many patients with ONFH require revision surgery. Given these serious outcomes, many studies have explored the clinical features of FNF patients in order to predict the risk of postoperative ONFH (9-12). Previous studies have clearly shown that fracture displacement is an important risk factor for ONFH (6-10). Traditional assessment of fracture displacement mainly relies on two-dimensional (2D) imaging such as X-ray films. These methods are limited because they mainly reflect angular displacement and cannot provide precise quantitative data for the fracture site. For example, the commonly used Garden and Arbeitsgemeinschaft für Osteosynthesefragen (AO) classification systems are effective in describing angular displacement and in classifying FNFs into displaced and non-displaced types based on 2D images (13-15). However, there are clear limitations with these systems. They lack precise quantification and do not provide clear criteria to distinguish between different fracture types, which can lead to subjective and less reliable classification (16). Computed tomography (CT) offers a more accurate way to measure displacement. Wu et al. (17,18) recently proposed a method based on three-dimensional (3D) CT reconstruction using Mimics software to evaluate spatial displacement in FNFs. Although this method is a useful step forward, it still focuses on angular displacement rather than true 3D displacement and requires bilateral hip CT scans. In reality, 3D displacement at the FNF site includes three components: vertical displacement in the Z-axis (up/down), anterior-posterior displacement in the X-axis (forward/backward), and separation displacement in the Y-axis (left/right). There is still no multi-factor prediction model that includes 3D displacement to quantify the risk of ONFH in FNF patients. Therefore, despite existing progress, an important gap remains. To address this issue, we developed a CT-based quantitative method to measure 3D displacement, including vertical, anterior-posterior, and separation displacement at the fracture site. The aim of this study is to (I) introduce this 3D quantitative measurement method for FNFs, and (II) build a predictive model that incorporates 3D displacement into the analysis of ONFH risk factors. This approach is expected to provide a more accurate and reliable assessment of ONFH risk, help improve patient outcomes, and support clinical decision-making. We present this article in accordance with the TRIPOD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2026-1-0108/rc). Methods Patient selection The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This retrospective study was approved by the ethics committee of Wuxi Ninth People’s Hospital Affiliated to Soochow University (No. KS2025-138-01; January 9, 2026). Written informed consent was waived due to the retrospective design and the use of de-identified data. According to previous reports, the sample size required for receiver operating characteristic (ROC) curve analysis is 15 times the number of observed indicators (19). Since we included 20 variables, a total of 302 participants were enrolled. Inclusion criteria: (I) traumatic intracapsular FNF patients who underwent CRIF in Wuxi Ninth People’s Hospital from January 2017 to March 2022; (II) age >16 years; (III) follow-up for more than 3 years; and (IV) available preoperative anterior-posterior (AP)/lateral X-ray and hip joint CT scans. Exclusion criteria: (I) pathological or old fractures; (II) pre-existing hip arthritis or immune diseases; (III) postoperative hormone use, alcoholism, smoking, or diabetes; (IV) presence of malignant tumors; (V) bilateral fractures; (VI) postoperative nonunion; and (VII) incomplete follow-up data. A total of 367 patients were screened, and 302 met the criteria and were included in the study (Figure 1). ONFH was diagnosed based on the updated 2019 ONFH grading system (20). Flowchart for patient inclusion. Data collection and processing General and clinical data were collected, organized, entered, and checked by two experienced and trained doctors. When there was disagreement between the two observers on count variables, a final decision was made through discussion. For continuous variables, the average of the two measurements was used. Statistical analysis was then performed on the dataset. Measurement of 3D displacement of FNF site The 3D displacement of the FNF site was measured using CT reconstruction on the AW Server 4.7 image 3D post-processing workstation (General Electric, GE Healthcare, Connecticut, USA) with CT (GE Optima CT 660, GE Healthcare). Measurement protocol: (I) the 3D reconstructed hip (Figure 2A) on the affected side was rotated by 15° to remove the anteversion angle of the femoral neck, which allowed us to obtain the anterior-posterior