If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Age, CCI, serum albumin, sodium, and hemoglobin are five identified risk variables for mortality after hip fracture surgery in geriatrics.
The novel nomogram established with them for stratifying the mortality risk after hip fracture surgery in geriatrics had good accuracy and usefulness.
External validation is warranted to confirm the performance of this nomogram.
Hip fracture is a significant public health problem, with associated high morbidity and mortality. Orthopedic surgeons are concerned to improve prognosis and stratify mortality risk after hip fracture surgery. This study established a nomogram that combines the Charlson Comorbidity Index (CCI) with specific laboratory parameters to predict mortality risk after hip fracture surgery in geriatrics.
The records of consecutive patients who underwent hip fracture surgery from January 2015 through May 2020 at one medical center were reviewed for perioperative factors and mortality. Patients with age ≥ 70 years who were diagnosed with intertrochanteric or femoral neck fractures were included. Patients who were diagnosed with pathological fracture, received only conservative treatment or lost to follow-up were excluded. A multivariate Cox proportional hazards regression model was used to identify risk factors. A nomogram was established with R software and evaluated using concordance (C)-index, area under receiver operating characteristic (AUC), calibration curves, and decision curve analysis (DCA).
In total, 454 patients were included with a mean age of 81.6 years. The mean follow-up and one-year mortality rate were 37.2 months and 10.4%, respectively. Five identified risk variables for mortality after hip fracture surgery in geriatrics comprised age (HR 1.05, 95% CI 1.01-1.08; P = 0.003), CCI (HR 1.38, 95% CI 1.24-1.54; P = 0.000), albumin (HR 1.78, 95% CI 1.31-2.43; P = 0.000), sodium (HR 1.59, 95% CI 1.18-2.15; P = 0.002) and hemoglobin (HR 1.46, 95% CI 1.07-2.00; P = 0.02). A nomogram was proposed and evaluated, showing a C-index of 0.76 ± 0.02. The AUCs for 6-month, 1-year, and 3-year mortality predictions were 0.83, 0.79, and 0.77, respectively. The calibration curve and DCA showed good discrimination and clinical usefulness.
This novel nomogram for stratifying the mortality risk after hip fracture surgery in geriatrics incorporated age, CCI, serum albumin, sodium, and hemoglobin. Internal validation indicated that the model has good accuracy and usefulness. This nomogram had improved convenience and precision compared with other models. External validation is warranted to confirm its performance.
]. Thus, a major concern of orthopedic surgeons is methods to improve prognosis and stratify and reduce the associated risk after hip fracture surgery.
Many models have incorporated various risk factors to predict the risk of mortality after hip fracture surgery. These risk factors included age, gender, comorbidities, American Society of Anesthesiologists (ASA) score, and other biomarkers. Orthopedic Physiologic and Operative Severity Score for the Enumeration of Mortality and Morbidity (O-POSSUM), Charlson Comorbidity Index (CCI), Nottingham Hip Fracture Score (NHFS), Estimation of Physiologic Ability and Surgical Stress (E-PASS), and others [
] were common models that had been evaluated for utility and accuracy. Although the areas under the receiver operating characteristic curves (ROC; AUCs) of most models were no less than 0.7, few had been greater than 0.8, which suggests poor performance [
]. It may be valuable to explore whether applying laboratory results in models that also utilize the CCI or ASA score could improve the predictive accuracy of the model.
This study constructed a nomogram that combines the CCI score with some special laboratory test results to predict the mortality risk after hip fracture surgery in geriatrics. The nomogram was evaluated at one single center.
Patients and methods
Patients and study design
A cohort of consecutive patients who underwent hip fracture surgery from January 2015 through May 2020 at Peking University First Hospital (Beijing, China) were recruited retrospectively. The inclusion criteria for patients were: (1) patients were diagnosed with intertrochanteric or femoral neck fractures; (2) the ages of the patients were no less than 70 years old; (3) patients received surgery therapy included plate/screw fixation, hemiarthroplasty, or total hip arthroplasty. Patients with any of the following were excluded: pathological fracture; received only conservative treatment; or lost to follow-up.
The initial search for cases of femoral neck or intertrochanteric fractures in our medical database yielded 658 patients (Fig. 1). Among them, 144, 41, and 15 were excluded for being, respectively, younger than 70 years, lost to follow-up, and duplicated for second hip fracture. No patient was diagnosed with pathological fracture. Ultimately, the records of 454 patients were collected for analysis.
