Development and Internal Validation of Predicting Models of Idiopathic Scoliosis Natural History and Treatment Outcomes Through the Use of Artificial Intelligence in a Large Clinical Database
Development and Internal Validation of Predicting Models of Idiopathic Scoliosis Natural History and Treatment Outcomes Through the Use of Artificial Intelligence in a Large Clinical Database
Scoliosis is a three-dimensional deformity of the spine. In its most common form (about 70% of cases), the causes are unknown, therefore it is called idiopathic scoliosis. In most cases, it is discovered after 10 years of age, and is defined diagnostically as a curve of at least 10°, measured on a standing x-ray using the Cobb method.
If scoliosis exceeds the critical threshold of 30° Cobb at the end of growth, there is a progressively greater risk of health and social problems in adult life. For this reason, the main aim of the treatment is to complete the growth period with a curve less than 30° and good sagittal balance, or at least well below 50°, which represents the surgical threshold.
Growth is a factor favouring the evolution of deformities, therefore patients are followed until the end of growth. This is why therapy can last many years, from the discovery of the presence of a deformity until bone maturation is achieved.
The early identification of parameters predictive of the outcome of the therapy to direct the least possible aggressiveness towards the result necessary for the patient's future, integrated with the evaluation of its effectiveness (monitoring), is one of the most important objectives in this field to minimize the burden of treatment in a particular phase of growth such as adolescent development, as well as to identify the subjects most at risk of worsening in adulthood.
The systematic collection of clinical data during the therapeutic process offers the possibility, through advanced analysis models, applied retrospectively, to identify predisposing factors and protective factors. When the data available is sufficiently large, it is possible to obtain predictive equations that assist clinicians in therapeutic choices and help patients understand the risks and benefits of available therapies. New technologies such as artificial intelligence techniques offer new and interesting ways of estimating risks and calculating the benefits and safety of some therapeutic choices compared to others.
This study aims to develop and internally validate data-driven stratification and prediction models to predict multiple end-of-care outcome measures that include curve magnitude, measured in Cobb degrees, measures determining the sagittal balance, and measures of quality of life and function measured through self-completion questionnaires.
Inclusion Criteria:
Exclusion Criteria: