Writing Healthcare IT From the Edge of the Missing Machine

Writing about Healthcare Information Technology [healthcare IT, the machinery by which clinical work is recorded, exchanged, billed, measured, audited, and occasionally understood] is a peculiar punishment because the subject is both everywhere and invisible, like plumbing in a mansion or corruption in a municipal tender. Nobody thinks about it until something leaks, smells, floods, or denies admission to an old man with a laminated folder full of reports from six different labs, each of them printed in a different font, each of them pretending to be medicine.

The trouble with a technical blog is not that the subject is too hard. Hard things can be explained. A bridge can be explained. A kidney can be explained. A database index can be dragged down from Mount Abstraction and made to sit on a plastic chair with tea. The trouble is that healthcare IT is not one thing. It is a compost heap of workflows, incentives, standards, procurement failures, billing habits, medical hierarchies, software vendors, clerks, doctors, patients, scanned paper, WhatsApp screenshots, expired passwords, and that ancient Indian institution, the man who knows the system because he has suffered under it long enough to become its unofficial priest.

From a bird’s-eye view, the subject looks almost respectable. We say Electronic Health Record [EHR, the clinical system used to document patient care], Health Information Exchange [HIE, the controlled sharing of clinical information across organizations], Fast Healthcare Interoperability Resources [FHIR, a modern standard for representing and exchanging healthcare data], analytics, patient engagement, population health, claims integration, artificial intelligence. The nouns arrive polished and hopeful, like visiting dignitaries in white shirts. But the moment you zoom in, the flooring changes. You find the crack under the linoleum. You find the nurse retyping a lab result because the interface dropped the observation code. You find the discharge summary that contains three diagnoses in prose and one coded diagnosis for billing, none of them exactly the same. You find “data quality” blamed for what is actually a failure of representation, governance, workflow design, and institutional honesty.

That is the fissure in writing about this subject. Stay high and you become useless, another fellow releasing helium balloons into a conference hall. Go low and you lose the passerby, who did not open a blog to inspect the semantic equivalent of infected boils under the undergarments of a collapsing system. Yet the boil is the system. The pus is architectural. The smell has a schema.

In India, healthcare IT is often less a functioning category than a placeholder for things that should exist but do not, things that exist but do not connect, and things that connect but do not mean what they claim to mean. Registration systems exist. Billing systems exist. Pharmacy systems exist. Laboratory systems exist. Radiology systems exist. Insurance portals exist. Government dashboards exist. Private hospital applications exist. But existence is not architecture. A bucket is not a water supply merely because it contains water for a while.

The public story says digitization. The operational story says clerks, printouts, duplicated identifiers, partial capture, fragile integrations, and a near-comic dependence on human memory. A patient becomes many patients because one system knows him by mobile number, another by hospital number, another by insurer identifier, another by spelling variation, another by the relative who paid the bill, and another by a smudged photocopy of an address proof. Then we wonder why analytics misbehave, as if the poor warehouse had been given clean wheat instead of municipal rubble.

The private sector is better in places, sometimes much better, but better is not the same as coherent. Private care may have shinier software, tighter revenue-cycle discipline, cleaner front desks, more aggressive billing integration, and faster diagnostic turnaround. It may also have islands of excellence floating in a sea of proprietary fragmentation. The chart may be digital inside the hospital and meaningless outside it. The lab result may be structured for internal display but useless for longitudinal analysis. The diagnosis may be captured for reimbursement, not clinical truth. The patient may still carry the burden of interoperability in a plastic file, like a medieval courier crossing kingdoms with a sealed message and a limp.

This is where the distinction between data transport and semantic meaning matters. Transport asks whether a message arrived. Meaning asks whether the receiving system understood the event in the same way the sending system intended it. Health Level Seven version 2 [HL7 v2, an older but still widely used healthcare messaging standard] can move a result from a laboratory system to an EHR. FHIR can represent a medication, condition, encounter, or observation with more modern elegance. But no transport standard can rescue an organization that has never agreed what counts as an active problem, whether “diabetes” means a clinical diagnosis, a billing artifact, a historical mention, a risk marker, or a family whisper copied forward by a junior doctor at midnight.

That is the non-obvious architectural insight: healthcare data is not merely produced by care; it is produced by the local bargain between care, payment, fear, habit, liability, staffing, and software. The data is organizational behavior fossilized into fields. A hospital’s database is not just a technical object. It is a diary written by a bureaucracy with a stethoscope.

This is why representation failures get mislabeled as data quality failures. “The data is bad” is the phrase people use when they have run out of patience, budget, or political courage. But often the data is faithfully recording a bad abstraction. If an EHR forces a clinician to choose from a diagnosis list that does not match the clinical situation, the resulting code is not dirty in the ordinary sense. It is cleanly wrong. If a discharge summary allows rich narrative but the analytics pipeline only extracts billing codes, the loss is not dirt. It is amputation. If a laboratory result travels without provenance, units, reference ranges, device context, or specimen timing, the problem is not that someone forgot to polish the data. The problem is that the system shipped a corpse and called it a patient.

Technical blogging is difficult because the honest version of this argument sounds gloomy to the uninitiated and obvious to the wounded. Healthcare IT people who have lived inside production systems know this already in their bones. Serious beginners need the scaffolding. General readers want the point before the soup cools. And the writer, poor fool, must somehow explain interface engines, coding systems, workflow-coupled data generation, identity matching, temporal ambiguity, source-of-truth conflict, and semantic drift without sounding like a man alphabetizing screws in a sinking ship.

