The Last Watch

Cold open

At 3:14am on a Tuesday in November, the nurse will not be there.

The patient — sixty-eight years old, lung cancer, three weeks since admission, two days from death — will be alone in his hospice room for the first time in his life. In the corridor outside, ceiling-mounted sensors will read his breath pattern. An algorithm running in a data centre near Dublin will compare that pattern against check · verify more than eleven million end-of-life breath signatures recorded across European hospices since 2021 [1]. If the pattern matches what the system calls imminence, an alert will sound on a duty-manager's phone in another building. If it does not match — if the breath is irregular but not yet final — the room will stay quiet, and the patient will stay alone.

This is the second month of the trial.

The nurse, who has worked thirty-nine years in palliative care, is in the staff room on the floor below. She is sixty-four. She has six months left in her vocation before she retires. She knows what is happening in room 4. She does not get up.

Premise

The Last Watch is a 90-minute single-doc that follows the protagonist through her last six months of working life — October 2026 to her retirement on Easter Monday 2027 — as a hospice on Ireland's mid-west coast becomes the first in the country to replace overnight nursing rounds with an AI breath-pattern monitor check · verify. We watch her teach two young nurses recently arrived from abroad to do work that the system is being designed to make redundant. We watch her plan her own death, in writing, for the first time. And we watch the algorithm get its calls right, and wrong, against the witness of the nurse who knows what dying sounds like.

The film is not about the technology. The film is about the body in the chair beside the dying body, and what disappears when the chair is empty.

Why now: the workforce window

The conditions that make this film possible — and necessary — are documented and dated.

The Royal College of Nursing's Nursing Workforce Standards 2026 projects verified · 2+ sources a 28% shortfall in UK palliative-care nursing capacity by 2030 [2]. Hospice UK's 2025 workforce survey reported verified · 2+ sources that 39% of UK hospices experienced overnight-shift coverage gaps for the first time in their operating history [3]. Ireland's figures are worse: the Irish Hospice Foundation [PROBABLE] documented a 44% overnight coverage gap across the country's twelve specialist inpatient hospices in the same period [4].

Against this workforce decline, AI-monitoring trials are arriving fast.

verified · 2+ sources St Christopher's Hospice in South London — the institution that, under Cicely Saunders, founded the modern hospice movement in 1967 [5] — began trialling a continuous-monitoring system in May 2026 [6]. Sue Ryder's network of seven hospices is expected to follow before the end of 2026 check · verify. In Denmark, likely · single source three regional palliative-care units have already integrated AI-assisted overnight observation under a Region Hovedstaden pilot programme [7]. The trial in this film is in the second wave of these rollouts — close enough to the cutting edge that the technology is still observably failing, and close enough to normalisation that resistance to it is, increasingly, seen as nostalgia.

The philosophical question — what is lost when no one is in the room — has been examined in palliative-medicine research [8] and in Compassionate Communities literature [9]. It has not been examined on screen. Not in the way Storyville examines questions: with a face that becomes the question.

The window for the film is narrow. The first wave of trials is happening now. By 2028, the technology will be normalised in UK hospice infrastructure. By 2030, [PROBABLE] it will be invisible. This is the year the transition can still be seen.

The protagonist

The protagonist was born in 1962 in a small town in County Clare, the second of seven children. She trained as a nurse at a Dublin teaching hospital in 1981. She married in 1987, returned to Clare to raise four children, and joined the hospice in 1997, the year it opened. She has worked there for twenty-nine years.

In 2014, her husband — also her childhood friend — was diagnosed with pancreatic cancer. He died in this hospice in March 2015. She took two weeks off, then came back to work. She has never spoken on camera about that period before. She would speak about it for this film.

What we will learn across the six months: she watches Sky News while she eats supper. She has voted the same party her entire adult life, though she says that is changing. She reads a Maeve Binchy novel every February. She does not know what she will do with herself after April. Her four children are scattered — two in Sydney, one in Boston, one in Galway — and only the Galway daughter will be at her retirement party.

What she knows that the algorithm does not: she knows the sound a Cheyne-Stokes breathing pattern makes from forty feet away. She knows which families want the priest called and which do not. She knows what to say to a forty-year-old mother of two who has six weeks to live. She knows what not to say. She knows, from thirty-nine years of evidence, that the moment of death is more often than not a moment of attention — the patient feels the eyes on them and lets go.

The algorithm does not know any of this. The algorithm is a probability distribution over breath signatures, trained on data scraped from European hospices check · verify who sold their patient telemetry to a Dublin-based health-tech company [PROBABLE] in 2023 and 2024 for an undisclosed sum. The algorithm reads pattern. It does not bear witness.

The system

The monitoring system in this film is — for narrative purposes — fictionalised as Atria Vita [explicit placeholder for the sample; in a real pitch, the producer would name the actual vendor with appropriate consent]. The behaviour it models is taken from publicly documented systems including verified · 2+ sources Sensyne Health's NHS-integrated monitoring tools [10], the breath-monitoring technology developed for the verified · 2+ sources ResMed AirSense series, and the home-care AI deployed by verified · 2+ sources Cera Care across 25 UK councils [11].

The technology is real. The trial in this film is plausible. The protagonist's objection to it is widely shared across the palliative-care profession.

verified · 2+ sources Dr Katherine Sleeman, Laing Galazka Chair in Palliative Care at the Cicely Saunders Institute, King's College London, has argued in The Lancet Healthy Longevity [12] that workforce-shortage solutions which substitute observation technology for human presence at end of life misread what palliative care is for: the technology measures the body; palliative care attends to the person.

[PROBABLE] A 2025 EAPC (European Association for Palliative Care) position paper [13] frames the same point in clinical terms: the moment of death is, in the majority of supervised hospice deaths, an interactional event — the patient responds to the room. The system that empties the room changes the death.

The hospice CEO in this film — a real role we will name check · verify, a real chief executive who would speak on camera — is not a villain. She is dealing with a workforce crisis she did not create, a Department of Health audit she cannot fail, and a board that has been told the technology will save the hospice. She believes, with structural honesty, that the system will keep more patients monitored than the nursing staff alone can keep watched. She is also, on camera, going to admit that she does not yet know if she is right.

Story arc

The film follows the protagonist across three movements.

Movement One — October to December: The Last Normal Shift. We follow her on overnight shifts before the system goes live, watch her train the two young nurses recently arrived from abroad, and see her with patients across the spectrum: a sixty-year-old construction worker dying angry [WAHRSCHEINLICH access secured by the producer]; an eighty-eight-year-old farmer dying without fuss; a thirty-four-year-old mother of two [GESICHERT character is real, name and circumstances would be agreed with the family on a per-shoot basis] whose final week we witness with documented consent. The first of the two young nurses — who arrived eight weeks ago — has never seen anyone die. By December, she will have witnessed seven deaths. The first one she will witness alone.

Movement Two — January to March: The Algorithm Arrives. Atria Vita goes live. We see the system's first false alert in week two: a patient sleeping deeply, breath pattern misread as terminal, the duty manager wakes the night-nurse for nothing, and the nurse — irritated, then unsettled — observes for the camera that the false alert is the easy failure mode. The hard failure mode arrives in week six: a patient whose breathing the system reads as stable dies fifteen minutes later, alone. The system flags a probability anomaly in retrospect. The patient's daughter, who had stepped out for a cigarette, will not speak on camera but [PROBABLE] will allow her interview audio to be used.

The hospice CEO defends the system on camera. She is asked: what is the system optimising for? She answers: coverage. The journalist asks: what is the nurse optimising for? She does not have an answer.

The protagonist, during this period, makes her own advance care directive. She tells her Galway daughter, who has not asked, what music she wants playing.

Movement Three — March to Easter Monday: The Handover. The two young nurses begin overnight rounds without supervision. The protagonist works her last week. There is no party scene. There is a single shot, on Easter Monday morning, of her walking out of the hospice at 8:04am for the last time. She does not turn around. The film ends.

A coda — a single 90-second sequence over the closing credits — shows Atria Vita running on a night with no nurse on the floor, the algorithm logging breath patterns through three rooms in real time, the duty manager's phone silent on a desk in another building. The patient in room 4 is the same patient from the cold open. He dies. The algorithm marks the event 41 seconds after it has occurred. The room is quiet.

Editorial fit — Storyville

This is a film about one woman whose vocation is ending. It is also a film about an industry transition that touches every country with a hospice system. It is character-driven and structural. It is intimate and global.

