Spatiotemporal models for periodontal disease monitoring and recall frequencies

 描述(由申请人提供)：牙周病监测和召回频率的时空模型牙周病(PD)引起的牙齿丢失在美国仍然是一个主要的公共卫生负担。随着牙科保险费的上涨，未来的专业牙科治疗计划将寻求根据患者的疾病风险来确定优先顺序 并投入更多资源监测和治疗高危患者。因此，有必要开发适当的统计模型和工具，以有效地评估PD、短期预后和牙周回忆间隔的风险，从而导致牙科治疗计划的成本效益。牙科数据集提出了许多有趣的统计挑战(非平稳性、非正态、空间相关性、非随机缺失、指示混淆、巨大的簇大小等)，这严重限制了当前可用的软件(如Patterson的EagleSoft(R)等)加载到牙周科医生的椅子端计算机中的潜力。目前可用的统计软件可能会在一个统一的范例内处理一些但不是所有这些挑战。目标：拟议的研究将开发统计工具，以(A)确定风险因素 对于PD进展，(B)快速有效地识别患者PD状态的变化，(C)使用 短期预测以指导牙周召回决定，以及(D)开发用户友好的软件来实施这些方法。研究对象：统计方法将使用来自HealthPartners HMO的一个丰富的8年纵向数据库来开发，该数据库由大约15,000名有随访的患者组成。可获得的数据和研究设计：纵向成组(CL)研究设计包括记录年龄、性别、种族、完整的修复体和牙周记录以及随访、吸烟状况、糖尿病状况、口腔卫生和其他基本参数的数据库。意义：这个项目有可能转化为牙科临床实践，因为它将为牙科从业者提供基于证据的标准，以指导‘个性化’牙周召回和治疗决策。预计产生的影响将是深远的，长期目标是将这些新方法纳入现有的椅子端牙科软件，从而开发出具有成本效益的治疗牙科计划，并保持谨慎的期望。

 DESCRIPTION (provided by applicant): Spatiotemporal models for periodontal disease monitoring and recall frequencies Tooth loss from periodontal disease (PD) remains a major public health burden in the US. With the rising cost of dental insurance premiums, future professional dental treatment plans will seek to prioritize patients based on their risk of disease and spend more resources monitoring and treating high-risk patients. Hence, there is a need to develop appropriate statistical models and tools for efficient risk assessment of PD, short-term prognosis, and periodontal recall intervals leading to cost-effectiveness of dental treatment plans. Dental datasets present many interesting statistical challenges (non-stationarity, non-normality, spatial dependence, non-random missingness, confounding by indication, huge cluster size, etc), which severely limit the potential of currently-available software (such as Patterson's EagleSoft(r), etc) loaded into the chair- side computer of a periodontist. Currently available statistical software might handle some, but not all of these challenges within a unified paradigm. Goals: The proposed study will develop statistical tools to (a) characterize risk factors for PD progression, (b) rapidly and efficiently indentify changes in a patient's PD status, (c) use short-term predictions to guide periodontal recall decisions, and (d) develop user-friendly software to implement these methods. Subjects: The statistical methods will be developed using a rich 8-year longitudinal database from the HealthPartners HMO, consisting of about 15,000 patients with follow-ups. Available data and study design: A clustered- longitudinal (CL) study design comprises the databases that recorded data for age, gender, race, complete restorative and periodontal records with follow-up, smoking status, diabetes status, oral hygiene, and other essential parameters. Significance: The potential translation to dental clinical practice for this project is strong because it will provide dental practitioners with evidence-based criteria to guid 'personalized' periodontal recalls and treatment decisions. The impact generated is expected to be far- reaching, and the long-term goal would incorporate these new methods into existing chair-side dental software leading to development of cost-effective treatment dental plans with prudent expectations.