Then comes the Indian complication. In India, the conversation is poisoned by a patriotic reflex that mistakes comparison for betrayal. Mention China and half the room reaches for the nearest flagpole. This is childish, but it is not harmless. A country that cannot compare cannot improve. A country that treats external evidence as insult becomes a committee of wounded uncles, each one convinced that pointing at a broken drain is an attack on civilization.

India and China were once close enough in rural texture that the comparison was not ridiculous. Villages, poverty, agrarian rhythms, vast populations, hungry states, postcolonial ambition, uneven literacy, uneven health. Then the paths diverged. China built with a ferocity that can be admired without becoming naïve about its costs. India argued, improvised, privatized fragments of competence, celebrated exceptions, tolerated civic decay, and learned the mystical art of calling delay democracy when delay was often just delay wearing a constitutional shawl.

The Indian chest-thumper hears this and says, “Go to China.” This is the great intellectual achievement of the insecure patriot: to convert diagnosis into deportation. But I did not come from China. I came from the north side of Calcutta, which, the last time I checked, had not been annexed by Beijing, though parts of it might benefit from a working footpath, less ceremonial garbage, and the occasional suspicion that public systems are meant to function.

The comparison matters because healthcare IT does not grow in a vacuum. It grows in the soil of civic capacity. If a city cannot maintain drains, traffic discipline, land records, air quality, building codes, municipal accountability, and basic administrative coherence, then its healthcare systems inherit that disorder. They may conceal it behind software, but software is a poor curtain. The database eventually smells of the street outside.

A nation’s healthcare IT maturity depends on boring virtues: identity, standards, governance, procurement competence, public trust, operational discipline, clean institutional ownership, and the ability to maintain systems after the ribbon-cutting photograph has yellowed. These are not glamorous. They do not produce slogans. Nobody writes a patriotic song about master data management. Yet without it, the grand digital health dream becomes another Indian spectacle: impressive portal, weak ground truth, heroic claims, exhausted users.

The bitter joke is that everyone claims to care about healthcare. Healthcare is sacred when someone is dying, marketable when someone is selling, political when someone is promising, and invisible when someone must fund maintenance, train staff, reconcile identifiers, govern terminology, or repair a broken interface at 2:17 in the morning. We love the hospital as symbol and neglect the system as organism. We applaud doctors, blame clerks, fear bills, worship machines, and ignore the data architecture that decides whether the next clinician sees the truth or a decorative lie.

This hypocrisy is not uniquely Indian, but India gives it a special flourish. We want world-class care without world-class plumbing. We want artificial intelligence before reliable patient identity. We want dashboards before documentation discipline. We want national platforms before local workflow repair. We want predictive analytics on top of records that cannot reliably say whether the same patient came back twice or two patients came once. We want the mango without the tree, the shade without the roots, the miracle without the maintenance contract.

The practical implication is blunt. Design must begin not with applications but with representational honesty. What entities does the system claim to know? Patient, encounter, order, result, diagnosis, medication, procedure, claim, referral, consent, provider, facility. Who owns each? When does it change? What is its source of truth? What is captured because care required it, what is captured because billing demanded it, what is captured because regulation threatened punishment, and what is captured because a tired human clicked the least-wrong option?

Governance must stop behaving like a festival committee and start behaving like architecture. Terminology mapping is not clerical housekeeping. It is semantic policy. Interface specifications are not vendor paperwork. They are contracts about reality. Data warehouses are not dumping grounds for exhausted transactional systems. They are interpretive machines, and every transformation inside them should carry provenance, latency expectations, loss markers, and an admission of what was guessed.

Implementation must also accept the dirty constraint: clean solutions are often impossible because the live system cannot be stopped. Hospitals do not pause for refactoring. Patients do not wait for ontology alignment. Legacy systems cannot always be retired. Vendors resist openness. Regulators demand reports before workflows mature. Procurement rewards visible features over invisible correctness. Clinicians are already overloaded. Clerks have private survival workflows that no architecture diagram shows. The old system keeps breathing through tubes while the new one is assembled beside it with borrowed screws.

So the architectural direction is not purity. Purity is for whiteboards and newly funded fools. The direction is staged realism. Stabilize identity. Capture provenance. Separate transport success from semantic confidence. Make transformations explicit. Preserve clinical narrative while extracting structured meaning carefully. Build canonical models where they help, not as imperial fantasies. Treat FHIR profiles and implementation guides as local agreements, not magic carpets. Design for reconciliation, not imaginary perfection. Audit not only whether data moved, but what meaning it lost in transit.

And write about it anyway.

Write the technical blog even if only seventeen people read it, and twelve of them are bots, and one is a recruiter searching for “FHIR architect” with the enthusiasm of a bored heron. Write because the public vocabulary is too thin. Write because India badly needs people who can tell the difference between digitization and systems thinking. Write because someone must say that a scanned PDF is not interoperability, that a portal is not governance, that a dashboard is not truth, that a hospital application is not healthcare IT maturity, and that national pride does not become infrastructure merely by shouting at skeptics.

The blog form is inadequate, of course. It is a small window cut into a large, damp, badly wired building. But a window is not nothing. Through it, a reader may see that the mess has structure. The boil has anatomy. The complaint has engineering behind it. And perhaps that is the only honest ambition left: not to make healthcare IT glamorous, but to make its hidden machinery visible enough that fewer people can pretend the machine is working when it is merely making noise.