One statistic about patient adherence gets repeated at almost every doctors’ meeting, and almost nobody asks what it actually measures.
One statistic travels faster than almost any other number I hear quoted in Indian medical circles. Seventy four percent of patients do not take their medicines as prescribed. I have heard it at conferences, inside coaching calls with doctors I work with, and more than once from a doctor repeating it with complete confidence and no source attached.
The number is real. It comes from genuine research on medication adherence across India. It is also the top of a range, not a single finding, and the full range tells a different story than the headline figure alone.
In the pillar on Habit-Based Treatment, I used this range briefly, as confirmation that the Habit Leak is a documented pattern and not a feeling. This piece goes where that one did not have room to go. Where the range comes from, why it moves so much between studies, and what hypertension and diabetes research in India specifically show once you look past the headline number.
The number everyone quotes, and what it actually measures
Start with where 18.7% to 74% comes from. It is not one study. It is the span across many Indian studies on medication adherence, using different measurement tools, on different conditions, in different parts of the country, over different years. A 2024 paper in JMIR Research Protocols lays out this exact range while reviewing the existing evidence on adherence in India, most of it assessed using scales originally built and validated on Western populations rather than Indian patients specifically (source).
A range this wide is not a flaw in the research. It is the research telling you something true. Patient adherence in India does not behave like one number. It behaves like dozens of different numbers, depending on what is being measured and how.
Most people who quote this statistic use only the top end, seventy four percent, because it is the more alarming figure, and alarming figures travel faster than accurate ones. The Habit Leak, one of the three invisible leaks I first named in an earlier piece, sits inside exactly this pattern. The real use of the range is not the shock value of its top end. It is the width itself.
Why the same question gets such different answers
The measurement tool changes the answer
Adherence research uses several different tools to ask the same basic question. Some rely on the patient’s own report, a simple interview or questionnaire. Some use pharmacy refill records, checking whether a prescription was renewed on time. Some use pill counts, physically checking what remains in a bottle at the next visit. Each tool catches a different kind of failure, and each tool can produce a different number from the same population.
| Measurement tool | What it actually captures | Typical direction of bias |
|---|---|---|
| Self-report | What the patient remembers and is willing to say | Tends to show better adherence than reality |
| Pharmacy refill records | Whether the prescription was renewed on schedule | Misses patients who refilled early or stockpiled |
| Pill counts | What remains in the bottle at the next visit | Misses timing, only catches total quantity |
A patient who reports taking medicine regularly may still be overdue for a refill. A full bottle at the next visit can tell you something a confident verbal answer never will. Studies that rely on self-report alone tend to show better adherence than studies using refill records or pill counts, not necessarily because the patients are different, but because the question is being asked differently.
Setting and geography change the answer
The same body of research that produces 18.7% to 74% also spans different settings. A study measuring adherence among working professionals attending a metro hospital, and a study measuring adherence among rural patients managing the same condition, are in effect answering two different questions wearing the same label. Region, income, distance to a pharmacy, and whether care is delivered in a hospital or a community setting all move the number before the patient’s own behaviour even enters the picture.
What hypertension and diabetes research specifically show
Hypertension
A 2025 meta-analysis in Current Hypertension Reviews pooled forty Indian studies covering more than eighteen thousand hypertension patients. The pooled non-adherence figure came out close to forty eight percent, roughly half. But the number moved sharply by setting. Non-adherence was markedly higher in the eastern region of the country than the national pooled figure, and higher again in rural and community-level care than in hospital-based care (source).
Read that correctly and the lesson is not “half of hypertension patients fail.” The lesson is that geography and the level of care a patient receives move the number more than the condition itself does.
Diabetes
A separate systematic review and meta-analysis focused on type 2 diabetes in India found individual study results ranging from eleven percent to sixty eight percent non-adherence, an even wider spread than the hypertension review, with a pooled estimate near forty eight percent among the subset of studies using a comparable scale (source). The National NCD Monitoring Survey, a separate and much larger government survey, found that only around a quarter of rural patients with diabetes were regularly taking their prescribed oral medication, a related but distinct measurement of ongoing medicine use rather than a scored adherence questionnaire (source).
Two different conditions. Two different review teams. Two pooled figures that land in roughly the same place, close to half, and two wide ranges underneath each pooled figure once you look at the individual studies making it up. The pattern across both is the same pattern this whole piece is making. The average hides more than it reveals.
