Introduction

Hip arthroplasty stands as a key surgical intervention for end-stage hip diseases, frequently employed for conditions like osteoarthritis, osteonecrosis of the femoral head, hip dysplasia, and hip fractures.¹ The ageing population and technological advancements have led to a consistent annual increase in hip arthroplasty procedures, with a particularly notable rise among elderly patients.² This group, characterized by distinct physiological traits and disease profiles, is more prone to adverse outcomes following surgery.³ Elderly patients frequently have multiple comorbidities, such as cardiovascular disease, diabetes, and osteoporosis, which markedly increase the risk of adverse postoperative outcomes.⁴ As ageing progresses, their physiological and functional reserves decline, along with a significant reduction in immune function, resulting in decreased tolerance to surgical trauma. This predisposition raises the likelihood of severe complications, including infection, deep vein thrombosis, and cardiovascular events.⁵ Moreover, diminished postoperative rehabilitation capacity, coupled with insufficient social and psychological support, further obstructs the functional recovery of elderly patients and adversely affects their long-term prognosis.⁶ Elderly patients are at a significantly higher risk of complications following hip arthroplasty than the general population, with studies reporting incidence rates of up to 20% to 40%.⁵ These complications considerably extend hospital stays, escalate healthcare costs, and heighten the risk of postoperative mortality.⁷ Furthermore, complications can cause delays in recovery, lower quality of life, and long-term negative impacts on physical and mental health.⁸ Therefore, preventing postoperative complications is essential for optimal postoperative management. A precise and practical risk assessment of complications in surgical patients is a critical step in implementing effective interventions.⁹

Research indicates a strong correlation between postoperative complications and factors such as demographics, comorbidities, and substance use.¹⁰ When combined, these factors may help predict the risk of postoperative complications.^11,12 The findings of Tanaka et al demonstrate that muscle strength (assessed via grip strength monitoring) and physical performance (evaluated through gait speed measurement) serve as valid indicators for evaluating functional recovery after hip arthroplasty.¹³ Several risk prediction models for surgical complications have been developed for preoperative counselling in first-time hip arthroplasty patients,^14–18 but only a limited number of studies have externally validated these models.^19,20 External validation is essential to evaluate the generalizability of predictive models, as differences in patient demographics and healthcare environments can significantly influence model performance.²¹ Research has shown that models for predicting complications after hip arthroplasty often lack sufficient predictive accuracy in new patient groups and varying healthcare settings, limiting their clinical utility.^13,20 This inadequacy underscores the necessity of creating more accurate, robust, and generalizable risk prediction tools for patients undergoing hip arthroplasty.

Significant progress has been made in predicting the risk of complications after hip arthroplasty. For instance, risk prediction models have been developed for specific complications, such as deep vein thrombosis (DVT),^22,23 postoperative infection,^24,25 and dislocation.^26,27 Much of the current research centres on postoperative complications, with comparatively less exploration of broader adverse outcomes, such as extended hospital stays, ICU admissions, and mortality, presenting considerable opportunities for innovation in this field.^28,29 This gap may result in insufficient assessment of overall patient risk, potentially compromising the comprehensiveness of clinical decision-making.²⁶ Investigating adverse postoperative outcomes, rather than isolated complications, allows for a more holistic evaluation of patient risk and offers a scientific basis for developing individualized perioperative interventions.³⁰ Thus, it is necessary to explore risk factors associated with adverse outcomes post-hip arthroplasty and establish predictive models.

Material and Methods

The TRIPOD checklist was followed to guide the reporting of this study.

Study Design and Participants

This case-control study retrospectively included patients who underwent hip arthroplasty in the Department of Orthopaedics at a tertiary hospital in Zhejiang Province from January 1, 2021 to December 31, 2023. During or after data collection, the authors could obtain information that could identify individual participants from January 1, 2021 to December 31, 2023.

There were 620 cases in the modelling group. The inclusion criteria were: (1) age ≥ 65 years; (2) meeting the indications for hip arthroplasty and being a first-time unilateral hip arthroplasty recipient; (3) no significant adverse events before admission, such as shock, altered consciousness, acute respiratory failure, or acute heart failure. The exclusion criteria were: (1) patients diagnosed with deep vein thrombosis (DVT) or pulmonary embolism (PE) before surgery or with embolism at other arterial or venous sites; (2) those with severe vital organ dysfunction; (3) patients with mental disorders; (4) patients with incomplete or missing medical records, laboratory data, or imaging results. The enrolled cases were divided into the No Adverse Outcomes Group and the Adverse Outcomes Group, based on whether patients experienced adverse outcomes. Adverse outcomes were defined as at least one of the following: surgical site infection, deep vein thrombosis (DVT), pulmonary embolism, prolonged hospital stay, transfer to the intensive care unit (ICU), or all-cause mortality. The sample size was calculated using the event/variable method (EPV),³¹ suggesting that the prediction model’s sample size should be 10 to 20 times the number of predictors. This study included 33 predictors, requiring a sample size of 330 to 660 cases. In total, 620 hospitalized elderly patients were included in the survey, with 212 in the Adverse Outcomes group and 408 in the No Adverse Outcomes group. The selection of patients in the modeling group is shown in Figure 1.

