Human factors testing in a novel automated peritoneal dialysis system.

AI for UI: Designing Error-tolerant Interfaces for Home Dialysis using Artificial Intelligence

GenAI-assisted processes can provide rapid, actionable design mitigations that reduce error likelihood and enhance patient autonomy, establishing a replicable pipeline for producing heuristic-driven design libraries across diverse medical device contexts.

Characterisation and management of therapy-interrupting system errors associated with the use of the HomeChoice cycler for automated peritoneal dialysis

Methodological and technical education of nurses is required to address challenges of technology integration in peritoneal dialysis (PD). We analysed the nature of therapy interrupting system errors in automated peritoneal dialysis (APD) cyclers of a German PD reference care centre and provide applicable resolution proposals. We documented therapy interrupting system errors of HomeChoice Pro and HomeChoice Claria cyclers (Baxter, BX, Unterschleißheim, Germany) including the years 2015-2023. We systematically characterise and suggest resolution strategies to these errors based on routinely collected health data and user experience to safely guide nephrology professionals and patients during troubleshooting. The ‘reporting of studies conducted using observational routinely collected health data’ (RECORD)-statement was considered. We identified 49 different therapy interrupting system errors which were assigned to eight major categories: Errors associated to the pressure chamber (N=18), the Digitalboard/EEPROM (N=18, Electrically Erasable Programmable Read Only Memory), air in set (N=3), opening of the set’s door (N=3), those associated to the energy system (N=3), the heater system (N=3) and a variety of errors resulting in an abrupt termination or non-commencement of the treatment session without clear group assignment (undefined, N=1) or without sufficient documentation. In total, 138 error incidents were documented during the observational timeframe. In frequential order errors were attributed to the pressure chamber (21.74%,), and air in set (19.57%) where specifically error 2240 occurred in 17.39% of total cases, the Digitalboard/EEPROM (17.39%), the energy or battery (16.67%; No. 1032 in 19/23 cases), the set’s door (7.25%) and the heater (4.35%). In 17 cases (12.32%) no error number was documented by the patient or staff. One error cause was registered as undefined (0.72%). Overall, errors were rarely occurring in <0.25% of treatment sessions performed in the observational timeframe. These findings may be useful to guide training purposes for nephrology personnel and persons doing APD. For the manufacturer, the information provided may be helpful to improve error management and information transmission to the persons doing APD with the HomeChoice devices.

Artificial Intelligence in Peritoneal Dialysis: Applications, Algorithms, and Future Directions.

BACKGROUND Peritoneal dialysis (PD) naturally lends itself to artificial intelligence (AI) integration due to its generation of dense, longitudinal and structured data. PD outcomes are strongly influenced by modifiable factors such as dialysate composition, dwell time, and fluid balance, making prediction directly actionable. SUMMARY Recent investigations have demonstrated the potential of machine learning (ML) and deep learning (DL) approaches across multiple domains of PD care. Applications include pre-dialysis patient stratification, prediction of technique failure, monitoring of dialysis adequacy, and assessment of fluid status. AI models have also been developed for the early detection of peritonitis and prediction of complications, hospitalizations, and mortality, frequently outperforming conventional statistical methods. In parallel, AI-driven chatbots and digital platforms have shown promise in enhancing patient education, engagement, and adherence. While these findings highlight significant potential, most studies to date are single-center, exploratory and limited in scale, underscoring the need for rigorous external validation and systematic clinical implementation. KEY MESSAGES AI holds considerable promises for improving outcomes in PD by enabling earlier risk identification, individualized therapy, and strengthened patient support. Future directions include integration with electronic health records, remote monitoring, and multi-omics data. Careful validation, ethical safeguards, and equitable deployment will be essential to realize AI's role as a transformative tool in PD care.

Assessing Patients' ability to learn medication protocols during a hemodialysis session.

The nephrologist in the present and near future: between algorithms and autonomy.

Frequency and Risk Factors for Intradialytic Hypotension in Patients Undergoing Maintenance Hemodialysis

Hemodialysis (HD) patients and providers have long struggled with intra-dialytic hypotension (IDH), a complication that continues to challenge the safety and effectiveness of HD. IDH is associated with chronic end-organ damage and remains an independent risk factor for mortality among HD patients. Persistent hemodynamic instability may also accelerate the decline of residual renal function, further worsening clinical outcomes. Because individuals on dialysis already face markedly increased cardiovascular mortality and morbidity, identifying the prevalence and contributing factors of IDH has become an important clinical and academic priority. This observational study was carried out in the Hemodialysis Unit of Nepal Medical College and Teaching Hospital over one year (Baishakh 2079 to Chaitra 2079 B.S, April 2022 to March 2023 AD) without interventions or the collection of patient material. Ethical approval was obtained from the Institutional Review Committee. During each dialysis session, patients were assessed for pre- and post-dialysis weight, blood pressure, and heart rate at multiple intervals, ultrafiltration volume, and any clinical events or medical interventions related to intra-dialytic hypotension. A total of 110 patients participated, contributing 5,232 dialysis sessions. The incidence of IDH varied widely depending on the definition used, measuring 14.47% and 9.42% under stricter criteria, and rising to 62.8% with a more liberal definition. Although HD remains a widely used and generally safe modality, complications such as IDH persist and can be life-threatening. As complications and their associated factors continue to change over time, a more vigilant and systematic approach is needed to identify, monitor, and address them effectively. This highlights the need for ongoing evaluation in HD practice.