view of the FNFs (Figure 2B). The built-in ruler in the imaging system was then used to measure the maximal vertical displacement at the fracture site. (II) The 3D reconstructed hip was further rotated to obtain the head-foot view of the FNFs. Based on this view, the maximal anterior-posterior and separation displacements were measured (Figure 2C,2D). The intraclass correlation coefficients (ICCs) for inter-observer and intra-observer reliability of the 3D measurements were 0.963, 95% confidence interval (CI): 0.949–0.974, and 0.962, 95% CI: 0.945–0.975, respectively (21). The schematic diagram of the measurement protocol for 3D displacement of FNFs. (A) Anterior-posterior view of the 3D reconstruction of the hips. (B) Anterior-posterior view of FNFs, used to measure vertical displacement. (C,D) Head-foot view of FNFs, used to measure anterior-posterior and separation displacements. Figure 2A,2B was adapted from (21) under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/). Observation variables The study included a range of variables, such as age, gender, body mass index (BMI), injury cause, Garden classification, Pauwels classification, fracture position, traction, osteoporosis, Charlson comorbidity index (CCI), comminuted status, 3D displacement, time to surgery, fixators, femoral neck shortening, malposition, and fixation removal. BMI was calculated as weight (kg) divided by height (m) squared and recorded after admission. Fracture position included subcapital and transcervical fractures. Injury cause was classified as high-energy or low-energy. Simplified Garden classification was assessed using preoperative X-rays and categorized FNFs into displaced and non-displaced types (16). Pauwels classification is used to assess the fracture line and is divided into three types (22). We used a modified method to measure the angle between the fracture line and the vertical axis of the femoral neck (23). Comminuted status was defined as more than three fracture fragments at the fracture site. 3D displacement and comprehensive displacement. 3D displacement included vertical, anterior-posterior, and separation. In addition, we measured comprehensive displacement, which was defined as the sum of vertical, anterior-posterior, and separation displacements. CCI is a method used to predict mortality by classifying or weighting comorbidities (24). Osteoporosis was diagnosed when the bone mineral density T-score was less than 2.5, based on dual-emission X-ray absorptiometry. Time to surgery was defined as the interval between injury and surgery. Fixators included three types of internal implants: cannulated screws, percutaneous compression plating (PCCP) (25), and femoral neck system (FNS). Malposition referred to non-anatomic reduction after surgery, which was indicated by an irregular S-shaped Lowell’s curve step formed by the convex femoral head and concave femoral neck (8,11). Fixation removal referred to the removal of implants when ONFH occurred. Neck shortening was evaluated by comparing anterior-posterior X-rays of the injured hip with those of the contralateral side during fracture healing. Traction referred to preoperative bone or skin traction. Statistical processing Statistical analysis was performed using SPSS Statistics for Windows, version 21.0 (IBM Corp., Armonk, NY, USA). Continuous variables with a normal distribution were expressed as mean ± standard deviation, and comparisons between two groups were made using t-tests. Count variables were presented as frequencies and percentages, and comparisons were performed using Chi-squared tests. ONFH was used as the dependent variable, and variables with statistical significance in the univariate analysis were included as independent variables in the multivariable logistic regression to identify independent risk factors for ONFH (8,12,26). For further analysis, the R Studio (4.0.2) packages “rms” and “rmda” were used. The dataset was randomly divided into a training set and a validation set at a ratio of 7:3. The training set was used to build the logistic regression model and nomogram, while the validation set was used to test the model using ROC curve analysis, calibration curve analysis, and decision curve analysis (DCA). All tests were two-sided, and statistical significance was set at α =0.05. Results The balance check result The balance check showed that there was no significant difference between the training and validation sets, except for malposition (Table 1). Table 1Balance check between the training and validation setsVariablesTotal (n=302)Test (n=91)Train (n=211)Statistic (t/χ2)P valueVertical displacement (mm)5.32±3.645.86±4.165.09±3.37t=1.550.123Separation displacement (mm)6.61±5.636.99±5.636.45±5.63t=0.770.442Anterior-posterior displacement (mm)2.15±1.842.21±1.922.12±1.80t=0.390.696Comprehensive displacement (mm)14.09±9.1114.97±9.6613.71±8.86t=1.100.271Age (years)52.27±13.2351.49±14.8652.61±12.48t=−0.630.533BMI (kg/m2)21.88±2.3122.03±2.3221.82±2.31t=0.730.463CCI1.48±1.911.29±1.821.57±1.95t=−1.180.238Time to surgery (days)5.42±1.695.43±1.625.42±1.72t=0.050.957Neck shortening (mm)3.11±2.053.23±1.943.06±2.09t=0.680.499Malposition (mm)0.45±0.940.27±0.750.53±1.00t=−2.430.016Fracture positionχ2=0.090.766 1 (subcapital)140 (46.36)41 (45.05)99 (46.92) 2 (transcervical)162 (53.64)50 (54.95)112 (53.08)Garden classificationχ2=0.010.907 1 (nondisplaced)71 (23.51)21 (23.08)50 (23.70) 2 (displaced)231 (76.49)70 (76.92)161 (76.30)Genderχ2=0.020.885 1 (man)128 (42.38)38 (41.76)90 (42.65) 2 (woman)174 (57.62)53 (58.24)121 (57.35)Injury causeχ2=0.070.793 1 (low energy)126 (41.72)39 (42.86)87 (41.23) 2 (high energy)176 (58.28)52 (57.14)124 (58.77)Pauwels classificationχ2=1.080.583 1 (type I)115 (38.08)33 (36.26)82 (38.86) 2 (type II)140 (46.36)46 (50.55)94 (44.55) 3 (type III)47 (15.56)12 (13.19)35 (16.59)Osteoporosisχ2=0.190.667 1 (no)78 (25.83)22 (24.18)56 (26.54) 2 (yes)224 (74.17)69 (75.82)155 (73.46)Comminutedχ2=0.860.354 1 (no)218 (72.19)69 (75.82)149 (70.62) 2 (yes)84 (27.81)22 (24.18)62 (29.38)Fixatorsχ2=1.080.582 1 (cannulated screw)154 (50.99)50 (54.95)104 (49.29) 2 (PCCP)81 (26.82)21 (23.08)60 (28.44) 3 (FNS)67 (22.19)20 (21.98)47 (22.27)Fixator removalχ2=0.130.716 1 (no)56 (18.54)18 (19.78)38 (18.01) 2 (yes)246 (81.46)73 (80.22)173 (81.99)Tractionχ2=0.070.792 1 (yes)44 (14.57)14 (15.38)30 (14.22) 2 (no)258 (85.43)77 (84.62)181 (85.78)Data are presented as mean ± standard deviation or n (%). BMI, body mass index; CCI, Charlson comorbidity index; FNS, femoral neck system; PCCP, percutaneous compression plating. Balance check between the training and validation sets Data are presented as mean ± standard deviation or n (%). BMI, body mass index; CCI, Charlson comorbidity index; FNS, femoral neck system; PCCP, percutaneous compression plating. Univariate analysis During the 3-year follow-up period, ONFH was identified in 68 cases, with an incidence rate of 22.5%. Univariate analysis showed that ONFH was significantly related to several variables, including Garden classification, age, time to surgery, osteoporosis, comminuted status, fixation removal, vertical displacement, separation displacement, and comprehensive displacement (P<0.05). In contrast, fracture position, gender, injury cause, Pauwels classification, fixators, traction, anterior-posterior displacement, BMI, CCI, neck shortening, and malposition were not significantly related to ONFH (P>0.05) (Table 2). Table 2Univariate analysis of ONFH after internal fixation for FNF in the training setVariablesβSEZPOR (95% CI)Fracture position 1 (subcapital)1.00 (reference) 2 (transcervical)−0.270.34−0.800.4250.76 (0.39–1.48)Garden classification 1 (nondisplaced)1.00 (reference) 2 (displaced)1.560.55−2.840.0040.21 (0.07–0.61)Gender 1 (man)1.00 (reference) 2 (woman)−0.140.34−0.420.6730.87 (0.44–1.69)Injury cause 1 (low energy)1.00 (reference) 2 (high energy)−0.100.34−0.300.7680.90 (0.46–1.77)Pauwels classification 1 (type I)1.00 (reference) 2 (type II)0.120.380.320.7481.13 (0.53–2.40) 3 (type III)0.580.471.230.2181.79 (0.71–4.50)Osteoporosis 1 (no)1.00 (reference) 2 (yes)1.230.502.450.0143.43 (1.28–9.20)Comminuted 1 (no)1.00 (reference) 2 (yes)0.930.352.700.0072.54 (1.29–5.00)Traction 1 (no)1.00 (reference) 2 (yes)−0.770.43−1.790.0740.46 (0.20–1.08)Fixators 1 (cannulated screw)1.00 (reference) 2 (PCCP)−0.120.41−0.290.7720.89 (0.39–2.00) 3 (FNS)0.300.410.730.4631.36 (0.60–3.05)Fixation removal 1 (no)1.00 (reference) 2 (yes)1.490.36−4.19<0.0010.23 (0.11–0.45)Vertical displacement0.320.065.04<0.0011.38 (1.22–1.57)Separation displacement0.190.045.21<0.0011.21 (1.13–1.30)Anterior-posterior displacement0.080.090.920.3581.09 (0.91–1.30)Comprehensive displacement0.140.035.41<0.0011.15 (1.09–1.21)Age−0.050.01−3.55<0.0010.95 (0.92–0.98)BMI0.050.070.660.5121.05 (0.91–1.21)CCI−0.000.09−0.000.9981.00 (0.84–1.19)Time to surgery0.560.114.91<0.0011.75 (1.40–2.19)Neck shortening−0.040.08−0.530.5980.96 (0.81–1.13)Malposition0.110.160.690.4911.12 (0.81–1.54)BMI, body mass index; CCI, Charlson comorbidity index; CI, confidence interval; FNF, femoral neck fracture; FNS, femoral neck system; ONFH, osteonecrosis of the femoral head; OR, odds ratio; PCCP, percutaneous compression plating; SE, standard error. Univariate analysis of ONFH after internal fixation for FNF in the training set BMI, body mass index; CCI, Charlson comorbidity index; CI, confidence interval; FNF, femoral neck fracture; FNS, femoral neck system; ONFH, osteonecrosis of the femoral head; OR, odds ratio; PCCP, percutaneous compression plating; SE, standard error. Multicollinearity diagnosis and logistic