Dr Arun Mitra · Community Medicine & Family Medicine
AIIMS Bibinagar
6 August 2026
Proximity is not access
What a drive-time map of stroke care leaves out
Dr Arun Mitra · Community Medicine & Family Medicine · AIIMS Bibinagar
Co-hosted with the Department of Neurology
One patient, Yadgir
Onset → someone acts
180 min
Reaches the nearest PHC
Workup at the PHC
130 min
Transfer dispatched
30 min
Door-to-needle after transfer
60 min
Total: 6 hours 40 minutes. The 4.5-hour window closed while he was still at the PHC.
This is not a story.
These are the median values in our model for a patient who does not recognise the stroke: the pathway 68% of patients take.
The map calls that taluk covered
Yadgir sits inside a drive-time band our national analysis scores as reachable.
I co-wrote the paper that produced the 26.3%.
This talk is about what that number does not say.
Two numbers, same state, same year
64% of strokes within a 2-hour IVT drive
11.8% of ischaemic strokes treatable in the 4.5-hour window
6,743 a year. Reperfusion does not apply to the other third of strokes.
Different denominators, same question. The drive-time measure overstates treatable access 5.4-fold.
Where this goes
What the national picture looks like
How we know, and how wrong we could be
What happens when the model has to respect the real pathway
Who carries the loss
What to build, and what building cannot fix
The gap is recognition and wrong-facility routing, not distance. That mechanism follows from how patients reach care, which is not unique to Karnataka.
Beat 2: The national picture
566 centres, and where they are not
566 IVT-capable centres, in 26 of 36 states and UTs
361 (63%) also EVT-capable
SCPM 0.41 (IVT-C) · 0.26 (EVT-C)
The national anchor
26.3%of India’s population within a 1-hour drive of an IVT-capable centre
20.6%within 1 hour of an EVT-capable centre
This is population inside a drive band, not patients treated.
It is the number the field quotes. Including us.
The spread behind the average
Beat 3: How we know
Four questions we had to answer
Where are the centres?
How long does it actually take to drive there?
What about the places we never measured?
How wrong could we be?
The last one is the reason this section exists.
Centres, and routes that failed
566 centres geocoded from the national roster
Routes the API could not return were re-run and corrected
We measured points. Patients live in between.
Every point is a measurement. Everything between them is an estimate.
We tested four ways and let the data choose
Nearest neighbour wins: ±29 minutes typical error, 81% better than guessing the average.
What we did with the disagreement
Nearest neighbour
Voronoi
IDW
Same data. Three methods. Three answers. The spread between them was carried forward as uncertainty, not averaged away.
Beat 4: The correction
Karnataka
87,806 strokes a year 95% CrI 77,259 – 102,083
~57,074 of them ischaemic
240 taluks. 229 have no stroke centre. Six have an EVT-capable centre.
Stage 1 — the drive-time view
What the pathway actually costs
Roughly three hours at the median, before anyone drives anywhere.
Stage 2 — add the clock
Still assuming every patient recognises the stroke and drives straight to a stroke centre.
Stage 3 — add the pathway
68% first present at a facility that cannot treat them.
Why the wrong door is fatal
Where the treatable patients come from
The map of it
Median taluk: 5.7%. 110 of 240 below 5%. Only five above 25%.
Is this robust?
Delay assumptions: perturbing every pathway delay one at a time moves the headline only between 8.8% and 14.1%, and only via the recognised patient’s clock.
Real facility geography: swapping the surrogate for observed KGIS facilities moves it less than 0.1 pp.
Interpolation: across the four methods, CV ≈ 16%.
What it means in patients
516additional good outcomes a year from IVT (mRS 0–1)
372from EVT (mRS 0–2)
At 75% recognition, the IVT figure roughly doubles.
Illustrative: time-stratified absolute-benefit estimates applied to modelled treatment times.
Beat 5: Who pays for it
The average was a lie too
1.4% Kalyana-Karnataka
14.0% the rest of the state
A region already recognised as backward under Article 371J.
Burden here, capacity there
The eight districts of Kalyana-Karnataka carry
17%of the state’s ischaemic stroke burden
and hold
2%of its treatable capacity
Stroke is one layer of a system-wide deficit
Yadgir, Raichur and Koppal sit at the bottom on every tier (PHC, CHC, taluk, district) before stroke capability is layered on at all.
Remember Yadgir
The patient we started with was not unlucky.
He was the modal experience of about 11 million people.
Hold that thought. The optimiser we are about to run has not seen this slide.
Beat 6: What to build
If distance isn’t the limit, is building pointless?
No. But what you build, and what it buys, both change.
IVT spokes: +4.1 points for fifteen
Nearly flat after the first site. And the first site is Gangawathi, in Koppal.
EVT hubs: +14.7 points for fifteen
Top five, all government medical colleges:
Gadag IMS
Tadikele Subbaiah IMS, Shivamogga
ESI Medical College, Kalaburagi
Vijayanagar IMS, Ballari
Adichunchanagiri IMS, Nagamangala
All five outside Bengaluru.
2.4× the leverage of a spoke: hubs are scarce, and each draws a wide drip-and-ship catchment.
Both hit the same ceiling
Adding hubs raises LVO coverage toward ≈30%, and then stops.
You cannot treat a patient who never recognises the stroke.
Supply is bounded by demand.
Unless recognition moves
Recognition
Baseline
After 5 hubs
After 10 hubs
15%
7.0%
11.7%
13.3%
32% (current)
15.0%
25.1%
28.3%
50%
23.5%
39.1%
44.2%
75%
35.2%
58.7%
66.3%
The same five hubs reach 58.7% if recognition reaches 75%.
Cheap and efficient, or expensive and necessary
IVT spoke
≈ ₹5 lakh per good outcome
EVT hub
≈ ₹78 lakh per good outcome
Hubs are justified by which strokes only they can treat.
The efficient answer is a mixed portfolio.
What actually binds: people
The optimiser says where a hub should go. It cannot staff one.
Beat 7: The ask
And Telangana?
30.4% within 1 hour of IVT
28.3% within 1 hour of EVT
Above the national average on both.
Above Karnataka on EVT.
Better served than most of the country.
The gap here is evidence, not access
Telangana appears in no stand-alone stroke systems-of-care study.
It appears only inside multi-state trial networks.
So does Karnataka. So do Tamil Nadu and Andhra Pradesh.
The number I am least sure of
Every headline in this talk rests on two borrowed values:
32% of patients recognise the stroke and travel directly
68% first present somewhere that cannot treat them
They come from a Kerala single-centre study and general Indian pre-hospital literature. Not from Karnataka. Certainly not from Telangana.
And the sensitivity analysis says almost nothing else moves the answer.
Three things this room can supply
1. A first-presentation audit. For consecutive stroke admissions: where did the patient go first, what tier was it, and how long did each step take? An addition to an existing registry, not a new study.
A Telangana counterpart analysis: the same pipeline, this state.
Clinical validation of the pathway parameters: you have seen these patients. I have priors.
Nearest neighbour
Voronoi
IDW
