What is Kinesiophobia? Kinesiophobia is an excessive, irrational fear of physical movement or activity due to a belief that movement will cause pain or re-injury. In postoperative lung cancer patients, kinesiophobia is a significant problem because it leads patients to avoid the physical activity and rehabilitation exercises that are essential for recovery, increasing complications and reducing quality of life.
Prevalence and Impact This study found that 43.74% of postoperative lung cancer patients experienced kinesiophobia - nearly half of all surgical patients. Patients with kinesiophobia had slower recovery, worse pulmonary function, greater fatigue, higher rates of depression, and poorer long-term quality of life compared to those without kinesiophobia.
Why Early Prediction Matters If healthcare teams could identify which patients are most likely to develop kinesiophobia before or immediately after surgery, targeted psychological and rehabilitation interventions could be implemented early. Currently, kinesiophobia is often not recognized until it has already impaired recovery.
Study Purpose This study developed and validated a machine learning risk prediction model for kinesiophobia in 519 postoperative lung cancer patients from Chinese hospitals. The model uses clinically available variables and provides interpretable predictions, making it practical for bedside use by nurses and rehabilitation specialists.
Patient Enrollment The study enrolled 519 postoperative lung cancer patients from hospitals in China. Inclusion criteria required patients to have undergone lung cancer surgery, be capable of completing questionnaires, and have no pre-existing severe psychiatric conditions. Patients were assessed for kinesiophobia using the Tampa Scale for Kinesiophobia (TSK), a validated 17-item questionnaire.
Tampa Scale for Kinesiophobia The TSK measures fear of movement and re-injury with scores ranging from 17 to 68. A score of 37 or higher indicates clinically significant kinesiophobia. The scale asks patients about beliefs such as whether physical activity could worsen their condition, whether they fear certain movements, and how much they avoid activity due to pain concerns.
Clinical Variables Collected A comprehensive range of potentially predictive variables were measured, including demographic factors (age, sex, income, education), surgical factors (type of surgery, surgical history, pain levels), psychological factors (coping style, anxiety, depression), social factors (social support from family and healthcare providers), and clinical factors (comorbidities, lung function).
Feature Selection with LASSO With many potential predictor variables, LASSO (Least Absolute Shrinkage and Selection Operator) regression was first used to identify the most important predictors while discarding redundant or noise variables. LASSO shrinks weak predictors to zero, automatically selecting a parsimonious set of the most informative features.
Models Tested Six machine learning algorithms were compared: Random Forest (RF), Logistic Regression, Decision Tree, Support Vector Machine (SVM), XGBoost, and K-Nearest Neighbors (KNN). Each was trained on the LASSO-selected features using the training portion of the dataset and evaluated on held-out test data.
Random Forest Selected as Best Model Random Forest achieved the highest performance with an AUROC of 0.893, outperforming all other models. Random Forest builds many decision trees using random subsets of data and features, then combines their predictions. This ensemble approach produces highly accurate and robust predictions while being resistant to overfitting.
Cross-Validation and Calibration All models were evaluated using 5-fold cross-validation to obtain stable performance estimates. Calibration plots and Brier scores assessed whether predicted probabilities matched actual kinesiophobia rates - a critical but often overlooked aspect of clinical risk model quality.
SHAP for Model Interpretability SHAP (SHapley Additive exPlanations) values were calculated for the Random Forest model to explain individual predictions. SHAP assigns each predictor variable a contribution value for each patient, showing not just which variables matter overall but how each variable influenced the specific prediction for each individual patient.
Positive Coping Style - Most Protective Factor Having a positive coping style - the tendency to proactively address challenges, seek information, and use problem-solving approaches - was the single most important predictor of kinesiophobia, according to SHAP analysis. Patients with positive coping styles had dramatically lower kinesiophobia rates, likely because they approach recovery challenges as manageable rather than threatening.
Pain Severity - Strongest Risk Factor Higher reported pain severity after surgery was strongly associated with kinesiophobia. This makes physiological sense: when movement causes significant pain, the brain forms associations between movement and danger, triggering protective fear responses. Effective postoperative pain management is therefore both a comfort measure and a kinesiophobia prevention strategy.