regression We first checked whether collinearity existed between variables. The Pearson correlation coefficient between comprehensive displacement and vertical displacement or separation displacement was >0.8. Comprehensive displacement also showed the lowest tolerance (<0.1) and the highest variance inflation factor (>10), indicating clear collinearity. After adjusting for other variables in the multivariable logistic regression model, vertical displacement, separation displacement, age, and time to surgery remained independently associated with ONFH (all P<0.05). The results showed that the risk of ONFH increased with greater vertical displacement, greater separation displacement, and a longer time from injury to surgery, while it decreased with increasing age. Specifically, each unit increase in vertical displacement, separation displacement, and time to surgery was associated with a 20%, 19%, and 87% increase in the odds of ONFH. In contrast, each additional year of age was associated with a 9.2% decrease in the odds. Other variables, including Garden classification, osteoporosis, injury cause, and comprehensive displacement, were not identified as independent risk factors for ONFH (Table 3). The regression equation from the model is as follows: Z = −3.09 + 0.18 × vertical displacement + 0.17 × separation displacement − 0.09 × age + 0.63 × time to surgery. Table 3Logistic regression result of multi risk factorsVariablesβSEZPOR (95% CI)Intercept−3.091.26−2.450.0140.05 (0.00–0.54)Vertical displacement0.180.092.060.0391.20 (1.01–1.42)Separation displacement0.170.053.56<0.0011.19 (1.08–1.30)Age−0.090.02−4.11<0.0010.92 (0.88–0.95)Time to surgery0.630.154.31<0.0011.87 (1.41–2.49)CI, confidence interval; OR, odds ratio; SE, standard error. Logistic regression result of multi risk factors CI, confidence interval; OR, odds ratio; SE, standard error. Nomogram of risk scoring model Based on the variables from the logistic regression results, a nomogram for predicting ONFH was established in Figure 3. It showed that the model provides clear clinical benefit when the risk threshold is between 0.1 and 0.9. In clinical use, a vertical line is drawn from the score of each variable to the “score” axis. The total score is then calculated by summing all points, and a vertical line is drawn down to the “risk of ONFH” axis. The point where it intersects indicates the probability of postoperative ONFH in FNF patients. In the nomogram, separation displacement contributed the highest score. Nomogram of predicting the likelihood of ONFH within 3 years after internal fixation of FNFs. FNF, femoral neck fractures; ONFH, osteonecrosis of the femoral head. Presentation and evaluation of the prediction model The area under the curve (AUC) of the ROC for the prediction model was >0.91, with sensitivity >81%, specificity >89%, and a Youden index of 0.769 (Figure 4 and Table 4). The ROC of the prediction model. AUC, area under the curve; CI, confidence interval; ROC, receiver operating characteristic. Table 4Confusion matrix for the training and validation setsDataAUC (95% CI)Accuracy (95% CI)Sensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)Cut offTrain0.95 (0.92–0.98)0.88 (0.83–0.92)0.88 (0.83–0.93)0.89 (0.79–0.98)0.97 (0.94–1.00)0.66 (0.54–0.78)0.231Test0.91 (0.85–0.97)0.84 (0.74–0.90)0.81 (0.71–0.90)0.92 (0.81–1.00)0.96 (0.92–1.00)0.63 (0.47–0.79)0.231AUC, area under the curve; CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value. Confusion matrix for the training and validation sets AUC, area under the curve; CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value. The calibration curves showed that the predicted probability was close to the actual probability (Figure 5). The Hosmer-Lemeshow goodness of fit (GOF) test showed χ2=14.355, P=0.919 in the training set and χ2=4.7309, P=0.685 in the validation set, indicating good model fit. Calibration curves for the training (A) and validation (B) sets. The DCA curves in Figure 6 showed that within the risk threshold range of 0.10–0.90, the clinical net benefit of using the model was higher than treating all patients or treating none. This suggests that the model provides useful net benefit within this range. DCA curves for the training (A) and validation (B) sets. DCA, decision curve analysis. Discussion ONFH following FNFs is mainly attributed to the disruption of the anastomosis of the lateral epiphyseal vessels, which reduces the blood supply to the femoral head and ultimately leads to avascular necrosis. In this study, among 302 FNF patients treated with minimally invasive internal fixation, the incidence of ONFH was 22.5%, which falls within the reported range of 20% to 37.9% in previous literature (9-12). Our results further show that fracture displacement is positively related to the