Applied Nephrology Master Series

Chapter 18 Remote Monitoring, Connected Cyclers and the Future of Peritoneal Dialysis

CHAPTER MISSION Convert digital PD from a device feature into a bedside care system: know what connected cyclers actually measure, distinguish monitoring from management, interpret remote signals within goal-directed PD, build safe alert workflows, recognize where evidence supports benefit, and define the safeguards required before data-driven treatment adaptation.

Figure 18.1 — Connected cycler, remote monitoring, remote patient management and adaptive PD are different layers of responsibility. The clinical intervention begins when information is reviewed and acted upon.

MASTER PRINCIPLE DATA TRANSMISSION IS NOT TREATMENT. A remote signal becomes clinical care only after data-quality validation, contextual interpretation, accountable human review, appropriate action and reassessment.

0. One-page chapter map

Table 18.1 — The eight decisions that govern digital PD.

Decision	Core question	Bedside output
1. Define	Connected cycler, RPM, remote management or adaptive pathway?	Correct intervention named
2. Acquire	Which treatment and patient data are available?	Signal inventory
3. Validate	Are data complete, plausible and correctly attributed?	Trusted information
4. Interpret	What mechanism could explain the trend?	Clinical hypothesis
5. Prioritize	Is the issue routine, time-sensitive or urgent?	Triage level
6. Act	Who owns the alert and what response is allowed?	Accountable intervention
7. Reassess	Did the patient and treatment trajectory improve safely?	Closed clinical loop
8. Govern	Are burden, equity, privacy, cost and model performance measured?	Sustainable digital service

Learning outcomes

Distinguish connected APD, remote patient monitoring, remote patient management, telehealth, adaptive PD and closed-loop treatment.
Interpret cycler data—treatment time, fills, drains, dwell behavior, ultrafiltration, alarms, bypasses and completion—without treating any single variable as a diagnosis.
Build a remote volume-status pathway that verifies measurement quality and integrates residual kidney function, intake, symptoms and PD prescription.
Use repeated drain alarms and incomplete treatments to trigger mechanism-based catheter troubleshooting rather than reflex prescription escalation.
Apply current randomized and real-world evidence without overstating observational or composite-endpoint results.
Explain why remote monitoring benefit depends on staffing, response pathways and patient/caregiver participation rather than connectivity alone.
Design alert ownership, escalation, documentation and after-hours safety rules.
Differentiate adjunctive technologies such as bioimpedance, lung ultrasound and effluent sensing from validated treatment triggers.
Assess AI/ML models by calibration, external validation and clinical utility—not discrimination alone.
Define a human-governed adaptive-PD pathway and explain why fully autonomous prescription change remains investigational.

EVIDENCE POSTURE ISPD goal-directed prescribing remains the clinical frame. Randomized evidence now includes the 2025 cluster trial of RM-enabled APD and a 2025 app-based fluid-management trial; smaller RCTs support patient satisfaction, workflow and selected quality-of-life effects. Systematic reviews and large observational cohorts generally favor hospitalization or time-on-PD outcomes but remain heterogeneous. A 2026 nationwide Taiwan cohort strengthens real-world evidence yet cannot establish causality. AI, effluent sensing and closed-loop adaptation remain emerging or investigational. [1–12]

1. Core concept: home therapy is continuous, clinic assessment is episodic

PD is delivered every day, but conventional follow-up samples the therapy intermittently. Between visits, treatment can shorten, drainage can deteriorate, ultrafiltration can drift, urine output can decline, symptoms can evolve, and the patient or caregiver can struggle with a regimen that looked acceptable at the last clinic encounter. Connected systems reduce this information gap by making parts of treatment delivery visible between visits. The clinical challenge then shifts from “Can we see the data?” to “Can we recognize a meaningful change early enough to improve a patient-important outcome?” [1–4]

Table 18.2 — Visibility does not equal diagnosis.

Remote finding	What it directly shows	What it does NOT prove
Low delivered UF	Cycler-measured net fluid removal fell	That the patient is volume overloaded
Slow drains	Drain duration/flow changed	That the catheter migrated
Lost treatment time	Prescribed exposure was not delivered	Why treatment was incomplete
Weight increase	Body weight rose	That the gain is extracellular fluid
Missed upload	No data reached the platform	That the patient missed PD
Repeated alarms	The machine encountered recurrent events	That the cause is patient behavior

BEDSIDE TRANSLATION Connected data are excellent at describing WHAT happened to treatment delivery. They are often weaker at explaining WHY it happened.

2. Definitions: separate technology from care

Table 18.3 — Digital-PD terminology.

Term	Operational meaning	Clinical responsibility
Connected cycler	Device stores and transmits treatment data	Makes treatment visible
Remote patient monitoring (RPM)	Remote review of transmitted device and/or patient data	Identifies patterns requiring assessment
Remote patient management	RPM plus communication, clinical intervention and follow-up	Turns information into care
Telehealth	Broader remote communication/consultation/education	May or may not include device data
Adaptive PD	Confirmed information triggers a prespecified, governed treatment adaptation	Requires protocol + clinician accountability
Closed-loop PD	Sensing, decision and treatment change occur with minimal case-by-case human approval	Investigational

These labels should not be used interchangeably. A connected device can upload data without any organized review. A remote-monitoring program can review data but fail to act in time. An adaptive pathway requires explicit confirmation, response rules, responsibility and reassessment. Fully closed-loop prescription change represents a substantially higher safety and governance threshold than current routine RPM. [3,4]

TERMINOLOGY RULE When evaluating evidence, ask what the intervention actually contained: connectivity, home measurements, nurse review, clinician feedback, prescription modification, education, or all of them together.

3. What a connected APD cycler can show

Figure 18.2 — Remote data are layered. Cycler and home measurements gain meaning only when combined with periodic membrane, residual-kidney-function, laboratory and clinical assessment.

Table 18.4 — Connected-cycler data dictionary.

Data stream	Useful clinical question	Common interpretive trap
Prescribed vs delivered time	Was planned treatment completed?	Assuming shortened therapy is intentional nonadherence
Fill volumes	Were programmed fills delivered?	Ignoring manual/daytime exchanges not captured
Drain volumes / profiles	Is outflow becoming slower or incomplete?	Calling every slow drain catheter migration
Dwell duration	Are alarms shortening effective dwell exposure?	Treating clock time as equivalent to transport adequacy
Ultrafiltration	Is delivered UF trending down?	Equating low UF with clinical hypervolemia
Bypass / lost dwell / lost volume	Is the cycler aborting or shortening therapy?	Ignoring position, constipation or setup problems
Alarm frequency/type	Is there a recurring treatment barrier?	Treating alert count as severity
Treatment completion	Was the session finished?	Ignoring data-transfer failure or planned deviations

MEASUREMENT RULE The platform should preserve the distinction between prescribed, delivered and clinically effective therapy.

4. Data quality comes before clinical interpretation

Remote care can fail before physiology is considered. Connectivity loss, incorrect device clock, duplicate records, scale calibration, blood-pressure technique, undocumented manual exchanges, caregiver-entered values, recent hospitalization or a changed prescription can create a false trajectory. Missingness itself may be informative, but it must first be classified as technical, operational, social or clinical. [3,13]

Table 18.5 — Data-quality checklist before acting.

Check	Question	If abnormal
Identity	Is this the correct patient and device?	Stop; correct attribution before action
Connectivity	Are expected nights uploaded?	Check network/device before labeling missed treatment
Prescription version	Does platform order match current clinical order?	Reconcile version and effective date
Completeness	Are manual exchanges/home measures missing?	Obtain missing context
Plausibility	Is the value physiologically/device plausible?	Repeat/verify measurement
Timing	Did acute illness, travel or admission alter the routine?	Interpret within event timeline
Technique	Were weight/BP measured consistently?	Retrain or remeasure

DO NOT AUTOMATE A data-quality failure must not become a treatment change. No adaptive pathway is safer than the data on which it acts.

5. The most useful remote target: treatment delivery and adherence barriers

Connected APD can identify shortened sessions, lost treatment time, repeated bypasses and missing treatment completion. This improves the accuracy of the question “What was delivered?” compared with relying on recall alone. However, a discrepancy should trigger curiosity rather than blame. Causes include alarms, pain, sleep disruption, caregiving responsibilities, electricity/network problems, misunderstanding, deliberate shared decisions, work/travel constraints, depression, cognitive decline or machine setup difficulty. [4,14]

Table 18.6 — Incomplete treatment: remote differential.

Pattern	Possible cause	Best first response
Repeated early termination	Alarm burden, sleep intolerance, symptoms	Ask what happened; inspect alarm sequence
No upload but patient reports treatment	Connectivity/platform failure	Verify device locally
Frequent bypass drain	Poor outflow, constipation, posture	Chapter 12 troubleshooting
Shortened cycles after prescription change	New intolerance or program error	Reconcile prescription + symptoms
Irregular missed nights	Life burden / support problem	Nonjudgmental barrier assessment
Progressive loss of delivered time	Mechanical or caregiver decline	Escalate early before technique loss

CLINICAL PEARL The most actionable remote “adherence” intervention is often solving the barrier that made the prescribed treatment unlivable.

6. Remote volume management: detect trajectories, not magic thresholds

Volume overload is a plausible high-value target because useful signals already exist: body weight, blood pressure, symptoms, urine output, delivered ultrafiltration, glucose/icodextrin use and prescription changes. The 2025 CKD-PD app randomized trial showed that monitored hydration metrics led to substantially more early clinical interventions for overhydration and fewer all-cause and volume-overload hospitalizations, without a significant survival or technique-failure difference. No universal remote threshold for weight, BP or UF has been validated across PD populations. [6]

Table 18.7 — Remote volume signal → confirmation.

Signal	Required context	Potential response
Weight rising over days	Scale quality, edema, intake, constipation, nutrition	Clinical review of sodium/fluid and prescription
BP rising	Technique, pain, medications, autonomic status	Confirm pattern; integrate volume phenotype
Delivered UF falling	Drain function, glucose exposure, membrane transport, recent prescription	Chapter 8 mechanism assessment
Urine falling	Collection reliability, diuretics, acute illness	Quantify RKF; reassess total therapy
Dyspnea / orthopnea report	Oxygenation, cardiac/lung cause, hydrothorax	Urgent assessment when clinically indicated
Repeated hypertonic use	Why stronger solutions are needed	Investigate sodium intake, RKF loss, membrane/flow/leak problems

Flowchart 18.2 — Remote volume-overload signal. A trend triggers verification and clinical phenotyping before intervention; severe symptoms bypass routine remote management.

CROSS-REFERENCE Use Chapter 8 for definitive volume, sodium and ultrafiltration-failure reasoning. Remote monitoring changes WHEN the problem may be recognized—not the underlying physiology.

7. Drain alarms and catheter dysfunction: remote pattern recognition

Repeated slow-drain alerts, initial-drain variance, bypasses, reduced effective dwell time and lost therapy volume can reveal evolving mechanical dysfunction before the next clinic visit. A 2026 observational study found that one alarm phenotype—marked initial-drain variance—was associated with later loss from PD, but such thresholds are platform- and study-specific and should not be generalized as universal clinical cutoffs. Alarm data are best used to prioritize assessment for constipation, posture, catheter position, fibrin, kinking, omental entrapment or other causes. [15]

Table 18.8 — Alarm phenotype → bedside hypothesis.

Remote pattern	Think first	Next step
Longer drain time only	Constipation/posture/low flow	Bowel + position review
Lost treatment time + slow drains	Repeated mechanical interruption	Chapter 12 structured pathway
Low UF + slow drains	Incomplete drainage can mimic UF failure	Correct flow before changing osmotic prescription
Alarms after new prescription	Fill/drain intolerance or programming issue	Review change timing and settings
Sudden multiple alarm types	Setup/power/device problem	Technical check before medical escalation
Persistent pattern with pain	Mechanical or intra-abdominal problem	Clinical assessment; image/procedure as indicated

Flowchart 18.3 — Repeated drain alarms and incomplete APD. The platform identifies a reproducible pattern; the clinician still localizes the mechanism.

