Health equity & access
Digital health can widen the gap it was built to close. The inverse care law, the digital divide, and why every design default is a decision about who gets left out.
In one line
The inverse care law: the availability of good medical care tends to vary inversely with the need for it. Julian Tudor Hart wrote that in 1971 about physical clinics. Digital health obeys it too — and often more sharply, because software's defaults are invisible.
The uncomfortable pattern
Digital health is usually sold as a leveller: reach the village, skip the queue, democratise the specialist. Sometimes it does. But run the reasoning honestly and a different pattern appears.
Who benefits most from a teleconsultation app? Someone with a smartphone, reliable data, enough literacy to navigate it, a language it supports, a private place to take the call, and the confidence to describe symptoms to a stranger on video.
Now describe who that isn't. It isn't the elderly woman in a village with a shared handset and no private room. It isn't the daily-wage worker whose data pack ran out. It isn't the person whose first language the app doesn't speak.
The people who most need better access are the least equipped to use the thing built to give it to them. So a service that helps everyone by 20% widens the absolute gap — because 20% of a lot is more than 20% of a little. Improvement and inequity are entirely compatible, and almost nobody measures the second.
The divide is not one thing
"Digital divide" flattens several separate barriers, and conflating them produces useless solutions:
- Device — no smartphone, or a shared family one that belongs to someone else.
- Connectivity — no data, expensive data, or coverage that fails when it matters.
- Literacy — reading at all, and then digital literacy, which is a separate skill.
- Language — an app in English in a country with twenty-two official languages.
- Disability — screen readers, contrast, motor control. Frequently an afterthought.
- Gender — in many contexts men control the household phone. This is a health-access issue that no amount of app design fixes.
- Trust — a legitimate reluctance to put your health on a government or corporate system, held most strongly by people with the most reason for it.
Each needs a different answer. Giving someone a phone doesn't fix language. Translating doesn't fix trust.
Where it shows in what we build
- Algorithms trained on well-resourced populations and deployed on everyone. The model didn't learn about the people who weren't in the data.
- Pulse oximeters calibrated on light skin, missing hypoxaemia in darker-skinned patients — a device-design decision that became a clinical harm.
- Assistive technology: ~1 billion people need it and lack it, while capital flows to surgical robots serving hundreds.
- Registration forms that assume a fixed address, a surname, a date of birth. Millions of people have none of those in the shape the form expects.
What India gets right, and what it doesn't
Genuinely right: Scan & Share. Its runaway adoption — 23+ crore OPD tokens — happened precisely because it asks almost nothing of the user. Point camera. Done. No literacy assumption, no app-switching, no understanding of federation required. The most equitable feature in ABDM is also its least sophisticated one, and that is not a coincidence.
Also right: eSanjeevani's hub-and-spoke arm, where a health worker sits with the patient. That design reaches people a pure app never will — because the last mile is a person, not an interface.
The open risk: an ABHA-anchored system is only as inclusive as ABHA enrolment, and identity systems have a long history of excluding exactly the people who most need services — through a missing document, a name spelled three ways, a biometric that fails on worn fingertips.
The design discipline
Ask who is absent. From the data, the test set, the user research, the room. That's where inequity enters, and it enters by default rather than by decision.
Treat the low-tech path as a feature, not a fallback. SMS, IVR, a printed QR, a person. These are not embarrassments to be replaced by an app later; they are frequently the only thing that reaches the last 30%.
Measure the gap, not just the mean. "Outcomes improved" can be true while the gap widened. If you only report the average, you will never find out — and you will keep congratulating yourself.