The physician who was quoting the wrong number
I worked with a general physician running a multi-specialty clinic in a tier two city, seeing a steady mix of hypertension and diabetes patients through most of his week. He had read the seventy four percent figure somewhere and used it often, in patient conversations, in a talk he gave at a local medical association meeting, and by his own admission, in his own head whenever a patient’s blood pressure reading disappointed him.
The number was doing real damage in a direction nobody intended. Patients heard “most people fail at this” and quietly concluded that failure was close to normal. The physician heard the same number and concluded his own patients were probably no different from the worst end of a national statistic, so he stopped looking closely at who in his own clinic was struggling and who was not.
“I was using the seventy four percent figure like it described my own patients. It never once occurred to me that I had never actually checked.”A General Physician, After Reviewing His Own Refill Records
We did something narrower than quoting any national figure at all. We pulled his own pharmacy partner’s refill data for his hypertension and diabetes patients across one quarter. His own hypertension number came out lower than the national pooled figure. His own diabetes number came out worse than the national figure, in the opposite direction from what the general statistic had led him to assume. Neither number matched the headline he had been repeating for two years.
The research was never wrong. It was being asked to do a job it was never built for, standing in for data about his own clinic that he had never collected.
What this means for your own clinic
None of this research tells you what is happening inside your own OPD. It was never built to. A pooled figure from forty studies across the country describes the country. It does not describe one physician’s patients, in one city, managing one condition, under the specific habits and systems unique to that clinic.
The honest use of this research is narrower and more useful than most doctors expect. It tells you that non-adherence is common enough, and documented thoroughly enough, that assuming your own patients are the exception without checking is the riskier assumption, not the more comfortable one. It does not tell you your own number. Only your own records can do that.
This is exactly the gap the first part of Habit-Based Treatment, visible tracking, is built to close. An Effort Doctor carries a national statistic in his head and treats it as a verdict on his own patients. An Ecosystem Doctor builds a visible record of what his own patients do between one visit and the next, the same structural instinct behind the thirty day dependency test I wrote about separately. Neither depends on remembering a number correctly. Both depend on the system doing the remembering.
- If I quoted an adherence statistic to a patient this month, where did that number come from?
- Do I know my own clinic’s adherence pattern for my most common condition, or am I estimating it from something I read or heard?
- Which condition in my own practice would most reward one real quarter of tracking before I assume anything further about it?
Questions doctors ask me about this
Is the seventy four percent figure accurate?
It is a real number from real research, but it is the upper end of a range, not a typical finding. Using it alone, the way most people do, overstates the picture for most conditions and most settings.
Which number should I use when I think about my own patients?
None of them, on their own. The published research tells you non-adherence is a real and common problem worth taking seriously. It cannot tell you your own clinic’s number. Only tracking your own patients can do that.
Why do the diabetes and hypertension figures vary so much between studies?
Mostly the measurement tool and the setting. Self-report, pharmacy refill records, and pill counts each catch a different kind of failure. Rural and community settings tend to show higher non-adherence than urban, hospital-based settings. Change either one and the number can move by a wide margin.
Should I ever quote these statistics to a patient directly?
Rarely, and carefully if at all. A frightening national number can make a struggling patient feel like a statistic rather than someone with a specific, fixable gap. It is more useful to you, as a planning tool, than it is to them, as something to hear in the room.
Does a high non-adherence figure mean my patients are not trying?
Not in my experience. Most of the patients behind these numbers intended to follow the plan completely. I have written separately about why intention and habit are not the same thing, and that gap explains far more of this pattern than a lack of effort does.
How do I start tracking my own clinic’s adherence instead of relying on national figures?
Start with one condition and one simple marker, a refill date, a tracking card, a follow-up call pattern, whatever you can sustain. A single quarter of your own data will tell you more about your own patients than any published range, including this one.
A statistic this wide was never going to settle the question of what is happening inside your own clinic. Its real value is narrower than the headline number suggests, and more useful once you stop asking it to do a job it was never built for.
Before your next clinic day, pick one condition you treat often, the same way I described in the pillar on Habit-Based Treatment, and ask yourself honestly whether you know your own adherence number for it, or whether you have been borrowing someone else’s.
Want an honest read on where your practice stands?
I offer a free 15 minute strategy call. 15 minutes. Zero sales pressure. Just a clear, honest plan for your practice, built around the Super Doctor by BK system and the Habit-Based Treatment thinking behind it.
Read the full framework in OPD to Ecosystem
I wrote the complete version of this thinking, the Habit Leak, the research behind it, and what closes it, in OPD to Ecosystem, across all nine chapters and the closing framework.
One Response