8. Peritonitis and catheter infection: monitoring can accelerate contact, not replace diagnosis

Cloudy effluent, pain, fever, exit-site abnormalities, altered drainage or a future turbidity sensor should prompt rapid communication and the established diagnostic pathway. Current ISPD peritonitis criteria and catheter-related infection definitions remain the standard; a negative device signal cannot exclude infection in a symptomatic patient. Emerging optical effluent sensing has shown the possibility of pre-symptomatic detection, but false alerts and unproven outcome impact currently limit routine treatment-directed use. [16–18]

Table 18.9 — Digital infection signals.

Signal	What it can do	What must still happen
Patient photo of cloudy effluent	Accelerate recognition/contact	Effluent cell count/differential/culture per Chapter 10
Exit-site photo	Support remote triage and follow-up	Direct exam/culture/ultrasound when indicated per Chapter 11
Cycler UF/drain change	Nonspecific early clue	Do not diagnose peritonitis from cycler data
Optical/turbidity sensor	May identify abnormal effluent earlier	Clinical validation; symptoms and ISPD testing still govern
Temperature/symptom entry	Adds context	Assess severity and alternative infection source

SAFETY BOUNDARY A negative sensor or “normal” cycler night must never overrule new abdominal pain, cloudy effluent or systemic illness.

9. Randomized evidence: what has actually been shown?

Table 18.10 — Key randomized evidence.

Study	Intervention	What improved	What did not establish
Jung et al., 2021	Multicenter RPM-APD vs traditional APD	Selected patient/staff satisfaction and some clinical/process measures	Definitive mortality/technique benefit
Uchiyama et al., 2022	Randomized crossover connected RPM	Treatment satisfaction, convenience, resource use; more prescription modification	Hard-outcome benefit; very small sample
Paniagua et al., 2025	21-hospital cluster-RCT RM-APD	CV/fluid/insufficient-dialysis composite and several secondary outcomes; fewer deaths/hospitalizations observed	First broad composite endpoint was not significantly different; connectivity alone not isolated
Anutrakulchai et al., 2025	CKD-PD app + monitored hydration metrics	More timely overhydration interventions; fewer hospitalizations	Survival or technique-failure benefit
Cely et al., 2026	MyPD mobile app pilot RCT	Engagement and usability	Outcome-proven disease-modifying effect

The Paniagua trial materially strengthened the field but should be taught with precision. It cluster-randomized 21 hospitals and enrolled 801 patients. Its first broad composite endpoint did not significantly differ, whereas a second cardiovascular/fluid-overload/insufficient-dialysis composite favored RM-APD and several secondary outcomes were also favorable. The intervention included organized clinical review and response, so the trial evaluates a remote-care pathway rather than “connectivity” in isolation. [5]

TRIAL INTERPRETATION Do not summarize the 2025 cluster trial as “all endpoints positive.” The evidence is stronger when the exact endpoints and care-pathway components are kept visible.

10. Observational evidence: useful, but confounding remains

Earlier multicenter and single-center cohorts repeatedly associated RPM with fewer hospitalizations, fewer unplanned visits, better technique retention or lower resource use. Meta-analyses also report favorable pooled associations. Yet remote monitoring is not randomly distributed: programs differ in staffing, training, patient selection, platform access and contemporaneous quality improvement. These factors can produce healthy-user, center and treatment-era effects. [8–12,19–22]

A 2026 nationwide Taiwan study of 2,477 incident APD patients associated RPM-enabled APD with lower all-cause mortality, less transfer to HD and fewer hospitalizations compared with an earlier conventional-APD era. The scale is important, but the nonconcurrent eras and observational design prevent a causal conclusion. [7]

Table 18.11 — How to read RPM evidence.

Design	Strength	Main limitation
Individual RCT	Best control of patient-level confounding	Often small and short
Cluster RCT	Tests real program-level workflow	Few clusters; contamination/attention effects
Contemporary cohort	Real-world effectiveness	Selection and residual confounding
Historical-era cohort	Large implementation signal	Treatment-era changes confound effect
Meta-analysis	Summarizes direction/heterogeneity	Can pool dissimilar telehealth interventions
Cost model	Policy planning	Highly assumption- and setting-dependent

11. Bioimpedance, lung ultrasound and home sensors: adjuncts, not remote prescriptions

Bioimpedance can quantify body-water compartments and reveal subclinical overhydration. PD randomized data show improvement in hydration measures with bioimpedance-guided care without consistent evidence of improved survival or technique outcomes. Lung ultrasound identifies B-lines and pulmonary congestion but B-lines are not specific for cardiogenic volume excess; PD-specific remote acquisition and action thresholds are not validated. Weight, BP, oxygen saturation, activity and symptom tools add context but inherit measurement error, adherence and specificity limitations. [23–25]

Table 18.12 — Multimodal sensing maturity.

Technology	Best current role	Why not a stand-alone trigger
Home weight/BP	Trend and context	Nonspecific; technique matters
Bioimpedance	Body-water phenotype	Physiologic improvement ≠ proven hard-outcome benefit
Lung ultrasound	Pulmonary-congestion phenotype	B-lines have multiple causes; remote acquisition variable
Wearables/activity	Functional trajectory / acute illness context	Device and behavior dependent
Pulse oximetry	Acute respiratory context	Not specific for fluid overload
Effluent optical sensing	Early abnormal-effluent signal	False alerts; clinical impact not yet proven

12. Remote prescription modification: the order must remain traceable

Some connected platforms allow a clinician to modify device settings remotely. This is operationally powerful because a validated change can reach the cycler without requiring the patient to reprogram a complex prescription. The clinical order nevertheless requires the same safeguards as any dialysis prescription: correct patient, indication, effective date, version control, authorization, patient/caregiver communication, confirmation that the device received the change, and a defined reassessment endpoint. [3,4]

Table 18.13 — Remote prescription-change safety.

