Collin Claire, Eyraud Clara, Martin Philippe, Michel Morgane, Le Roux Enora, Alberti Corinne
Université Paris Cité, Inserm, ECEVE, Paris, F-75010, France.
AP-HP, Nord-Université Paris Cité, Inserm, Hôpital Universitaire Robert Debré, Clinical Epidemiology Unit, CIC 1426, Paris, F-75019, France.
BMC Med. 2025 Jul 2;23(1):400. doi: 10.1186/s12916-025-04245-1.
Determining the success of population health interventions often involves assessing multiple, multidimensional outcomes rather than a single one, which presents significant methodological challenges under the evidence-based medicine paradigm. This scoping review examines outcome selection, analysis, and interpretation, and the accuracy of conclusions in complex digital health interventions promoting health among adolescents and young adults (DHI-AYA).
A comprehensive search of PubMed, EMBASE, ClinicalTrials.gov, PsycINFO, and CINAHL identified DHI-AYA implemented between 2017 and 2023. Studies were categorised by methodological choice regarding outcome hierarchical position: unique primary, multiple primary, or non-hierarchised outcomes. Outcomes were further classified into effectiveness, process, or economic categories. The authors' conclusions on intervention success were compared with conclusions drawn by the research team based on the reported outcome analysis strategy. Secondly, four analytical strategies were applied to a subset of selected interventions to illustrate the impact of outcome hierarchical position and number on conclusions about intervention success.
Analysis of 100 studies linked to 26 DHI-AYA identified 251 distinct outcomes: 164 effectiveness, 78 process, and 9 economic outcomes. Seven interventions were evaluated using a unique primary outcome, 10 using multiple primary outcomes, and 9 using multiple non-hierarchised outcomes. Primary and secondary outcomes were predominantly effectiveness endpoints. The research team reclassified nine interventions (35%) deemed successful by authors as non-conclusive due to statistically conflicting results across outcomes. Most interventions deemed non-conclusive by the research team were evaluated using non-hierarchised outcomes (7/10, 70%). The choice of outcome analysis strategy substantially affected conclusions on intervention success.
Discrepancies in intervention success assessments highlight the need for enhanced transparency, robustness, and trustworthiness in conclusion-drawing processes. In response, five methodological proposals are formulated: (1) developing core outcome sets specific to population health intervention research (PHIR), (2) collaboratively selecting multidimensional outcomes through a steering committee that accounts for stakeholder preferences and existing theoretical models, (3) exploring multi-criteria decision analysis and consensus-driven methods to transparently combine outcomes, (4) enhancing methodological reporting through intervention development and evaluation to improve scientific integrity and reproducibility, and (5) increasing PHIR expert involvement in ethics, funding, and evaluation committees to improve recognition of evidence produced in this field.
CRD42023401979.
确定人群健康干预措施的成功与否通常涉及评估多个多维结果而非单一结果，这在循证医学范式下带来了重大的方法学挑战。本范围综述考察了在促进青少年和青年健康的复杂数字健康干预措施（DHI - AYA）中结果的选择、分析和解释，以及结论的准确性。
全面检索了PubMed、EMBASE、ClinicalTrials.gov、PsycINFO和CINAHL，以确定2017年至2023年期间实施的DHI - AYA。根据结果层次位置的方法选择对研究进行分类：唯一主要结果、多个主要结果或非层次化结果。结果进一步分为有效性、过程或经济类别。将作者关于干预成功的结论与研究团队根据报告的结果分析策略得出的结论进行比较。其次，对一部分选定的干预措施应用了四种分析策略，以说明结果层次位置和数量对干预成功结论的影响。
对与26项DHI - AYA相关的100项研究进行分析，确定了251个不同的结果：164个有效性结果、78个过程结果和9个经济结果。7项干预措施使用唯一主要结果进行评估，10项使用多个主要结果，9项使用多个非层次化结果。主要和次要结果主要是有效性终点。研究团队将作者认为成功的9项干预措施（35%）重新分类为非结论性，原因是不同结果之间存在统计学上相互矛盾的结果。研究团队认为非结论性的大多数干预措施使用非层次化结果进行评估（7/10，70%）。结果分析策略的选择对干预成功的结论有重大影响。
干预成功评估中的差异凸显了在得出结论过程中提高透明度、稳健性和可信度的必要性。为此，制定了五项方法学建议：（1）制定针对人群健康干预研究（PHIR）的核心结果集；（2）通过一个考虑利益相关者偏好和现有理论模型的指导委员会共同选择多维结果；（3）探索多标准决策分析和共识驱动方法以透明地整合结果；（4）通过干预开发和评估加强方法学报告，以提高科学完整性和可重复性；（5）增加PHIR专家参与伦理、资助和评估委员会，以提高对该领域所产生证据的认可度。
PROSPERO注册号：CRD42023401979。
