Digital design makes community care more reliable; what it does to the patient's time is still unmeasured

Across five domains this month, the measures that a protocol, a structure or a device can move are moving: an assessment rate, a documentation rate, an uptake figure, a data-completeness figure. The measures a patient would recognise are flat, untested or unmeasured: readmission, a gain in function that clears the clinically important difference, a death at home, time at home. Remote monitoring does not remove the work of watching a patient; it changes who does it and whether anyone is paid for it. And no source this month measures patient time in value-added terms at all. On this evidence, digital design improves the reliability of care, and its effect on the person’s time is unknown. Synthesis

Process measures move while patient-outcome measures stay flat, untested or unmeasured, in five domains

The shape recurs from the weeks of 11 September to 2 October, and no source cites another. In acute frailty, two emergency department tools, the Clinical Frailty Scale and a multidimensional prognostic index, predicted 30- and 90-day mortality in 402 patients, but neither predicted readmission (Gregersen et al., PMID 42806214; Primary Study; synthesis). In palliative care, a 41-trial narrative review found that interventions to promote goals-of-care conversations raise documentation and conversation quality, with inconsistent effect on utilisation (Thomas et al., DOI 10.1177/10966218261485018; Synthesis; signal). In geriatric surgery, a verification programme across 20 hospitals raised vulnerability assessment by 4.0 percentage points, from 43.4% to 47.3% (Yap et al., DOI 10.1097/XCS.0000000000002207; Audit/Data; signal). Frailty-targeted cardiovascular rehabilitation improved proximal function but not mortality, with very low certainty (Wang et al., PMID 42772577; Meta-analysis; signal). Home-based pulmonary rehabilitation in interstitial lung disease improved six-minute walk distance and health status, but on average not to the minimum clinically important difference (Joseph et al., PMID 42813303; Primary Study; synthesis).

A digital system that triggers comprehensive geriatric assessment from the Clinical Frailty Scale completed frailty assessment within four days in 71% of patients, with no patient outcome measured (Chou et al., PMID 42763222; Audit/Data). It was a retrospective observational pilot in one medical centre in Taiwan between March and November 2025, and its structured assessment agreed only modestly with clinician-led assessment (ρ=0.38 for total abnormal domains), which means it routes patients but has not been shown to substitute for the assessment. Pattern

The same gap appears between domains that never refer to each other. In 465 patients seen by liaison psychiatry across Spanish hospitals, frailty, not psychiatric variables, was independently associated with early mortality, and older age with lower specialist follow-up (Narvaiza-Grau et al., PMID 42785108; Primary Study; signal). In a Japanese emergency department cohort, a Clinical Frailty Scale score of 5 or more raised the odds of hospitalisation in lower-acuity triage categories (odds ratios 1.54 and 2.01), though the global interaction was not significant, so no difference between categories is claimed (Hata et al., PMID 42811477; Primary Study; synthesis). Risk, or the conversation, is better identified; downstream outcome is not shown.

Technology in community-facing care is delivering governance and workflow gains, not yet patient-outcome gains

Read against international practice, the NHS position is thinly evidenced. The strongest synthesis this month, a systematic review of 46 studies of digital post-discharge follow-up in adults aged 65 and over, finds the tools feasible and well liked but with mixed results on rehospitalisation and mental health (Sekhon et al., PMID 42679838; Synthesis; signal). The UK’s own deployment data are sparse and sobering: of 22 palliative virtual wards surveyed, 15% use remote-monitoring technology, three-quarters depend on charity and two have closed (May-Miller and Swann, PMID 40306765; Audit/Data; signal).

Where technology has been scaled fastest it is administrative: ambient voice documentation has been procured for more than 1,200 GP practices and over 70,000 clinicians, with the King’s Fund calling for a national value-based procurement strategy and documenting hallucinated medication details (King’s Fund, 23 July 2026; Guidance). The driving case there is clinician administrative burden, not patient monitoring or admission avoidance.

The month’s best positive evidence is about design, not devices: a Taiwanese digital trigger that standardises who receives comprehensive geriatric assessment, and a verification programme that raised surgical vulnerability assessment by 4.0 points. Both are process results. The domain’s signature claim, that digital and structural scaffolding converts evidence into delivered care, is well supported as a claim about reliability and unsupported as a claim about outcomes. Synthesis

Remote monitoring relocates clinical work because the evaluation boundary excludes where it lands

What generates this finding in different countries without a shared research programme? A monitored patient emits data continuously; data need interpreting; interpretation is clinical work. A bed absorbed that work inside its own staffing model. A remote service must either fund it explicitly or let it land on whoever is nearest. This month shows three landing sites.