Before change	During change	After change
Confirm problem and mechanism	Use authorized clinical workflow	Verify transmission and applied settings
Confirm patient/device identity	Record exact prescription version	Confirm patient/caregiver understanding
Review BP/volume/RKF/symptoms	Avoid simultaneous untraceable changes	Review next treatment data
Define expected benefit + harm	Keep rollback/override path	Document response and adverse effects

VERSION-CONTROL RULE A remote prescription is not complete when the clinician clicks “send.” It is complete when the correct patient receives the intended settings and the clinical response is verified.

13. Alert architecture: every signal needs an owner

Figure 18.3 — Safe remote alerting requires validation, contextual interpretation, accountable action, reassessment and learning. Alerts without ownership create risk rather than safety.

Table 18.14 — Minimum alert-governance fields.

Field	Question the program must answer	Failure if absent
Owner	Which role reviews this alert?	Nobody assumes responsibility
Review cadence	When is the dashboard actually checked?	Patient assumes 24/7 surveillance
Severity tier	Which patterns are routine vs urgent?	Alert overload or delayed escalation
Response protocol	What confirmation/action is allowed?	Inconsistent care
After-hours plan	What happens outside staffed review?	Unsafe false reassurance
Documentation	Where are contact/action/outcome recorded?	No audit trail
Reassessment	When is response evaluated?	Open loop
Escalation	When does remote care become in-person/ED care?	Remote management beyond safe limits

ALERT-FATIGUE RULE More alerts do not equal safer care. Low-value, duplicate or non-actionable alerts consume the same finite attention needed for high-risk events.

14. Digital equity: inability to connect is not nonadherence

Remote care can reduce travel and improve access, but it can also create a new eligibility barrier. Broadband reliability, electricity, device ownership, language, literacy, vision, hearing, dexterity, cognition, caregiver availability and technical confidence determine whether a digital pathway is usable. Programs should maintain alternatives—telephone, paper, community or clinic review, store-and-forward images, professional assistance—rather than making connectivity a prerequisite for safe PD. [26,27]

Table 18.15 — Equity screen for digital PD.

Barrier	How it appears	Adaptation
Unreliable connectivity	Missing uploads / delayed data	Offline fallback + telephone pathway
Language/literacy	Low app use / misunderstood alerts	Translated/simple interface + teach-back
Visual/dexterity limitation	Difficulty entering data or operating device	Care-partner/assisted workflow
Cognitive impairment	Inconsistent tasks / alarm response	Simplify + professional/family assistance
Financial/resource barrier	Device/data-plan burden	Program-funded access where feasible
Rural distance	Hard in-person escalation	Local network + defined emergency pathway

EQUITY RULE Digital monitoring should expand supportability, not redefine the least connected patients as “poor candidates” for PD.

15. Economics and workload: count the work moved, not only visits avoided

Economic studies suggest that RPM can be cost-effective in some health systems and may reduce travel, in-person visits and hospital use. A 2023 Polish model estimated an ICER below its national willingness-to-pay threshold; a 2025 Korean home-care analysis also favored cost-effectiveness. Such conclusions depend on platform price, staffing, hospitalization effect, reimbursement and geography. Digital care can save patient travel while increasing nurse review, messaging and alert triage. Both sides of the ledger should be measured. [28,29]

Table 18.16 — Digital-PD workload ledger.

Potential saving	Potential new workload	Measure both
Fewer routine trips	Dashboard review time	Patient + staff time
Fewer urgent visits	Alert triage/messages	Visits avoided vs contacts created
Earlier troubleshooting	More detected low-grade issues	Time to resolution + alert burden
Remote prescription change	Version control/documentation	Number of safe changes + errors
Patient reassurance	Expectation of continuous surveillance	Patient experience + anxiety

16. AI and prediction models: prediction is not intervention

Machine-learning studies in PD now target mortality, technique loss, peritonitis and other outcomes. A 2024 systematic review found heterogeneous, mainly retrospective studies with limited external validation. More recent work is beginning to use temporal validation and platform-derived alarm features. These advances can improve prioritization research, but a model that predicts risk does not establish which action helps the patient. Clinical use requires calibration, transportability, decision utility, subgroup performance, model-drift monitoring and prospective impact evaluation. [30–33]

Table 18.17 — Model maturity ladder.

Stage	Required evidence	Clinical status
Development	Internal discrimination + calibration	Research
Temporal validation	Performance in later patients	Still research
External validation	Different center/platform/population	Candidate tool
Silent prospective phase	Real-time predictions without changing care	Safety/feasibility testing
Impact trial	Model-guided care vs comparator	Tests clinical utility
Post-implementation monitoring	Drift, overrides, subgroup harms, outcomes	Required for sustained use

Flowchart 18.4 — Prediction model to adaptive-PD research. External validation and prospective impact must precede treatment-directed automation.

AI RULE An AUROC is not a treatment effect. No prediction model should bypass the clinical question: “What action improves this patient’s outcome?”

17. Adaptive PD: the future is a governed learning loop

Figure 18.4 — The future of PD is disciplined adaptation: validated information, human oversight, patient-important outcomes and auditability. Fully autonomous closed-loop prescription change remains investigational.

Adapted APD already exists as a prescription concept: cycles can use different dwell durations and fill volumes to individualize therapy. Digital “adaptive PD” is a broader idea—using confirmed longitudinal information to select or adjust therapy within a predefined pathway. The safest near-term form is clinician-governed adaptation, not autonomous control. [34,35]

Table 18.18 — Human-governed adaptive-PD pathway.

Step	Required question	Safety gate
Acquire	What new information was captured?	Data completeness
Validate	Is the signal real?	Technical/measurement check
Interpret	Which PD mechanism explains it?	Clinical context
Prioritize	How urgent and consequential is it?	Severity rules
Act	What evidence-based response is appropriate?	Named clinician/protocol
Reassess	Did benefit occur without harm?	Defined endpoint
Learn	Should threshold/workflow change?	Audit + CQI
Automate later?	Has prospective benefit and safety been proven?	Override + auditability + equity

FUTURE BOUNDARY Closed-loop treatment belongs in a research framework until control limits, failure modes, human override, auditability and patient-important benefit are prospectively demonstrated.

18. Major clinical algorithms

Flowchart 18.1 — Daily connected-cycler review. Data validation precedes interpretation; clinically meaningful abnormalities enter a targeted chapter pathway.