Figure 1 Flowchart of patient selection.

Additionally, 264 post-hip arthroplasty patients from another tertiary care hospital’s orthopaedic department were selected as the external validation group between January 1, 2024 and December 31, 2024.³² During or after data collection, the authors could obtain information that could identify individual participants from January 1, 2024 to December 31, 2024.

Candidate Predictors

In this study, following a literature review, nine orthopaedic physicians and six orthopaedic specialist nurses, each with intermediate or higher professional titles, were invited to provide expert evidence.³³ This collaborative effort led to the establishment of 33 candidate predictors for adverse outcomes, categorized as (1) Demographic data: age, sexuality, BMI; (2) Patient history: smoking, drinking, hypertension, diabetes, craniocerebral disease, cardiac disease, malignancy, osteoporosis; (3) laboratory indexes: red blood cell (RBC), haemoglobin (HB), platelets (PLT), albumin (ALB), C-reactive protein (CRP), glutamate aminotransferase (ALT), glutamate oxaloacetate transaminase (AST), direct bilirubin (DBil), indirect bilirubin (IBil), international normalized ratio (INR), D-dimer (D-D), prothrombin time (PT), activated partial thromboplastin time (APTT); (4) Surgical factors: surgical procedure, type of anaesthesia, ASA grade, operation duration, intraoperative blood loss, intraoperative blood transfusion, changeover surgery status (consecutive hip arthroplasty procedures performed by the same surgical team within a turnover interval ≤30 minutes); and (5) postoperative indicators: postoperative pain scores and the use of drains.

Adverse Outcomes

Adverse outcomes include surgical site infection, deep vein thrombosis (DVT), pulmonary embolism, prolonged hospital stay, ICU transfer, and in-hospital death.^34,35 Variations in surgical procedures and care across regions, patient populations, and healthcare settings have led to differing definitions of prolonged length of stay (PLOS).^6,36 In this study, PLOS is defined as the 75th percentile of length of stay among all patients, which is 22 days.³⁷ In-hospital death is defined as all-cause death during hospitalization, while ICU transfer is defined as any transfer (including injury-related).³⁸

Data Collection

The five researchers responsible for data collection from the electronic case system received standardized data collection training from the project leader. The researchers subsequently retrieved demographic data, patient history, laboratory indexes, surgical factors, and postoperative indicators from the hospital’s electronic medical record system and documented them on a pre-designed form. The data were further reviewed by another researcher to identify missing items, duplicates, and other anomalies, ensuring the completeness and accuracy of the information.

Ethical Considerations

Approval for the study was granted by the Ethics Committee (reference number: 20241025–4). As the study was retrospective and posed minimal risk to participants, the requirement for mandatory informed consent was waived.

Statistical Methods

Statistical analysis was conducted using SPSS (version 25.0, IBM Corporation, USA) and R (version 4.2.2, R Foundation for Statistical Computing). The Shapiro–Wilk (SW) test was performed to assess normality distribution. For count data, frequency and percentage were calculated, and the chi-square test was applied for intergroup comparisons. Normally distributed measurement data were presented as mean±standard deviation, and the t-test was used for intergroup comparisons. Non-normally distributed measurement data were described by the median and interquartile range, with the rank-sum test used for intergroup comparisons. Multifactorial logistic regression was then performed to construct the prediction model. The receiver operating characteristic (ROC) curve was plotted, and the area under the curve (AUC) was calculated to evaluate the model’s discriminatory ability. The calibration of the model was assessed using the Hosmer-Lemeshow test. Finally, the practical applicability of the model was evaluated by calculating sensitivity, specificity, and accuracy using clinical application data.

Results

Demographic Data

A total of 884 valid cases were collected, including 620 in the modelling and 264 in the validation groups. In the modelling group, there were 264 males (42.58%) and 356 females (57.42%); aged 65–98 (74.79 SD 7.39) years. In the validation group, there were 108 males (40.91%) and 156 females (59.09%); aged 65–95 (74.37 SD 6.32) years.