Nurse practitioners: in a Dutch regional network, remote heart-failure management raised nurse-practitioner consultations from 5.3 to 15.3 per patient and consultation time from 121 to 233 minutes, with no change in cardiologist time or readmissions (n=131, 34 monitored; Nieuwenhuys et al., PMID 42691076; Primary Study; synthesis). Alert desks: Australian and New Zealand cardiac-device clinics spend 30.2% of workload on alert management, 90.5% of it by cardiac physiologists (Sheahen et al., PMID 41737930; Audit/Data). Carers: in a Sydney virtual hospital, patients and carers took on onboarding, equipment and troubleshooting roles, and carers substituted where digital literacy was low (Ward et al., PMID 41937437; Synthesis; signal).

The same Australia and New Zealand data show the other side of the ledger: remotely monitored patients attend in person less often (follow-up interval 10.0 against 7.2 months). That is a shift of contact, not a measured net increase, so net workload is genuinely unknown. This extends the published pattern on remote monitoring redistributing clinical work, which stands on the Dutch comparison alone. Pattern

The generator is the unit of evaluation. Services are costed per episode against a bed-day, so the cost of alerts, nurse time and unpaid carer labour sits outside the boundary by design. What would have to change at root is the boundary, not the technology: any business case for remote monitoring would need alert-handling time and carer support inside the same cost frame as the bed-days it claims to save. Nothing in this month’s evidence does this. Synthesis

Remote care reaches dispersed populations where it is built around distance, and loses reliability where a ward-validated device is sent home

A Quebec hub-and-spoke virtual inpatient unit preserved 786 hospitalisation-days for a semirural hospital, with 14 patients in the first two weeks, no transfers and no on-site registered nurses (Malas et al., PMID 42615595; Audit/Data). Remote monitoring use was highest in remote (88.4%) and rural (85.3%) regions of Australia and New Zealand against metropolitan regions (80.3%), but those are implanted devices that transmit passively (Sheahen et al., PMID 41737930). Where a device was built for the ward and sent home, data failed: in a single-centre prospective study of 45 patients, oxygen-saturation data availability from a wearable vital-sign monitor was 48% in hospital and 39% at home, and respiratory-rate availability was 46% at home (Jonkers et al., PMID 42348131; Primary Study). A scoping review of rural telehealth in western Canada and the western United States points to hybrid, locally anchored access, with low confidence that emergency department use falls durably (Karczewski et al., PMID 42355916; Synthesis; signal). Pattern

The reading is conditional. Reach is evidenced where the model is designed around the distance problem and the technology is passive; reliability is evidenced as lost where a device validated on a ward is sent to a home. Neither tests a patient-operated, non-passive technology in a remote population.

Social and material disadvantage lengthens time in the system and lowers recovery

In Ontario, patients with recent homelessness were designated as awaiting alternative care in 9.1% of cases against 2.0% of matched patients from the most materially deprived neighbourhoods (rate ratio 4.51; 51,377 patients; Richard et al., PMID 42467431; Primary Study; signal). In a US prospective cohort of 164 adults receiving outpatient orthopaedic physiotherapy, greater social-need burden independently lowered function at discharge (B=−2.35), with only 112 completing the discharge assessment (Butcher et al., PMID 42816756; Primary Study; synthesis). Both are non-UK, and neither tests a UK deprivation gradient. Pattern

Technology studies measure uptake and data completeness, while patients and carers report communication and continuity

The technology literature counts what a system can see: devices distributed, transmitters in use, the proportion of vital signs captured, satisfaction scores. Remote monitoring use rose by 55.8% against 10.2% growth in implants (Sheahen et al., PMID 41737930); oxygen-saturation availability from the wearable was 39% at home (Jonkers et al., PMID 42348131); and 96.5% of 887 patients in a Madrid telemedicine hospital-at-home service reported satisfaction (Mir Montero et al., PMID 42126191; Audit/Data; signal).