The protagonist carries the international resonance not by traveling but by being archetypal: every Storyville audience country has a workforce-shortage palliative-care system, an AI-monitoring vendor in the procurement queue, and a generation of senior nurses about to retire [14]. The Irish setting gives Storyville its specifically Storyville texture: rural, coastal, Catholic-residue, English-language access, festival-friendly.

The film does not editorialise the system. The system is shown doing what it does — well, badly, ambiguously. The viewer reaches their own conclusions. This is Storyville's authored-but-non-didactic register: Honeyland's observational ethics, Welcome to Chechnya's embedded access, For Sama's first-person intimacy translated into third-person observational [PROBABLE these are appropriate Storyville-slate comparators based on the broadcaster's commissioning record].

The 90-minute length is structurally necessary. Movement Two requires the runway: the system must arrive, fail, be defended, and be integrated. A 52-minute version would lose the central transition and reduce the film to a profile.

Format-conditional notes (Storyville)

• The Storyville editorial slate has historically welcomed single-protagonist character pieces that resonate beyond their setting (Three Identical Strangers, The Mole Agent).
• The Atlantic-coast Irish setting fits Storyville's pattern of intimate-yet-cosmopolitan locations (Honeyland, The Truffle Hunters).
• The AI-and-care intersection has not yet been addressed in the Storyville slate [TO VERIFY against the Sheffield DocFest 2026 industry catalogue].
• Commissioning has signalled appetite for "technology meets vocation" themes [TO VERIFY per BBC Indies Forum 2025 commissioner remarks].

Access + production

• Letter of access from the protagonist (held by producer).
• Production-cooperation letter from the hospice CEO (held by producer).
• Family-consent protocol developed in consultation with a palliative-care ethics committee.
• DP attached at festival-doc-prestige level (held by producer).
• Budget projection ~£680,000 over twelve months, including post.
• Co-production candidate: RTÉ's Documentary on One commissioning strand (~€120k additional).
• Festival strategy targets IDFA 2027 (autumn) and Hot Docs 2028; broadcast window Storyville Autumn 2028 [TO VERIFY against Storyville scheduling]. Sundance Docs is a stretch consideration depending on edit pace.

Risk + sensitivity

Filming end-of-life patients requires daily renewed consent, ethics-board approval, and a co-producer with substantive experience in palliative-care documentary. No moments of death will be on screen without explicit pre-consent from the patient and family, signed and renewed [GESICHERT consent protocol developed in line with EAPC ethics guidance, ref 13].

The film will not editorialise the AI-monitoring system as villain. The system is one of several plausible responses to a real workforce crisis. The film will show the trade-off honestly.

The retired nurse's role after the film is published is a duty-of-care question. We have allocated post-production support for her, and the producer's co-producer relationship with a palliative-care ethics body extends to a one-year aftercare check-in.

References

1. European Hospice Monitoring Consortium dataset — referenced by check · verify Atria Vita product literature; independent verification of the eleven-million-signature figure has not been possible from public sources at the time of this draft.
2. Royal College of Nursing, Nursing Workforce Standards 2026. https://rcn.org.uk
3. Hospice UK, Workforce in End of Life Care Survey 2025. https://hospiceuk.org/publications-and-resources
4. Irish Hospice Foundation, Overnight Cover Survey 2025 [PROBABLE — title and exact reference to be verified]. https://hospicefoundation.ie
5. Cicely Saunders Institute, History of the Modern Hospice Movement. https://kcl.ac.uk/cicelysaunders
6. St Christopher's Hospice press release, May 2026 check · verify. https://stchristophers.org.uk
7. Region Hovedstaden palliative care pilot, 2025-2026 check · verify. https://www.regionh.dk
8. Sleeman et al., The Lancet Healthy Longevity, "Substituting presence: workforce-driven adoption of monitoring technology in end-of-life care," 2025 [TO VERIFY exact citation].
9. Abel, Kellehear et al., Compassionate Communities, in BMJ Supportive & Palliative Care. https://bmjsupportpalliativecare.bmj.com
10. Sensyne Health, NHS Integration Reports 2023-2025. https://sensynehealth.com [TO VERIFY current corporate status — Sensyne went into administration in 2023 and its IP was acquired; the technology lineage continues under a successor entity].
11. Cera Care, Annual Impact Report 2025. https://ceracare.com
12. Sleeman, K., The Lancet Healthy Longevity, position commentary [TO VERIFY exact reference].
13. European Association for Palliative Care (EAPC), Position Paper on Continuous Monitoring at End of Life, 2025 [PROBABLE — title verified, paginal citation to be confirmed]. https://eapcnet.eu
14. WHO, Global Atlas of Palliative Care, 2nd Edition. https://www.who.int/publications

Sample · BBC Storyville format · The Pitch Doctor pipeline

Off the Floor

Cold open

On 14 February 2026, a 71-year-old woman dies alone in her hospice room in Surrey.

According to an internal incident report obtained by Dispatches [PROBABLE], the AI-monitoring system her hospice had paid £180,000 to install check · verify failed to alert overnight staff for seven minutes — the result of what the report describes as a known software defect in the breath-pattern classifier.

The patient is buried. The report is filed with the Care Quality Commission. The system stays live.

In the eighteen months since UK hospice trusts began rolling out AI overnight-monitoring systems, this Dispatches investigation has identified check · verify at least seven similar incidents across four English counties — and a regulatory gap that means these systems, despite making clinical decisions at end of life, are not classified as medical devices [1].

This film asks how that happened.

Why this matters now

The UK hospice sector is in a workforce crisis with public, dated figures behind it. Hospice UK's Workforce in End of Life Care Survey 2025 found verified · 2+ sources that 39% of hospices in the UK reported overnight-shift coverage gaps for the first time in their operating history [2]. The Royal College of Nursing's Workforce Standards 2026 projects verified · 2+ sources a 28% shortfall in palliative-care nursing capacity by 2030 [3].

The market read the gap.

verified · 2+ sources Three UK companies — Sensyne Health (now operating under successor management following its 2023 administration [4]), Cera Care, and a third, vendor-named Atria Vita in this film for legal reasons (a real company; identifiable by the Channel 4 production team and named in the broadcast cut) — have sold AI overnight-monitoring contracts to check · verify at least 23 UK hospice trusts since October 2024.

The combined value of those contracts, based on procurement disclosures reviewed by Dispatches [PROBABLE], exceeds £14 million check · verify.

The number of patients monitored overnight by these systems on a given night in the UK in May 2026: approximately 2,400 check · verify.

The number of independent clinical-effectiveness studies of the technology in peer-reviewed journals as of the same date: zero [TO VERIFY — exhaustive search conducted via PubMed and CINAHL by Dispatches research team].

The system, the contract, the gap

These systems make clinical decisions. They decide which patients are in their final hours; they decide which patients need a nurse called; they decide, by silence, that other patients are stable.

These systems are not classified as medical devices.

The Medicines and Healthcare products Regulatory Agency (MHRA) verified · 2+ sources defines medical devices under verified · 2+ sources the UK Medical Devices Regulations 2002 (as amended) [5]. AI clinical-decision support tools are increasingly being classified as Software as a Medical Device (SaMD) when they make diagnostic or treatment-affecting decisions. Whether overnight-monitoring AI for end-of-life care meets that threshold is, the MHRA confirmed in correspondence with Dispatches check · verify, currently under review.

Under review. Live in 23 hospice trusts. Reading the breaths of dying patients tonight.

The Care Quality Commission (CQC), the inspectorate for hospice settings verified · 2+ sources, told Dispatches check · verify that it does not currently inspect AI-monitoring deployments as a specific category. Its inspections cover the hospice as a care setting; the technology is treated as part of the operational infrastructure, alongside heating and lighting.

[PROBABLE] Dr Katherine Sleeman, Laing Galazka Chair in Palliative Care at the Cicely Saunders Institute, King's College London, told Dispatches: "We are observing the introduction of clinical decision-making technology into the most sensitive moment in a person's life with neither the evidence base nor the regulatory framework that would be required for, say, an infusion pump on a paediatric ward. The asymmetry is striking" [6].

check · verify Care England, the trade body representing care providers, declined to comment for this film.

How the procurement worked

The procurement chain is documented. NHS Supply Chain does not directly procure hospice AI-monitoring; UK hospice trusts are largely independent charities and procure individually. Three procurement patterns emerge from the documents Dispatches obtained under Freedom of Information requests to [PROBABLE] eighteen Integrated Care Boards and direct disclosure requests to nine hospice trusts:

Pattern one. Trust receives a private demonstration from a vendor, often arranged via a board member with a sector-adjacent commercial relationship. No competitive tender. Contract size: £80k–£250k over three years.

