How Cities Can Use Housing Data to Predict COVID-19 Hotspots: Lessons from Chelsea, MA

By Katharine Robb • June 17, 2020

Introduction: COVID-19 and Housing

“Sheltering-in-place” to protect ourselves and our communities during the COVID-19 pandemic requires just that: a safe place to shelter. But for millions of Americans, home is not a safe place. Their living situations are threatened by physical and psychological hazards as well as uncertain tenure and the threat of eviction. Substandard housing – including overcrowded conditions – is a growing public health problem across the US. These risks can stem from physical aspects of the environment, such as insect infestations, lead paint, and fire hazards – to social aspects, such as overcrowded conditions that contribute to strained relationships, poor mental health, and child development problems.

The pandemic is bringing the problems associated with substandard housing into sharp relief. A nationwide study of the impact of county-level housing conditions and COVID-19 infection found that as the proportion of households with poor housing (defined as overcrowding, rent greater than 50 percent of income, or incomplete kitchen or bathroom facilities) increased, so did the incidence and mortality of COVID-19. This association persisted even after controlling for demographics, income, comorbidities, and access to healthcare. Housing is much more than a physical structure – it has a profound impact on our health.

The Problem: Overcrowding in Chelsea, MA

Chelsea, Massachusetts, is emblematic of how the nation’s affordable housing crisis interacts with and amplifies the pandemic. High rents and low wages force many into unsafe and unstable housing conditions. Chelsea is a small, densely populated city, with over 40,000 residents on just 1.8 square miles of land. The majority of residents are people of color, recent immigrants, and low-income families, and the housing stock is in poor condition. Faced with skyrocketing rental costs, many families have no realistic alternative to substandard housing and overcrowded conditions. Residents double- or triple-up with other families or live in illegal “conversion apartments” (e.g., basements, closets, or porches) in order to avoid homelessness and stay close to jobs. These conditions were like tinder that set the pandemic ablaze—in Chelsea, rates of coronavirus cases are six times the state average.

In 2018, I spent a year working closely with housing inspectors in Chelsea to develop and implement a novel social service referral program within the Inspectional Services Department. Housing inspectors occupy a unique role – their job is to enforce the state sanitary code for housing—but in the course of their work, inspectors come face-to-face with some of the most hidden and dire housing and health problems. When I accompanied inspectors on their rounds, we came upon entire families subletting single rooms within a home. In one single room, a family of four slept, cooked meals on a hot plate, and carried out their activities of daily life. In some cases, residents only had access to bathroom facilities during certain hours of the day. In other cases, inspectors described even more precarious living conditions, like people using flattened cardboard boxes as bedding in unfinished basements; families living in closets without proper ventilation or means of escape from fire; a family living on a porch without heat, running water, or toilets. With living conditions like these, people’s mental and physical health is already strained – add in the coronavirus pandemic, and you can see why following recommendations like practicing good hygiene and isolating sick individuals is impossible.

Housing conditions affect coronavirus infection risk through many mechanisms. Overcrowding within homes not only facilitates greater transmission because of greater person-to-person contact, but is intimately intertwined with other risk factors such as poverty and underlying health conditions such as asthma, cardiovascular disease, and chronic stress. Comorbidities increase vulnerability to severe outcomes from COVID-19. People living in substandard housing are also more likely to be working low-wage jobs – the kind deemed essential during the pandemic, which unfortunately carry a higher exposure risk. Once infected, conditions within overcrowded homes often preclude people from being able to follow hand hygiene recommendations. They may lack reliable access to kitchen or bathroom facilities or control over the cleanliness of these environments. Isolating sick individuals or practicing social distancing is impossible in close quarters.

However, these precarious living conditions, and the people who inhabit them, have largely been hidden. For one thing, physical structures are often purposefully hidden because they are unlawful (for example, an illegal conversion of an unfinished basement; subdivision of a living room with a makeshift barrier to form an additional bedroom). Further, people living in overcrowded conditions are often themselves hidden due to immigration status, English-language proficiency, and poverty.

The pandemic is shining a light on these tenement-like conditions that many thought had vanished at the end of the 19th century. It is exposing an urgent need for both better crisis response and also for better data to understand how the nation’s housing crisis intersects with the pandemic to inform long-term policy solutions to address both.