risk of ONFH. More importantly, this study provides a quantitative link between factors such as 3D displacement and the risk of ONFH. To make these findings easier to use in practice, we developed a nomogram to predict ONFH in FNF patients. The nomogram is simple and easy to use. For example, for a 45-year-old patient with an FNF, a time to surgery of 6 days, a vertical displacement of 12 mm, and a separation displacement of 15 mm, the corresponding scores are 50, 40, 35, and 57. The total score is 182, which corresponds to an estimated 70% risk of ONFH within 3 years after CRIF. This study presents a predictive model that combines several key clinical factors, including quantified 3D displacement at the FNF site, to assess the risk of ONFH. The model shows good predictive performance and can help with early risk assessment. It may also support clinical decisions, such as closer follow-up or preventive measures for patients at high risk. In this study, univariate analysis identified several factors associated with ONFH, including Garden classification, age, time to surgery, osteoporosis, comminuted status, fixation removal, vertical displacement, separation displacement, and comprehensive displacement. However, multivariable logistic regression analysis showed that only vertical displacement, separation displacement, age, and time to surgery were independent risk factors for ONFH. The findings can be explained as follows: 3D displacement: we observed that 3D displacement at the FNF site is more distinctly visualized in targeted 3D reconstructions of the fracture itself compared to standard hip joint reconstructions or scans. Therefore, we measured the maximal 3D displacement values from the 3D reconstruction of the FNFs. Fracture displacement has a clear impact on blood supply to the femoral head and is an important factor related to ONFH (8-10). Our results show that displacement in different directions has different effects on ONFH. Vertical displacement reflects up-down movement and shortening of the lower limb caused by axial force. Anterior-posterior displacement reflects front-back movement caused by limb weight and muscle contraction. Separation displacement reflects external rotation deformity. Compared with traditional 2D imaging, the 3D quantitative measurement method provides a more accurate evaluation of displacement and shows that vertical and separation displacements carry higher risk. The analysis indicates that both vertical and separation displacements are significant risk factors for ONFH. For each unit increase in vertical and separation displacement, the odds of ONFH increase by 20% and 19%, respectively. This shows a clear positive relationship, where larger displacement leads to higher risk. At the same time, we found collinearity between comprehensive displacement and its components, vertical and separation displacement. For this reason, comprehensive displacement was not kept as an independent factor in the final model. 3D displacement: we observed that 3D displacement at the FNF site is more distinctly visualized in targeted 3D reconstructions of the fracture itself compared to standard hip joint reconstructions or scans. Therefore, we measured the maximal 3D displacement values from the 3D reconstruction of the FNFs. Fracture displacement has a clear impact on blood supply to the femoral head and is an important factor related to ONFH (8-10). Our results show that displacement in different directions has different effects on ONFH. Vertical displacement reflects up-down movement and shortening of the lower limb caused by axial force. Anterior-posterior displacement reflects front-back movement caused by limb weight and muscle contraction. Separation displacement reflects external rotation deformity. Compared with traditional 2D imaging, the 3D quantitative measurement method provides a more accurate evaluation of displacement and shows that vertical and separation displacements carry higher risk. The analysis indicates that both vertical and separation displacements are significant risk factors for ONFH. For each unit increase in vertical and separation displacement, the odds of ONFH increase by 20% and 19%, respectively. This shows a clear positive relationship, where larger displacement leads to higher risk. At the same time, we found collinearity between comprehensive displacement and its components, vertical and separation displacement. For this reason, comprehensive displacement was not kept as an independent factor in the final model. Time to surgery: delayed surgery is a well-known risk factor for ONFH (9,10,20,24). In this study, longer time to surgery was significantly associated with higher risk. For each additional day before surgery, the probability of ONFH increased by 87%. Therefore, the