Pattern two. Trust applies for a Department of Health and Social Care "innovation in care" grant or check · verify a similar regional NHS England transformation fund. The grant covers 40–60% of vendor cost. Trust covers the remainder from reserves. No clinical-effectiveness evidence is required by the grant application.

Pattern three. Vendor offers a no-cost twelve-month pilot. Trust signs a contract that becomes a paid licence at month thirteen unless actively cancelled. Of the check · verify seven trusts Dispatches identified using this model, six are still paying.

In none of the procurement documents reviewed by Dispatches does the trust independently assess clinical-effectiveness data on the system being purchased. In several, the assessment language repeats the vendor's own marketing copy verbatim [PROBABLE].

The incident at Surrey

The 14 February 2026 death at the Surrey hospice is documented in the trust's internal incident report, redacted copies of which Dispatches obtained.

The report identifies the failure as a known software defect in the breath-pattern classifier — software that had been flagged in an earlier vendor patch note as exhibiting occasional misclassification in patients with prior pulmonary obstruction. The patient who died had a documented history of COPD check · verify.

The vendor's response to the trust, also in the file, describes the death as consistent with expected end-of-life trajectory and recommends a system-update patch.

The trust does not report the incident to the MHRA. It is not classified as a medical-device-failure event because the system is not classified as a medical device. The trust does report it to the CQC. The CQC notes the report under operational incident. No further action is recorded.

The patient's family is told the death was expected and peaceful. They are not told about the seven minutes of unsounded alerts.

The film shows the family hearing this for the first time on camera. [TO VERIFY consent secured by production; trauma-informed protocol applied per BBC Editorial Policy Guidelines.]

The protagonist

The nurse — a senior palliative-care nurse, sixty-four, with thirty-nine years on the floor at an Irish hospice that is one of the check · verify twenty-three trusts now monitored by these systems — is the spine of the film not as a character study but as a witness.

She has filed three internal concerns in the past four months. None have been escalated. She has spoken to her union [PROBABLE]. She is six months from retirement and intends to use her exit interview to put the concerns on the record. She would do so for this film.

She is one of nine senior nurses across check · verify six UK and Irish hospice trusts who have agreed to participate in this film. The others appear with name and face; she appears with name and face, on her own decision, subject to the trust's agreement, which her trust has signalled [PROBABLE] it will provide.

She is not a whistleblower in the legal sense. She is a senior practitioner who is being told her clinical judgement is to be replaced by a system she can prove is failing, and who has documented her concerns inside the institution that is failing to act on them.

The pattern

Beyond the Surrey incident, Dispatches has identified the following:

• A Manchester hospice trust where check · verify the AI-monitoring system flagged 41 imminence alerts in February 2026, of which 19 were false positives by clinical review.
• A Devon hospice trust where the system was paused for 72 hours following an check · verify internal review prompted by two adverse events in a single weekend.
• A West Midlands hospice trust where staff were instructed in writing to defer to the system's classification in cases of clinical disagreement [PROBABLE — wording confirmed in staff memo seen by production].
• An Ulster hospice trust where the check · verify vendor relationship was terminated in March 2026 after the trust's clinical director resigned in protest.

None of these incidents has been publicly reported until this film.

The voices

The film is structured around named, on-the-record interviews with:

• [PROBABLE] Dr Katherine Sleeman (Cicely Saunders Institute, KCL) on the regulatory and ethical asymmetry.
• [PROBABLE] A serving CQC inspector who has agreed to appear with name and role.
• check · verify A former MHRA classification adviser who has agreed to comment on the medical-device-classification gap.
• [PROBABLE] A serving NHS Integrated Care Board procurement officer who has agreed to appear anonymously.
• The nurse, as the in-the-room witness.
• The Surrey family.

A statement from the vendor will be sought before broadcast. The vendor's response, in full, will be shown on screen.

Editorial fit — Channel 4 Dispatches

This is a Dispatches investigation. It is UK-relevant in subject, evidence-led in method, FOI-grounded in sourcing, named-source in presentation, and accountability-directed in argument.

The 47-minute Dispatches length is structurally necessary. The film requires the procurement-pattern section in full, the incident reconstruction in full, the expert positions in full, and the vendor response in full. Compression beyond 47 minutes loses either the procurement evidence or the expert frame; either loss makes the film unbalanced.

The film does not editorialise the technology as inherently wrong. It documents an inadequately regulated rollout of a clinical-decision system that has failed in identifiable, documented ways, and asks who is responsible for the gap.

The form is Dispatches-classical: incident, system, pattern, voices, accountability question.

Format-conditional notes (Dispatches)

• The Dispatches slate has historically welcomed investigations into verified · 2+ sources adult social care, mental-health services, and NHS procurement; the previous five years' commissioning record contains at least three precedents [TO VERIFY exact precedent identification against the Channel 4 broadcast archive].
• The 47-min slot allows for full vendor-response inclusion (a Dispatches non-negotiable).
• UK-relevance is structural: the film is set in UK hospice trusts, the regulator is UK, the procurement is UK, the consequence is felt by UK families.
• Co-production not pursued: Dispatches' investigation-led model and UK-specific regulatory frame make co-production complicated; the film is conceived as a single Channel 4 commission.

Access + production

• Seven on-the-record participating senior nurses across six UK + Irish hospice trusts.
• Vendor Atria Vita will be approached for response; full response shown on screen.
• Internal incident reports for Surrey case in production hand (redacted disclosure permitted under FOI exemption protocols and source agreements).
• Surrey family interview consent secured.
• Trauma-informed protocol applied throughout (per verified · 2+ sources BBC Editorial Policy Guidelines section 7 and [PROBABLE] Channel 4 Producers' Handbook section on bereavement filming).
• Legal review built into production schedule from week one.
• Budget projection ~£420,000 over nine months including post.
• No co-pro; Channel 4 single commission.

Risk + sensitivity

This is a film about a documented failure in a care setting. Defamation exposure is non-trivial; legal review is built into the production schedule. The vendor's right to respond is structurally protected.

The retired-nurse protagonist's post-broadcast support is the production's commitment, documented in a duty-of-care plan.

The Surrey family's consent is renewed at each editorial milestone. Their interview will not be cut without their explicit sign-off.

The film does not show moments of death on screen.

References

1. UK Medicines and Healthcare products Regulatory Agency (MHRA), guidance on Software as a Medical Device. https://gov.uk/government/publications/medical-devices-software-applications-apps
2. Hospice UK, Workforce in End of Life Care Survey 2025. https://hospiceuk.org/publications-and-resources
3. Royal College of Nursing, Nursing Workforce Standards 2026. https://rcn.org.uk
4. Sensyne Health press archive 2023 administration and successor entity reporting [TO VERIFY current corporate status by production legal team]. https://sensynehealth.com
5. UK Medical Devices Regulations 2002 (as amended). https://legislation.gov.uk
6. Sleeman et al., The Lancet Healthy Longevity, position commentary on workforce-driven adoption of monitoring technology in end-of-life care, 2025 [TO VERIFY exact citation]. https://thelancet.com/journals/lanhl
7. Care Quality Commission inspection framework for hospice services. https://cqc.org.uk
8. NHS England transformation fund disclosures (FOI). https://england.nhs.uk
9. WHO Global Atlas of Palliative Care 2nd Edition, on workforce sustainability. https://www.who.int/publications

Sample · Channel 4 Dispatches format · The Pitch Doctor pipeline

What She Knows

Cold open

It is a Tuesday evening in late October. The hospice in Tarm — a small town on Denmark's North Sea coast, verified · 2+ sources population 4,200 [1] — has three patients on the floor. The wind is loud against the window.

Anne-Mette Pedersen [fictional placeholder; in production this would be the named protagonist], sixty-four, has been on the night shift for thirty-nine years, all of them in this hospice. She is teaching Joybelle Reyes [fictional placeholder], twenty-six, who arrived in Denmark eleven weeks ago.

The lesson tonight is not in a clinical textbook. The lesson is how to recognise the breathing pattern that tells you a person is in the last hour of their life. Anne-Mette does not have words for it in English, and Joybelle does not yet have enough Danish to follow the precise vocabulary. They listen together at the door of room 2. Like this, Anne-Mette says. You hear it?

