综合评价青少年多维度健康结局的方法构建研究

Development and validation of a comprehensive evaluation method for multidimensional health outcomes in adolescents

摘要

目的 比较潜在剖面分析、K均值聚类与高斯混合聚类在评价青少年多维度健康结局指标中的应用表现,为评价多维度健康结局分析策略的选择提供参考依据。方法 数据来源于一项孕期微营养素干预的整群随机对照试验的后续前瞻性队列随访的断面数据,纳入的健康结局指标包括行为问题得分、认知得分、肺活量、心率、体质指数、收缩压和舒张压,分别对其进行潜在剖面分析与聚类分析,将人群分成不同健康模式的亚组；通过Davies-Bouldin指数和Calinski-Harabasz指数评价各方法的聚类效果。最后采用Logistic回归探究青少年不同健康模式的共享病因影响因素。结果 纳入研究对象共1 767人,中位年龄12(11,12)岁,其中男性1 050人(59%)；性发育已经启动1 416人(83%)。潜在剖面分析、K均值聚类与高斯混合聚类均将青少年分为3类,且各亚组的健康模式表征类似。以潜在剖面分析分类结果为例,3类健康模式可命名为“代谢问题高风险组”(29.2%)、“健康问题低风险组”(60.2%)和“行为问题高风险组”(10.6%)。内部评价指标显示,潜在剖面分析与K均值聚类效果优于高斯混合聚类。Logistic回归影响因素分析结果显示,较大年龄、性发育已经启动的青少年潜在代谢问题危险性更高；年龄越大、男性的青少年也更容易发生情绪行为问题。结论 潜在剖面分析等识别健康模式的方法可以较好地识别青少年未来的健康风险。青少年不同维度的健康结局可能相互影响,并表现出聚集性趋势,形成一定的健康模式。针对健康模式的病因学研究,可以识别不同健康结局的共享危险因素,为不同干预措施的“打包”式精准定制提供科学依据,最大化干预措施对多维度健康结局的协同效应。

Abstract

Objective To compare the performance of latent profile analysis (LPA), K-means clustering, and Gaussian mixture model (GMM) clustering in evaluating multidimensional health outcome indicators among adolescents so as to provide evidence-based guidance for selecting appropriate analytical strategies. Methods The data were derived from cross-sectional analyses of a prospective cohort follow-up of a cluster-randomized controlled trial of prenatal micronutrient supplementation. The health outcome indicators consisted of behavioral problem scores, cognitive scores, vital capacity, heart rate, body mass index (BMI), systolic blood pressure, and diastolic blood pressure. LPA and clustering analyses were made separately to classify adolescents into subgroups with distinct health patterns. The performance of each method was assessed using the Davies-Bouldin index and Calinski-Harabasz index. Logistic regression was then applied to explore the shared etiological factors associated with different health patterns. Results A total of 1 767 adolescents were recruited in the analysis, with a median age of 12 years (interquartile range: 11-12); 1 050 (59%) were male, and 1 416 (83%) had initiated puberty. All the three methods—LPA, K-means, and GMM—identified three distinct subgroups with similar health pattern profiles. Based on the LPA results, the subgroups were labeled as: “high risk of metabolic problems” (29.2%), “low risk of health problems” (60.2%), and “high risk of behavioral problems” (10.6%). Internal validation metrics indicated that both LPA and K-means clustering outperformed GMM clustering. Logistic regression revealed that older adolescents and those who had entered puberty were more likely to fall into the metabolic risk group. Additionally, older age and male sex were associated with an increased likelihood of behavioral and emotional problems. Conclusion Methods such as latent profile analysis, which identify health patterns, can effectively detect future health risks in adolescents. These outcomes tend to cluster and influence each other, forming distinct health profiles. Etiological investigations based on health patterns can help identify shared risk factors across different outcomes and provide scientific evidence for the bundled and precision-tailored design of intervention strategies, thereby maximizing synergistic effects on adolescent health.

关键词

潜在剖面分析 / 聚类分析 / 多维健康结局 / 青少年

Key words

latent profile analysis / clustering analysis / multidimensional health outcomes / adolescent

参考文献

[1] CHEN T J, DONG B, DONG Y, et al. Matching actions to needs: shifting policy responses to the changing health needs of Chinese children and adolescents[J]. Lancet, 2024, 403(10438): 1808-1820.

[2] DONG B, ZOU Z, SONG Y, et al. Adolescent health and healthy China 2030: a review[J]. J Adolesc Health, 2020, 67(5S): S24-S31.

[3] CHARZYŃSKA E, HABIBI SOOLA A, MOZAFFARI N, et al. Patterns of work—related stress and their predictors among emergency department nurses and emergency medical services staff in a time of crisis: a latent profile analysis[J]. BMC Nurs, 2023, 22(1): 98.

[4] 张晓楠, 焦珂笛, 苏生利, 等 . 基于潜在剖面分析的男男性行为人群HIV感染者生存质量及相关因素研究[J]. 中华流行病学杂志, 2023, 44(8): 1290-1295.