Basic perioperative demographic (age, gender) and clinic characteristics were collected through the electronic medical record database. Clinic characteristics included ASA score, CCI, preoperative wait time, length of stay, fracture type, anesthesia type, transfusion, and laboratory test results. Specifically, fracture options according to the inclusion criteria were either femoral neck or intertrochanteric fracture. Anesthesia was either general, or combined spinal epidural. The laboratory tests consisted of the following: albumin, sodium, serum creatinine (SCr), hemoglobin, and neutrophil-to-lymphocyte ratio (NLR).
Serum albumin level was considered normal if >35 g/L. Hypoalbuminemia (low serum albumin) was classified as mild, moderate, or severe (30–35, 25–30, or < 25 g/L, respectively.) A normal sodium level was considered 135–145 mmol/L; hyponatremia was stratified as mild or moderate/severe (130–135, or < 130 mmol/L). The hemoglobin status was classified based on level and gender of the patient, with normal being ≥ 120 g/L and ≥ 110 g/L for men and women, respectively. Anemia (low hemoglobin) was considered mild (men, 120–90 g/L; women, 110–90 g/L), or moderate/severe (both, <90 g/L).
For follow-up, the recruited patients were contacted by telephone. The minimal follow up duration was 9 months after surgery. If the patient had died before the follow-up time, the date of death was recorded.
Categorical variables were recorded as numbers, and analyzed using the chi-squared or Fisher's exact test. Continuous variables were recorded as mean ± standard deviation and analyzed using Student's t-test or the Wilcoxon rank-sum test. Survival curves were depicted using the Kaplan-Meier method. The risk factors for mortality after hip fracture surgery were initially identified using a univariate Cox proportional hazards regression model with P < 0.1. The chosen covariates were further analyzed by backward stepwise multivariate Cox proportional hazards regression, and variables with P < 0.05 were considered as final predictors. SPSS 23 for Windows was used in statistical analyses.
A nomogram was constructed by rms package in R (version 4.1.0) according to the results of the multivariate Cox analysis. The performance of the nomogram was evaluated by concordance index (C-index) and AUC curve over time [
]. (The C-index and AUC directly reflect the accuracy of the model—the C-index and AUC rise with the accuracy). After internal validation by bootstrapping, a calibration curve was used to assess the calibration of the nomogram by comparing the actual risk and predicted risk. Based on the net benefit and threshold probabilities, decision curve analysis (DCA) was used to estimate the clinical usefulness of the nomogram.
Overall, the study population comprised 454 patients with a mean age of 81.6 years (Table 1). Among them, 274 and 180 had diagnosed femoral neck fracture and intertrochanteric fracture, respectively. The mean follow-up time was 37.2 months, and the one-year mortality rate was 10.4% (Fig. 2).
Table 1Demographic and clinic characteristics of the included patients.
The initial screening for the prediction model by univariate Cox regression with P < 0.1 identified several candidate variables including gender, age, CCI, ASA score, transfusion, albumin, sodium, serum creatinine, hemoglobin, and NLR. Multivariate Cox proportional hazards regression identified the following 5 variables that had the strongest association with overall mortality: age; CCI; serum albumin, sodium, and hemoglobin (Table 2).
Table 2Cox proportional hazards regression of variables for mortality after hip fracture surgery in geriatrics.
A new nomogram for predicting the mortality risk after hip fracture surgery in geriatrics was constructed with 5 variables: age, CCI, albumin, sodium, and hemoglobin (Fig. 3). The accuracy of the nomogram was evaluated internally. The C-index of the nomogram was 0.76 ± 0.02. The AUC curve values for predicting mortality at 6 months, 1 year, and 3 years were 0.83, 0.79, and 0.77, respectively (Fig. 4). The C-index and AUC showed that this model had good discrimination and predictive accuracy. There was good consistency between the model's observed and predicted probabilities for overall survival rates at 6 months, 1 year, and 3 years (Fig. 5). The DCA indicated that the nomogram could be a good tool for prediction of overall mortality after hip fracture surgery in geriatric patients (Fig. 6).
