The care gap in the Kingdom

"You might want to grab a mask for while you wait," suggested the woman at the reception pointing me to the waiting room. "There's a nasty new flu variant going around, and this is the place to get it."

By the time I checked in at the ExpressCare at the beginning of March, my immune system had already mastered the entire alphabet of flu, covid, and other variants that I came across every other week since Thanksgiving. For better or worse, I was finally clear of those seasonal illnesses and moving on to new maladies.

My visit was pleasant enough, and produced a prescription that seemed promising. But as relieved as I was to get a prescription, I braced myself for the question that has gotten more frustrating in recent years. "What pharmacy should we send this to?"

I live in Hardwick, Vermont. A small town of 3,000 in the state's rural Northeast Kingdom. All my life the town had a pharmacy. When I was a kid it was a Brooks, which at some point turned into a Rite Aid. It transformed again into a Walgreens before finally, much beleaguered, closing its doors in 2024. This is just one of thousands of pharmacies across the country that have shuttered in recent years.

Losing the pharmacy has left a critical gap in care in our community. Our nearest pharmacy is now in Morrisville, a 20 minute drive from the village of Hardwick. This has left thousands of people in the area in a pharmacy desert.

Bigger than Hardwick

A few weeks ago I said that Hardwick's most critical need is a pharmacy. I still think that's true, but the problem is even bigger than Hardwick. The drive is even farther for others outside of the village center and in surrounding towns like Greensboro and Walden. Elsewhere in the NEK people are even worse served: there isn't any pharmacy at all in Essex County.

The Northeastern Vermont Regional Hospital was recently awarded a grant to address the care gap, and in particular to bring a pharmacy back to Hardwick.

Selfishly, I welcome the news. But I've been wondering, what does this care gap look like across the rest of the region? Where would a new pharmacy benefit the most people?

Getting there from here

The road network in Vermont is built up, over, and around steep hills and winds down through sinuous river valleys. Many of the roads around here are unpaved and, especially this time of year, can quite literally bog down your trip.

To understand the gap in services around here, you can't look at linear distance alone. You can't even just look at the map, as tourists who book stays for their Stowe ski trips in Cambridge quickly discover. You have to simulate driving point to point from places that people live, on roads that are actually passable.

The maps above show the scope of the care gap using travel time. You can see above how important the Hardwick pharmacy was. Losing it doubled travel times for thousands of people.

The weight of the problem

There's a quip that most attempts to show geographic trend maps are really just population maps. Ironically, the maps above _don't_ show population, but really ought to.

As the maps show, the worst served area is Essex County. A naive analysis might recommend putting a pharmacy in downtown Granby, which is rather far from any pharmacy. Perhaps the 81 residents of the town would welcome it. From a utilitarian and economic standpoint, there are clearly more practical options.

Drawing the population density gives a slightly more informed view of the problem. The map below shows population density, with regions over 20 minutes from a pharmacy highlighted in red.

The case for Hardwick

Granby is not the most practical place for a pharmacy. Is Hardwick? You can see in the map above that there is more density in the area. More people stand to benefit from a pharmacy here.

To identify the optimal place for a new pharmacy, I re-ran my drive time simulation with a hypothetical new pharmacy in every town in the area, one by one. Since in Vermont a "town" is a sort of umbrella administrative division and not meaningful as a specific geographic point, I took a list of 237 villages and other named places that could conceivably host a new pharmacy.

For example, in Hardwick, we have the obvious location of the Hardwick village, the next-most-plausible one of East Hardwick village, and also some low-probability sites like Mackville and Hardwick Center. For the sake of argument, I considered them all.

To account for population, I used a metric of human-minutes saved. In other words, multiplying the drive time to the new pharmacy by the number of residents who would use it, and then finding the aggregate time saved over their current travel time. Of course this isn't perfect for many reasons, since there are other variables we don't know that affect who is going to a pharmacy, how often, and which is most convenient. But this metric at least tries to estimate where a new pharmacy would do the most good for the most people.

Sure enough, the Hardwick village rises to the top as the place that a new pharmacy would do "the most good."

Where else is in need?

If you sort the table by "people served," you see that Walden, not Hardwick, might be the best location in terms of number of people served.

The issue here goes back to population density and the road network. Hardwick benefits from being at the intersection of routes 14, 15 and 16, and being relatively well-developed for the region. Walden is more central to the care gap in Caledonia County, but has less population density, slower roads, and little infrastructure.