[5] ZHANG X N, JIAO K D, SU S L, et al. A study on quality of life and related factors among HIV—infected men who have sex with men based on latent profile analysis[J]. Chin J Epidemiol, 2023, 44(8): 1290-1295.

[6] ALTMAN N, KRZYWINSKI M. Clustering[J]. Nat Methods, 2017, 14(6): 545-546.

[7] 赵杰, 路露, 张晓娜, 等 . 聚类分析在护理领域的常用方法及应用进展[J]. 现代临床护理, 2021, 20(9): 75-80.

[8] ZHAO J, LU L, ZHANG X N, et al. Use of cluster analysis in nursing research: a literature review[J]. Mod Clin Nurs, 2021, 20(9): 75-80.

[9] GREVE B, PIGEOT I, HUYBRECHTS I, et al. A comparison of heuristic and model—based clustering methods for dietary pattern analysis[J]. Public Health Nutr, 2016, 19(2): 255-264.

[10] 何玉林, 李旭, 贺颖婷, 等 . 基于最大均值差异的子空间高斯混合模型聚类集成算法[J]. 计算机应用, 2025, 45(6): 1712-1723.

[11] HE Y L, LI X, HE Y T, et al. Subspace gaussian mixture model clustering ensemble algorithm based on maximum mean discrepancy[J]. J Comput Appl, 2025, 45(6): 1712-1723.

[12] ZENG L, DIBLEY M J, CHENG Y, et al. Impact of micronutrient supplementation during pregnancy on birth weight, duration of gestation, and perinatal mortality in rural western China: double blind cluster randomised controlled trial[J]. BMJ, 2008, 337: a2001.

[13] ZHU Z, CHENG Y, ZENG L, et al. Association of antenatal micronutrient supplementation with adolescent intellectual development in rural western China: 14—year follow—up from a randomized clinical trial[J]. JAMA Pediatr, 2018, 172(9): 832-841.

[14] LEUNG P W L, KWONG S L, TANG C P, et al. Test—retest reliability and criterion validity of the Chinese version of CBCL, TRF, and YSR[J]. J Child Psy Psychiatry, 2006, 47(9): 970-973.

[15] 张厚粲. 韦氏儿童智力量表第四版(WISC—Ⅳ)中文版的修订[J]. 心理科学, 2009, 32(5): 1177-1179.

[16] ZHANG H C. The revision of WISC—IV Chinese version[J]. Psy Sci, 2009, 32(5): 1177-1179.

[17] WOHLFAHRT—VEJE C, MOURITSEN A, HAGEN C P, et al. Pubertal onset in boys and girls is influenced by pubertal timing of both parents[J]. J Clin Endocrinol Metab, 2016, 101(7): 2667-2674.

[18] ZENG L, YAN H, CHEN Z. Measurement of the living standards of family in rural area and relationship between wealth index and perinatal care status[J]. J Hyg Res, 2008, 37(6): 714-717.

[19] LIU F, YANG D, LIU Y, et al. Use of latent profile analysis and K—means clustering to identify student anxiety profiles[J]. BMC Psychiatry, 2022, 22(1): 12.

[20] 何选森, 何帆, 樊跃平, 等 . 高维数据聚类数量可视化确定模式[J]. 沈阳航空航天大学学报, 2024, 41(3): 71-84.

[21] HE X S, HE F, FAN Y P, et al. Visualized determination mode for clustering quantity of high—dimensional data[J]. J Shenyang Aerospace Univ, 2024, 41(3): 71-84.

[22] CHRIS FRALEY, ADRIAN E RAFTERY. Model—based clustering, discriminant analysis, and density estimation[J]. J Am Stat Assoc, 2002, 97(458): 611-631.

[23] 付春龍, 盛腾. 常用聚类内部评价指标的对比分析[J]. 电脑编程技巧与维护, 2023(4): 3—5, 9.

[24] FU C L, SHENG T. Comparative analysis of commonly used internal evaluation indicators for clustering[J]. Comput Program Skills Maint, 2023(4): 3—5, 9.

[25] WEI J, LIU S, CHENG Y, et al. Association of infant physical development and rapid growth with pubertal onset among girls in rural China[J]. JAMA Netw Open, 2021, 4(5): e216831.

[26] TIAN J, ZHU Y, LIU S, et al. Associations between life—course household wealth mobility and adolescent physical growth, cognitive development and emotional and behavioral problems: a birth cohort in rural western China[J]. Front Public Health, 2023, 11: 1061251.

[27] 宋月萍, 胡曦元, 龙修琛. 免费营养午餐政策对农村青少年超重的影响———基于中国教育追踪调查初中队列数据[J]. 中华流行病学杂志, 2022, 43(6): 885-891.