What patients and carers report in the same month is of a different kind. Two Australian qualitative studies of 15 interviews each independently find that patients value continuity, coordination and goal-centred care over professional boundary or model, and that “things are different once home” even when a transition programme is sound on paper (Peiris et al., PMID 42773690; Shannon et al., DOI 10.1017/S1463423626101716; Primary Study). Nursing-home staff using telehealth for palliative care call it a bridge to, not a replacement for, in-person care, with limits for rapport and sensitive conversations (Alexander et al., PMID 42817853; Primary Study; synthesis). The Quebec unit that preserved 20 semirural beds did so with remote nurses plus local hands-on staff, twice-daily huddles and executive sponsorship, and patients valued staying close to home (Malas et al., PMID 42615595).

The measurable components of the technology literature (uptake, completeness, satisfaction) are properties of the system’s contact with a patient. What the person experiences is whether someone knows their situation when they are at home. A measurement regime built on the first cannot detect failure of the second, which is why high satisfaction and high uptake sit alongside flat readmissions and carer strain. Synthesis

The consensus that structure drives delivered practice has been earned for process and not for outcome

The professional consensus now forming, visible in verification programmes, consensus statements and national virtual-hospital models, is that deliberate design converts good intentions into delivered practice. It is plausible, and the evidence is real, but read closely it stops short. The geriatric surgery verification programme raised assessment by 4.0 percentage points in 6,631 patients in a retrospective, 20-hospital study (Yap et al., DOI 10.1097/XCS.0000000000002207): a modest change in a rate of assessment, not in an outcome. The digital assessment trigger reached assessment in 71% within four days, with modest agreement with clinician-led assessment (Chou et al., PMID 42763222). Saudi Arabia’s Seha national virtual hospital is described by its own authors as successful as a governance and workforce-integration exercise, while independent evidence on mortality, readmission, length of stay, cost-effectiveness and equity “remains limited” years after launch (Shujaat et al., PMID 42330317; Synthesis; signal).

Structure is probably necessary; that it is sufficient has not been shown. A consensus that moves on process measures can feel validated by its own scoreboard while patient outcomes stay flat, and this month three domains do exactly that.

The best explanation is that community technology is commissioned as a purchase, not as a service redesign

Four explanations were weighed against this month’s evidence only. Technology immaturity is eliminated as a general account: the Australian and New Zealand devices are mature, implanted and passively transmitting, with national reimbursement, and still generate 30.2% of clinic workload in alerts (Sheahen et al., PMID 41737930), while the Quebec model used ordinary virtual-nursing technology successfully (Malas et al., PMID 42615595). Evaluation immaturity is partly true and not eliminated; it explains why outcomes are unmeasured but not why workload landing sites are consistently outside the frame. Workforce shortage is a cause in the Quebec case and a consequence everywhere else, not an independent explanation.

The favoured explanation is that technology is bought as equipment and commissioned as a service without the service. Three-quarters of 22 UK palliative virtual wards run on charity while 15% use remote-monitoring technology, and two have closed (May-Miller and Swann, PMID 40306765; Audit/Data; signal); ambient voice technology is procured at scale ahead of a national procurement strategy (King’s Fund, 23 July 2026); Seha scaled ahead of independent evaluation; and the one success with documented design is Quebec, where governance, escalation pathways and an exit to a local virtual ward team were specified from day one. What would change the conclusion is a commissioned remote-monitoring specification that funds alert handling and carer support and then reports outcome parity or gain against a matched comparison. Synthesis

What is presented as more certain than it is

• The headline international outcome figures. The adjusted odds ratio of 0.6 for 30-day mortality in Israel’s national hospital-at-home programme (the headline “40% lower”) comes from a retrospective matched comparison with ward admissions, not a trial, and the paper does not isolate any technology.
• Satisfaction as benefit. 96.5% satisfaction (887 patients, one hospital, retrospective) is acceptability. It says nothing about outcome, workload or equity.
• Staff-effectiveness gains in Quebec (47% to 87.5%) rest on 8 of 17 and 7 of 8 survey respondents; the 14-patient early cohort is a feasibility result.
• Cost reductions of 45–65% in a Taiwanese nursing-home pilot rest on 60 patients matched to 120 over a year, with outcomes similar between groups (Chou et al., PMID 41459007; Audit/Data; signal).
• The workload finding itself. The Dutch result is 34 monitored patients in one network. Calling it a law of remote monitoring would repeat the error it warns against.
• Understated: the consistency of the process-versus-outcome shape across five domains and several countries, none of which cite each other.

