Trajectories/Predictors of Oral Health-Related Quality of Life to Early Adulthood

项目总结/摘要 在美国，医疗保健费用继续呈指数级增长，口腔疾病仍然是最主要的疾病之一。 10个类别的个人医疗保健支出。为了应对不断上涨的护理成本， 不必要的治疗，越来越强调以病人为中心的护理，包括病人的看法， 和健康相关的生活质量评估作为医学和牙科研究的重要健康结果。 口腔健康状况对个人的身体和心理有影响，并影响他们的生活质量- 它们如何生长，如何看，如何说话，如何咀嚼，如何社交。解决口腔健康相关的生活质量（OHRQoL）是 重要的是，将有助于提高护理质量，最大限度地减少口腔健康差异，提高患者满意度， 和整体生活质量，并降低成本。尽管OHRQoL的重要性， 青少年/青年期OHRQoL的纵向和无轨迹研究。理想情况下，这些研究将 确定纵向因素和模式/轨迹，以更全面地了解OHRQoL的发展， 个体进入成年期。充分应对纵向数据的挑战，并创建预测性 模型捕捉许多重要的轨迹决定因素，有必要使用高性能的算法 比如机器学习，一种人工智能。我们的研究将是第一个开发机器学习的 使用纵向数据预测OHRQoL的工具。我们选择机器学习是因为它可以 适应高维数据，以准确预测个人的OHRQoL轨迹。我们将 利用来自我们爱荷华州氟化物研究的独特纵向数据， 年龄23岁。OHRQoL轨迹将使用在17岁、19岁、18岁和19岁时测量的三个因变量定义， 和23：1）儿童感知问卷，2）整体口腔健康，和3）视觉模拟生活质量评分。由于 针对数据的复杂性和高维性，我们将使用无监督机器学习（K-means for 纵向数据）和监督机器学习（LASSO回归，随机森林和极端 梯度推进模型）分别用于轨迹分析和结果预测。具体目标 研究的目的是：1）使用以下方法确定从青春期后期到成年早期的OHRQoL轨迹 无监督机器学习，以及2）使用有监督机器学习来识别轨迹组成员关系的预测器。 机器学习这项研究将大大有助于我们对青少年的了解。 OHRQoL轨迹和决定因素。这些结果将为临床医生和政策制定者奠定基础， 过渡到一个更以病人为中心的护理模式，这将改善口腔健康的结果，减少口腔 健康差异，降低成本，提高患者满意度。我们的研究将介绍和展示 机器学习在口腔健康研究中的作用。从长远来看，我们将开发基于Web的应用程序 临床医生和政策制定者可以用来更好地设计干预和治疗，以适应口腔健康 个人和群体的需求。

PROJECT SUMMARY/ABSTRACT Healthcare costs continue to grow exponentially in the United States and oral diseases remain one of the top 10 categories in terms of personal health care expenditures. To tackle the rising costs of care and minimize unnecessary treatment, there is increasing emphasis on patient-centered care, by including patient perceptions and health-related quality of life assessments as important health outcomes in medical and dental research. Oral health conditions have physical and psychological effects on individuals and influence their quality of life - how they grow, look, speak, chew, and socialize. Addressing Oral Health-Related Quality of Life (OHRQoL) is important and will help improve the quality of care, minimize oral health disparities, improve patient satisfaction and overall quality of life, and reduce costs. Despite the importance of OHRQoL, there have been few longitudinal and no trajectory studies of OHRQoL in adolescence/young adulthood. Ideally, such studies would identify longitudinal factors and patterns/trajectories to more fully understand development of OHRQoL as individuals enter adulthood. To adequately address the challenges of longitudinal data and create a predictive model capturing the many important trajectory determinants, it is necessary to use a high-performing algorithm like machine learning, a type of artificial intelligence. Our study will be the first to develop machine learning tools for prediction of OHRQoL using longitudinal data. We chose machine learning because it can accommodate the high-dimensional data to accurately predict individuals’ OHRQoL trajectories. We will leverage unique longitudinal data from our Iowa Fluoride Study, with data from subjects followed from birth to age 23 years. OHRQoL trajectories will be defined using three dependent variables measured at ages 17, 19, and 23: 1) Child Perception Questionnaire, 2) global oral health, and 3) visual analog quality of life scores. Due to the complexity and high dimensionality of the data, we will use unsupervised machine learning (K-means for longitudinal data) and supervised machine learning (LASSO regression, random forest and extreme gradient-boosting model) for the trajectory analysis and outcome predictions, respectively. The specific aims of the study will be to 1) determine the OHRQoL trajectories from late adolescence to young adulthood using unsupervised machine learning, and 2) identify predictors of trajectory group membership using supervised machine learning. The study will contribute significantly to our knowledge of adolescents’/young adults’ OHRQoL trajectories and determinants. The outcomes will set the stage for clinicians and policymakers to transition to a care model that is more patient-centered, which will improve oral health outcomes, reduce oral health disparities, reduce costs, and increase patient satisfaction. Our research will introduce and showcase the usefulness of machine learning in oral health research. Long term, we will develop a web-based application that clinicians and policymakers can use to better design interventions and treatments to suit the oral health needs of individuals and populations.