American Academy of Orthopaedic Surgeons (AAOS) clinical practice guidelines suggest early CRIF for FNF patients under 70 years old. This helps reduce vascular pressure and improve blood supply, which may lower the risk of ONFH (27). Time to surgery: delayed surgery is a well-known risk factor for ONFH (9,10,20,24). In this study, longer time to surgery was significantly associated with higher risk. For each additional day before surgery, the probability of ONFH increased by 87%. Therefore, the American Academy of Orthopaedic Surgeons (AAOS) clinical practice guidelines suggest early CRIF for FNF patients under 70 years old. This helps reduce vascular pressure and improve blood supply, which may lower the risk of ONFH (27). Age: previous studies have shown that younger patients (<60 years) have a higher risk of ONFH compared with older patients (8-10,26). Our results also support this finding. Age acted as a protective factor, with each additional year reducing the probability of ONFH by 9.2%. This is in line with earlier reports. Age: previous studies have shown that younger patients (<60 years) have a higher risk of ONFH compared with older patients (8-10,26). Our results also support this finding. Age acted as a protective factor, with each additional year reducing the probability of ONFH by 9.2%. This is in line with earlier reports. Other factors: some studies have reported that injury cause, Garden or Pauwels classification, osteoporosis, comminuted status, preoperative traction, fixation removal, and posterior inclination angle of the femoral head (>20°) are related to ONFH (9,10,27-29). It is generally accepted that ONFH occurs more often in high-energy injury and non-osteoporotic FNF patients than in low-energy injury and osteoporotic patients (9). However, in our analysis, these factors were not significant after adjusting for other variables. This may be related to sample differences. Other factors: some studies have reported that injury cause, Garden or Pauwels classification, osteoporosis, comminuted status, preoperative traction, fixation removal, and posterior inclination angle of the femoral head (>20°) are related to ONFH (9,10,27-29). It is generally accepted that ONFH occurs more often in high-energy injury and non-osteoporotic FNF patients than in low-energy injury and osteoporotic patients (9). However, in our analysis, these factors were not significant after adjusting for other variables. This may be related to sample differences. This study developed a predictive model for ONFH that includes 3D displacement. It provides a simple and practical way to assess patient risk. The nomogram allows visual and numerical estimation of multiple factors, which can help guide treatment decisions and improve outcomes. The ROC curve showed an AUC of more than 0.91, with sensitivity above 81%, specificity above 89%, and a Youden index of 0.769. The calibration curves showed good agreement between predicted and observed results, indicating stable model performance. The DCA curves showed that the model provided clinical benefit when the risk threshold ranged from 0.10 to 0.90, which supports its practical value. This study has several limitations. It was conducted at a single center with a relatively small sample size and a follow-up of only three years. Biochemical indicators were not included. Therefore, selection bias, performance bias, detection bias, and reporting bias may exist. In addition, the comprehensive displacement in this article is not the straight-line distance calculated using the Euclidean distance formula in 3D space. Future studies should include multiple centers, larger samples, and more variables to further improve the model and make it more widely applicable. In addition, the model was only internally validated and lacks external validation. The relationship between ONFH and displacement in different directions also needs further investigation, including animal studies. Finally, there may be a risk of overfitting, considering the number of predictors relative to the number of ONFH events. Conclusions Our results show that ONFH is related to factors such as Garden classification, age, time to surgery, osteoporosis, comminuted status, fixation removal, and vertical, separation, and comprehensive displacement. Among these, age, time to surgery, and vertical and separation displacement are independent risk factors, with separation displacement showing the highest risk. The predictive model developed in this study includes 3D displacement of the fracture site and can quantify the risk of postoperative ONFH. This model may help improve current prediction methods and support clinical management in FNF patients. However, further external validation is still needed. Supplementary The article’s supplementary files as