Single-Factor Analysis of Postoperative Adverse Outcomes in Model Group

The 620 patients in the modelling group were divided into the no adverse outcomes group and the adverse outcomes group, with 212 cases in the adverse outcomes group and the incidence of adverse outcomes was 34.19%. After single factor analysis, 21 variables were statistically significant (p < 0.05). Those 21 variables included age, smoking, hypertension, diabetes, craniocerebral disease, cardiac disease, malignancy, surgical procedure, HB, ALB, ALT, CRP, D-dimer, PT, APTT, INR, ASA grade, operation duration, intraoperative blood loss, intraoperative blood transfusion, changeover surgery status. The results of the univariate analysis are presented in Table 1. The 21 identified influencing factors were subjected to multicollinearity diagnostics, revealing that the variance inflation factor (VIF) values ranged from 1.044 to 1.423, and tolerance values ranged from 0.703 to 0.958, indicating the absence of multicollinearity among the variables.

Table 1 Risk Factors of Postoperative Adverse Outcomes and Univariate Analysis Results (n = 620)

Prediction Model of Adverse Outcomes Risk After Hip Arthroplasty

Twenty-one variables with statistical significance mentioned above were subjected to binary logistic regression. The results showed that age, malignancy history, surgical procedure, ALB, PT, ASA grade, operation duration and changeover surgery status were associated with adverse outcomes, as shown in Table 2. The model was constructed as follows: logistic (P) = 0.692 + 0.056× age + 1.200×1(malignancy history) - 0.749×1(surgical procedure) - 0.350×ALB + 0.169×PT + 2.036×1(ASA grade) + 0.041×operation duration + 0.927×1(changeover surgery status). The visualised column line graph is plotted in Figure 2.

Table 2 Multivariate Regression Results of Postoperative Adverse Outcomes (n = 620)

Figure 2 Predictive nomogram for adverse outcomes after Hip arthroplasty in elderly patients.

Prediction Effect of Adverse Outcomes Prediction Model After Internal Verification of Hip Arthroplasty

The Hosmer-Lemeshow test was conducted for the predictive model, showing p = 0.712 and χ2 = 5.418. ROC curve was used to test the sensitivity and specificity of the patient model in predicting actual adverse outcomes. The maximum Jordon index was the best critical value for the predictive model. The AUC for the modelling group was 0.902 (refer to Figure 3). The Youden index is 0.668, with a sensitivity of 0.84 and a specificity of 0.828 (refer to Table 3).

Table 3 Internal Validation Data (n = 620)

Figure 3 Receiver operating characteristic curve of the modeling group for predicting the risk of adverse outcomes after Hip arthroplasty in elderly patients.

External Validation of Adverse Outcomes Prediction Model After Hip Arthroplasty

External validation of the model was performed on 264 patients. Postoperative adverse outcomes are observed in 85 cases, corresponding to an incidence rate of 32.20%. The model predicts 93 cases, accurately identifying 67, achieving a sensitivity of 78.82%. Out of 179 cases without adverse outcomes, the model predicts 171, with 153 correct predictions, resulting in a specificity of 85.47%. The total accuracy of the model was (67 + 153)/ 264 = 83.33%, as shown in Table 4.

Table 4 Clinical Validation Data (n = 264)

Discussions

In this study, data from 884 hip arthroplasty patients are analyzed. Risk factors for adverse outcomes are identified through univariate and multivariate logistic regression analyses. The model’s validity is confirmed through clinical validation, where the Hosmer-Lemeshow test indicates p = 0.712, the ROC curve area is 0.902, and the prediction accuracy reaches 83.33%. These results demonstrate the model’s stability, predictive capability, and clinical utility. Therefore, the model can be used to predict the high-risk group for adverse outcomes after hip arthroplasty and provide a reference for the use of preventive measures.

Serum biomarkers are of great significance for predicting and diagnosing complications after hip arthroplasty. The findings of Moldovan’s study demonstrate that the systemic inflammation index (SII) and the neutrophil lymphocyte ratio (NLR) are the most accurate biomarkers for diagnosing periprosthetic joint infection (PJI) following THA.³⁹ The research by Watanabe et al demonstrated that high serum homocysteine has a significant impact on the incidence of intraoperative periprosthetic fracture.⁴⁰ Lin et al’s study demonstrates that serum levels of total cholesterol (TG), triglyceride (TC), and low-density lipoprotein (LDL-C) are risk factors for postoperative delirium (POD) after hip arthroplasty, whereas high-density lipoprotein (HDL-C) concentration acts as a protective factor.⁴¹

This study identifies ASA grade as a significant risk factor for adverse outcomes post-hip arthroplasty, corroborating the findings of Kılınç and Pazarcı.⁴² ASA grade is a key indicator of a patient’s general health and surgical tolerance, where higher grades often reflect more comorbidities or reduced physiological reserves, thereby heightening the risk of adverse outcomes.⁴³ Additionally, higher ASA grades correlate with increased anaesthetic and surgical risks during the perioperative period, further raising the probability of complications.⁴⁴ Consequently, enhanced perioperative management is crucial for patients with elevated ASA classifications. This includes comprehensive preoperative optimization, tailored anaesthesia and surgical plans, and rigorous postoperative monitoring and early intervention to reduce risks and enhance patient outcomes.⁴⁵