Flowchart 18.2 — Remote volume signal. Verify the trajectory and phenotype before changing PD; severe symptoms require direct assessment.

Flowchart 18.3 — Repeated drain alarms or incomplete treatment. Separate data/technical problems from functional catheter problems, then use Chapter 12.

Flowchart 18.4 — Prediction model to adaptive-PD research. The evidence ladder prevents a promising algorithm from becoming an unsafe prescription engine.

19. Retention tables: pattern recognition

Table 18.19 — If you see this remotely, think this first.

Remote pattern	First hypothesis	Immediate next step
Three nights of low UF	Data/drain/prescription/volume issue—not a diagnosis	Verify data + Chapter 8 phenotype
Progressively longer drains	Constipation/posture/catheter mechanics	Bowel + Chapter 12 review
No upload for 2 nights	Connectivity or treatment gap	Contact; do not label nonadherence
Frequent early terminations	Alarm burden, symptoms or life burden	Ask cause; inspect event sequence
Weight ↑ + BP ↑ + urine ↓	Possible congestion + RKF loss	Confirm measurements + clinical review
Cloudy-effluent photo	Possible peritonitis	Immediate Chapter 10 pathway
New exit-site photo abnormality	Possible ESI	Chapter 11 assessment
Model flags high risk	Risk estimate only	Verify applicability; no automatic treatment change

Table 18.20 — What not to confuse.

Do not confuse	With	Correction
Connected cycler	Remote management	Data upload ≠ clinical response
Delivered UF	Euvolemia	Integrate weight, urine, symptoms and cardiovascular context
Missing data	Nonadherence	Check connectivity and workflow
Alarm count	Diagnosis	Localize mechanism
Prediction	Clinical utility	Test whether model-guided action improves outcomes
Bioimpedance/LUS abnormality	Prescription order	Adjunctive phenotype only
Remote prescription capability	Autonomous care	Clinical order still requires governance
Digital access	PD eligibility	Provide non-digital alternatives

20. Clinical pearls

1. The best connected-cycler question is often “What was actually delivered last night?”

2. Remote monitoring is a surveillance process; remote patient management adds action.

3. Low UF on a cycler is not synonymous with volume overload.

4. A missed upload can be a modem problem, not a missed dialysis session.

5. Repeated drain alarms can reveal a mechanical trajectory before the clinic visit.

6. There is no validated universal remote weight, BP or UF threshold for prescription change.

7. The 2025 cluster trial tested a care pathway with organized response—not connectivity alone.

8. The CKD-PD app RCT supports timely fluid-management intervention, not autonomous fluid prescriptions.

9. A photo can accelerate triage; it does not replace effluent culture or tunnel assessment when indicated.

10. Every alert needs a named owner and after-hours safety statement.

11. Digital equity is a clinical safety issue, not only an access issue.

12. Prediction models need calibration, external validation and impact trials before treatment-directed use.

13. Remote prescription changes require version control and post-change verification.

14. The future target is disciplined adaptation with human override, not sensor-to-prescription automation.

21. Common pitfalls — and the correction

Table 18.21 — Pitfall → why it fails → correction.

Pitfall	Why it fails	Correction
Treating a dashboard as a diagnosis	Signals are often nonspecific	Validate and phenotype clinically
Calling missing data nonadherence	Connectivity and workflow failures are common	Contact and verify first
Acting on one-night low UF	Normal variability / drain artifact	Use trends + clinical context
Using universal alert cutoffs	Thresholds vary by platform/population and are unvalidated	Local validation + severity pathway
Assuming RPM benefit is device effect	Studies bundle team review and intervention	Describe intervention components
Ignoring alert burden	Too many alerts degrade attention	Measure false/duplicate alerts and review time
Remote-only peritonitis management	Diagnosis requires effluent assessment and microbiology	Use Chapter 10 pathway
Automatic dose escalation when urine falls	RKF loss needs quantification and whole-patient reassessment	Measure before changing
Deploying ML after internal AUROC only	Poor transportability/calibration can harm	External validation + impact trial
Making digital access mandatory	Excludes vulnerable patients	Offer assisted/non-digital pathways

22. Mini-cases: decisions, not trivia

Case 1 — Low UF trend

Scenario	Decision
A connected APD patient shows falling UF for four nights and 1.8-kg weight gain. BP is slightly higher, urine output is lower, and no drain alarms are present.	Verify weight technique and intake, assess edema/orthopnea and quantify RKF. Use Chapter 8 volume/UF reasoning before changing osmotic strength.

Case 2 — “Nonadherence” alert

Scenario	Decision
No cycler data upload for 48 hours. The patient says both treatments were completed normally.	Treat this first as a connectivity/data-quality problem. Confirm device logs locally before documenting missed therapy.

Case 3 — Slow drains

Scenario	Decision
A patient has rising drain times and repeated bypass alerts but stable weight and UF. They report constipation.	Treat constipation and review posture. If the pattern persists, use Chapter 12 catheter-dysfunction evaluation.

Case 4 — Cloudy photo

Scenario	Decision
A CAPD patient sends a photo of newly cloudy effluent through the clinic app but reports little pain.	Remote contact accelerates care; it does not lower the diagnostic threshold. Initiate Chapter 10 effluent testing/treatment pathway promptly.

Case 5 — Remote prescription

Scenario	Decision
A clinician increases fill volume remotely after low clearance. The next morning the patient reports abdominal pressure and new genital swelling.	Stop assuming the change was successful because it transmitted. Reassess for pressure-related leak (Chapter 12), revert/modify safely and document the outcome.

Case 6 — High-risk AI score

Scenario	Decision
A research model flags imminent peritonitis risk despite normal symptoms and clear effluent.	Do not prescribe antibiotics from a risk score. Confirm model validation status; use enhanced observation/research protocol only if prospectively approved.

Case 7 — Digital exclusion

Scenario	Decision
An older patient on assisted APD has repeated missing home BP/weight because the caregiver cannot use the app.	Redesign the monitoring pathway—telephone or paper reporting, professional assistance, simplified data capture—rather than declaring the patient unsuitable for PD.

