Thinking · Method

The Continuity Cohort Method. A named cohort, fixed at the trigger, closed at day twenty-one.

Abhijith Magal · 4 October 2026 · 8 min read

This page is the method's canonical home on the web. A published version of the full method note can be found on Zenodo at DOI 10.5281/zenodo.22861246, and every claim below is sourced there in full.

The method rests on a distinction that sounds small and is not. Evacuation asks where people are. Continuity asks who they were in the middle of being treated for. Only one of those questions gets asked before landfall.

The gap the method closes

Disaster response in India has become very good at the thing it set out to do. Warnings reach districts. People are moved before impact at a scale few systems anywhere match. In Odisha, where the 1999 super cyclone killed 9,885 people by the official toll, deaths from later cyclones of comparable force have been held to a small fraction of that figure, a change the UN Office for Disaster Risk Reduction attributes to two decades of investment in preparedness. None of this argues with any of that.

It argues with what happens next. Hertelendy and colleagues, reviewing healthcare system resilience to tropical cyclones in 2025 in The Lancet Regional Health – Americas, set out seven challenges that decide whether a health system holds. The one they put first is continuity of care for vulnerable populations, and they observe that while immediate cyclone deaths are falling, delayed illness and death from disrupted access to care are rising. Read against India, that first challenge is the one with no owner, no measure and no line item.

A person mid-way through tuberculosis treatment, a woman at thirty-six weeks: the disaster does not cancel their entitlement to free medicines and care. The entitlement never breaks. The route to it does, and nobody counts the people who fell off it.

Every input already exists

The method adds one component and no institutions. Each of the four inputs below is built, national, and running under a named government owner today.

The hazard signal comes from the India Meteorological Department's four-stage cyclone warning. The second stage, the cyclone alert, goes out at least 48 hours ahead and names the coastal districts likely to be affected, addressed to the Chief Secretary of the state and the Collectors of those districts by name. The advisory channel is SACHET, the Common Alerting Protocol system run by the National Disaster Management Authority: operational in all 36 States and Union Territories, with more than 134 billion SMS alerts disseminated in more than 19 Indian languages as reported by the Ministry of Communications in April 2026. The state can already draw a boundary around a warned area and reach every phone inside it. It cannot yet reach a named patient inside it, because the list of names sits in a different system.

The fourth input is the entitlement itself. Ayushman Arogya Mandirs are mandated to deliver NCD and TB care with free essential drugs: the Government of India supports 106 medicines and 14 tests at sub-health centres, 172 medicines and 63 tests at primary health centres. The warning reaches the Collector 48 hours out. The alert system can reach every phone in the warned area. The registers hold, by name, who is on insulin, who is in the intensive phase of TB treatment, and who is due to deliver this month. The medicines are free. Every piece exists. They sit in four departments, and nothing joins them inside the window where joining would change anything.

The gap is not the data. It is the join.

What Cyclone Dana showed

Part of the join has already been made once. Ahead of Cyclone Dana's landfall on the Odisha coast in October 2024, the state moved pregnant women from the projected track into health facilities and maternity homes before impact. The operation worked and deserves to be named as the precedent it is: a register question, who is in late pregnancy inside this track, was asked before landfall and acted on at scale.

Then look at the counting.

Every figure counts people moved or deliveries conducted. They differ between speakers, between days, and within one report, and none of them can say how many of the women on the pre-landfall list were receiving antenatal or postnatal care against their own record three weeks later. A denominator fixed once at the trigger and never revised is what makes that question answerable. That is not a criticism of the response. It is the specification for the measurement the response deserves.

The method, in three moves

The interface does one job in three moves. It fixes a named cohort the moment the trigger fires: every individual held in the NCD, Ni-kshay or RCH register with active treatment or care status and a registered address inside the warned area, resolved to one identity per person, frozen and never revised. It carries that list into the response, so the people on it are found rather than waited for. And it reconciles at day twenty-one: a person counts as continuous if there is documented clinical contact, a recorded dose or refill against their original identifier, or medicine still on hand from the last dispensing. The result is the day-21 continuity rate, read against the same blocks at undisturbed index dates, with the baseline printed alongside and the finding being the gap between the two.

Three rules keep the number honest. The freeze rule: a cohort assembled afterwards silently drops the people who were worst affected. The supply rule: twenty-one days does not clear a monthly dispensing cycle, so a person who collected thirty days of medicine the day before landfall is fully covered and must not be counted as lost, or the measure punishes the patients who were best prepared. And the write-back rule: a resumed treatment entered as a new record makes a recovery read as growth, so every resumed contact is written against the person's original identifier and no new record is created for anyone already in the cohort.

Relief reporting cannot produce this measurement. It counts services delivered to whoever presented, and presenting is precisely what failed. Everyone who did not come is invisible in it.