[28] SONG Y P, HU X Y, LONG X C. Influence of free nutrition lunch policy on overweight of rural adolescents: based on China education panel survey junior high school cohort data[J]. Chin J Epidemiol, 2022, 43(6): 885-891.

[29] 马宁, 钟盼亮, 党佳佳, 等 . 中国9～18岁汉族儿童青少年感知运动益处与障碍及其与校内体育活动时间的关系[J]. 中华流行病学杂志, 2023, 44(3): 422-429.

[30] MA N, ZHONG P L, DANG J J, et al. Perceived exercise benefits and barriers and their association with physical activity time in Chinese Han children and adolescents aged 9 to 18 years[J]. Chinese J Epidemiol, 2023, 44(3): 422-429.

[31] 尹奎, 彭坚, 张君. 潜在剖面分析在组织行为领域中的应用[J]. 心理科学进展, 2020, 28(7): 1056-1070.

[32] YIN K, PENG J, ZHANG J. The application of latent profile analysis in organizational behavior research[J]. Adv in Psy Sci, 2020, 28(7): 1056-1070.

[33] 游展鸿, 孙国珍, 卢静, 等 . 心房颤动患者身体活动的潜在剖面分析及影响因素研究[J]. 中国全科医学, 2025, 28(12): 1486-1493.

[34] YOU Z H, SUN G Z, LU J, et al. Latent profile analysis and its influencing factors of physical activity in patients with atrial fibrillation[J]. Chin Gen Pract, 2025, 28(12): 1486-1493.

[35] GARTSTEIN M A, PROKASKY A, BELL M A, et al. Latent profile and cluster analysis of infant temperament: comparisons across person—centered approaches[J]. Dev Psychol, 2017, 53(10): 1811-1825.

[36] GAO C X, DWYER D, ZHU Y, et al. An overview of clustering methods with guidelines for application in mental health research[J]. Psychiatry Res, 2023, 327: 115265.

[37] 张新宇, 徐慧琼, 王陈芳, 等 . 熵权逼近理想解排序法综合评价安徽省四地区学生体质健康[J]. 中华预防医学杂志, 2023, 57(7): 997-1003.

[38] ZHANG X Y, XU H Q, WANG C F, et al. Application of entropy weight TOPSIS comprehensive method in the evaluation of students' physical health level[J]. Chin J Prev Med, 2023, 57(7): 997-1003.

[39] 刘励. 儿童青少年体质健康的综合评价及影响因素研究[D]. 武汉: 华中科技大学, 2009.

[40] LIU L. A study on synthetical evaluation of children and adolescents health and its influencing factors[D]. Wuhan: Huazhong University of Science and Technology, 2009.

[41] CADMAN T, ELHAKEEM A, VINTHER J L, et al. Associations of maternal educational level, proximity to green space during pregnancy, and gestational diabetes with body mass index from infancy to early adulthood: a proof—of—concept federated analysis in 18 birth cohorts[J]. Am J Epidemiol, 2024, 193(5): 753-763.

[42] ABOU EL ELLA S S, BARSEEM N F, TAWFIK M A, et al. BMI relationship to the onset of puberty: assessment of growth parameters and sexual maturity changes in Egyptian children and adolescents of both sexes[J]. J Pediatr Endocrinol Metab, 2020, 33(1): 121-128.

[43] HSU P C, HWANG F M, CHIEN M I, et al. The impact of maternal influences on childhood obesity[J]. Sci Rep, 2022, 12(1): 6258.

[44] REINEHR T, ROTH C L. Is there a causal relationship between obesity and puberty?[J]. Lancet Child Adolesc Health, 2019, 3(1): 44-54.

[45] LAKSHMAN R, ZHANG J, ZHANG J, et al. Higher maternal education is associated with favourable growth of young children in different countries[J]. J Epidemiol Community Health, 2013, 67(7): 595-602.

[46] ADHIKARI S, MA J, SHAKYA S, et al. Self—reported emotional and behavioral problems among school—going adolescents in Nepal—a cross—sectional study[J]. PLoS One, 2023, 18(6): e0287305.

[47] ABOOBAKER S, JANGAM K V, SAGAR K J V, et al. Predictors of emotional and behavioral problems among Indian adolescents: a clinic—based study[J]. Asian J Psychiatr, 2019, 39: 104-109.

[48] CHRISTIAN P, SMITH E R. Adolescent undernutrition: global burden, physiology, and nutritional risks[J]. Ann Nutr Metab, 2018, 72(4): 316-328.

[49] TWIG G, TIROSH A, DERAZNE E, et al. Cognitive function in adolescence and the risk for premature diabetes and cardiovascular mortality in adulthood[J]. Cardiovasc Diabetol, 2018, 17(1): 154.

基金资助

国家自然科学基金资助项目(82103867)

中央高校基本科研业务费专项资金资助项目(xxj032025066)

陕西省重点研发计划资助项目(2024SF-YBXM-310)

陕西省创新能力支撑计划资助项目(2023-CX-PT-47)