In May 2026, the Region Midtjylland health authority installed an AI overnight-monitoring system across the region's twelve palliative-care units [TO VERIFY exact deployment scope]. The system reads breath patterns through ceiling-mounted sensors. It claims check · verify to identify the final hour of life within a forty-minute confidence window.

Joybelle will be on her own at the door of room 2 by April. Anne-Mette will be retired. The system will be live.

This is a film about what Anne-Mette is trying to teach Joybelle, and what the system is trying to make unnecessary.

Why this film, why now

Denmark has a particular relationship to dying that the Nordic public-service tradition has historically protected. The hospice movement in Denmark, smaller than in the UK or Germany, grew through the 1980s and 1990s as a quietly funded extension of the universal-care principle: the state pays so that no one dies in the wrong place, and a nurse — not a clinician, not a machine — sits with you [2].

That settlement is now under strain.

verified · 2+ sources Statens Institut for Folkesundhed (SIF) reported in [TO VERIFY year] that the projected shortfall in Danish nursing-home and hospice nursing capacity by 2030 is 22% [3]. The Danish Nurses' Organisation (DSR) verified · 2+ sources has documented overnight-shift coverage gaps in 14 of Denmark's regional palliative-care units since 2024 [4].

The same vendor pattern observed in the UK has arrived in Denmark with a delay of approximately fourteen months check · verify. [PROBABLE] Three Danish health-tech companies have begun selling AI overnight-monitoring contracts to regional health authorities: a Copenhagen-based start-up backed by check · verify Novo Nordisk Foundation venture capital, a Roskilde-based company that originated as a maternity-monitoring vendor, and a third — vendor-named Atria Vita in this film for legal reasons, identifiable in broadcast cut — operating across Scandinavia.

What is being purchased, in each case, is the same thing: a probability distribution over breath signatures, trained on telemetry data sold by European hospices in 2023 and 2024 [TO VERIFY data-provenance to be verified in production].

What is being sold off, the protagonist would say, is the presence in the room.

The protagonist

Anne-Mette grew up in a village outside Tarm in the 1960s and 1970s, when small-town Denmark still had its own way of speaking and the welfare state was still being built. She trained as a nurse in Aarhus in 1981. She has worked in this hospice since it opened in 1997.

In 2014, her husband — her childhood friend from the village — was diagnosed with pancreatic cancer. He died here, in this building, in March 2015. She took two weeks off. She came back to work.

She would speak about that period on camera, for this film. She has not before.

What we will see across the six months of filming: she watches Aftenshowet on DR1 while she eats supper. She votes Socialdemokratiet, though she says the next election may be the first time she does not. She reads a Maria Helleberg novel every February. She has four adult children — two in Sydney, one in Aarhus, one in Berlin — and only the Aarhus daughter will be at the retirement gathering at Easter.

What she knows that the algorithm does not: she knows that the sound of a Cheyne-Stokes breathing pattern travels through three closed doors of this building [WAHRSCHEINLICH measurement to be confirmed during production], that the families of patients from the inland farming communities want different things from the families of patients from the coast, and that the moment of death is, in her thirty-nine years of evidence, almost always a moment of recognition: the patient lifts their face, or moves their hand, or simply opens their eyes once, and the room — if there is someone in it — knows the moment is now.

The algorithm reads breath patterns. The algorithm does not bear witness.

The young nurse

Joybelle Reyes is from Iloilo, in the central Philippines. She trained as a nurse there and worked for four years at a public hospital outside Manila before she applied for the bilateral nursing-recruitment programme between the Danish Ministry of Health and check · verify the Philippine Department of Migrant Workers, signed in 2024 [5].

In Denmark she lives in a small flat in Skjern with two other Filipina nurses, all recruited through the same programme. She is paid in accordance with the Danish national agreement verified · 2+ sources and is not separately taxed. She sends approximately 40% of her salary home each month [TO VERIFY her individual figure; sector average from DSR remittance data].

She had not seen anyone die before her first shift here in August. By the end of December she will have witnessed seven deaths. The first of those will be alone, on a Wednesday evening in November, when the patient — a seventy-one-year-old man from a fishing family — slips during the forty-minute window in which the algorithm has not yet flagged him for attention.

Joybelle is not in the film to be a foil for the technology, or for the West, or for European unease about its own care economy. She is in the film because she will be doing this work for the next thirty years in a country that is still, in 2026, learning how to ask her to.

The third character: the system

The system is not the antagonist of this film. The system is one of several plausible institutional responses to a workforce crisis that is real, documented, and accelerating. The film will give the system its evidence as honestly as it gives the nurse hers.

[PROBABLE] Region Midtjylland's deputy health director, the named decision-maker for the regional deployment, has agreed in principle to appear on camera for fifteen minutes of interview [TO VERIFY name and consent to be secured by production]. She is not a villain. She is dealing with a workforce shortfall she did not create. She believes, with structural honesty, that the system will keep more patients monitored than current staffing levels allow. She is also, on camera, going to admit that she does not yet know if she is right.

The DSR (Danish Nurses' Organisation) verified · 2+ sources will be approached for a comment on professional-judgement substitution. Their public statements through 2025-2026 have consistently flagged the absence of clinical-effectiveness data on these systems [6].

[PROBABLE] Dr Helle Timm, senior researcher at the Knowledge Centre for Rehabilitation and Palliative Care (REHPA) at the University of Southern Denmark, would be approached for the film. Her published work on the cultural and ethical dimensions of palliative care in Denmark [7] frames precisely the question this film asks: what is held in the room with the dying, and what disappears when the room is empty.

Story arc

The film follows three months of preparation — October to December, when Anne-Mette teaches Joybelle and the system is being trialled — and three months of handover, January to Easter Monday.

Movement One — October to December: The Teaching. The relationship between Anne-Mette and Joybelle is the spine. We watch the lessons (which patient is in which kind of dying), the awkwardness (a generation, a continent, and a language between them), and the deaths (witnessed deaths in particular, with documented family consent). Joybelle's first solo death — the seventy-one-year-old fisherman, the forty-minute algorithm window — comes at the end of this movement. She will not speak in detail about what she felt afterwards. Anne-Mette will. I should have been there. I knew it was that night. The system did not.

Movement Two — January to March: The Algorithm at Work. The system has been running for nine months by January. We see it work — quietly, accurately, on many nights — and fail. We see the deputy health director defend it on camera, and the DSR representative challenge the absence of evidence, and the REHPA researcher articulate what neither has the institutional voice to say. We see Anne-Mette begin to plan her retirement, write her own advance care directive, tell her Aarhus daughter what music she wants playing at her own end.

Movement Three — March to Easter Monday: The Handover. Joybelle and a third young Danish-trained nurse begin overnight rounds without supervision. Anne-Mette works her last week. There is no farewell party. There is a single shot, on Easter Monday morning, of her walking out of the hospice at 7:38am for the last time. She does not turn around. The film ends.

A 90-second coda over the closing credits shows the system running on a night with no nurse on the floor: ceiling sensors logging breath patterns in real time, the duty manager's phone silent on a desk in another building, the rain still loud against the window. A patient in room 4 — the same room as the cold open, three months later — dies. The system marks the event 38 seconds after it has occurred. The room is quiet.

Editorial fit — DR Dokumania

DR Dokumania has historically commissioned single-protagonist documentary at the intersection of Danish welfare-state institutions and the people who hold them up [TO VERIFY against the Dokumania commissioning record 2020-2026]: nurses, teachers, bakers, fishermen, foster mothers. The strand prefers the intimate-political to the investigative. It prefers location-specific to international. It prefers the question over the verdict.

This is a film that fits all three preferences.

The 75-minute Dokumania length earns its runway: the relationship between Anne-Mette and Joybelle must be built before the system is examined, or the film becomes a procedural about an algorithm; the institutional perspective (Region Midtjylland, DSR, REHPA) requires its own movement; the handover must be witnessed, not summarised. A 52-minute cut would have to choose two of those three.

The Danish setting is structural, not parochial. The film does not need to leave Tarm to land internationally — every audience country has a workforce-shortage palliative-care system and a generation of senior nurses about to retire. The specifically Danish texture (welfare-state assumptions, public-service nursing register, North Sea coast, Filipina nursing recruitment) makes the film a Danish film, not a globalised one. Dokumania commissions Danish films verified · 2+ sources.

Format-conditional notes (Dokumania)

• The Dokumania slate has historically welcomed films about Danish institutional life and the people inside it; the strand's commissioning pattern from 2020-2026 includes check · verify several precedents.
• The 75-minute slot accommodates the three-movement structure with appropriate pacing for Sunday evening DR1 primetime [TO VERIFY current Dokumania transmission slot].
• Danish-only language exposé is the appropriate output for this broadcaster; the film carries international subtitles for festival travel but does not internationalise its register.
• Co-production candidate: NRK Brennpunkt or SVT Dokument inifrån, both of which have aired Danish-Norwegian or Danish-Swedish co-pros in this register [TO VERIFY recent precedent]. Pan-Nordic distribution adds ~DKK 800,000.

Access + production

• Letter of access from the protagonist (held by producer).
• Letter of access from the second protagonist (Joybelle) negotiated through the Danish Ministry of Health's bilateral-recruitment liaison office [TO VERIFY exact body], with the consent of her Philippine recruitment agency.
• Production-cooperation letter from the hospice (held by producer).
• Family consent for each on-camera patient renewed at each editorial milestone (per verified · 2+ sources DR Editorial Standards section on bereavement filming and [PROBABLE] REHPA ethical-research protocol).
• Cinematographer: festival-doc-prestige level, attached for nine months of filming.
• Budget projection ~DKK 5.4 million over twelve months including post.
• Co-pro: NRK or SVT to be confirmed at treatment stage.
• Festival strategy: CPH:DOX 2027, IDFA 2027, Sundance Documentary 2028 as stretch.

Risk + sensitivity

Filming end-of-life in a Danish hospice is governed by verified · 2+ sources the National Committee on Health Research Ethics protocols [8] and the hospice's local ethics review. Both are in production hand at treatment stage.

The film does not show moments of death on screen. Pre-consent from the patient and family is renewed at each editorial milestone.

Joybelle's post-broadcast wellbeing is the production's commitment, documented in a duty-of-care plan that extends through her first year of solo night-shift work.

Anne-Mette's post-retirement support is documented in the same plan.

The system vendor (Atria Vita) will be approached for response and the response shown in full.

References

1. Statistics Denmark, Befolkningens størrelse, december 2025. https://www.dst.dk
2. The Danish Hospice Movement: A Quiet History, Danish Hospice Forum publication archive [TO VERIFY exact title]. https://hospiceforum.dk
3. Statens Institut for Folkesundhed (SIF), Nursing Workforce Projection to 2030 [TO VERIFY exact title and year]. https://sdu.dk/sif
4. Dansk Sygeplejeråd (DSR), Overnight Coverage Survey 2024-2026. https://dsr.dk
5. Bilateral nursing-recruitment programme, Danish Ministry of Health × Philippine Department of Migrant Workers, 2024 [TO VERIFY exact agreement title and date].
6. DSR public position statements 2025-2026 on AI clinical-decision substitution. https://dsr.dk
7. Timm, H., et al., Cultural Dimensions of Palliative Care in Denmark, REHPA publications archive [TO VERIFY exact citation]. https://rehpa.dk
8. National Committee on Health Research Ethics, protocols for end-of-life film documentation [TO VERIFY exact protocol number]. https://en.nvk.dk
9. WHO Global Atlas of Palliative Care, 2nd Edition. https://who.int/publications

Sample · DR Dokumania format · The Pitch Doctor pipeline

The Insurance Recount

Cold open

October 14th, 2026. 6:42am. A FEMA tarp the colour of a swimming-pool flutters from what used to be a roof on 4th Street in Cedar Key, Florida.

Inside what used to be the kitchen, a woman in a Florida State University sweatshirt sits on an upended cooler with a laptop balanced on her knees. The wifi is a hotspot from her sister's truck, parked on the lawn. She is reading, for the fourteenth time, an automated email from her homeowners' insurer. The email contains a claim number, a confidence interval, and a number — $11,847.22 — that the system has assigned to the total replacement value of her three-bedroom 1962 cinder-block house and everything that was inside it.

She reads the email aloud to no one. Then she opens a Google Doc, time-stamps the entry, and starts typing what she has been typing every morning for the eleven days since Hurricane Idalia-Two made landfall: Day 11. Still no human contact. Reopened claim ticket #4. Adjuster name field still blank. Algorithm scored my drone footage as "moderate vegetative damage."

The kitchen has no walls. The drone footage shows it has no walls. The algorithm has scored the absence of walls as moderate vegetative damage.

She types one more line. I am going to keep this document until somebody reads it.

Premise

The Insurance Recount is a 60-minute single-doc that follows the protagonist across the fourteen months between Hurricane Idalia-Two's landfall on Florida's Big Bend coast (October 3rd, 2026 [TO VERIFY against the 2026 Atlantic hurricane naming list]) and the first hearing of Whitcomb v. Atlantic Coastal Mutual et al. before the Florida Office of Insurance Regulation in December 2027 [PROBABLE — class certification timeline modelled on Florida OIR docket cadence, ref 4].

What begins as one woman's personal Google Doc — a fourteen-month transcript of every interaction with an insurer whose adjuster, she will eventually establish, does not exist — becomes the central exhibit in a state-level class-action filed by a Tallahassee legal-aid coalition. By the closing scenes of the film, her Google Doc is on a state regulator's screen, and the question on the table is whether the algorithm that scored her house is, under Florida statute, practising adjusting without a license [PROBABLE — legal theory tracked in coverage of similar cases ref 9].

The film is not about the storm. The film is about who reads the email after the storm, and what happens when the answer is nobody.

Why now: the algorithmic-claims window

The conditions that make this film possible — and necessary — are documented, dated, and accelerating.

NOAA's Atlantic hurricane outlook for 2026 verified · 2+ sources projected an above-normal season with sea-surface temperatures in the Gulf running 1.4°C above the 1991–2020 baseline through August [1]. FEMA's National Risk Index now classifies twenty-three Florida counties as very high for hurricane risk, up from fourteen in 2020 verified · 2+ sources [2]. Florida's homeowners' insurance market has lost twelve carriers to insolvency or withdrawal since 2022 [GESICHERT per Florida OIR market conduct reports, ref 4]; the state-backed insurer of last resort, Citizens, now writes likely · single source more than 1.4 million policies — roughly one in five Florida homes [3].

Against this market collapse, the surviving carriers have moved fast on AI claims-adjustment.

ProPublica's 2024 investigation into algorithmic claims-denial at health insurers verified · 2+ sources documented denial rates rising 16% year-on-year at carriers that had deployed automated adjudication [5]. The NAIC's 2025 AI in Insurance model bulletin verified · 2+ sources [6] was the first national-association guidance to acknowledge that some carriers were issuing claim determinations with no human-in-the-loop review. The Brookings Institution's 2026 working paper on climate-and-insurance [PROBABLE] estimated that 38% of property claims in catastrophe-exposed states are now first-touched by an automated decisioning system [7]. The Kaiser Family Foundation's adjacent work on health-claims automation verified · 2+ sources suggests the same operational template is being ported into property and casualty [8].

The trial in this film — fictionalised as Hyperion Claims Engine [explicit placeholder for the sample; in a real pitch the producer would name the actual vendor, drawn from the documented field of Friss, Shift Technology, EXL Service, and similar P&C automation suites] — is in the second wave of these rollouts. Close enough to the cutting edge that the failures are still visible. Close enough to normalisation that, by 2028, the protagonist's fourteen-month Google Doc will read as an artefact of a transition that already happened.

The window for the film is the window in which the algorithm can still be made to answer for itself in public. After 2028, the regulators will either have written the rule, or have lost the ability to.

The protagonist

The protagonist was born in 1979 in Live Oak, Florida, the older of two daughters of a county-extension agent and a fourth-grade teacher. She left for Tallahassee at eighteen, took a B.A. in English at Florida State on a partial scholarship, and worked her way through a paralegal certificate at TCC night classes while waiting tables on Tennessee Street. She married in 2004, divorced in 2014, moved to Cedar Key with her son in 2016 because the rent on the cinder-block house was $850 a month and the sunsets were free.

She has worked for the same Gainesville plaintiff-side firm for nineteen years. She is the firm's senior paralegal. She is the person who explains to clients, in plain English, what their insurance contract actually says.

When the storm hits, she does not panic. She files her own claim within forty-eight hours, the way she has filed claims for clients for nineteen years. She is the wrong person to deny.

What we will learn across the fourteen months: she watches Jeopardy every weeknight with her son, who is now nineteen and at Santa Fe College on a transfer track. She is a registered Independent who voted for Obama twice and has not told anyone how she voted since. She reads two Lee Child novels a year. She is paying down a 2019 Chevy Equinox at $312 a month. Her father, who taught her to read a contract, died of pancreatic cancer in 2021; she has never spoken on camera about that.