Many models have been designed to predict the mortality risk after hip fracture surgery, but their accuracy and convenience are not satisfactory. To improve their shortcomings, the present novel nomogram model combined certain laboratory tests results with the original CCI. The 5 variables identified through Cox regression as reflecting potential risk were age, CCI, serum albumin, sodium, and hemoglobin levels. These factors were incorporated in a new nomogram, and the model was subjected to internal validation; the C-index was 0.76, and the AUC values for prediction of mortality were 0.83, 0.79, and 0.77, respectively, for 6 months, one year, and 3 years after surgery for hip fracture.
The CCI, which includes a series of weighted comorbidities to predict 1-year mortality [
]. Together, the CCI and ASA are commonly used tools for risk stratification. According to the present multivariable Cox proportional hazards regression model, the CCI, but not the ASA score, was an independent factor associated with mortality. This is consistent with Varady et al. [
] was limited to one year after surgery and shorter compared with our study. The ASA score is proposed to optimize perioperative care, while CCI is used to measure the burden of specific comorbidity and prognosis of diseases. This might be the cause of the difference.
The present analysis chose the CCI, rather than ASA score, as an important predictor in the nomogram model. Yet, as a single factor to predict the mortality risk of patients with hip fracture, the CCI did not perform well. Varady et al. [
]. found that the CCI had low power of prediction, with an AUC of 0.71 (0.65–0.77) to predict 30-day mortality. To improve the predictive power, more factors might be combined with the CCI.
The present study found that serum albumin, sodium, and hemoglobin, which are easily accessible and convenient laboratory tests, reflected well the risk of mortality of patients after hip fracture surgery. This is consistent with previous studies. A meta-analysis [
] found that anemia at admission was associated with greater 30-day postoperative morbidity and mortality in geriatrics with hip fracture. Albumin, sodium, and hemoglobin can be easily tested before surgery, and according to the above studies, their significance is stable and convincing. Thus, albumin, sodium, and hemoglobin were chosen as important predictors in the design of the present nomogram.
The C-index and AUC value are parameters that indicate the predictive accuracy of a model. The accuracy of prediction for mortality of the present novel nomogram, which combines the CCI with the specified laboratory test results, was good: with a C-index of 0.76, and AUCs at 6 months and 1 and 3 years of 0.83, 0.79, and 0.77, respectively. Other models had been used to stratify the risk of patients with hip fracture [
]; the ASA, CCI, NHFS, E-PASS, and O-POSSUM are the most frequently investigated and used. Yet, compared with these models, the nomogram presented here showed a higher C-index and AUC value. The NHFS was designed specifically for patients with hip fracture, with a promising performance in predicting short-term mortality risk: the AUC is 0.72 in patients with hip fracture [
] assessed the predictive accuracy of 30-day mortality of 6 models (CCI, O-POSSUM, E-PASS, a risk model by Jiang et al., NHFS, and a model by Holt), and determined that none had the AUC necessary for excellent discrimination (i.e., > 0.80). Models developed respectively by Cenzer et al. [
] established a new model by machine learning algorithm and achieved an AUC of 0.75 for prediction of 1-year mortality.
Compared with these models, the present novel nomogram showed a better performance to stratify the mortality risk in geriatrics with hip fracture, and the CCI and albumin, sodium, and hemoglobin can be obtained conveniently and preoperatively. Thus, theoretically, this model could be applied with improved convenience and precision. However, external validation is needed to confirmed its performance before it could be used in clinical practice.
This study has several limitations. First, although the nomogram performed well as evaluated by internal validation, external validation was not applied. The results warrant external validation, conducted in other clinical centers or with new patient cohorts. In addition, this was a retrospective study and the study period was relatively long, so that some differences regarding perioperative interventions may now exist. Thus, in future a cross-sectional study might be conducted.
This study determined that age, CCI, serum albumin, sodium, and hemoglobin were independent risk factors for mortality after hip fracture surgery in geriatrics no younger than 70 years. These parameters are practical and convenient to obtain in a clinical setting, and were incorporated in the novel nomogram designed herein. The model was carefully evaluated and internal validation showed good accuracy and usefulness. This nomogram for predicting the mortality risk after hip fracture surgery in geriatrics warrants confirmation by external validation.
Declaration of Competing Interest
The authors declare that they have no conflict of interest.
The authors thank the doctors in the Department of Orthopedics, Peking University First Hospital for their help and advices in the process of data collection, including Tianyue Zhu, Licheng Wen, Weibing Chai, Hongzhang Lu, Jun Li, Zhenning Liu, Yilin Ye, Daojian Zhang, Xin Yang, and Zhichao Meng.