Case 8 — Composite trial claim

Scenario	Decision
A trainee says “the 2025 RCT proved RPM reduces everything—death, admissions and adverse events.”	Correct the evidence: the first broad composite was not significantly different; the second CV/fluid/efficiency composite and several secondary outcomes favored RM-APD.

23. Active recall

1. What is the difference between a connected cycler and remote patient management?

2. Which APD data can be reviewed remotely?

3. Why is low delivered UF not equivalent to clinical hypervolemia?

4. What are the first three checks before acting on a remote alert?

5. How should missing data be classified before calling it nonadherence?

6. Which chapter should a persistent slow-drain alert ultimately enter?

7. What did the 2025 Paniagua cluster trial show—and what did its first composite not show?

8. What did the 2025 CKD-PD app trial primarily demonstrate?

9. Why can a negative effluent sensor never exclude symptomatic peritonitis?

10. What minimum fields make an alert system governable?

11. What is the maturity ladder from ML development to clinical impact?

12. Why does a remotely transmitted prescription require version control?

13. What should a digital-equity screen include?

14. What distinguishes human-governed adaptive PD from closed-loop PD?

15. What evidence would be required before autonomous prescription change becomes routine?

MUST MEMORIZE

Concept	Memory anchor
Connected cycler	Makes treatment visible
RPM	Organizes remote review
Remote management	Adds contact + intervention
Adaptive PD	Confirmed signal → governed change → reassessment
Closed loop	Investigational automation
Low UF	Signal, not volume diagnosis
Missing upload	Check connectivity first
Alarm governance	Owner + timing + action + escalation + documentation
AI model	Validation + calibration + utility + impact
Equity	Alternative pathway must exist

USE AS REFERENCE: exact platform alarm definitions; device-specific remote-programming steps; locally validated thresholds; regulatory/privacy requirements; staffing/after-hours arrangements; current product manuals.

24. Flashcards: spaced repetition

Q: Connected cycler vs RPM?	Answer
	A: The cycler records/transmits; RPM organizes remote review of those data.

Q: RPM vs remote management?	Answer
	A: Remote management adds communication, clinical action and follow-up.

Q: Most direct cycler question?	Answer
	A: What treatment was actually delivered?

Q: Low UF means overload?	Answer
	A: No. It requires clinical volume phenotyping.

Q: Missing upload means missed treatment?	Answer
	A: Not until connectivity/device/local logs are checked.

Q: Repeated drain alarms suggest?	Answer
	A: A mechanical/functional trend requiring Chapter 12 localization.

Q: Can a photo diagnose peritonitis?	Answer
	A: No; it can accelerate contact and effluent testing.

Q: Universal remote threshold for weight/UF?	Answer
	A: No validated universal threshold.

Q: 2025 cluster RCT key nuance?	Answer
	A: Broad composite not significantly different; CV/fluid/efficiency composite and several secondary outcomes favored RM-APD.

Q: CKD-PD app trial?	Answer
	A: More timely overhydration interventions and fewer hospitalizations; no significant survival/technique difference.

Q: Why alert owner?	Answer
	A: Unowned alerts create false surveillance and delayed action.

Q: AI AUROC enough?	Answer
	A: No—calibration, external validation, clinical utility and impact are required.

Q: Remote prescription safety?	Answer
	A: Identity, authorization, version control, confirmation, patient understanding, reassessment.

Q: Digital equity principle?	Answer
	A: Connectivity should expand support—not become a PD eligibility barrier.

Q: Closed-loop PD today?	Answer
	A: Investigational.

25. Rapid differential / troubleshooting

Table 18.22 — Remote abnormality troubleshooting.

Problem	Common cause	Important alternative	Next step
No data	Network/device issue	Missed treatment / hospitalization	Verify contact + local device log
Low UF	Drain artifact / prescription	Congestion / membrane change / leak	Clinical volume + Chapter 8 review
High weight	Salt/water gain	Scale error / nutrition / constipation	Repeat + symptoms + urine + BP
Slow drains	Constipation/posture	Migration/omentum/kink	Chapter 12 pathway
Lost treatment time	Alarms/setup	Symptoms/treatment burden	Inspect sequence + patient interview
Frequent alerts	Poor thresholds/workflow	True recurrent clinical problem	Audit value + prioritize
Cloudy photo	Peritonitis	Blood/fibrin/other mimic	Chapter 10 immediate assessment
Exit-site photo change	ESI/trauma/dermatitis	Tunnel disease	Chapter 11 pathway
High model risk	Statistical risk	Model miscalibration/drift	Do not treat automatically; verify evidence

26. Final revision sheet

CORE CONCEPT Digital PD is a clinical pathway, not a dashboard. Acquire → validate → interpret → prioritize → act → reassess → learn.

One-minute revision

Connected APD shows delivered treatment between visits.
RPM adds remote review; remote management adds intervention.
No single remote metric replaces goal-directed PD assessment.
Validate identity, connectivity, prescription version and measurement quality before acting.
Volume signals require weight/BP/symptoms/urine/RKF/prescription context.
Drain alarms enter a catheter-mechanics differential, not an automatic prescription change.
Peritonitis and ESI still require current ISPD diagnostic/source-control pathways.
Randomized evidence supports selected benefits of organized remote-care pathways, not data transmission alone.
Every alert needs ownership, response timing, escalation and documentation.
Prediction models need external validation and impact trials.
Digital equity and caregiver burden are part of safety.
Closed-loop prescription change remains investigational.

TEN TAKE-HOME RULES

1. Name the digital intervention precisely.

2. Validate the data before interpreting physiology.

3. Interpret trends, not isolated numbers.

4. Use remote data to enter the correct clinical chapter pathway.

5. Never equate missing data with nonadherence without verification.

6. Treat alerts as work requiring ownership, not as decoration.

7. Remote prescription changes need the same traceability as clinic orders.

8. Separate randomized evidence from observational association.

9. Demand external validation and clinical utility from AI/ML.

10. Advance from monitoring to adaptation only with human oversight, safety measurement and reassessment.

FINAL MENTAL MODEL A connected PD system becomes valuable when it reduces the time between meaningful change and safe clinical action. The future is not “more data”; it is better governed decisions.

