When the Program Works

In 2006 the Nobel Peace Prize went to Muhammad Yunus and the Grameen Bank for lending small sums to poor people, most of them women, whom banks had turned away. By then microcredit had spread across dozens of countries, and its lenders reported repayment rates ordinary banks would envy. As an operation it worked: the money went out, the borrowers were found, and the loans came back.

Then the evaluations arrived. Researchers ran randomized trials of microcredit expansions in six countries, from Bosnia and Herzegovina to Mongolia, and in 2015 the results were published together. The editors summarized them as "a consistent pattern of modestly positive, but not transformative, effects." Borrowing rose where the lenders arrived and some small businesses grew, but average household income and consumption barely moved.1

That finding is instructive because of the order in which things happened. Suppose microcredit had failed as an operation, with lenders unable to find borrowers or loans going bad. Then the disappointing results would have had a dozen explanations: poor management, too little capital, corruption, a drought. Because the lending worked, none of those explanations was available. Credit had been extended competently and repaid, and poverty stayed roughly where it was. The limit belonged to the premise, the idea that lack of credit was what kept most poor households poor.

Success can teach what failure cannot. A failed program is evidence about the program. A program that succeeds on its own terms and leaves the need standing is evidence about the arrangement around it, and that evidence can be read two ways. Sometimes, as with microcredit, the success shows that the premise was wrong. Sometimes it shows the premise was right and the program keeps working only because the route it serves leaves the same work to be done for every person who comes through. Institutions find the second reading harder to hear, and patient navigation is a case of it.

Whose terms

The case does not sit cleanly, and the place where it resists belongs in the argument. Microcredit's operational success was measured largely by repayment, and repayment is the lender's measure. It can be produced by group liability, by public meetings where a defaulter is named, and by borrowers taking a second loan to pay off the first. In Andhra Pradesh in 2010, reports of coercive collection and of borrowers driven to suicide led the state government to restrict microfinance lenders by ordinance, and repayment there collapsed.2 The repayment figures had been accurate. They recorded the success of the lending, which is a different thing from the success of the borrowers.

The method needs a qualifier, because success on its own terms rules out the execution excuses only if the terms were the right ones. Before reading a success as evidence about the arrangement, ask whose success it was: the program's, measured by what the program controls, or the people's, measured by what happened to them.

The fork

Reform tends to run through a recognizable sequence. Someone identifies an exclusion: a group lacks credit, care, representation, protection. An intervention improves outcomes within the existing arrangement, and more people get in, get help, get through. Then, if the intervention works, it reaches a point where the success itself exposes something. It might be a dependency that remains, a burden that moved instead of shrinking, or a step that every person still has to be walked through.

At that point the reform can go two ways. It can repeat the intervention, with more access, more assistance and better delivery, and leave the underlying relationship as it was. Or it can treat the success as a finding and ask what it found. Repetition is easier to fund, easier to measure and easier to celebrate. Reading the success as a finding requires admitting that the intervention that worked was compensating for something that should not need compensation.

Income tax shows both branches. One answer to the burden of filing is more help with filing: preparers, software, volunteer clinics. Denmark took the other. It moved the assembling to the employers, banks and mortgage lenders who already held the numbers, and since 1988 most Danes have found their return already done, with nothing left to do but correct what is wrong.3

The navigator

Harold Freeman, a surgeon at Harlem Hospital, started the first patient navigation program there in 1990. As national president of the American Cancer Society he had led hearings on cancer among poor Americans. Those hearings found poor people meeting obstacles at every step between a suspicious finding and treatment. Navigators were hired to walk patients through those steps one by one and to take down the barriers Freeman named: cost, communication, the medical system itself, fear and distrust. Freeman later reported the result for the hospital's breast cancer patients. Five-year survival rose from 39 percent for those diagnosed between 1964 and 1986 to 70 percent for those diagnosed between 1995 and 2000.4

Freeman's comparison is before and after, and the program arrived alongside free screening, so the figure credits navigation with more than it can prove alone. A later multi-site trial found a moderate benefit on how quickly abnormal findings were resolved and treatment begun, and none at all in the first ninety days.5 The argument here rests on where the effect comes from, whatever its size, and the same trial pointed to it. Navigation helped most at the centers where usual care was slowest to follow up. Wherever navigation helps, it helps by getting a person through a step the system has made hard to get through, and every such success locates the problem in the route.

What followed took the fork in one direction. Congress passed the Patient Navigator Outreach and Chronic Disease Prevention Act in 2005. The American College of Surgeons' Commission on Cancer wrote a patient navigation process into the standards it uses to accredit cancer programs. Navigators now have their own training programs and credentials. Each step was a reasonable response to a program that saved lives, and together they add up to the answer "more navigation." It is harder to find anyone whose job is the other answer, which asks what steps the navigators spend their days getting patients through and why those steps exist. Freeman framed navigation as the removal of barriers, but the barriers themselves are the finding.

A health system that adds a navigation profession looks more developed, but the trades that grow up to guide people through a process are sized by the work the process leaves to each person, and development proper removes the steps instead.6

A successful reform should eventually make some of the virtues it once celebrated unnecessary. The navigator's persistence and the advocate's relentlessness are admirable, and the people who have them have earned every word of praise. Their continued necessity is still a finding about the arrangement they work inside.

An institution built on a need that should shrink

An organization that becomes excellent at a function whose necessity ought to diminish has an awkward relation to its own success. Its measures rise with the need: patients navigated, claims filed, families served. Its funders pay for visible service. Its staff build careers and professional identities on the work, and its standing comes from helping people through an arrangement it has little authority to change.

None of this requires cynicism, and most of it happens without anyone choosing it. Paul DiMaggio and Walter Powell showed that organizations in a field converge on the structures that confer legitimacy in that field, whether or not those structures best serve their purposes.7 Once a navigation program marks a serious cancer center, serious cancer centers will have one. Why their patients should need it then falls outside what anyone is evaluated on.

The compensating institution also holds the best evidence anyone has about the defect. Navigators know which step loses patients. Advocates know which form fails, and which office loses documents. Case by case, they have been mapping the arrangement's failures at the scale of a single person, which is the scale at which nothing has to change.8

Give that knowledge a second outlet and the institution acquires a second product. Imagine a navigation program that reported each quarter the steps its navigators spent the most time on, with a duty on the hospital to answer for each, and a trigger: a step that tops the list quarter after quarter puts the rule behind it under review. Such a program would be turning repair into evidence a rule has to face.9 A second number would measure it honestly: the share of patients who got through the route without needing a navigator at all. That share should rise toward whatever the irreducible remainder turns out to be.

What stays

Some navigation is irreducible, and the argument would be wrong to pretend otherwise. Cancer is frightening, its treatment is complicated, and a person who guides someone through it is giving care that no redesign of the route would make unnecessary. Freeman's barriers included fear, which no scheduling reform removes. So the residual will not reach zero, and nobody should want it to. The useful question is how much of a navigator's day goes on the disease and how much goes on the route. Time spent on the disease is what a guide is for, while time spent on the route is what the program found.

This archive has argued that a rule's truest record is its remainder, the cases it did not fit and the work someone did to carry them.10 Its successes, read correctly, are a second record, and in one respect a better one, because they arrive with the easy explanations already ruled out. A program that has worked has earned a harder question than whether to do more of it. Now that the program is no longer the explanation, what is still standing between people and the thing it helped them reach?