Factors contributing to 1-year dissatisfaction after total knee arthroplasty: a nomogram prediction model

The hospital research and ethics committee approved the study protocol. From January 2017 to December 2019, 574 patients who underwent TKA for knee arthritis at our department were recruited for this study; 208 were finally included in our analyses. The inclusion criteria were primary TKA for arthritis (primary osteoarthritis [OA], rheumatoid arthritis [RA], and traumatic osteoarthritis), posterior stabilized rotating platform without patellar replacement, devices implanted using a ligament balancing technique, able to participate in the evaluation program, and completion of all study procedures and follow-up visits. The exclusion criteria were as follows: huge bony defects; history of a fracture that resulted in a deformity after healing that might change the bony alignment, such as high tibial or distal femoral osteotomies; partial knee replacement (unicompartmental, bicompartmental, or patellofemoral joint replacement); patellectomy; arthroscopy within 1 year; use of blocks, long stems, or special implants, such as constrained condylar knee implants; juvenile rheumatoid or psoriatic arthritis; systemic lupus erythematosus; significant neurological or musculoskeletal disorder or disease that may adversely affect gait or function recovery; lower back pain and pain in other joints; complications related to TKA within 1 year postoperatively; readmission for TKA complications, such as pulmonary embolism, hospital-acquired urinary tract infection, delayed wound healing, suture abscess, wound dehiscence; and the need for manipulation under anesthesia during the postoperative course because of arthrofibrosis, periprosthetic joint infection, recurvatum instability requiring revision, and prosthesis fracture.

Candidate variables were prospectively collected preoperatively and included age [20], gender [21], body mass index (BMI) [22], primary diagnosis, comorbidities [23], degree of anxiety [24], education level [25], marital status [25], smoking status [26], radiographic degree of OA in the femorotibial joint using the Kellgren–Lawrence scale [27] and coronal alignment of the lower limb (hip–knee–ankle [HKA] angle) measured by weight-bearing long-length radiographs [28]. Perioperative variables were the use of a tourniquet, anesthesia methods, local infiltration analgesia, and cosmetic closure. Postoperative variables included the range of extension, range of flexion, degree of pain, any other type of noise resulting from knee movement, and the postoperative coronal HKA angle.

A standard perioperative prophylactic antibiotic protocol was used for all patients. The type of anesthesia (general anesthesia or combined spinal–epidural anesthesia) was chosen based on the patient’s medical condition, concomitant spinal pathology, and vascular condition of the lower limbs. All patients included in this study underwent the same surgical technique for TKA. The same senior surgeon performed all surgical procedures. The standard surgical procedure for all patients included the medial parapatellar approach after anesthesia was administered. The ATTUNE® Knee System (DePuy Synthes, Warsaw, IN) or posterior stabilized rotating platform was implanted using a ligament balancing technique, and physical therapy was initiated on the first postoperative day for all patients. The decision to perform cosmetic suturing was based on the suturing difficulty.

The Knee Society Score (KSS, 2011) is a comprehensive and validated measurement determined based on a clinician-reported objective subscale (not further evaluated in this study) and patient-reported subscales (including symptoms, satisfaction, and functional activity subscale). The higher the score, the better the outcomes of all subscales [29] The KSS takes into consideration patient reports of both expectations and satisfaction [30]. One year after TKA, we measured patient satisfaction by asking an overall satisfaction assessment question based on the 2011 KSS patient satisfaction subscale. As Van Onsem indicated [14], the satisfaction subscale consists of five items. Each item is scored using a scale from 0 to 5 points, resulting in a maximum KSS of 40; a score ≥ 20 indicated that the patient was satisfied and a score < 20 indicated that the patient was dissatisfied [31]. We adopted the same criteria. Finally, if patients underwent TKA multiple times, then we asked them to report the most recent surgery.

The pain degree (EQ-5L: pain) was rated as follows: I have no pain or discomfort = 0 points; I have slight pain or discomfort = 1 point; I have moderate pain or discomfort = 2 points; I have severe pain or discomfort = 3 points; and I have extreme pain or discomfort = 4 points [32].

Anxiety or degree of depression (EQ-5L: anxiety) was rated as follows [31]: I am not anxious or depressed = 1 point; I am slightly anxious or depressed = 2 points; I am moderately anxious or depressed = 3 points; I am severely anxious or depressed = 4 points; and I am extremely anxious or depressed = 5 points.