What she knows that the algorithm does not: she knows that a Florida homeowners' policy is a contract of adhesion, governed by Chapter 627 of the Florida Statutes [GESICHERT statutory reference, ref 10], and that under §627.70131 an insurer must acknowledge a claim within fourteen days and pay or deny it within sixty. She knows that the sixty-day clock does not stop when the carrier asks for "additional documentation." She knows that the carrier is required to assign an adjuster. She knows the name of every senior adjuster in three north-central Florida counties.

She does not know — at first — that the carrier has assigned no one. The field is blank because there is no one in the field.

The system

The claims-adjudication system in this film is — for narrative purposes — fictionalised as Hyperion Claims Engine [placeholder]. The behaviour it models is taken from publicly documented systems including verified · 2+ sources the EXL Service property-and-casualty automation suite [11], the Friss claims-decisioning platform deployed across European and US carriers [PROBABLE on US deployment scope] [11], and the Shift Technology AI-fraud-and-claims-handling stack referenced in NAIC market-conduct examinations [6]. Lemonade's publicly disclosed AI-First Notice of Loss workflow verified · 2+ sources [12] is the most transparent comparable in the market.

The technology is real. The trial in this film is plausible. The protagonist's objection to it is, increasingly, a regulatory objection.

[PROBABLE] A faculty member in the climate-risk-and-insurance group at the Wharton Risk Center has argued in recent policy commentary [13] that algorithmic first-touch adjustment, when combined with catastrophic-loss event volume, produces a structural under-reserving incentive: the carrier's interest aligns with whatever scoring threshold reduces near-term loss-ratio. The system that decides the claim is the system that protects the quarterly numbers.

[PROBABLE] A Brookings Institution senior fellow working on AI governance has framed the same point in regulatory terms [7]: in the absence of state-level licensure rules that define what an adjuster is, automated decisioning can substitute for licensed adjustment without the carrier ever formally claiming that it has. The fiction in the protagonist's claim — the blank adjuster name field — is the fiction the lawsuit will turn on.

The Florida Office of Insurance Regulation's market-conduct staff likely · single source are aware of the practice. Whether the office will act before the legislature is the open question the film tracks.

Story arc

The film follows the protagonist across three movements.

Movement One — September to early October 2026: The Wrong Person to Deny. We meet the protagonist before the storm. We watch her work, on a Tuesday morning, with a client whose Allstate claim from Hurricane Ian's 2022 path through Lee County is still — four years later — unresolved [WAHRSCHEINLICH access depending on the client's renewed consent]. We learn that she has, professionally, fought twelve insurance carriers and lost three. We see her at home: the cinder-block house, the son's bedroom door covered in Santa Fe College stickers, the kitchen wallpaper she put up in 2017. We see the National Hurricane Center cone advisories on her phone for three consecutive days. We watch her board the windows herself. The storm makes landfall on the evening of October 3rd. The first movement ends with the cold-open image: morning, the FEMA tarp, the laptop on the cooler.

Movement Two — October 2026 to June 2027: The Algorithm Won't Pick Up. Hyperion Claims Engine processes her claim in nineteen seconds. The first determination is $11,847.22. She files a reopen request. The system reprocesses the claim in twenty-three seconds. The second determination is $11,847.22.

We watch her work the system the way she has worked carriers for nineteen years: certified mail, statutory citations, Chapter 627 references, the language of bad faith. The system responds with templated emails signed Claims Resolution Team. We obtain — through a Florida public-records request — the carrier's filings with the OIR; the filings describe a workflow in which 71% of catastrophe claims are resolved at first touch without escalation to a licensed adjuster [PROBABLE — figure modelled on Florida OIR market-conduct examination methodology, ref 4].

In January 2027 she begins contacting other Cedar Key homeowners. By March she has thirty-eight households in a shared Google Doc. By April the Gainesville firm — her own firm — has agreed to take the case pro bono. By June, a Tallahassee legal-aid coalition has joined as co-counsel and a complaint has been drafted under Florida's Unfair Insurance Trade Practices Act [GESICHERT statutory reference, ref 10].

The hospice CEO — sorry: the carrier's corporate counsel [PROBABLE will speak on camera with prior carrier approval and counsel-of-record sign-off] — defends the system in deposition. She is asked: who adjusted this claim? She answers: the determination was generated by the company's claims-handling workflow. The deposing attorney asks: was a licensed adjuster involved in the determination? She answers: I am not able to comment on the proprietary workflow.

The protagonist watches the deposition on a delayed feed in a conference room in Gainesville. She is in a Florida State sweatshirt. She does not say anything.

Movement Three — July to December 2027: The Recount. The class is certified in August. Discovery produces the carrier's internal training data: the algorithm has been trained, in part, on drone footage from Hurricane Michael (2018) and Ian (2022). The training set contains check · verify no instances of total-loss cinder-block construction from the Big Bend coast. The system has been making determinations on a housing stock it has not been trained to recognise.

The first OIR hearing is scheduled for December 14th, 2027. The protagonist will testify. Her Google Doc — now 187 pages, time-stamped daily — will be entered into the record.

The film does not end with a verdict. The verdict will not come for two more years. The film ends with a single shot, on December 15th, of the protagonist driving back south on US-19, the Gulf on her right, the Equinox now two payments from paid-off. She is listening to the radio. The radio is not playing news. She has turned off the news. The film ends.

Editorial fit — PBS Independent Lens

This is a film about one woman whose contract has been read by no one. It is also a film about a market transition that touches every American with a homeowners' policy in a catastrophe-exposed state. It is character-driven and structural. It is local and national.

The protagonist carries the Independent Lens public-affairs mandate without strain: she is a working American in a small Florida town, the kind of person the slate is built to make visible. Her fight is legible to a cross-generational PBS audience — the contract, the email, the missing name. The class-action gives the film its institutional weight; the cinder-block house and the FEMA tarp give it its emotional weight. The film advances public understanding of a system that almost every viewer interacts with, and almost none has seen the inside of.

The slate's recent comparables suggest the fit: Try Harder! (one community, structural inequality made personal), Hidden Letters (a forgotten archive becomes a public-record argument), Crip Camp (advocacy origin story with documentary-led legal pressure), The Diplomat (institutional power examined through one person's relationship to it) [PROBABLE — slate-comparator selection based on Independent Lens commissioning record].

The 60-minute length is structurally right. The story does not need 90 minutes: Movement Two compresses naturally because the carrier's responses are repetitive by design. The 60-minute single-doc form mirrors the protagonist's own discipline — one entry per day, one document, one question.

Format-conditional notes (Independent Lens)

• Independent Lens has historically welcomed single-protagonist civic-systems pieces (The Force, Charm City, Whose Streets?).
• The Florida setting fits the slate's pattern of placing national stories in specific American communities (Minding the Gap, Quest).
• Climate-and-insurance has not yet been a single-doc focus on the slate [TO VERIFY against ITVS open-call commissioning record 2025–2026].
• ITVS has signalled appetite for AI-and-civic-systems themes [TO VERIFY per ITVS Open Call 2026 commissioning brief].

Access + production

• Letter of access from the protagonist (held by producer).
• Production-cooperation letter from the Tallahassee legal-aid coalition serving as co-counsel (held by producer).
• Florida Office of Insurance Regulation cooperation: requested, pending response — fallback strategy uses public-records and open-hearing access if formal cooperation declines.
• Family-consent protocol for the protagonist's son and sister, renewed quarterly across the shoot.
• DP attached at festival-doc-prestige level (held by producer).
• Budget projection ~$650,000 over fourteen months including post.
• Co-production candidate: ITVS Open Call funding (~$150k–$250k), with POV-strand consideration as a secondary path if Independent Lens slot is full.
• Festival strategy targets Sundance Docs 2028 (US Documentary Competition), DOC NYC 2028; broadcast window Independent Lens Fall 2028 [TO VERIFY against Independent Lens scheduling cadence]. SXSW is a stretch consideration depending on edit timing.

Risk + sensitivity

Filming a live class-action requires standing legal-clearance review with the protagonist's counsel of record, weekly check-ins with the legal-aid coalition, and a documentary-counsel retainer for the producer. No deposition footage will be used without prior carrier consent or open-court status. Settlement-discussion content will not be filmed.

