Objective  To investigate the risk factors for re-fracture after percutaneous kyphoplasty (PKP) in elderly patients with osteoporotic thoracolumbar compression fractures and to construct a line graph prediction model.    Methods  One hundred and eighty-two elderly patients with osteoporotic thoracolumbar compression fractures treated with PKP from January 2016 to November 2019 were selected for the study, and the patients were continuously followed up for 3 years after surgery. Clinical data were collected from both groups; Receiver operating characteristic (ROC) curve analysis was performed on the measures; Logistic regression analysis was performed to determine the independent risk factors affecting postoperative re-fracture in PKP; the R language software 4.0 “rms” package was used to construct a predictive model for the line graph, and the calibration and decision curves were used to internally validate the predictive model for the line graph and for clinical evaluation of predictive performance.   Results  The differences between the two groups were statistically significant (P<0.05) in terms of bone mineral density (BMD), number of injured vertebrae, single-segment cement injection, type of cement distribution, cement leakage, difference in vertebral body height before and after PKP, and change in posterior convexity angle. The area under the curve (AUC) for BMD, number of injured vertebrae, single-segment cement injection volume, cement leakage, pre-and post-PKP vertebral height difference, and posterior convexity change were 0.772, 0.732, 0.722, 0.801, and 0.813, respectively, and the best cutoff values were -3.1, 2, 3.9 ml, 0.4 mm, and 8.7°, respectively. BMD, number of injured vertebrae, single-segment cement injection volume, cement leakage, pre-and post-PKP vertebral height difference, and posterior convexity change were independent risk factors for re-fracture after PKP in elderly patients with osteoporotic thoracolumbar compression fractures. The calibration curve of the column line graph prediction model was close to the original curve and the ideal curve with a C-index of 0.818 (95% CI: 0.762-0.883), and the model fit was good; the threshold value of the column line graph prediction model was >0.22, which could provide a net clinical benefit, and the net clinical benefit was higher than the independent predictors.   Conclusion  BMD, number of injured vertebrae, single-segment cement injection, cement leakage, pre-and post-PKP vertebral height difference, and posterior convexity angle change are independent risk factors affecting the recurrent fracture after PKP in elderly patients with osteoporotic thoracolumbar compression fracture, and this study constructs a column line graph model to predict the recurrent fracture after PKP in elderly patients with osteoporotic thoracolumbar compression fracture as a predictor for clinical. This study provides an important reference for clinical prevention and treatment, and has clinical application value.