Rapid oral viva

Table 18.23 — Viva prompts.

Prompt	One-line answer
What does a connected cycler add?	A near-daily record of treatment delivery.
What makes RPM clinically active?	Organized human review linked to response.
Best use of low-UF alert?	Trigger verification and volume/UF phenotyping.
Can RPM prevent peritonitis by itself?	Not proven; it may accelerate recognition/contact.
What did randomized evidence improve?	Selected fluid/CV, hospitalization, satisfaction and process outcomes depending on intervention.
Why not automate prescription change?	Signals are nonspecific and action safety is not prospectively established.
What makes AI clinically credible?	External validation + calibration + utility + impact trial + monitoring.
Digital-equity safeguard?	Maintain usable non-digital/assisted pathways.

SAFETY BOUNDARY Acute symptoms, severe dyspnea, hypotension, suspected peritonitis, tunnel infection, major leak or surgical abdomen must bypass routine dashboard workflow and enter direct clinical assessment.

27. Selected authoritative references

1. Brown EA, Blake PG, Boudville N, et al. International Society for Peritoneal Dialysis practice recommendations: Prescribing high-quality goal-directed peritoneal dialysis. Perit Dial Int. 2020;40(3):244–253. https://doi.org/10.1177/0896860819895364. PMID: 32063219.

2. Chow JSF, Brunier G, Figueiredo AE, et al. Teaching peritoneal dialysis: A position paper for the International Society for Peritoneal Dialysis. Perit Dial Int. 2025;45(6):327–343. https://doi.org/10.1177/08968608251375512. PMID: 40966019.

3. Li L, Perl J. Can Remote Patient Management Improve Outcomes in Peritoneal Dialysis? Contrib Nephrol. 2019;197:113–123. https://doi.org/10.1159/000496306. PMID: 34569512.

4. Crepaldi C, Giuliani A, Milan Manani S, et al. Remote Patient Management in Peritoneal Dialysis: Impact on Clinician's Practice and Behavior. Contrib Nephrol. 2019;197:44–53. https://doi.org/10.1159/000496317. PMID: 34569511.

5. Paniagua R, Ramos A, Avila M, et al. Remote monitoring of automated peritoneal dialysis reduces mortality, adverse events and hospitalizations: a cluster-randomized controlled trial. Nephrol Dial Transplant. 2025;40(3):588–597. https://doi.org/10.1093/ndt/gfae188. PMID: 39165115.

6. Anutrakulchai S, Tatiyanupanwong S, Kananuraks S, et al. Effect of the Chronic Kidney Disease-Peritoneal Dialysis (CKD-PD) App on Improvement of Overhydration Treatment in Patients on Peritoneal Dialysis: Randomized Controlled Trial. J Med Internet Res. 2025;27:e70641. https://doi.org/10.2196/70641. PMID: 40397925.

7. Chung MC, Yu TM, Cheng BC, et al. Remote Patient Monitoring-Enabled Automated Peritoneal Dialysis and Clinical Outcomes: A Nationwide Real-World Cohort Study from Taiwan. Clin J Am Soc Nephrol. 2026. https://doi.org/10.2215/CJN.0000001112. PMID: 42447366.

8. Jung HY, Jeon Y, Kim YS, et al. Outcomes of Remote Patient Monitoring for Automated Peritoneal Dialysis: A Randomized Controlled Trial. Nephron. 2021;145(6):702–710. https://doi.org/10.1159/000518364. PMID: 34515160.

9. Uchiyama K, Morimoto K, Washida N, et al. Effects of a remote patient monitoring system for patients on automated peritoneal dialysis: a randomized crossover controlled trial. Int Urol Nephrol. 2022;54(10):2673–2681. https://doi.org/10.1007/s11255-022-03178-5. PMID: 35362819.

10. Biebuyck GK, Neradova A, de Fijter CWH, Jakulj L. Impact of telehealth interventions added to peritoneal dialysis-care: a systematic review. BMC Nephrol. 2022;23(1):292. https://doi.org/10.1186/s12882-022-02869-6. PMID: 35999512.

11. Nygard HT, Nguyen L, Berg RC. Effect of remote patient monitoring for patients with chronic kidney disease who perform dialysis at home: a systematic review. BMJ Open. 2022;12(12):e061772. https://doi.org/10.1136/bmjopen-2022-061772. PMID: 36600376.

12. Ali H, Mohamed MM, Fulop T, Hamer R. Outcomes of Remote Patient Monitoring in Peritoneal Dialysis: A Meta-Analysis and Review of Practical Implications for COVID-19 Epidemics. ASAIO J. 2023;69(4):e142-e148. https://doi.org/10.1097/MAT.0000000000001891. PMID: 36867191.

13. Dazzarola MP, et al. Integrating algorithm-driven analytics and clinical decision support in remote monitoring in automated peritoneal dialysis: patient and clinician experiences. Blood Purif. 2026. https://doi.org/10.1159/000552265.

14. Chang MY, Chi PJ, Wang HH, et al. Evaluation of the Impact of Remote Monitoring Using the Sharesource Connectivity Platform on Adherence to Automated Peritoneal Dialysis in 51 Patients. Med Sci Monit. 2023;29:e939523. https://doi.org/10.12659/MSM.939523. PMID: 37020409.

15. Huang KC, Chao JY, Liu KH, et al. Prognostic Value of Remote Monitoring Alarms for Technique Failure in Automated Peritoneal Dialysis. Blood Purif. 2026 Apr 2:1–9. https://doi.org/10.1159/000551828. PMID: 41926476.

16. Li PKT, Chow KM, Cho Y, et al. ISPD peritonitis guideline recommendations: 2022 update on prevention and treatment. Perit Dial Int. 2022;42(2):110–153. https://doi.org/10.1177/08968608221080586. PMID: 35264029.

17. Chow KM, Li PKT, Cho Y, et al. ISPD Catheter-related Infection Recommendations: 2023 Update. Perit Dial Int. 2023;43(3):201–219. https://doi.org/10.1177/08968608231172740. PMID: 37232412.

