From Data to Better Service Care: Building a quality management system for mealtime safety from five years of routine observation data in NGO residential care for older adults with visual impairment
从数据到更好的照护：基于五年常规观察数据，为视障长者住宿照护机构建立膳食安全质量管理体系

摘要

提供住宿式长期照护服务的非政府组织（NGO）往往积累了大量常规临床记录，却未能将其转化为真正运作的质量管理体系。本文报告了一家在香港经营两所视障长者专科住宿照护院舍的NGO所开展的服务发展项目：该机构分析了8,834份结构化的进餐观察记录（涵盖333位住客、86位职员观察者，时间跨度为2020年10月至2026年8月），以了解进食安全、营养摄入及吞咽风险状况。通过四层分析——发生率、关联性、多变量调整及住客层面变异与趋势分析——研究发现吞咽风险征象往往呈聚集性而非独立出现，嗜睡/疲倦与不良坐姿是与低摄入量相关的最强可改善因素，而机构层面的平均值往往掩盖了住客之间的巨大差异。这些发现被纳入一套三层次质量架构（质量控制、质量管理、质量改进）、一个五层管治架构、一套平衡的八项指标体系、一个四级风险分层与升级路径，以及一个六阶段数据决策流程，并通过一次小型计划-实施-研究-行动（PDSA）试验及一份为期十二个月的路线图加以落实。由于所有依据来自单一机构的回溯性观察服务数据集，本文从概念阐释与实践改进的角度进行撰写，属于服务发展报告而非对照研究。超出原始资料所支持范围的陈述，均在文中标注为‘潜在含义’。文末就服务规划、人力培训、跨专业协作及管治等方面，向希望将常规前线数据转化为具纪律性、可持续学习循环的医疗及社会照护管理者提出建议。

Abstract

Non-governmental organisations (NGOs) providing residential long-term care often accumulate large volumes of routine clinical documentation without converting it into a functioning quality management system. This article reports a service-development initiative from an NGO operating two specialist residential homes for older adults with visual impairment in Hong Kong, which analysed 8,834 structured mealtime-observation records (333 residents, 86 staff observers, October 2020–August 2026) to understand feeding safety, nutritional intake, and swallowing risk. A four-layer analysis — prevalence, association, multivariable adjustment, and temporal/resident-level variation — found that swallowing-risk signs cluster rather than occur independently, that drowsiness/fatigue and poor positioning are the strongest modifiable correlates of low intake, and that organisational averages conceal wide resident-to-resident variation. These findings were embedded into a three-discipline quality architecture (quality control, quality management, quality improvement), a five-tier governance cascade, a balanced set of eight measures, a four-tier risk-stratification and escalation pathway, and a six-stage data-to-decision flow, tested through a small Plan-Do-Study-Act (PDSA) pilot and a twelve-month roadmap. Because the underlying evidence is a single organisation’s retrospective, observational service dataset, the article is framed as a conceptual and practice-improvement account rather than a controlled study, and statements beyond what the source material supports are labelled as potential implications. Recommendations are offered for healthcare and social-care managers seeking to convert routine frontline data into a disciplined, sustainable learning cycle covering service planning, workforce development, interdisciplinary collaboration, and governance.