以患者为中心视角下基于数据挖掘的慢性病精准预防与管理模型构建

The prevention and management of chronic diseases face three major challenges in China: the heavy burden of disease, the high management complexity, and the insufficient matching of service and residents' health needs. In response to these challenges, this study is designed to use a data-driven approach to identify disease risks among different subgroups of populations from a patient-centered perspective, and then stratifies disease risks among subpopulations so as to match and prioritize appropriate prevention services to different subgroups...Aim 1 uses predictive data mining to assess different disease risks among subgroups of population. Correlation-based feature selections will be used to select the subset of features that affect the disease risks. Rotation forest and alternating decision tree will be used as classification and predicting techniques. Aim 2 builds a Markov model based on life expectancy, to match appropriate and prioritized prevention services to subpopulations. Aim 3 implements chronic disease scenario models to simulate and evaluate the effect of our constructed precision prevention model for chronic diseases...The core of this study is to provide predictable, targeted and evidence-based chronic disease identification and prevention services to the population, to cope with the complexity and uncertainty of chronic disease management, and thus to make as accurate prediction and decision-making as possible. We expect the model constructed in this study will be able to responds to the health needs and expectations of population, and achieve the goal of patient-centered care.

在我国，慢性病的预防和管理面临三大挑战：疾病负担重、管理复杂性高、服务资源与居民健康需求匹配不足。应对挑战，本研究从患者为中心的角度出发，运用数据挖掘手段识别不同人群的疾病风险，据此对人群进行疾病风险的分层，进而匹配适宜的预防服务并排序优先级，从而做到预防与管理的精准化。第一部分采用预测型数据挖掘评估人群的疾病风险，运用基于关联规则的特征选择算法对影响疾病风险的特征子集进行选择，并采用旋转森林和交互式决策树作为分类和预测技术。第二部分构建基于预期寿命的马尔科夫模型，匹配适宜的预防服务并排序服务的优先级。第三部分采用慢性病情境模型对慢性病精准管理模式进行模拟和效果评价。本研究核心在于为人群提供可预测、有针对性、循证的慢性病识别和预防服务,克服慢性病管理的复杂性和不确定性，做出尽可能准确的预测和决策；并从居民的健康需求出发，真正实现以患者为中心的理念，满足我国居民多层次、个性化的健康需求。

我国已经进入慢性病的高负担期，在慢性病的预防和管理领域有重大的管理创新需求。当前慢性病的预防和管理面临三大挑战，包括疾病负担重、管理复杂性高、服务资源与居民健康需求匹配不足。本研究从患者为中心的角度出发，运用数据挖掘手段识别不同人群的疾病风险，据此对人群进行疾病风险的分层，进而匹配适宜的预防服务并排序优先级，从而做到预防与管理的精准化。本研究以精准健康管理经典理论作为指导，聚焦各人群子集的疾病风险进行及早和有目的性的干预，促进对慢性病危险因素的早期识别和预防，从根源防控慢性病的发生和蔓延。研究主要包括以下四方面的内容：(1) 人群健康的影响因素分析；(2) 数据挖掘模型构建及在健康管理中的应用；(3) 慢性病患者的健康管理服务；(4) 健康管理政策的分析。本研究核心在于为人群提供可预测、有针对性、循证的慢性病识别和预防服务,克服慢性病管理的复杂性和不确定性，做出尽可能准确的预测和决策；并从居民的健康需求出发，真正实现以患者为中心的理念，满足我国居民多层次、个性化的健康需求。研究成果所构建的模型可进一步转化成为具有应用性的开源软件，被各类医疗卫生机构利用，以鉴别人群的各类慢性病风险，对人群进行疾病风险的分层，及时采取有针对性的预防和管理措施、提升患者对干预的依从性、并大幅度提升有限的慢性病防控资源的合理配置。