研究背景：近年来，随着社会节奏加快与生活方式变化，中年人群面临日益严重的健康问题，普遍呈现出睡眠质量下降等亚健康状态，且慢性病发病呈年轻化趋势。作为社会和家庭的中坚力量，中年人不仅健康风险较高，且对健康管理有着强烈需求。移动健康应用程序（mHealth APP）因其便捷性和智能化特征，已成为中年人群日常健康管理的重要工具，并在慢性病预防和控制中展现出积极成效。然而，当前mHealth APP数量众多、质量参差，现有评价体系多从技术或医疗专业视角出发，缺乏从中年用户实际使用体验出发的评价标准，导致用户难以识别和筛选，影响其持续使用与健康管理效果。目前现有mHealth APP可用性评价工具忽视了中年群体的认知特征和主观感受，缺乏以用户体验为核心的系统性评价框架。鉴于中年亚健康人群对个体化健康管理的高度需求，亟需构建一套科学、系统且贴合用户体验的mHealth APP评价工具，以指导其科学选择与持续使用APP，从而促进中年人群生活质量的提升与健康水平的改善。

研究目的：1. 构建符合中年亚健康用户需求的mHealth APP评价工具。2. 对构建的mHealth APP评价工具进行信效度检验，确保其在该用户群体中的可靠性。

研究方法：本研究基于用户体验蜂巢模型，前期通过文献回顾和半结构式访谈法初步构建移动健康APP评价量表的维度和条目池。采用德尔菲法选取12名护理信息学等领域专家对初步形成的量表进行两轮修订。根据指标选取标准，当某指标的得分均数＜3.5分，或条目满分率＜20%，或变异系数＞0.25时，该指标将予以删除，以完善并形成量表初稿。采用方便抽样法选取北京市45至59岁的中年亚健康人群，对量表初稿进行信效度验证。通过Cronbach's α系数评估量表的信度，采用探索性因子分析（EFA）进行因子结构探索，验证性因子分析（CFA）进行模型拟合验证。数据分析采用SPSS 26.0及AMOS 24.0软件。

研究结果：

1. 条目池的构建：基于理论用户体验蜂巢模型、前期文献回顾和15名受访者半结构式访谈，形成条目池共包含125个条目。经过课题组讨论，初步形成8个一级指标、13个二级指标和49个三级指标作为初始评价量表。

2. 德尔菲法：（1）专家基本情况：根据专家遴选标准，选取护理信息学领域专家10名，医学信息领域专家1名，信息与通信工程领域专家1名，12名专家均参与两轮函询，专家平均年龄为40.83 ± 8.68岁。（2）两轮专家（n = 12）咨询积极性较高，专家咨询表回收率均为100%，两轮的专家权威系数Cr分别为0.946、0.955；在第一轮专家函询的结果中，各指标重要性均值位于4.0~5.0之间，标准差处于0.0~1.24之间，变异系数值在0.0~0.28之间，满分率均高于40%，肯德尔和谐系数W = 0.175，显著性检验结果χ² = 113.652，P = 0.001。经过第一轮专家函询，剔除3个二级指标，增加4个二级指标，剔除8个三级指标，增加13个三级指标，修改35个三级指标。在第二轮专家函询的结果中，各指标均值均不低于4.5，指标重要性程度较高，且变异系数低于0.25，肯德尔和谐系数W = 0.215，χ² = 199.120，P＜0.001。经过第二轮专家咨询，无增加和剔除指标，因语言描述、措辞修改5个三级指标。（3）经过两轮专家咨询，结合专家意见和建议，共剔除3个二级指标，增加4个二级指标，剔除8个三级指标，增加13个三级指标，修改40个三级指标，形成8个维度，54个条目的移动健康APP评价量表（初稿）。

