In Chennai, the economist Frank Schilbach once gave some cycle-rickshaw drivers a choice. They could be paid 150 rupees for each visit to his study office, no matter what. Or they could be paid 120 if they showed up sober and 60 if they didn’t. The first deal paid more on every kind of day. Roughly a third of the men preferred the second one.
Schilbach checked that they understood the offers. Even on a sober day, they were giving up 30 rupees, about a tenth of their reported daily earnings. They wanted to preserve a financial reason to stay sober, and were willing to accept less money to do it. The experiment is here.
I could have years of excellent data about a man’s alcohol purchases and know nothing about the years he spent trying to stop. A purchase leaves a receipt, which makes that preference unusually easy to establish. The request for help is evidence too, even if it never becomes a transaction.
Suppose a man buys alcohol on Monday and calls a clinic on Tuesday because he wants to stop. I would be reluctant to tell him that Monday revealed his authentic preferences and that Tuesday requires a lecture about personal responsibility. He has already done something fairly responsible: found someone he thinks can help and asked them.
My instincts about this are libertarian. Adults get to take risks and enjoy things I would not choose for myself, and I am suspicious of people who are eager to explain what other adults really want. Giving someone authority over which pleasures other adults may pursue requires much more justification than discovering that pleasure sometimes ends badly.
What interests me here is what happens when the person himself asks for something different.
Consider what happens when the man orders his alcohol. A shop has inventory. A driver has a vehicle and an afternoon. A customer has an address, a payment method, and a desire to avoid putting on shoes. Software brings these things together, handles the transaction, and guides the driver through the identity checks required for delivery.
Now follow his Tuesday call. Imagine he finds a clinic in his insurer’s directory, only to learn that it no longer accepts his plan. Another clinic asks him to call during business hours, which are also the hours his employer would prefer he spend working. Eventually he finds an appointment and rearranges work to attend it. Then somebody needs a record from somewhere else, and he has to find out who has it and how to get them to send it.
“Just call them” sounds like a small request until it is the third person who has said it to you and you are trying to remember which office closes for lunch.
The delivery company has organized itself around the customer’s decision. The clinic expects the customer to organize himself around its business.
If he gives up somewhere in this process, I would want to know where. Did he reach anyone who could book him? Did the available time work? Did we ask him for something he couldn’t get? I’m wary of explanations that arrive at his lack of motivation before answering these questions, particularly when he is the one who made the first call.
We already know how to investigate this kind of problem in a consumer business. If enough people abandon a checkout, someone looks at the checkout. They might find a confusing field or a payment method that doesn’t work. They might discover that the customer is being asked to stop what they are doing, retrieve something from another website, and come back. None of this tells us how to treat addiction, but it ought to make us curious about what happens between asking for treatment and receiving it.
There is a complication here that a checkout comparison can obscure. The man may want help cutting down when the clinician thinks he should stop. He may ask for help and then change his mind. If he misses an appointment, perhaps he no longer wants to come; perhaps the appointment was at a time he couldn’t get away from work. Those possibilities call for different responses, and a system designed to get people to return won’t necessarily distinguish them. I want more people to get care, so it is worth admitting that I have an interest here that could make me too quick to interpret his reluctance.
Still, when someone is actively trying to book an appointment, there is quite a lot we can do before we need to resolve any of that.
Some of the difficulty is legitimate. A clinician needs to decide whether a medication is appropriate, and sometimes the conversation needs to take longer than the patient expected. I can accept all of that and still want an explanation for why the patient is chasing a record we need. The difficulty of making a clinical judgment doesn’t explain every other difficulty we have attached to the visit.
Some requirements have already changed. Until the end of 2022, clinicians generally needed a special federal waiver to prescribe buprenorphine for opioid use disorder, with limits on how many patients they could treat. Congress eliminated the waiver and those patient caps. Standard DEA registration requirements and applicable state laws remained.
A patient still needs a clinician willing to see them, a way to pay, and a pharmacy that can fill the prescription. There are things people are allowed to obtain that they have a hell of a time actually obtaining.
At Grata we can see some of this directly. By September, 2,000 people had requested insurance-reimbursed care in states where we had no clinicians and no payer contracts. They had told us what they needed. We had not yet recruited the clinicians or secured the contracts required to provide it. Improving a form would only have got them to the same disappointment faster.
In the states where we can provide care, we also see people stop at the insurance check and at scheduling. Those observations alone cannot tell us why. We need to find out what happens at those steps before deciding whether the problem is the form, the cost, the appointment times, or something else.
This leaves us with several kinds of work, some of which software can help with and some of which require us to hire people, negotiate contracts, or get someone on the phone. I can explain why these things take time, but I can’t reasonably describe the resulting gap as a failure on the patient’s part. Recruiting clinicians and getting them into payer networks is work our company has to do.
A lot of what I want us to get better at is unglamorous in this way. Find out whether we can see someone before sending them through a process that assumes we can. Make the information they have already given us available to the person who needs it. Get a prescription to the intended pharmacy. Follow up when something hasn’t happened, including when the thing that hasn’t happened was our responsibility.
The men in Chennai were willing to give up money for a better chance of carrying out their own decision. I want that wish to attract the kind of ambition we have brought to selling alcohol: people willing to spend years figuring out how to make something difficult easy for a customer, with all the clinical work and administrative bullshit that entails.
The man making that Tuesday call has something he wants to do with his life. Needing our help does not make it any less his.