Social Support Level Patients with stronger social support from family, friends, and healthcare providers had lower kinesiophobia rates. Social support may buffer fear by reassuring patients that they are not alone in recovery and providing encouragement during painful rehabilitation activities. Healthcare professional support was particularly protective.
Personal Income, Surgical History, and Gender Lower personal income was associated with higher kinesiophobia, potentially reflecting reduced access to rehabilitation resources or greater anxiety about inability to work during extended recovery. Prior surgical history was protective, possibly because experienced patients have realistic expectations about recovery. Female patients showed slightly higher kinesiophobia rates, consistent with broader literature on pain catastrophizing by gender.
Excellent Discrimination The Random Forest model achieved an AUROC of 0.893, indicating excellent ability to distinguish patients who will develop kinesiophobia from those who will not. By convention, AUROC above 0.8 is considered very good for clinical prediction models, and above 0.9 is excellent.
Sensitivity and Specificity Balance At the optimal clinical threshold, the model achieved high sensitivity (correctly identifying most true kinesiophobia cases) while maintaining reasonable specificity (avoiding excessive false positives). For a screening tool where missing cases has greater cost than unnecessary follow-up, the model was tuned to prioritize sensitivity.
SHAP Summary and Dependence Plots SHAP summary plots visually ranked all six predictors by importance and showed the direction of their effects. Dependence plots revealed the precise relationship between each variable and kinesiophobia probability - for example, showing the pain severity threshold above which kinesiophobia risk rises sharply.
Decision Curve Analysis Decision curve analysis assessed the clinical net benefit of using the model across a range of risk thresholds. The Random Forest model provided positive net benefit over a wide range of clinically reasonable thresholds, confirming that using the model to guide interventions would improve patient outcomes compared to treating all or no patients.
Targeted Nursing Interventions The model can be applied by nursing staff at hospital admission or immediately after surgery to flag high-risk patients. Identified high-risk patients could receive targeted psychological support, pain education, enhanced social support coordination, and structured rehabilitation guidance before kinesiophobia develops or worsens.
Pain Management as Prevention Since pain severity is a key modifiable risk factor, optimizing postoperative analgesia is both a comfort measure and a kinesiophobia prevention strategy. Multimodal pain management protocols that minimize opioid use while providing effective pain control could simultaneously address kinesiophobia risk.
Coping Style Enhancement Interventions Because positive coping style was the most protective factor, psychological interventions that strengthen patients' coping resources could reduce kinesiophobia. Cognitive behavioral therapy approaches, acceptance and commitment therapy, and structured rehabilitation education programs have evidence for improving coping and reducing movement fear.
Social Support System Building Healthcare teams can systematically assess and strengthen social support systems for high-risk patients - involving family members in rehabilitation education, connecting patients with peer support programs, and ensuring regular check-ins from physiotherapists and nurses during the critical early recovery period.
Single Geographic Context The study was conducted exclusively in Chinese hospitals, limiting generalizability to Western healthcare contexts with different cultural attitudes toward pain, surgery, and rehabilitation. Cultural factors strongly influence kinesiophobia, and model performance should be validated in other countries before widespread adoption.
Cross-Sectional Assessment Kinesiophobia was assessed at a single postoperative time point rather than tracked longitudinally. Future studies should measure kinesiophobia trajectories over time - identifying when it develops, peaks, and resolves - to better define optimal intervention windows.
Missing Biological Factors The model relies entirely on psychological, social, and clinical variables without incorporating biological factors such as inflammatory markers, neurological pain pathway variants, or pharmacogenomic factors affecting pain medication response. Integrating biomarkers could improve prediction accuracy.
Implementation Studies Needed While the model was statistically validated, its real-world impact on patient outcomes has not been measured. Randomized trials comparing kinesiophobia rates and recovery outcomes between standard care and model-guided intervention are needed to demonstrate whether model-directed interventions actually improve patient outcomes.