3. 量表信效度验证：（1）探索性因子分析：①共回收270份有效量表，其中纸质版量表78份，电子版量表192份，每份量表平均填写时间为9.87±2.15分钟。②采用主成分分析法和Kaiser正态化最大方差旋转法进行因子提取与旋转，结果显示KMO = 0.952，Bartlett球形检验结果显著（Bartlett’s χ² = 11326.508, df = 1431.000, P＜0.001），共提取8个因子，旋转前后的累积方差解释率为70.810%，表明量表具有较好的结构效度。③信度检验：整体量表的Cronbach's α为0.970，各个维度的Cronbach's α系数分布在0.851~0.955之间，表明量表信度较好。（2）验证性因子分析：①共回收454份有效量表，其中纸质版量表101份，电子版量表353份。每份量表平均填写时间为10.21±2.36分钟。②量表得分情况：易用性维度平均得分3.483±0.954分，可用性维度3.462±0.968，有用性维度3.559±0.964，可获得性维度3.500±0.933，可找到性维度3.470±1.019，可靠性维度1.582±0.403（因采用两点计分法，该维度得分低于其他维度），满意度维度3.458±0.990，价值性维度3.462±1.005。③信度检验：整体量表的Cronbach's α为0.967，各个维度的Cronbach's α系数分布在0.875~0.951之间，表明量表信度较好。④CFA结果显示，所有指标的非标准载荷系数均达到显著性水平（P＜0.001），且标准载荷系数均大于0.74。结构方程模型拟合参数结果为χ²/df = 1.118，RMSEA = 0.016，RMR = 0.035，CFI = 0.991，NFI = 0.918，均符合要求，表明量表具有较好的区分效度。

研究结论：本研究构建的mHealth APP评价量表具有良好的信效度和科学性，可用于中年亚健康人群评价并筛选高质量APP并指导开发者优化设计，为我国mHealth APP评价提供了新工具。但本研究样本局限于北京地区，未来可考虑扩大研究范围至全国，探讨其在其他年龄段的适用性。

Background: In recent years, with the acceleration of societal pace and changes in lifestyle, middle-aged individuals are facing increasingly severe health challenges, commonly exhibiting sub-health conditions such as declining sleep quality, while the onset of chronic diseases tends to occur at younger ages. As the backbone of both society and families, this population bears a high health risk and demonstrates a strong demand for health management. Mobile health applications (mHealth APPs), due to their convenience and intelligent features, have become important tools for daily health management among middle-aged individuals, showing positive outcomes in the prevention and control of chronic diseases. However, the current mHealth APP market is saturated with a wide variety of applications of uneven quality. Existing evaluation systems mainly adopt technical or professional medical perspectives, lacking assessment criteria based on the actual user experience of middle-aged users. This limits users' ability to select appropriate applications and affects their long-term engagement and health management outcomes. Presently, available usability evaluation tools for mHealth APPs largely overlook the cognitive characteristics and subjective experiences of the middle-aged population, and fail to provide a systematic, user experience-centered evaluation framework. Given the high demand for personalized health management in sub-healthy middle-aged users, there is an urgent need to develop a scientific and systematic mHealth APP evaluation tool centered on user experience, to support their informed selection and sustained use of such applications, thereby improving their quality of life and health status.

Objective: 1. To develop an mHealth APP evaluation tool that meets the needs of sub-healthy middle-aged users. 2. To test the reliability and validity of the developed mHealth APP evaluation tool to ensure its applicability to this user group.

Methods: This study, guided by the User Experience Honeycomb model, initially constructed the dimensions and item pool of the mHealth app evaluation scale through a literature review and semi-structured interviews. The Delphi method was then employed, involving two rounds of expert consultations with 12 nursing informatics experts. Indicators were selected based on predefined Delphi criteria: items with a mean score < 3.5, a full score rate < 20%, or a coefficient of variation > 0.25 were removed. The preliminary scale was finalized accordingly. Purposeful sampling was used to recruit middle-aged individuals aged 45~59 years with suboptimal health in Beijing for reliability and validity testing of the scale. The reliability of the scale was assessed using Cronbach's α coefficient, while validity was examined through exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). Data were analyzed using SPSS 26.0 and AMOS 24.0 software.

Results: 

1. Item Pool Construction: Based on the theoretical User Experience Honeycomb Model, a preliminary literature review, and semi-structured interviews with 15 participants, an initial item pool comprising 125 items was established. Following discussions within the research team, an initial evaluation scale was formulated, consisting of 8 primary indicators, 13 secondary indicators, and 49 tertiary indicators. 