18. Mehrotra R, Williamson DE, Betts CR, et al. A Prospective Clinical Study to EvaluAte the AbiliTy of the CloudCath System to Detect Peritonitis During In-Home Peritoneal Dialysis (CATCH). Kidney Int Rep. 2024;9(4):929–940. https://doi.org/10.1016/j.ekir.2024.01.033. PMID: 38765568.

19. Sanabria M, Buitrago G, Lindholm B, et al. Remote patient monitoring program in automated peritoneal dialysis: impact on hospitalizations. Perit Dial Int. 2019;39(5):472–478. https://doi.org/10.3747/pdi.2018.00287.

20. Corzo L, Wilkie M, Vesga JI, et al. Technique failure in remote patient monitoring program in patients undergoing automated peritoneal dialysis: a retrospective cohort study. Perit Dial Int. 2022;42(3):288–296. https://doi.org/10.1177/0896860820982223.

21. Chaudhuri S, Han H, Muchiutti C, et al. Remote treatment monitoring on hospitalization and technique failure rates in peritoneal dialysis patients. Kidney360. 2020;1(3):191–202. https://doi.org/10.34067/KID.0000302019.

22. Milan Manani S, Baretta M, Giuliani A, et al. Remote monitoring in peritoneal dialysis: benefits on clinical outcomes and on quality of life. J Nephrol. 2020;33(6):1301–1308. https://doi.org/10.1007/s40620-020-00812-2.

23. Tian N, Yang X, Guo Q, et al. Bioimpedance Guided Fluid Management in Peritoneal Dialysis: A Randomized Controlled Trial. Clin J Am Soc Nephrol. 2020;15(5):685–694. https://doi.org/10.2215/CJN.06480619. PMID: 32349977.

24. Kharat A, Tallaa F, Lepage MA, Trinh E, Suri RS, Mavrakanas TA. Volume Status Assessment by Lung Ultrasound in End-Stage Kidney Disease: A Systematic Review. Can J Kidney Health Dis. 2023;10:20543581231217853. https://doi.org/10.1177/20543581231217853. PMID: 38148768.

25. Morelle J, Stachowska-Pietka J, Öberg C, Gadola L, La Milia V, Yu Z, et al. ISPD recommendations for the evaluation of peritoneal membrane dysfunction in adults: classification, measurement, interpretation and rationale for intervention. Perit Dial Int. 2021;41(4):352–72. https://doi.org/10.1177/0896860820982218. PMID: 33563110.

26. Oliver MJ, Abra G, Bechade C, et al. Assisted peritoneal dialysis: Position paper for the ISPD. Perit Dial Int. 2024;44(3):160–170. https://doi.org/10.1177/08968608241246447. PMID: 38712887.

27. World Health Organization. Global strategy on digital health 2020-2027. WHO; current digital-health implementation framework.

28. Augustynska J, Lichodziejewska-Niemierko M, Naumnik B, et al. Automated Peritoneal Dialysis With Remote Patient Monitoring: Clinical Effects and Economic Consequences for Poland. Value Health Reg Issues. 2024;40:53–62. https://doi.org/10.1016/j.vhri.2023.09.011. PMID: 37976660.

29. Kim K, Kim TH, Shin J, et al. Home care program with telemonitoring for patients undergoing peritoneal dialysis in South Korea: a cost-utility analysis. Kidney Res Clin Pract. 2025;44(4):651–663. https://doi.org/10.23876/j.krcp.24.246. PMID: 40264373.

30. Mushtaq MM, Mushtaq M, Ali H, et al. Artificial intelligence and machine learning in peritoneal dialysis: a systematic review of clinical outcomes and predictive modeling. Int Urol Nephrol. 2024;56(12):3857–3867. https://doi.org/10.1007/s11255-024-04144-z. PMID: 38970709.

31. Wang Q, Luo Q, Ding Y, Wan S, Zhang Y, Xiong F. Prediction of Imminent Peritoneal Dialysis-Associated Peritonitis Using Time-Updated Electronic Health Records and Machine Learning: A Temporal Validation Study. J Inflamm Res. 2026;19:595197. https://doi.org/10.2147/JIR.S595197. PMID: 42022267.

32. Zhao Z, Li Y, Quan Q, Wang H, Zhang W, Zhang X. Risk prediction models for peritoneal dialysis-associated peritonitis: a systematic review and meta-analysis. Int Urol Nephrol. 2026;58(3):967–979. https://doi.org/10.1007/s11255-025-04795-6. PMID: 40975843.

33. Collins GS, Moons KGM, Dhiman P, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378. https://doi.org/10.1136/bmj-2023-078378.

34. Current clinical practice in adapted automated peritoneal dialysis (aAPD)-A prospective, non-interventional study. PLoS One. 2021;16(12):e0258440. https://doi.org/10.1371/journal.pone.0258440. PMID: 34882678.

35. Slon-Roblero MF, Sanchez-Alvarez JE, Bajo-Rubio MA. Personalized peritoneal dialysis prescription-beyond clinical or analytical values. Clin Kidney J. 2024;17(Suppl 1):i44-i52. https://doi.org/10.1093/ckj/sfae080. PMID: 38846417.

36. Quinn RR, Johnson DW, Mehrotra R, et al. 2026 ISPD Position Statement on tracking and reporting loss from PD therapy. Perit Dial Int. 2026. https://doi.org/10.1177/08968608261461639. PMID: 42307417.

37. Cely J, Doria C, Dazzarola MP, et al. Engagement and usability of a mobile health app for peritoneal dialysis patients: a pilot randomized controlled trial. Perit Dial Int. 2026;46(3):223–232. https://doi.org/10.1177/08968608251339578. PMID: 40398855.

SOURCE NOTE Guideline status and key digital-PD evidence were checked 3 September 2026. Commercial platform capabilities, alarm labels, remote-programming functions, connectivity requirements and regulatory/privacy obligations vary by device, jurisdiction and software version; current manufacturer documentation and local governance take precedence for operational details. No universal remote alert threshold or fully autonomous PD prescription algorithm is endorsed in this chapter.