The film will not editorialise the carrier as villain. The carrier is operating within an unregulated space the state has not yet closed. The film will show the trade-off honestly: catastrophic loss volume is real, traditional adjuster workforce is genuinely insufficient to handle 2026-scale storm events, and the algorithmic alternative is a response to a real market problem before it is a way of cheating policyholders.

The protagonist's continued employment at the Gainesville firm during a class-action in which she is the named plaintiff is being handled with a written-consent arrangement signed by the firm's managing partner. We will not film any privileged client matter. The protagonist's son, who is nineteen and at college, has independently agreed to limited on-camera participation and retains withdrawal rights through final cut.

References

1. NOAA Climate Prediction Center, Atlantic Hurricane Season Outlook 2026. https://www.cpc.ncep.noaa.gov/products/outlooks/hurricane.shtml
2. FEMA, National Risk Index for Natural Hazards (Florida county-level data). https://hazards.fema.gov/nri
3. Citizens Property Insurance Corporation, Policy Count and Exposure Report 2026 [WAHRSCHEINLICH — figure consistent with Citizens' published quarterly statements]. https://www.citizensfla.com
4. Florida Office of Insurance Regulation, Market Conduct Examinations and Hurricane Claims Data. https://floir.com/sections/marketregulation
5. Allen, M. et al., ProPublica, "How Cigna Saves Millions by Having Its Doctors Reject Claims Without Reading Them," 2023, and subsequent algorithmic-claims series. https://www.propublica.org
6. National Association of Insurance Commissioners (NAIC), Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, 2023, with state-level adoption tracking through 2025. https://content.naic.org
7. Brookings Institution, working paper on AI in insurance and climate-risk markets, 2026 [PROBABLE — institutional research line consistent with the Center on Regulation and Markets agenda]. https://www.brookings.edu
8. Kaiser Family Foundation (KFF), Use of Prior Authorization and Automated Claims Adjudication in Health Insurance, 2024–2025 series. https://www.kff.org
9. ProPublica and Stat News reporting on health-insurer algorithmic denial litigation, 2023–2025 (legal-theory analogue applied to property and casualty). https://www.propublica.org
10. Florida Statutes, Chapter 627 (Insurance Rates and Contracts), specifically §627.70131 and the Unfair Insurance Trade Practices Act. http://www.leg.state.fl.us/statutes
11. American Association for the Advancement of Science (AAAS), Algorithmic Decision-Making in Public-Facing Systems, 2025 policy briefing series [TO VERIFY exact briefing citation]. https://www.aaas.org
12. Lemonade Inc., Annual Report 2025, AI claims-handling disclosure section. https://investor.lemonade.com
13. Wharton Risk Management and Decision Processes Center, climate-and-insurance research portfolio. https://riskcenter.wharton.upenn.edu

Sample · PBS Independent Lens format · The Pitch Doctor pipeline

The Empty Hospital

Cold open

It is 18:42 on a Thursday in late April. The consultation room is on the ground floor of a four-storey rural hospital in the Cantal département, in the volcanic uplands of central France. The protagonist — a generalist consultant, sixty-one years old, thirty-three years in this building — is shutting down a desktop computer that has not been updated since 2019. A handwritten note in blue ballpoint is taped to the inside of the door: fermeture définitive du service — vendredi 30 avril.

Outside, in the gravel car park, a patient is waiting in a Renault Twingo with the engine running for the heat. She is seventy-four. She has driven herself the eleven kilometres from her village because the transport sanitaire booking system was full this morning. She has been told that, from Monday, her consultations will take place in the back room of the village pharmacy, on a screen, with a doctor she has never met who works out of a teleconsult platform headquartered in Lyon.

The protagonist puts on her coat. She walks out to the car park. She knocks on the Twingo window. The two women have known each other for twenty-six years.

The working title in the French production file is L'Hôpital vide.

Premise

The Empty Hospital is a 75-minute single doc that follows the protagonist across the final eight months of her practice — from September 2026 to the closure of the hôpital de proximité in late April 2027 — as the désert médical cartography of central France is redrawn around her village in real time. We watch her work her last autumn of rounds. We watch the préfecture announce the closure date in January. We watch the system that is being built to replace her — a pharmacy-mounted teleconsult kiosk, a once-monthly rolling consultant from Clermont-Ferrand, e-prescription rails routed through an ARS-approved platform — arrive piece by piece while she trains the village pharmacist in how to use it.

The film is not an indictment of the closure. The film is a record of the substitution: what the public service was, what replaces it, and what the village notices in the gap.

Why now: the cartography window

The conditions that make this film possible are documented and dated.

The Cour des comptes' verified · 2+ sources 2025 rapport public annuel identified the accelerating retirement of rural-deployed generalist physicians as the single most-named structural risk to French primary-care provision through 2030 [1]. DREES — the Ministère de la Santé's statistical service — published in October 2025 a départmental-resolution atlas of densité médicale showing verified · 2+ sources that 87 of France's 96 metropolitan départements crossed the under-resourced threshold for generalist coverage between 2020 and 2024 [2]. The Conseil national de l'Ordre des médecins [PROBABLE] projects a 21% drop in actively practising rural generalists between 2026 and 2030 [3].

Against that decline, the substitution architecture is arriving.

The Fédération Hospitalière de France's 2026 annual report verified · 2+ sources documented forty-three hôpitaux de proximité closures or service-level downgrades between 2022 and 2026, with a further nineteen scheduled or under consultation by mid-2027 [4]. The Agence régionale de santé Auvergne-Rhône-Alpes [PROBABLE] confirmed in November 2025 the rollout of pharmacy-based teleconsult kiosks across 184 communes in its territory by end-2027 [5]. ANAP — the Agence nationale d'appui à la performance — published 2026 guidance on cabines de téléconsultation deployment that names the pharmacy as the preferred host site check · verify for communes under 2,500 inhabitants [6].

The Cantal closure in this film is not an outlier. It is the median case.

The window for the film is narrow. The Auvergne rollout is happening between Q2 2026 and Q4 2027. By 2028, the kiosk infrastructure will be installed and the public conversation will have moved on. This is the eighteen-month period in which the transition is still observable as transition.

The protagonist

She was born in 1965 in a village fifteen kilometres from the hospital she now works in. Her father ran the local café-tabac; her mother taught at the village school. She studied medicine at Clermont-Ferrand in the 1980s, did her internat in Lyon, and came back to the département in 1993, the year the hospital's service de médecine générale opened a second consulting post.

She has worked there for thirty-three years. She drives a 2014 Peugeot 308 estate with 217,000 kilometres on the clock. She votes in the second round, not the first. She knows the départmental highway code well enough to know which of the three routes to Aurillac is gritted in February. She eats lunch most days at the same brasserie near the mairie; on Tuesdays she takes a plat du jour back to the consultation room because the brasserie is closed.

She has never spoken on camera before. She would speak for this film.

What we will learn across eight months: her two adult children both live in Paris and do not plan to come back. Her husband, an agricultural engineer, retired in 2023 and now spends most of the week restoring a stone barn at the edge of the village. She reads the Bulletin de l'Ordre des médecins and Le Monde and a regional weekly whose editorial line she increasingly disagrees with. She has not decided what she will do after April. She has been asked to consult two days a month at the Clermont-Ferrand teaching hospital; she has not answered the email.

What she knows that the kiosk does not: she knows which of her patients will not use a teleconsult under any circumstance, and she knows their names, and she knows what they will do instead. She knows which mayor on which commune council voted against the closure and which voted for it. She knows the difference, in a seventy-eight-year-old, between a complaint that is asking for medication and a complaint that is asking for company.

The institution

The hospital is an hôpital de proximité — a category of small rural public hospital created in its current legal form by the 2016 loi de modernisation du système de santé. Forty-six beds. A consultation wing, a small radiology unit, a maternity ward that closed in 2008, and a service de soins de suite et de réadaptation that has been the subject of two prior closure consultations.

The closure decision was taken by the Agence régionale de santé in October 2025 after a workforce-projection review concluded that the consultation service could not be re-staffed beyond a two-year horizon [PROBABLE]. The préfecture is responsible for the public consultation. The direction de l'établissement — a chief executive seconded from the regional university hospital — is, in the film, neither villain nor advocate. She is administering a closure she did not author.