What is known, what is inferred, what is uncertain

Known: remote-monitoring workload lands on nurses, alert desks and carers in at least three documented settings; ward-validated wearables lose data at home; frailty tools predict mortality but not readmission in the cohorts tested; and self-administered outpatient antibiotic therapy in a 1,064-episode UK cohort is not significantly worse overall than nurse-delivered or clinic delivery, with failures clustering in older, comorbid, diabetic patients and those with peripheral vascular disease or central venous access (Durojaiye et al., PMID 42796749; Primary Study; signal). Reasonably inferred: that the system-wide cost of remote monitoring is understated wherever alert time and carer labour are outside the evaluation boundary. Uncertain, and stated as such: whether remote monitoring is net workload-saving once reduced in-person visits are counted; whether the Dutch result, from a single regional network and one condition, generalises to frailty virtual wards; and whether any of the favourable international outcome figures would survive randomisation.

Few UK remote-monitoring outcome studies cleared the bar, and no study measured patient time in value-added terms

• Patient time measured in value-added terms: none. A search for studies measuring it found none. Every time measure in the month’s evidence is bed-days, length of stay or consultation minutes. A hypothesis, not a finding: value-added time is not a unit anyone is paid, regulated or audited on.
• No matched NHS-versus-named-international outcome comparison for virtual wards or remote monitoring, for a third consecutive month. This is a structural absence, not silence.
• UK remote-monitoring evidence is thin by volume. Four consecutive weekly sweeps found no standalone UK remote-monitoring outcome study clearing the bar. UK content was a national survey of palliative virtual wards (May-Miller and Swann, as above), an advanced clinical practitioner-led palliative heart-failure model in which 29% of telehealth-eligible patients took part (Kirkland et al., DOI 10.1093/eurjcn/zvag058; Primary Study), and a pilot randomised trial that did not reach its target, with 38 patients randomised (Khan et al., DOI 10.1136/heartjnl-2026-327782; Primary Study). A hypothesis: NHS remote monitoring is rolled out as an operational programme and evaluated as service evaluation, which rarely reaches indexed journals.
• Digital mental-health evidence in a technology month: none cleared the bar. This may reflect screening on remit grounds rather than a gap in the literature.
• Policy. The Department of Health and Social Care committed in May to an interim product for the dementia and frailty framework in September and the full framework by the end of the calendar year (Department of Health and Social Care, 22 May 2026; Guidance); no citable date for the interim product was found as of 2 October, and it is now overdue against its own commitment. The consultation on the allied health professions framework for 2027 to 2032 closed on 18 September with no outcome published at 14 days, and a professional body’s notice says the final framework is due in 2027 (Society of Radiographers, 11 August 2026; Guidance; see the September edition).

Hybrid rather than substitute: four domains arrive at the same design

Palliative care (nursing-home telehealth as “a bridge to, not a replacement for” in-person care, in 22 homes, 14 of them rural; Alexander et al., PMID 42817853), acute frailty (Taiwanese direct home admission, in which 10.8% of home-treated patients escalated to hospital; Wang et al., PMID 42816724; Primary Study; synthesis), OPAT (self-administered therapy works for selected patients and needs monitoring for others) and therapies (a rural occupational-therapy codesign concept that names “multiple delivery modes” and task sharing, with no outcome data and not peer reviewed; Jesus et al., PMID 42818959; [Preprint]; synthesis) each end at hybrid delivery rather than substitution. The rural telehealth scoping review reaches the same conclusion independently (Karczewski et al., PMID 42355916). Pattern

The unit of account is also the wrong one. Every favourable economic or capacity figure this month is counted in bed-days or length of stay (Taiwan 9.72 against 11.85 days, Wang et al., PMID 42816724; Quebec 786 days, Malas et al., PMID 42615595), while the costs the month’s other evidence documents, alert-handling time, nurse-practitioner time and carer technical labour, sit outside the evaluation boundary. No source costs both together, so none can say whether the system-wide net is positive. Synthesis

International benchmark: Israel’s national hospital-at-home programme measures outcomes at scale, but not time at home

Candidates were the named comparators in the month’s evidence: Clalit (Israel), Taiwan’s national-insurance hospital-at-home pilot, Quebec’s hub-and-spoke virtual unit, Australian and New Zealand device remote monitoring, a Madrid telemedicine hospital-at-home service, a Dutch regional network and Saudi Arabia’s Seha national virtual hospital. Each was screened on a hard patient outcome against a comparator, sample size and design, and transferability. Fame was not a criterion; Seha is the best-known flagship in the set and is the one disqualified.

