The need for early detection and intervention for age-related health deterioration via home-based health monitoring systems is growing with global population aging. Health deterioration is typically evaluated using questionnaires, but obtaining frequent responses is difficult. Although stair walking involving a high physical load is suitable for assessing health deterioration, uniform assessment using home-based systems is difficult because different stair designs result in different movements. There is currently insufficient research on differences in evaluation performance between different stair designs. The current study aimed to develop an objective index to assess difficulties in activities of daily living (ADLs) as an indicator of health deterioration by observing stair walking, and to investigate the effects of stair dimensions on the relationship between the index and health deterioration. We hypothesized that knee joint positions during stair ascent would be associated with health deterioration, and developed the “Stepping in Level” index, reflecting the knee joint position when stepping in, obtained using a markerless depth information-based system. Seventy elderly people ascended two types of stairs (steep: 33.7°; shallow: 28.2°) and completed a questionnaire regarding ADL difficulty. We used a generalized linear mixed model to investigate the ability of “Stepping in Level” to explain ADL difficulty. The results revealed that “Stepping in Level” was able to explain ADL difficulty, and steep stairs were more suitable than shallow stairs for assessment. Our system enables daily assessment of difficulty in ADLs and detection of subtle changes in daily life. The current findings could inform design guidelines for health monitoring.

随着家庭人口老化，通过家庭健康监测系统对与年龄有关的健康恶化的早期检测和干预的需求正在增长。通常使用问卷调查评估健康状况恶化，但是很难获得频繁的回复。尽管涉及高物理负荷的楼梯行走适合评估健康状况恶化，但是使用基于家庭的系统进行统一评估仍然很困难，因为不同的楼梯设计会导致不同的运动。当前，关于不同楼梯设计之间的评估性能差异的研究不足。当前的研究旨在建立一个客观指数，以评估日常生活活动中的困难（ADL），以通过观察楼梯行走来评估健康状况的恶化，并研究楼梯尺寸对指数与健康恶化之间关系的影响。我们假设在楼梯上升过程中膝关节位置会导致健康状况恶化，并开发了“步入水平”指数，该指数反映了使用无标记深度信息系统获得的步入时膝关节的位置。七十位老年人登上两种类型的楼梯（陡峭：33.7°；浅坡：28.2°），并完成了有关ADL困难的问卷调查。我们使用广义线性混合模型来研究“逐步学习”来解释ADL难度的能力。结果表明，“阶梯式进阶”能够解释ADL的困难，陡峭的台阶比浅台阶更适合评估。我们的系统可以每天评估ADL的难度，并检测日常生活中的细微变化。当前的发现可以为健康监测的设计指南提供依据。