Does digital transitional care management reduce hospital readmissions in cancer patients?

A 10-day digital layer added to existing oncology care

The paper asks whether a scalable, digitally delivered transitional care program can lower readmissions after a cancer-related hospital stay. Transitional care management here means the handoff from hospital to home. Both arms kept access to the institution's existing oncology, nursing triage, after-hours and Symptom Care Clinic pathways; the intervention added a 10-day digital layer on top.

That layer had three parts. Patients received daily electronic symptom assessments through the secure patient portal, covering pain, fatigue, nausea, vomiting, diarrhea, constipation, shortness of breath, decreased oral intake and fever. Their answers triggered real-time self-management education, and moderate or severe symptoms were routed to clinical teams as yellow or red alerts. Red alerts required a nurse to call the patient the same day; responses to yellow alerts were left to nursing judgment.

Ten days was chosen because it matches the median time to readmission among oncology patients cited by the authors. The assessment window ran 9 am to 5 pm, seven days a week, reflecting nursing resources available to answer alerts; after hours, patients were told to use the existing symptom triage phone line. Usual care meant symptom monitoring at the discretion of the treating oncology team, with no universal proactive nursing outreach.

Allocation by record-number parity, not randomization

Between April 19 and September 19, 2024, 2,554 patients were discharged from medical oncology services at Memorial Sloan Kettering Cancer Center; 1,713 met eligibility criteria and were allocated by medical record number parity, with the assignment rule alternating by month. That produced 894 patients in the DTCM arm and 819 in usual care.

Eligibility required age 18 or older, discharge home, a primary medical oncologist at the center and enrollment in the patient portal. Patients discharged to rehabilitation, nursing facilities or home hospice were excluded, as were those already in another remote monitoring program or an investigational treatment, because those settings involve different teams or structured symptom reporting.

Baseline characteristics were reported as balanced: median age 65.6 years, 52.9% female, 66.0% white, 10.9% Hispanic, with gastrointestinal (25.5%), thoracic (10.7%) and genitourinary (10.3%) cancers most common. The design was pragmatic and embedded in routine operations, and the institutional review board deemed it exempt, so patients were automatically enrolled and could opt out by not completing assessments. That choice of design is the main reason the results describe an association within one health system rather than a randomized causal effect.

Readmissions fell at 30 and 90 days

The primary outcome was 30-day readmission, analyzed with cumulative incidence functions and death treated as a competing event, meaning death before readmission was handled as a separate outcome rather than a censoring event. Cumulative incidence of death before readmission within 30 days was similar between arms: 3.8% in the intervention arm versus 2.6% in usual care, with a P value of 0.15.

Thirty-day readmission was 30% (95% CI 27–33%) with DTCM versus 37% (95% CI 34–40%) with usual care, P = 0.001, an absolute risk reduction of 7 percentage points. The gap persisted at 90 days: 48% (95% CI 44–51%) versus 54% (95% CI 50–57%), P = 0.004. Emergency department visits without admission within 30 days were also less frequent, 12% (95% CI 10–15%) versus 17% (95% CI 14–20%), P = 0.007.

These are cumulative incidences, not simple proportions of patients ever readmitted, so they should not be read as counts of individual readmission events. The authors also report that patients with breast cancer, lymphoma, sarcoma and gastrointestinal malignancies showed the largest differences between arms, based on separate models by diagnosis.

Only 56.6% of the intervention arm engaged

Uptake is the number that most changes how the headline result should be read. Among the 894 patients allocated to DTCM, 56.6% (506) completed at least one electronic symptom assessment. Engaged patients completed a median of 3 assessments (interquartile range 1–7; mean 4.3), totaling 2,150 assessments over the study period.

Engagement was uneven. 63% of non-Hispanic white patients completed at least one assessment versus 47% of non-white patients (P < 0.001), and 59% of patients whose preferred language was English engaged versus 37% of those with another preferred language (P < 0.001). The paper does not report whether the readmission benefit differed across these groups.