Exemplar: Clalit Health Services’ national Hospital-at-Home programme. Outcome data: 3,335 hospital-at-home patients (pneumonia, heart failure, urinary tract infection, cellulitis; 2022) were matched 1:1 to general medical ward admissions. 30-day mortality was 5.8% against 9.1% (adjusted odds ratio 0.6, 95% confidence interval 0.49–0.73) and 30-day readmission 13% against 16% (adjusted odds ratio 0.8, 0.70–0.92); 84% preferred home care for future need (Lustman et al., PMID 41423642; Audit/Data; signal). Limits: retrospective, matched not randomised; selection into the programme is not fully observable; four diagnoses; and the paper does not separate any technology from the model. Transferability: the programme is an integrated payer-provider’s funded, protocolised alternative admission route, not a pilot bolt-on. For an NHS service the equivalent conditions are a commissioned admission-substitute pathway with its own funding, a defined selection protocol, and the data capability to produce matched national outcomes. What the NHS evidence lacks is that last capability: no UK source reports an equivalent matched outcome dataset for virtual wards. The first two conditions are commissioning choices; the third is infrastructure. Synthesis Use value: the paper measures mortality, readmission and preference. It reports nothing on time at home, carer burden or patient-reported outcome, so Clalit is an exemplar of measured outcome at scale, not a demonstrated exemplar of value-added time.

Second exemplar, for one dimension only: Australian and New Zealand device clinics, for measuring remote-monitoring workload. It is the only source this month that quantifies alert-handling load at national scale (30.2% of clinic workload; 90.5% by physiologists; Sheahen et al., PMID 41737930). It carries no patient-outcome data, so it is an exemplar of accounting, not of benefit. Transferability requires clinic-level workload surveys and manufacturer data linkage, and because the technology is passive, patient digital literacy is not tested.

Disqualified as exemplars. Seha: its own authors state that independent evidence on outcomes, cost-effectiveness, equity and workforce sustainability “remains limited” (Shujaat et al., PMID 42330317). Taiwan’s pilot: shorter episodes (6.6 against 11.8 days) and 45.5–65.1% lower cost with similar mortality and readmission, but with 60 patients against 120 (Chou et al., PMID 41459007); a promising second source, not yet an exemplar. Quebec: a design exemplar, with no on-site registered nurses, specified escalation pathways and handover to a local team, with outcomes limited to bed-days preserved and a 14-patient early cohort (Malas et al., PMID 42615595).

The gap. No matched NHS-versus-international outcome comparison exists in this month’s evidence, for the third consecutive month, so the gap is defined in design and measurement terms. Funding: Taiwanese and Australian and New Zealand programmes run on insurance or national reimbursement, while 75% of UK palliative virtual wards run on charity. Governance: a Quebec hub with specified escalation, and an Italian inter-society heart-failure pathway embedding telemedicine in one hospital-to-community pathway (Mallardo et al., DOI 10.1016/j.ejim.2026.106881; Guidance). Measurement: Israel produces national matched outcomes; the NHS evidence produces surveys of uptake. Closing the gap would require commissioning and data infrastructure more than technology.

Implication for practice

Well supported: cost alert-handling time, nurse-practitioner time and carer technical support into any remote-monitoring business case (Dutch, Australian and New Zealand, and Sydney evidence); evaluate remote services on a patient outcome and time at home, not uptake or satisfaction; and design remote models around a named access problem, as hybrid, locally anchored care (Karczewski et al., PMID 42355916). Directional, not yet evidenced: a national matched-outcome dataset for virtual wards, and a commissioning gate that funds the safeguards. Speculative: that hub-and-spoke models would transfer to dispersed community nursing.

