The Hubli-Dharwad district health office sits on the second floor of a building that smells of phenyl and old paper. I was there on a Tuesday in March, waiting to speak with the district epidemiologist about a cluster of dengue cases near the railway tracks. Dr. Suresh Hiremath—a man in his late forties with a practiced, unhurried manner—was not at his desk. He was sitting on a plastic chair in the corridor, a clipboard balanced on his knee, writing.

The Integrated Disease Surveillance Programme portal had been down since the previous evening. The state server in Bengaluru had a maintenance window that overlapped with the reporting deadline, and nobody had thought to adjust the schedule. So Dr. Hiremath was drafting his outbreak investigation report by hand, on a form printed from a PDF designed to be filled in online. The form’s fields were too small for the information he needed to record. He was writing in the margins.

I sat with him for forty minutes. In that time, I watched him make a series of quiet decisions about what to include and what to leave out. The settlement where the cases had appeared did not have a formal name. It appeared in no municipal register. He wrote unauthorized settlement near Hubli Junction because the form required a location descriptor that matched the revenue map. A child of nine had been hospitalized with severe dengue. The child’s mother had told the ASHA worker that the house had no window screens and that the open drain behind the house had not been cleared in months. Dr. Hiremath wrote probable vector breeding site identified and moved on. He did not write about the drain. Not the housing conditions. Not the mother’s testimony. The form did not have a field for that.

What I was watching was not a failure of clinical knowledge or epidemiological method. It was a writing problem. And by writing problem, I do not mean style or grammar. I mean the capacity to structure complex, multi-source evidence into a document that a health system will read, understand, and act on. This is intellectual labor that is rarely acknowledged, never trained, and almost never resourced. Its absence has consequences for who gets counted, who gets protected, and who gets left out of the story health systems tell themselves about what is happening.

The Ledger and the Form

In my work with informal settlement communities in Bengaluru—particularly in settlements off Sarjapur Road and in the Lakkasandra area near the lake—I have spent time with ASHA workers and community health volunteers who maintain handwritten ledgers. These ledgers are not official documents. They are working notebooks, often school exercise books, in which the workers record what they see during their rounds: who has a fever, who is coughing, whose child has not been eating, whose house flooded in the last rain, whose landlord has threatened eviction.

Sannamma, an ASHA worker in a settlement near Electronic City whom I have known since 2023, showed me her ledger one afternoon. The entries were in Kannada, written in blue ink, with dates and household numbers. Some entries had small drawings—a sketch of a blocked drain, an arrow pointing to where a water tap had stopped working. “I write what I see, not what the form asks,” she told me. “The form asks for symptoms. I see symptoms. But I also see the water. I see the garbage. I see the children playing near the construction site where the mud collects. The form does not want to know about the mud.”

The form does not want to know about the mud. That sentence has stayed with me for two years. It captures something precise about the gap between lived experience and official documentation. Sannamma is not making a philosophical argument. She is describing a practical constraint: the reporting template she is given has fields for symptoms, dates of onset, hospital visits, and household composition. It does not have a field for environmental conditions, for housing tenure status, for the proximity of construction activity, or for the drainage infrastructure that the municipal corporation is responsible for maintaining. When she converts her ledger into the weekly reporting format, the mud disappears.

This translation—from ledger to form—is the first layer of narrative loss. It is also the layer where the most skilled writing labor happens, because the worker has to decide what to foreground and what to omit, how to render a complex environmental exposure in a vocabulary that the surveillance system recognizes, and how to flag something urgent without triggering a response the community may not want—a demolition notice, a fogging operation that disrupts daily life. This is not transcription. It is interpretation. It is writing.

The Surveillance Vocabulary

The second layer is the vocabulary of public health surveillance itself. The IDSP and similar systems use a specific register: cases, suspected cases, lab-confirmed cases, outbreak thresholds, attack rates. This vocabulary is designed for aggregation. It lets a district health officer compare numbers across weeks and across areas. But it is not designed for context. When a community health worker in a Pune settlement records that three children in adjacent houses have developed skin rashes after a local waste sorting yard changed its operating hours, the surveillance system can capture the three cases. It cannot capture the causal hypothesis the community is proposing: that the extended hours mean more plastic burning in the evening, when children are playing outside.

I worked with a waste picker collective in Pune’s Hadapsar area in 2024 on a community-based health assessment. The collective—mostly women—had been tracking respiratory complaints among their members for over a year, using a simple register maintained at the sorting center. The register was not a health document in any formal sense. It was a record of who had coughed, who had missed work, who had gone to the local dispensary, and who had been told by the doctor to take steam and rest. When we analyzed the register alongside air quality data from a low-cost monitor the collective had installed with support from a local NGO, the pattern was clear: respiratory complaints spiked on days when open burning was visible from the sorting yard.

But when we tried to present this finding to the municipal health department, we were told the register was not a valid data source. The cases had not been confirmed. The exposure had not been measured according to protocol. The collective’s documentation—months of careful, consistent, community-based observation—could not be translated into the language of the surveillance system without losing the very thing that made it valuable: the connection between the burning, the breathing, and the bodies.

The translation work required to bridge this gap is enormous, and it falls on people who are already overburdened. A community health worker who wants the municipal health department to act on open burning near a waste sorting yard needs to write a report that does three things at once: present the health complaints in clinical language the system recognizes; document the environmental exposure in terms that meet regulatory standards, which may require data the community does not have; and tell the story of the community’s experience in a way that is compelling without being dismissed as anecdotal. This is a writing task of considerable sophistication. It is also a task for which almost no training, support, or time is provided.