2. Delphi Method: (1) In accordance with expert selection criteria, 10 experts in nursing informatics, 1 expert in medical informatics, and 1 expert in information and communication engineering were selected, totaling 12 experts. The average age of the experts was 40.83±8.68 years. (2) Both rounds of expert consultations (n = 12) exhibited high participation, with a 100% response rate. The expert authority coefficients (Cr) for the two rounds were 0.946 and 0.955. In the first round, the mean importance scores of the indicators ranged from 4.0 to 5.0, with standard deviations between 0.0 and 1.24, and coefficients of variation ranging from 0.0 to 0.28. The full-score rate exceeded 40% for all indicators. The Kendall's W coefficient of concordance was 0.175, with a significance test result of χ² = 113.652, P = 0.001. Based on the initial round of expert consultation, the evaluation framework underwent systematic refinement involving the elimination of 3 secondary indicators, incorporation of 4 novel secondary indicators, removal of 8 tertiary indicators, addition of 13 new tertiary indicators, and substantive modification of 35 existing tertiary indicators. In the second round, the mean importance scores of all indicators were not lower than 4.5, indicating high importance, while the coefficient of variation remained below 0.25. The Kendall's W coefficient of concordance was 0.215, with χ² = 199.120, P < 0.001. After the second round of expert consultation, no new indicators were added or removed; however, five tertiary indicators were revised in terms of language and wording for improved clarity and precision. (3) After two rounds of expert consultation, incorporating expert feedback and suggestions, a refined evaluation scale was developed, consisting of 8 dimensions and 54 items.

3. Reliability and Validity Testing: (1) Exploratory Factor Analysis: ① A total of 270 valid questionnaires were collected, including 78 paper-based and 192 electronic versions. The average time to complete each questionnaire was 9.87 ± 2.15 minutes. ② Principal component analysis with Kaiser normalization and varimax rotation was applied for factor extraction and rotation. The Kaiser-Meyer-Olkin (KMO) value was 0.952, and Bartlett's test of sphericity was significant (Bartlett's χ² = 11326.508, df = 1431.000, P < 0.001). Eight factors were extracted, with a cumulative variance contribution rate of 70.810% before and after rotation, indicating good structural validity of the scale. ③ Reliability analysis showed that the overall Cronbach's α coefficient was 0.970, and the Cronbach's α coefficients for each dimension ranged from 0.851 to 0.955, indicating high internal consistency reliability. (2) Confirmatory Factor Analysis: ① A total of 486 questionnaires were distributed, with 454 valid responses collected (101 paper-based and 353 electronic questionnaires), achieving a response rate of 93.42%. The average completion time for each questionnaire was 10.21 ± 2.36 minutes. ② Scores of the scale: The average scores for each dimension were as follows: Usability (3.483 ± 0.954), Usefulness (3.462 ± 0.968), Effectiveness (3.559 ± 0.964), Accessibility (3.500 ± 0.933), Findability (3.470 ± 1.019), Reliability (1.582 ± 0.403, due to binary scoring), Satisfaction (3.458 ± 0.990), and Value (3.462 ± 1.005). ③ Reliability analysis: The overall Cronbach's α for the scale was 0.967, with Cronbach's α values for each dimension ranging from 0.875 to 0.951, indicating good reliability. ④ CFA results demonstrated that all unstandardized factor loadings were statistically significant (P < 0.001), with standardized loadings exceeding 0.74. Structural equation modeling fit indices were χ²/df = 1.118, RMSEA = 0.016, RMR = 0.035, CFI = 0.991, and NFI = 0.918, all meeting the required standards, demonstrating good construct validity of the scale.

Conclusion: The mHealth APP evaluation scale developed in this study demonstrates good reliability, validity, and scientific rigor. It can be used to assess and select high-quality apps for middle-aged users and guide developers in optimizing app design, providing a new tool for mHealth APP evaluation in China. However, this study was limited to samples from Beijing. Future research should consider expanding the sample size nationwide and exploring its applicability to other age groups.