Early signals: screening tools agree only modestly with clinicians, advance care planning is reported as success at near-floor levels, and self-administered OPAT failures cluster

A digitally triggered, FINDNEEDS-based comprehensive geriatric assessment correlated at ρ=0.38 with clinician-led assessment in the retrospective one-hospital pilot described above (Chou et al., PMID 42763222). Signal

New ReSPECT and DNACPR documentation rose from 2.55% to 5.57% across 10,950 hospital-at-home admissions in an uncontrolled quality-improvement report (Zhang et al., PMID 42710076; Audit/Data). The 41-trial review finds no reliable utilisation gain from goals-of-care interventions (Thomas et al., as above), and a Swedish care-home training study found its validated index unchanged despite qualitative impact (Mikaelsson et al., PMID 42732293; Primary Study). Signal

In a UK cohort of 1,064 outpatient antibiotic episodes, failure was 9.1% (26 of 286) for self-administered therapy, not significantly different from home-nurse delivery (adjusted relative risk 0.72) or clinic delivery (0.94), with possibly more vascular-access events; within self-administered therapy, older age, comorbidity, diabetes, peripheral vascular disease and central venous access predicted failure (Durojaiye et al., PMID 42796749). There was no rurality analysis. Signal

Three signals from the September edition have not moved and are dropped: the paid-caregiver-burden paradox, the finding that apathy predicts poorer recovery after hip fracture, and the national virtual hospital without independent outcome evaluation.

The next ninety days and twelve months

Against the September edition’s scan: the interim product due in September was not published as of 2 October and is now overdue against its own commitment; the allied health professions framework consultation closed on 18 September and its outcome is pending; the technology strand sharpened rather than reversed; and no validated-measurement requirement has been attached to virtual-ward commissioning.

• Medium confidence: a final national palliative and end-of-life care framework is published by year end. Whether it names carer capacity or technology funding is the test of the published pattern on safeguards identified and then left unbuilt.
• Medium-low: an outcome or response to the allied health professions framework consultation (no timeline published).
• Low-medium: the dementia and frailty interim product appears (overdue; no date).
• Low: a national commissioning specification attaching measurement, carer-capacity or equity gates to virtual wards or remote monitoring (none in four months of looking).

Over twelve months: medium-high confidence that there will still be no published matched NHS-versus-international outcome comparison for virtual wards or remote monitoring, since no mechanism for producing one has appeared in four months of evidence. Medium: further national-scale outcome reports from jurisdictions already running funded programmes; the Taiwanese authors state that their pilot informed a national acute-care-at-home programme (Wang et al., PMID 42816724), so national data are the most likely next source. Medium-low: the final palliative framework names carer capacity or technology funding. Low: a national commissioning specification funding alert handling or carer support.

Over five years, three scenarios. Scale without gates: virtual wards and remote monitoring expand on bed-pressure grounds, and alert and carer work is absorbed by community nursing and unpaid carers; early indicators would be more charity-funded services closing, nurse consultation time rising in published evaluations, and carer-burden findings replicated. Evaluated infrastructure: a national outcomes dataset with alert workload costed, as Israel’s integrated payer-provider has produced; early indicators would be a commissioning specification requiring validated outcome measurement and carer assessment, and a UK matched cohort. Hub consolidation: Quebec-style and Seha-style hubs absorb monitoring work for dispersed sites; early indicators would be hub-and-spoke contracts and evidence of the spoke taking over, as Quebec’s own site launched a local virtual ward team after five months. The key variables are who funds alert time, whether outcome measurement is required, and whether carer capacity is assessed.

Digital design improves the reliability of care; its effect on the person’s time is unknown

That shape is visible only when emergency department frailty cohorts, goals-of-care trials, a surgical verification programme, cardiac and pulmonary rehabilitation, and remote-monitoring workload studies are read as one body of evidence; each is individually unremarkable and individually honest about its limits. Held together they say something that the sector’s own account of technology does not: that “technology gives time back” is being asserted on the strength of evidence that technology gives the system reliable process, while the one place the month’s evidence counts time directly, in alert desks, nurse-practitioner minutes and carer labour, shows it relocating.

The finding applies to the published patterns themselves. Three of the nine established patterns, on identification, exercise delivery and investment, were written at the level where evidence is strongest, not where it matters, and each narrowed when tested against evidence already published. The same discipline applied outward says that a pattern is not strengthened by recurrence if every recurrence is at process level. A patient’s time is value-added only if what follows the score, the alert or the call is delivered to them, and no source this month measures that. Until a system reports time at home, carer burden and alert workload alongside its outcome, the strongest honest claim is that digital design improves the reliability of care and that its effect on the person’s time is unknown. Israel’s national programme shows that outcomes at scale can be measured; it does not yet show what that scale does to the people who hold the home end of it. Synthesis