The Narrative Arc of an Outbreak Report

The third layer is the one that matters most for health equity: the narrative arc of the final report. When Dr. Hiremath in Hubli-Dharwad writes his outbreak investigation report, he is not just recording facts. He is constructing a narrative that will determine how the health system responds. If the report frames the dengue cases as isolated incidents—individual patients who happen to live near the railway tracks—the response will be clinical: test, treat, follow up. If the report frames them as evidence of a systemic environmental failure—stagnant drainage, absent vector control, inadequate housing—the response should be structural: clear the drains, repair the infrastructure, address the conditions that make the settlement vulnerable.

The difference between these two framings is not a matter of clinical evidence. It is a matter of how the evidence is narrated.

This is where the parallel with incident documentation in other fields becomes instructive. Google’s Site Reliability Engineering practices, as documented in their SRE Book, include a chapter on Postmortem Culture: Learning from Failure that makes an argument directly relevant to health reporting: the way an incident is written up determines whether systemic causes surface or get buried as isolated events. The SRE approach to blameless postmortems is designed specifically to shift attention from individual error to systemic conditions. An outbreak report that followed the same logic—asking not who failed to use a mosquito net but what drainage infrastructure failed and why—would produce a fundamentally different public health response. The SRE Book also addresses what it calls toil: repetitive, manual work that adds no enduring value and crowds out higher-impact effort. The parallel to frontline health documentation is exact. When a community health worker spends three hours filling in a form by hand because the digital system is down, or when a district epidemiologist writes the same case information into three different formats for three different departments, that is toil. Labor that produces no new knowledge and actively prevents the kind of careful documentation that would reveal structural causes.

Writing as Health Infrastructure

Here is where I want to make a claim that may sound strange coming from a public health researcher: writing is health infrastructure. Not metaphorically. Literally. The reports, the ledgers, the case histories, the community assessments—these are the documents that determine whether a health system can see what is happening in the communities it is supposed to serve. When those documents are incomplete, when the writing labor that produces them is undervalued and under-resourced, the health system becomes blind to the very people it most needs to see.

The Authors Guild, in their AI Best Practices for Authors, makes an argument that resonates here more than I expected: they write about preserving human voices and the thinking that goes into writing, and warn against quality writing becoming a rare luxury good representing only a minority of views. The parallel to health documentation is uncomfortable but precise. When the capacity to write a complete, context-rich, structurally informed health report is available only to well-resourced research institutions and private consultants, the communities with the least power are also the communities whose health experiences are least likely to be documented in a way that prompts action. The writing gap becomes the equity gap.

This is not an argument against digital reporting systems or standardized templates. Standardization has value: it allows comparison, aggregation, and trend analysis. But standardization without investment in the writing capacity of the people who feed the system produces a specific kind of failure: reports that are technically complete and substantively hollow. They count the cases. They miss the causes.

What Would It Take?

If we took the writing labor of frontline health workers seriously, what would change?

First, training. Community health workers are trained in clinical skills—recognizing symptoms, conducting tests, providing first-line care. They are almost never trained in documentation as a skill: how to structure a narrative, how to describe environmental conditions in language a surveillance system can process, how to flag a pattern without overstating the evidence. This training does not require a medical school. It requires a workshop, a mentor, and time—resources that are rarely allocated.

Second, tools. The forms that frontline workers fill in are designed for data extraction, not for narrative construction. A form that included a single open field—describe the conditions you observed in this household and neighborhood this week—would not replace the structured fields, but it would create a space for the information that currently falls through the cracks. The mud. The drain. The landlord’s threat. The child who plays near the construction site. For those managing complex documentation across multiple sources—field notes, community testimony, environmental observations—having a structured drafting tool can reduce the cognitive load of assembling a coherent narrative from fragments. Something as straightforward as an AI writing app that helps structure multi-source drafts could reduce the time it takes to turn a week of handwritten observations into a report a district health officer will actually read, without replacing the judgment and local knowledge only the worker possesses. The point is not to replace the writer. It is to reduce the toil that prevents the writing from happening at all.

Third, recognition. The writing that community health workers do is invisible labor. It does not appear in their job descriptions. It is not counted in their workload. It is not acknowledged in their pay. When Sannamma spends an extra hour at the end of her rounds converting her ledger into the weekly reporting format, she is doing unpaid writing labor that the health system depends on. Recognizing this labor—naming it, counting it, compensating it—would be a structural intervention in health equity, because it would begin to close the gap between what communities experience and what the system can see.

The Report That Got Written

I went back to the Hubli-Dharwad health office a week later. Dr. Hiremath had submitted his report. The IDSP portal was back online, and he had transcribed his handwritten notes into the digital form. I asked him what had happened to the information about the drain, the housing conditions, the mother’s testimony. He looked at me with the expression of a man who has been doing this work for twenty years and has stopped expecting the question to lead anywhere.

“It’s not in the report,” he said. “There was no field for it.”

Two months later, the settlement near the railway tracks had another dengue outbreak. The same drain had not been cleared. The same houses had no screens. The same children were playing in the same mud. The report on the second outbreak was, as far as I could tell, identical to the first. Same cases. Same location descriptor. Same missing context.

The report that never got written is not a metaphor. It is a document that exists in the gap between what a community experiences and what a health system can see. That gap is partly clinical, partly a data problem, and partly political. But it is also a writing problem. And writing problems, unlike many health problems, are solvable with modest investment, targeted training, and a willingness to take seriously the intellectual labor of people who are currently treated as data collectors rather than as writers.

If you work in public health, in urban planning, in community organizing, or in policy—the next time you read a health report, read it as a piece of writing. Ask what it includes, what it omits, and whose voice is absent. Then ask yourself: what would it take to write a report that tells the whole story? That question is not about style. It is about who gets seen, who gets protected, and who gets left out of the narrative that determines whether a drain gets cleared, a house gets screened, and a child stops getting sick.