Five calibration parameters give the method its meaning: the trigger stage and the authority to fire it, the identity-matching rule across three registers, the medicine-on-hand supply rule, the day twenty-one close-out including how a death, a transfer and a refusal are each recorded, and the baseline window that makes two events comparable. They are named in the method note and deliberately not valued there, because a figure computed under different parameters is a different figure, comparable with nothing. The study protocol fixes them in advance of any data, and the operating specification a district would run from is maintained by Syntropy Earth as Continuity Cohort Method v1.0, released under licence.

Who holds what

Nobody gives up control of anything they hold today. The government system owns the data and the consent; the registers are its, and the continuity number is its to publish. The district authority owns the trigger: it fires the early action decision, runs the response with the continuity list travelling alongside relief, and signs off the count. The institution that funds or runs preparedness owns the reach: pre-positioning against the named list, finding people instead of waiting for them, closing every case with a recorded reason, including the hard ones, and carrying the capability into the next event. The interface owns the calibration only. It hands over one comparable number, then steps back.

A capability, not a project

A funded project ends when the funding ends, and the learning leaves with the team. This is built the other way round. The registers are national, so once the interface is calibrated for one district, the marginal cost of the next is close to nothing. It adds no field footprint, competes with no implementing partner, and survives staff turnover because it is a protocol rather than a team.

The output is a public good. A continuity rate read against a district's own baseline is comparable between events and between districts, which makes it publishable rather than proprietary to whoever paid for it. An institution that runs it twice holds evidence for a preparedness indicator that state authorities and the national system could adopt as standard. That is the difference between funding a programme and establishing a measure, and it is where this work is headed.

Where it is published

The method is published as two Zenodo records, each under a CC BY-NC-ND 4.0 licence. The method note defines the method and sources every input: The Continuity Cohort Method: holding treatment continuity through a disaster response, DOI 10.5281/zenodo.22861246, first circulated August 2026. The study protocol sets out a pre-randomised, event-activated cluster trial of the method in coastal districts of India, with a retrospective feasibility phase on Cyclone Dana: DOI 10.5281/zenodo.22880413. The protocol states its own status plainly: not yet registered in the Clinical Trials Registry, India, not yet ethically approved, no participants enrolled and no data collected, and no study activity begins before those two are in place. One question the method note names as open, and the protocol treats with its statutory references, is whether and under what consent a disaster process can reach a health register at all. That is a question of law and departmental authority, stated for counsel and the departments concerned rather than guessed at from outside.

The method sits inside a wider body of work on the continuity dimension of climate-access: the evidence that displacement breaks chronic care, and a register design that counts reconnection, in when the emergency ends and the access gap does not, and the argument the practice is built on, in climate change is a healthcare story.

Common questions

What is the Continuity Cohort Method?

A published method for holding treatment continuity through a disaster response. It fixes a named cohort of people on continuous treatment from government registers at the moment the early action trigger fires, carries that list through the response so people are found rather than waited for, and reconciles documented clinical contact at day twenty-one against the same area's undisturbed baseline.

What does it measure?

One number: the day-21 continuity rate, the share of the frozen pre-landfall cohort with documented contact, a recorded dose or refill, or medicine still on hand by day twenty-one, read against the same area's own undisturbed baseline. The finding is the gap between the two readings, with the baseline printed alongside.

Why fix the cohort before landfall rather than after?

A cohort assembled after the event silently drops the people who were worst affected, because they are the ones who never presented. Fixing the denominator at the trigger and never revising it is what makes the day-21 reading honest, and what makes two events comparable.

Does it need new data or a new institution?

No. In India every input already exists under a named government owner: the IMD warning addressed to Collectors, the SACHET alert system, the NCD, Ni-kshay and RCH registers, and free essential medicines at Ayushman Arogya Mandirs. The method adds one component, the interface that joins them inside the warning window, and no institutions.

Who owns the continuity number?

The government system holds the registers, and the number is theirs to publish. The district authority signs off the count. The institution carries the capability into the next event and the next district. The interface owns the calibration only, then steps back.

Where is it published?

Two records on Zenodo, each under CC BY-NC-ND 4.0: the method note (DOI 10.5281/zenodo.22861246) and the study protocol (DOI 10.5281/zenodo.22880413). This page is the method's canonical home on the web.

Abhijith Magal

Founder, Syntropy Earth. Nine years across two global pharmaceutical multinationals in patient access and commercial roles, with health-equity work alongside the WHO-Foundation and UNICEF. He works on climate-access: where climate disruption breaks the link between patients and care.

No modelled, estimated or simulated figures appear on this page. Figures attributed to officials through press reporting are labelled as such. No government body has endorsed the method described here, and nothing on this page is legal advice.

Last updated: 4 October 2026