Statistical analyses were performed using SPSS (version 23.0; IBM, Armonk, NY, USA) and R 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria). Categorical variables were evaluated using the chi-square test. Continuous variables of the two groups were evaluated using the independent Student’s t test or Mann–Whitney U test. Univariate and multivariate logistic regression analyses were performed to identify the independent predictors of postoperative dissatisfaction. The data of the training cohort were analyzed using the least absolute shrinkage and selection operator (LASSO) method to screen suitable and effective risk factors for dissatisfaction. The results of univariate and multivariate regression analyses were visualized using forest plots. Finally, the results of the logistic regression analysis were used to construct a nomogram prediction model. The receiver operating characteristic (ROC) curve was used to quantify the discrimination performance of the nomogram, and calibration curves and the Hosmer–Lemeshow test were used to evaluate the calibration of the nomogram. A decision curve analysis was performed to determine the clinical applications of the risk prediction model. Bootstraps for 1000 resamples were performed using the ROC curve, calibration curve, and decision curve analysis to reduce the overfitting deviation. All statistical tests were two-sided, and P < 0.05 was considered significant.

The univariate analysis (Fig. 3a) and multivariate Cox regression analysis (Fig. 3b) were performed for 154 TKA patients in the training cohort. The results showed that gender, primary diagnosis, postoperative residual pain, poor postoperative range of motion (ROM), wound healing, and the rate of change in the degree of coronal lower limb alignment (HKA angle) were considered independent predictors of dissatisfaction after TKA (p < 0.05).

We combined the six independent predictors into a predictive model and displayed them in the form of a nomogram (Fig. 4), which is a quantitative and convenient tool. To obtain the individualized risk of patient dissatisfaction 1 year after TKA, the point value for each variable was assessed; then, all values were summed to obtain the total score to determine the risk.

The ROC curve was used to evaluate the discriminatory ability of the prediction model. The pooled area under the curve of the nomogram was 0.9206 (95% CI: 0.8785, 0.9627), with sensitivity and specificity of 100% and 72.73%, respectively, for the training cohort (Fig. 5a). For the validation cohort, the pooled area under the curve was 0.9662 (95% CI: 0.9231, 1.0000), with sensitivity and specificity of 92.31% and 92.68%, respectively (Fig. 5), indicating moderately good performance (Table 3). The calibration curve and Hosmer–Lemeshow test indicated that the prediction model showed good fit for both the training and validation cohorts (Fig. 6a and b). Furthermore, the Hosmer–Lemeshow test indicated good agreement between the predicted and actual probabilities for both the training and validation cohorts. The decision curve analysis of the training cohort (Fig. 7a) and validation cohort (Fig. 7b) indicated that the application of the model for TKA patients to predict the risk of dissatisfaction 1 year after surgery was more effective than intervention for all patients. Each patient was separated into a high-risk group or low-risk group according to a cutoff value of 50% predicted by the nomogram.

The decision curve showed that it would be more accurate to use the nomogram in the development cohort to predict the risk of dissatisfaction when the risk threshold probability was between 1 and 99%, and in the validation cohort, when between 1 and 71% (Fig. 7a and b). We applied the nomogram in TKA patients as an example. The first patient had postoperative range of motion, postoperative pain, wound healing, primary diagnosis, rate of change in the degree of coronal lower limb alignment after TKA (%), the gender of the patients. These were between 90 and 100° (21.9 points), with moderate pain or discomfort (41.6 points), without cosmetic closure (12.9 points), with rheumatoid arthritis (RA) (10.3 points), (80%) (44.4 points), and a female (14.6 points), respectively. The total calculated nomogram score was 145.7, and the 1-year risk of dissatisfaction after TKA was more than 95%. This patient had a high risk of dissatisfaction 1 year after TKA. The second patient had postoperative range of motion, postoperative pain, wound healing, primary diagnosis, rate of change in the degree of coronal lower limb alignment after TKA (%), and the gender of the patients. There were between 110 and 120° (7.4 points), with slight pain or discomfort (20.8 points), with cosmetic closure (0 points), with osteoarthritis (OA) (0 points), (20%) (11.1 points), and a male (0 points), respectively. The total calculated nomogram score was 39.3, and the 1-year risk of dissatisfaction after TKA was less than 5%. This patient had a very low risk of dissatisfaction 1 year after TKA.