Among patients who completed assessments, 46.6% (236 of 506) reported at least one severe symptom, and 28.3% (143 of 506) reported severe pain, the most common severe symptom. In other words, the digital layer surfaced a substantial burden of serious symptoms among those who used it, which is the mechanism the authors point to for why alerts could prompt earlier outpatient management.

Engaged patients had lower early readmission, descriptively

An exploratory comparison inside the intervention arm found 30-day readmission of 28% (95% CI 24–32%) among patients who engaged versus 31% (95% CI 27–36%) among those who did not, P = 0.22. At 90 days the rates were 48% and 47%, P = 0.87.

The early gap was larger: cumulative readmission was 4.7% versus 11% at day 5 and 10% versus 17% at day 10 among engaged and nonengaged patients. The authors explicitly label these comparisons descriptive, because engagement happened after allocation and an early readmission could itself prevent a patient from ever engaging. This analysis therefore cannot say whether engaging lowered readmission or early deterioration blocked engagement.

The primary comparison remains the allocated arms, and that comparison is what carries the 30% versus 37% result. The engagement numbers describe who used the tool, not a proven dose-response effect.

More after-hours triage calls, not more clinic visits

The paper also tracked care activity to probe how the effect might work. There were no differences between arms in the cumulative incidence of a first nursing call, a first physician clinic visit or a first Symptom Care Clinic visit within 30 days. The intervention arm did have a higher cumulative incidence of a first after-hours triage call, 27% (95% CI 24–30%) versus 23% (95% CI 20–26%), P = 0.01.

Among patients who received at least one nursing call within 30 days, the mean number of calls per patient was 4.9 in both arms. Within the 10 days of active digital monitoring, however, the intervention arm averaged more nursing calls per patient, 2.53 (95% CI 2.41–2.65) versus 2.29 (95% CI 2.18–2.40), P = 0.007.

The authors read this pattern as consistent with their proposed mechanisms: structured symptom reporting, repeated automated check-ins and faster routing to the primary oncology team. They also note that weekend monitoring and triage support may have helped, and recommend future work extend monitoring beyond 9 am to 5 pm for 24-hour symptom capture. These are the authors' interpretations of the utilization pattern, not separately tested causal pathways.

The digital layer cannot be isolated from structured care

Because the intervention bundled digital symptom assessments, automated education, clinician alerts and structured monitoring, the study cannot separate the effect of the digital modality from the effect of simply introducing structured transitional care. It also cannot say whether a similarly structured non-digital program would have produced comparable outcomes. The authors state this limitation directly and call for studies comparing digital and non-digital approaches and formally evaluating staffing requirements and implementation costs.

Two further boundaries matter for transfer. First, the setting is a single cancer center with an existing patient portal, oncology nursing triage, after-hours lines and a Symptom Care Clinic, all available to both arms; the added value of the digital layer in a system without those pathways is untested here. Second, the population excludes patients discharged to rehabilitation, nursing facilities or home hospice, and those in other remote monitoring or investigational treatment, so the results do not extend to those groups.

The authors also note that the intervention's clinical effect cannot be established independently of the digital platform's role in making structured transitional care operationally feasible at health-system scale, without universal staff-initiated outreach or additional clinical full-time equivalents.

The 7-point gap and the engagement gap together

The most decision-relevant reading of this paper is a pair of numbers held together. Allocating patients to DTCM was associated with a 7-percentage-point lower 30-day readmission cumulative incidence (30% versus 37%) and a 5-point lower rate of 30-day emergency department visits without admission (12% versus 17%), sustained at 90 days (48% versus 54%). At the same time, 43.4% of the intervention arm never completed a single assessment, and engagement was markedly lower among non-white patients and those whose preferred language was not English.

For a reader weighing whether such a program would work in their own setting, the paper's own variables set the priority: the benefit appeared in a population with existing oncology triage infrastructure, the intervention ran only 10 days and only during daytime hours, and the largest early readmission differences in the descriptive engagement comparison appeared in the first 10 days. The authors' own next step is to test digital against non-digital structured care and to measure staffing and cost, which is the comparison this study leaves open rather than settles.

Sources used in this answer