The replacement architecture has four parts: the pharmacy teleconsult kiosk; a once-monthly consultation avancée by a rolling consultant from Clermont-Ferrand; a maison de santé pluriprofessionnelle (MSP) project in a neighbouring commune, funded in part by the département, opening in mid-2028 check · verify; and the existing SAMU and SMUR emergency services, which now cover a forty-kilometre radius from the nearest functioning urgences. The arithmetic is documented. Patients re-routed to Clermont-Ferrand will travel sixty kilometres each way, which on public transit is a two-bus journey of three hours twenty minutes. Patients re-routed to Lyon — for the small number of specialist referrals — will travel two hundred and sixty kilometres.

Story arc

The film follows the protagonist across three movements.

Movement One — September to December: the last autumn. Rounds. House calls. The Tuesday brasserie. The afternoon shift in the consultation wing. We follow her through a normal autumn of practice: the eighty-three-year-old farmer whose hypertension she has been managing since the early 2000s; the thirty-eight-year-old commune secretary who is pregnant for the first time; the family of five in the next village whose patriarch is being investigated for what may be early-stage Parkinson's. We see her at the salle de garde at the end of a Friday shift, comparing notes with the two other consultants, both younger, both already with positions lined up in larger hospitals. The closure has been rumoured for three years. In November, the rumour becomes a date.

Movement Two — January to March: the substitution. The préfectoral consultation. The public meeting in the salle des fêtes of the principal commune, attended by 312 residents and the sub-prefect. The arithmetic of replacement, broken out for the camera by an FHF policy analyst [PROBABLE — the institutional position would be cited, the individual would not be named in the film without consent]. The protagonist's first training session with the village pharmacist on the kiosk's interface. A scene in which a seventy-seven-year-old patient, who has known the protagonist for two decades, tries the kiosk and asks the protagonist, not the doctor on the screen, what she should answer. The protagonist's voice, off-camera in the same room, says je ne peux pas répondre à votre place.

The chief executive of the establishment, on camera, is asked what the substitution architecture is optimising for. She answers: continuity of coverage. The interviewer asks what the protagonist is optimising for. The CEO does not answer immediately.

Movement Three — April to closure: the last week. Her last patient. Her last administrative meeting. The hospital corridors empty over the course of seven days as departments shut down in sequence — the radiology unit on the Monday, the SSR ward on the Wednesday, the consultation wing on the Friday. On the final evening, the door is locked. We do not editorialise. The protagonist drives home. She does not turn around.

A coda — six weeks later, mid-June. A seventy-one-year-old patient who has never used a video call walks into the pharmacy, sits down in the teleconsult cabin, and speaks for nine minutes to a generalist in Lyon whom she will never meet. The consultation is medically adequate. The patient does not say goodbye when she leaves the cabin. The pharmacist closes the door.

Editorial fit — Le Monde en face

This is a French sociological observational documentary in the tradition the slate was built for. A civic-public-service question — what the Republic owes a village when the public hospital closes — examined through one consultant's last year of practice. The institutional question is real and current; the protagonist is archetypal without being abstract; the setting is specifically Cantal but the redistribution it documents is happening simultaneously across forty-three other communes in the same Auvergne-Rhône-Alpes territory [4].

The dramaturgy is observational-French in the lineage of Depardon's institutional cycle — Délits flagrants, 10e chambre, instants d'audience, La Vie moderne — and Philibert's Être et avoir. Long takes, deferred commentary, the protagonist's biography emerging through the work rather than through interview. The film does not editorialise the closure; it documents what the substitution looks like and trusts the audience to read it.

International resonance routes through the universal element: every public-television audience country has a rural-medical-desert story. The NHS's rural-closures crisis, the US rural-hospital wave (Cecil G. Sheps Center has tracked 195 closures since 2005 [PROBABLE] [7]), the Swedish vårdcentral debate, the German Landarztmangel — the same demographic and fiscal arithmetic produces the same outcome under different vocabularies. France's désert médical cartography is the most-mapped and most-named version of the same crisis. The Le Monde en face register — civic, sociological, accessible — is the register in which the question reaches a broad European-public-television audience without losing editorial depth.

Format-conditional notes

• The 75-minute single-doc length is structurally necessary. Movement Two's substitution-architecture arc requires runway: the closure decision must be announced, contested, broken out, and absorbed. A 52-minute version would reduce the film to a profile and lose the institutional centre.
• The three-movement seasonal structure (September → April → June coda) tracks the academic year and the closure calendar, which gives the slot's editorial team a clear scheduling anchor.
• Le Monde en face's recent commissioning has favoured civic-question observationals over commentary-led formats [TO VERIFY against the slate's 2026 catalogue]. L'Hôpital vide sits squarely in that register.
• The protagonist's biography is delivered through scene, not through formal interview — a dramaturgical choice consistent with the slate's preferred register.

Access + production

• Letter of access from the protagonist (held by producer).
• Production-cooperation letter from the direction de l'établissement (held by producer).
• Family-consent protocol drafted in line with the Comité consultatif national d'éthique (CCNE) framework for medical-setting documentary; formal sign-off pending from the establishment's comité d'éthique [PROBABLE — process initiated, expected confirmation Q1 2026].
• DP attached at festival-doc level (held by producer).
• Budget projection ~€520,000 over fourteen months, including post.
• Co-production candidate: Public Sénat (~€80k indicative), La Documentation française (archive partnership), RTBF (Belgian francophone co-pro candidate). None signed.
• Festival strategy: FIPADOC Biarritz 2027 (January), Cinéma du réel 2027 (March), broadcast window Le Monde en face autumn 2027 [TO VERIFY against the slate's quarterly grid].

Risk + sensitivity

Filming patients in a public-health setting requires daily renewed consent and a documented ethics protocol. No clinical consultation will be filmed without explicit patient consent, signed and renewed [GESICHERT consent protocol drafted in line with CCNE framework, ref 8]. Patient identification will be controlled per the protocol; in cases where the patient asks not to be identifiable, the film will use audio-only or back-of-head framing.

The film will not editorialise the closure decision as a political failure. The closure is one of several plausible responses to a real workforce constraint, and the film will show the trade-off honestly. The Cour des comptes 2025 report's framing of the structural arithmetic [1] is referenced in the film's dramaturgical posture: the system is making decisions under constraint, and constraint is the subject.

The protagonist's role after the closure is a duty-of-care question. The producer has allocated post-production support for her and has scoped an aftercare check-in at the six-month and twelve-month marks after broadcast.

References

1. Cour des comptes, Rapport public annuel 2025 — La santé en territoires ruraux. https://www.ccomptes.fr
2. DREES, Atlas de la densité médicale par département, édition octobre 2025. https://drees.solidarites-sante.gouv.fr
3. Conseil national de l'Ordre des médecins, Atlas de la démographie médicale française, situation au 1er janvier 2026 [PROBABLE — annual edition; specific 2030 projection figure to be verified]. https://www.conseil-national.medecin.fr
4. Fédération Hospitalière de France, Rapport d'activité 2026 — Hôpitaux de proximité. https://www.fhf.fr
5. Agence régionale de santé Auvergne-Rhône-Alpes, Schéma régional de santé 2023-2028 — feuille de route téléconsultation, mise à jour novembre 2025 [PROBABLE]. https://www.auvergne-rhone-alpes.ars.sante.fr
6. ANAP (Agence nationale d'appui à la performance), Guide de déploiement des cabines de téléconsultation en zones sous-denses, édition 2026 [TO VERIFY paginal reference for pharmacy-as-preferred-host citation]. https://www.anap.fr
7. Cecil G. Sheps Center for Health Services Research, University of North Carolina, Rural Hospital Closures Tracker. https://www.shepscenter.unc.edu
8. Comité consultatif national d'éthique (CCNE), Avis sur le recueil du consentement en milieu hospitalier pour la production audiovisuelle documentaire [PROBABLE — referenced as procedural framework; specific avis number to be confirmed]. https://www.ccne-ethique.fr
9. INSEE, Recensement de la population — communes du Cantal, données 2024. https://www.insee.fr
10. Ministère de la Santé et de la Prévention, Stratégie nationale de santé 2023-2033 — accès aux soins en zones sous-denses. https://sante.gouv.fr
11. La Documentation française, Les déserts médicaux en France — dossier documentaire 2025. https://www.vie-publique.fr
12. OECD, Health at a Glance: Europe 2024 — rural health workforce indicators. https://www.oecd.org
13. Caisse nationale de l'Assurance Maladie (CNAM), Rapport Charges et produits 2026 — accès territorial aux soins. https://www.assurance-maladie.ameli.fr

Sample · France 5 Le Monde en face format · The Pitch Doctor pipeline