A Potential Solution: Data-Driven Approaches

Cities increasingly have access to data that can be used to improve housing-related health. These datasets may come from police and fire departments, tax assessor’s information, code enforcement, utilities, the census, and other sources linked by address. Taken together, this data can identify high-risk properties for inspection and areas in need of services such as rental assistance, food, and information. Using city data, combined with the accumulated knowledge of housing inspectors and in partnership with community organizations, cities can detect patterns and trends, prioritize services, and evaluate the impact of initiatives to improve housing-related health. While direct data on overcrowding and other precarious living conditions is often not available due to the underground nature of the problem, existing data can still be used to make this problem visible.

Together with research assistants Nicolas Diaz Amigo and Ashley Marcoux, using housing code violation data from proactive inspections in Chelsea (an initiative to inspect every rental property in the city), we found that over half (54 percent) of homes inspected had at least one housing code violation. This finding is indicative of the risks many face while sheltering in place. Further, 29 percent of homes inspected had locks installed on internal room doors. According to housing inspectors in Chelsea, internal locks are an indicator of unrelated individuals living together and are a reliable marker of overcrowding. Homes with a housing code violation for internal room locks were significantly more likely to harbor more housing code violations compared with homes that didn’t have this (an average of 12 violations in homes with internal locks compared with six violations in homes without internal locks). These homes were also more likely to be older, on smaller land areas, and less likely to be owner-occupied.

Using data from across city departments, we applied machine learning to assign a risk score to each residential property in the city for the likelihood of overcrowding (as measured by internal room locks). More information on this work, including the code, is available on Github. Using a training and testing dataset of homes previously inspected, we were able to predict the internal locks violation with 73 percent accuracy. We optimized the model to identify as many of the truly overcrowded properties as possible and achieved a recall of 37 percent. The map below shows areas of the city with higher risk of residential crowding in red. Maps like these can be used to identify hotspots for COVID-19 risk and other public health concerns. We are developing similar maps for other high-risk-to-health housing code violations and working with Chelsea’s Inspectional Services Department to trial risk-based inspections (currently on hold due to the pandemic).

Heat map of Chelsea with greater density clustered in the southern halfOnce overcrowded living situations are identified, solutions must be put in place. In the context of COVID-19, where household transmission is thought to be responsible for the bulk of new cases, sick individuals in overcrowded homes need to be housed elsewhere to halt transmission. The City of Chelsea took this step by making hotel rooms available, free of charge, for people who test positive for COVID-19 and are not able to safely isolate at home. People who are not sick also need support to stay safely housed, including access to emergency assistance for food and hygiene products, which has also been underway in the city. Further, as many living arrangements are informal and aren’t protected by eviction moratoriums, access to emergency housing is critical to keep people off the streets and out of shelters. Knowing where services are most needed, and what other problems people may be facing, can help cities prioritize and align limited resources.

Conclusion: Carrying forward lessons from the past

Looking toward the future, let’s not miss the brutal lesson the pandemic is teaching us about the dangers of overcrowding and substandard housing. Past pandemics shaped the housing codes that are in place today. These codes originated over a century ago and mandated minimum conditions for habitability including adequate ventilation and sanitary conditions to reduce the spread of infectious disease. As a result, homes became safer and infectious disease rates plummeted.

Starting today, cities can again be agents of change by creatively leveraging housing policies and partnerships to address housing conditions afflicting people in the 21st century. For example, cities can update ordinances and formalize existing rooming-house-like conditions so that people can live more safely in high-density housing that is subject to minimum standards and inspections. Housing policy solutions must be pursued in combination with policies that support employment and other social services so that loss of jobs or other economic hardship doesn’t necessitate moving into dangerous conditions or sentence a family to homelessness. Analysis of city data can be of service to this goal to understand the nature, location, and magnitude of housing-related health problems and inform response activities and partnerships.

The pandemic has shined a bright light on a second, hidden epidemic: substandard housing conditions that impact millions of Americans in cities like Chelsea. Now is the time to bring novel, data-driven solutions to the problem of poor housing conditions that have plagued cities since the dawn of urbanization.

About the Author

Katharine Robb

Katharine Robb studies urban environmental health in the US and low-resource settings abroad. She holds a Doctor of Public Health (DrPH) degree from the Harvard T.H. Chan School of Public Health and a Masters in Global Environmental Health from Emory University. She is currently a Postdoctoral Research Fellow at the Harvard Kennedy School Ash Center for Democratic Governance and Innovation.

Email the Author