Drug overdose deaths in the U.S., including those linked to opioids, have spiked during the Covid-19 pandemic. The rise in opioid-related overdoses in particular has had a wide-reaching impact across the country, with stark increases across geographies and age groups, including sharp surges within the Black and Latino communities.
But behind the layers of despair, a glimmer of hope is beginning to shine through. In February, after years of intense litigation, 46 states and thousands of local governments agreed to a $26 billion settlement with four of the country’s largest drug distributors to end all civil liability for their role in that opioid epidemic. These funds will be made available across the country for addiction treatment, prevention services, and social services.
But how can we ensure that the money will actually make an impact? Unfortunately, widespread evaluations across the social sector reinforce what sociologist Peter Rossi hypothesized in his Iron Law of Evaluation decades ago: Program effects, when taken to scale, tend toward zero. (Jon Baron collected some empirical support for this idea in his 2013 testimony to the House Committee on Ways and Means, and various researchers and practitioners have tried to mitigate the Iron Law for years.)
Paying for outcomes is intriguing but often expensive
For the past decade, governments have funded a series of experiments testing funding models that pay for results, rather than services. The underlying idea is far from new and its underlying principles are reflected in the rise of value-based care. But the Pay for Success movement has taken these ideas and run with them, at times linking 100% of payment to long-term, policy-relevant outcomes such as improvements in housing stability or economic mobility.
These projects are ambitious, powerful, complicated, expensive, and unpredictable. A central challenge that the field faces is epistemological: If you’re going to pay for the outcomes of programs, you have to really understand those outcomes. What’s more, data tracking, outcomes measurement, and evaluation are slow, expensive processes.
For example, in early 2017, Social Finance launched a project with the Commonwealth of Massachusetts, JVS Boston, and others to help limited English speakers find better jobs. State funds were to be released in large part on the basis of a randomized controlled trial, conducted carefully by a professional evaluator, to understand the program’s impact. For nearly four years, while we knew a lot about who was enrolling and how much of the program they participated in but we knew little about their outcomes.
Often, the outcomes we care about most—overdose reduction, incarceration, economic mobility—take place over long periods. Data is often hard to access, and can lag by months or years. We’re typically reliant on brilliant but expensive third-party evaluators to access, clean, and interpret this data. In a standard project, outcomes-based contracts can spend 10-20% of their budgets just to know what happened, after it all happened. That burdens projects with too much cost and undermines our ability to understand the real-time impact of programmatic changes.
However, new tools are emerging to address these inefficiencies at a low cost. And wastewater monitoring is one of those tools.
Focusing on population-level outcomes with wastewater
A typical drug test screen determines if a person is using a substance by looking for chemical markers in their urine. Wastewater monitoring applies this same principle to whole communities, analyzing the waste within sewer systems to measure collective patterns of substance use.
This technology offers us the opportunity to build better, simpler outcomes-based funding tools. Wastewater monitoring is already widely used in the U.S. to track Covid-19, and the data provides a measure of community disease burden that is independent of testing and access to health care.
Additionally, wastewater data can be generated in time-bound and geographic scales most relevant to policies of interest. For example, opioid overdose statistics are released by the CDC on a quarterly basis for the previous year, creating a one-year delay between a policy intervention and the ability to quantify its impact. In contrast, wastewater data on opioid consumption can be reliably generated monthly and likely even faster. Wastewater monitoring also allows for flexible geographic granularity: Sampling can be established at a wastewater treatment plant to reflect an entire community, or it can be deployed more strategically at upstream manhole sites in a sewer network to capture data for more precise geographies, including individual neighborhoods.
And wastewater monitoring is a comprehensive, direct measure of population-level health outcomes. Say, for example, that a program aims to reduce total amounts of opioid use in a community. Opioid overdose statistics may not reflect the success of the program, because only a small proportion of all people with opioid use disorder overdose. Prescription monitoring programs can track changes in the number of opioid prescriptions but they cannot measure whether people are actually taking more or less of those prescriptions. By contrast, wastewater monitoring holistically measures rates of actual substance use, even among the 99% of users who do not overdose.
That kind of speed and comprehensiveness is a radical change from the way Pay for Success projects—and, for that matter, the evaluations for many social sector programs—work today.
Imagine instead an outcomes-based contract focused on reducing opioid misuse in long-term care (LTC) facilities. A state—say, Massachusetts—could agree to pay for a reduction in illicit opioid use among LTC residents. There are 375 nursing homes in the state, and 50 have over 100 residents. Through a Pay for Success contract, we could test an opioid misuse intervention in half of them, and then place wastewater sensors to test levels of heroin and fentanyl in each, maintaining anonymity among residents, but closely tracking aggregate impact. We could get weekly progress reports on the progress at each facility, allowing us to make changes to the program and immediately understand their effects—and allowing the state to pay only if those programs make an appreciable impact.
We could set up sensors across California’s 115 jails and measure whether the introduction of an in-jail methadone treatment program in 10 of them lowers rates of illicit substance use versus comparison populations. We could set up a wastewater sensor across 15 different neighborhoods in Austin to measure whether prescription drug take-back days lower prescription drug use in the weeks thereafter, or which locations for syringe services programs make the biggest impact on a community.
This is a remarkable opportunity to create lasting insights for a range of public health policy strategies. Combined with outcomes-based contracts based on the Pay for Success approach, wastewater monitoring provides a powerful tool to test programs quickly and inexpensively, paying only to the extent that those programs successfully achieve community goals.
Social Finance and Biobot hope to work together in the future but are not currently collaborating on any projects.
Claire Duvallet, Ph.D., is a founding data scientist at Biobot Analytics. Duvallet did her Ph.D. in the Biological Engineering department at MIT in Eric Alm’s lab, where she studied the relationship between the human microbiome and health and disease.
Jake Segal is a Vice President at Social Finance and leads the firm’s California office. In this role, Segal works with state and local governments, foundations, and nonprofit service providers to create innovative, data-driven partnerships that deliver better results.