The results of this study showed that a history of malignancy was a risk factor for adverse outcomes after hip arthroplasty, which is similar to the findings of Karam et al.⁴⁶ The study concluded that patients with malignant tumours are usually accompanied by immunocompromised, malnourished, and abnormal coagulation and may have a weakened ability to recover from the organism due to having received radiotherapy, chemotherapy, or other anti-tumour treatments. These factors may increase the risk of postoperative infection, deep vein thrombosis, and other complications.⁴⁷ In addition, the postoperative recovery process of patients with malignant tumours may be potentially affected by tumour recurrence or metastasis, further exacerbating the likelihood of adverse outcomes.⁴⁸ Therefore, in clinical practice, patients with a history of malignancy should be more comprehensively assessed and prepared during the perioperative period, including optimisation of preoperative nutritional support, strengthening of postoperative anticoagulation and infection prevention and control measures, and development of an individualised follow-up and recovery plan, in order to minimise the risk of adverse outcomes and to improve the patient’s postoperative quality of life.⁴⁹

Our data suggest that albumin is a protective factor, which aligns with the findings of Kishawi et al.⁵⁰ This may be attributed to the fact that albumin levels reflect the nutritional status and immune function of the patient, with higher albumin levels contributing to improved disease resistance and tissue repair.⁵¹ Preoperative albumin has a strong potential as a valuable indicator for assessing patients’ overall health.⁵² Therefore, optimizing patients’ nutritional status preoperatively, such as by increasing albumin levels through nutritional support, maybe a key intervention to reduce the risk of adverse outcomes in clinical practice.

The results of this study indicate that age, prothrombin time (PT), and operation duration were all risk factors for adverse outcomes after hip arthroplasty. Elderly patients are at a significantly increased risk of postoperative complications due to the presence of more comorbidities, reduced physical function and decreased tissue repair capacity.¹⁹ Prolonged PT may reflect abnormal coagulation or underlying hepatic impairment, which may increase the probability of postoperative complications such as bleeding, haematomas and infections.⁵³ Prolonged surgical time is usually associated with increased surgical complexity and prolonged exposure time, which may lead to an increased risk of infection and increased physiological stress on the patient.⁵⁴ In addition, this study identifies whether the surgery is a changeover procedure as an independent risk factor for adverse outcomes after hip arthroplasty. However, this finding has not been adequately reported in previous studies. Therefore, this study highlights the potential significance of changeover procedures in postoperative risk assessment. Changeover procedures may increase the risk of postoperative infections, tissue damage, and other complications due to factors such as the complexity of operating room resource allocation, increased fatigue of the surgical team, and extended sterile operating time.⁵⁵ Further research into the mechanisms linking changeover procedures to postoperative adverse outcomes will not only help optimize surgical scheduling and resource utilization but also provide new perspectives for clinical perioperative management.⁵⁶ In addition, this study found that choosing total hip arthroplasty (THA) reduced the risk of postoperative adverse outcomes compared with hemiarthroplasty (HA). However, other studies have shown no significant differences in postoperative complication rates between THA and HA patients.^57,58 This may be attributed to the heterogeneity of the study populations, including variations in patients’ underlying health conditions, surgical indications, and the experience of the surgical teams. Therefore, more high-quality studies are needed in the future to investigate further the long-term effects of different surgical modalities on adverse outcomes after hip arthroplasty in order to provide a more reliable and evidence-based basis for the choice of surgical modalities.

Compared to similar studies, this study has several strengths, although it also has some limitations. As the data were collected retrospectively, there may be information bias and potential confounders. Furthermore, while the modelling data were obtained from a single centre, the external validation of the model used an independent dataset from another hospital, partially assessing the model’s applicability in a cross-centre setting. Although the validation process enhances the model’s reliability, its generalizability to other healthcare settings requires further validation through additional studies, as the modelling data remain limited to a single centre.Additionally, since this study focused solely on adverse outcomes during hospitalization and did not include follow-up, it was not possible to assess the risks associated with long-term prognosis after patient discharge. Future studies could incorporate postoperative follow-up data to provide a more comprehensive assessment of the risk of adverse outcomes after hip arthroplasty, thereby further enhancing the clinical value and applicability of the mode.

Conclusions

In conclusion, this study found that independent risk factors for adverse outcomes after hip arthroplasty included age, malignancy history, PT, ASA grade, operation duration and changeover surgery status, whereas surgical procedure and ALB were protective factors. The nomogram for predicting the risk of adverse outcomes has good consistency and differentiation in predicting the risk and thus can help clinical practitioners assess adverse outcomes after hip arthroplasty with good clinical application.