In other words, while a pharmacy in Walden would be somewhat more convenient for more people, a pharmacy in Hardwick would be, on average, a significant convenience for almost as many people.

There are gaps in care that a pharmacy in Hardwick (or Walden) won't fix. While I bet Craftsbury residents will use it, they will still be over 20 minutes away. Southern Caledonia County around Peacham and all of Essex county will still be in need. The maps above show areas still deep in the red in every scenario.

When I go to Morrisville these days, I find it almost comical how many pharmacies there are within a mile of one another. But the truth is that redundancy is also important. As we've experienced in Hardwick in the past few years, a single pharmacy is a point of failure. The Kinney Drugs in Barton is currently playing a critical role in the NEK, as will the new pharmacy in Caledonia County, wherever it ends up.

Methods and limitations

The numbers here are estimates. I think they're fairly reliable, but there are some caveats.

Area of interest

I looked at the traditional Northeast Kingdom (Essex, Caledonia, and Orleans counties), but also included Washington and Lamoille. Washington and Lamoille skew the numbers in some important ways: they are more dense and more developed than the NEK. I chose to do this for a couple reasons.

One is that I am a proud Woodbury Bear and I can attest that Woodbury and surrounding parts of northern Washington County benefit from Hardwick's (Caledonia County) infrastructure. Such is also Hardwick's relationship with Morrisville (Lamoille County), especially with regard to the pharmacies right now. So, including these counties at the margins makes sense.

The second reason is that I think it adds a nice contrast to the map, to see how many more services exist in Lamoille and Washington counties than in all of the NEK.

The downside is that including these counties somewhat masks the criticality of the problem. For example, the statistic of the percent of people living in a pharmacy desert would be quite a bit higher if it excluded Washington and Lamoille counties, perhaps double.

Address points

I pulled addresses using a tool I built a couple years ago. I used addresses from five counties in Northeastern Vermont: Essex, Caledonia, Orleans, Lamoille, and Washington counties. Many of these are *not* residential addresses! So, many travel times are computed from, say, a business instead of a home. Of course, people often aren't driving directly from home to a pharmacy, and we don't try to estimate any behavioral patterns here, either.

Population estimates

I weighted addresses using Census blockgroup population data. The addresses can be thought of as representing `w` people, as the population divided by the number of addresses in the blockgroup. Again, there is the limitation that not all addresses we have are primary residential addresses. Plus, not all households within a blockgroup have a uniform number of residents.

Pharmacies

I found active pharmacies anywhere in Northern Vermont and Canada from Google Maps. These looked reasonable to me, though it's true sometimes that the specific function of a place can't be determined from Google Maps. For example, perhaps a doctor's office is incorrectly marked as a pharmacy. Or perhaps an establishment specializes in a certain type of prescription medical equipment, such as AdaptHealth.

I suspect there might be a few of these cases in my pharmacy list, but I think it probably does not change the overall gist of this rough analysis.

Drive time simulation

I estimated the driving time from every address to every pharmacy using OSRM to find the minimum travel time to a pharmacy across the entire region.

I am certain there are inaccuracies in the routing. OSM is great for a free resource, but (as a contributor myself) I know that it is not fully accurate. People still get stuck following bad GPS advice around here, and I imagine OSRM has a few of these issues, too.

The drive time model is also not perfectly accurate. I compared a few thousand routes against Google Maps and noticed the estimates were a little different. I believe Google's routing is better calibrated, but it'd be prohibitively expensive to run the same analysis with Google's API. It should be possible to calibrate OSRM somewhat using a sample of Google's routes, but I will leave that for a future exercise.

Demographics and behavior

The analysis assumes every person has the same needs. I think this once again leads to understating the problem in this analysis. The rural parts of the NEK--the parts least well served here--skew older. Older adults tend to require medical care more than younger adults. So it could well be that areas with more young adults like Montpelier, Barre, and Waterbury are on average making less use of the pharmacies they have, while older communities like Greensboro could use one relatively more.

There are other limitations here, too, like the fact that I can't account for convenience and behavior. People are often dragged in different directions here due to work. For example people who live in Hardwick but work in Montpelier might feel less inconvenienced by lacking a pharmacy at home than someone who lives in Hardwick and works in town.