Supplementary Material for: Prognostic Value of Remote Monitoring Alarms for Technique Failure in Automated Peritoneal Dialysis

Introduction: While remote patient monitoring (RPM) is associated with decreased hospitalization and technique failure in automated peritoneal dialysis (APD), the prognostic utility of individual alarm patterns remains unclear.
Methods: We performed a retrospective, single-center cohort study of adult APD patients managed with the Sharesource RPM platform at National Cheng Kung University Hospital. Three predefined red-flag alarms, including initial drain variance > 50%, lost therapy volume > 10%, and lost treatment time > 30 min, were recorded from September 2019 to May 2020. Alarm burden was expressed as episodes per patient-month and categorized into tertiles. Patients were followed for three years for death-censored technique failure, defined as permanent transfer to hemodialysis. Associations were evaluated using logistic regression. Fine and Gray competing-risk models and generalized estimating equations (GEE) served as sensitivity analyses.
Results: Sixty-seven patients (mean age 45 ± 13 years; 61.2% male) generated 272 alarm episodes. The initial drain variance accounted for 54.8% of alarms, lost treatment time for 39.3%, and lost therapy volume for 5.9%. Technique-failure occurred in eight patients (11.9%), with rates rising across alarm-frequency tertiles (4.5%, 9.1%, and 21.7%; P for trend = 0.074). In multivariable analysis, initial drain variance > 50% independently predicted technique failure (adjusted odds ratio [OR] 1.48; 95% confidence interval [CI] 1.13–2.04; P = 0.006). Results were consistent in sensitivity analyses (Fine and Gray subdistribution hazard ratio 1.55, 95% CI 1.07–2.25; GEE OR 2.58, 95% CI 1.00–6.62). The other two alarm types were not significantly associated with outcomes.
Conclusion: An RPM alarm for initial drain variance > 50% emerged as an early marker of technique failure in APD. Incorporating alarm-pattern analysis into routine RPM review may facilitate targeted surveillance and timely intervention for high-risk patients.

引言：尽管远程患者监测（Remote Patient Monitoring, RPM）与自动化腹膜透析（Automated Peritoneal Dialysis, APD）患者的住院率降低、技术失败率下降相关，但单个警报模式的预后应用价值仍不明确。
方法：本研究为回顾性单中心队列研究，纳入国立成功大学医院内使用Sharesource远程患者监测平台管理的成年自动化腹膜透析患者。研究记录了2019年9月至2020年5月期间的3项预定义红色警戒警报：初始引流偏差>50%、治疗丢失量>10%及治疗丢失时间>30分钟。以每患者月的警报发作次数表示警报负荷，并将其分为三分位组。对患者进行为期三年的随访，终点为死亡删失的技术失败（定义为永久转为血液透析）。采用逻辑回归评估警报与结局的关联，并使用Fine和Gray竞争风险模型及广义估计方程（Generalized Estimating Equations, GEE）进行敏感性分析。
结果：共纳入67例患者（平均年龄45±13岁；男性占比61.2%），总计发生272次警报发作。其中初始引流偏差相关警报占54.8%，治疗丢失时间相关警报占39.3%，治疗丢失量相关警报占5.9%。共有8例患者（11.9%）出现技术失败，且该发生率随警报频率三分位组升高而上升（分别为4.5%、9.1%和21.7%；趋势检验P=0.074）。多变量分析显示，初始引流偏差>50%可独立预测技术失败（校正后优势比[OR]=1.48；95%置信区间[CI]=1.13~2.04；P=0.006）。敏感性分析结果与之一致（Fine和Gray亚分布风险比=1.55，95%CI=1.07~2.25；GEE分析OR=2.58，95%CI=1.00~6.62）。其余两类警报与结局无显著关联。
结论：针对初始引流偏差>50%的远程患者监测警报，可作为自动化腹膜透析患者技术失败的早期预测标志物。将警报模式分析纳入常规远程患者监测评估流程，有助于为高危患者开展针对性监测与及时干预。