Months before the pandemic, the Boston Globe Spotlight Team chose to tackle a subject few want to think about: death, dying, and the limited options for most people to die in peace, especially those from marginalized groups.
They found that in Massachusetts, a state that boasts some of the world’s greatest hospitals, income and race are factors in not only where and how people die, but in how long they live. The difference can be 15 years. They found that the region’s nursing homes were ill-prepared to handle the pandemic, and uncovered factors responsible for staggering COVID-19 fatality rates there.
In a year saturated with bleak news, the stories in “Last Words” were among the most well-read in the Globe, lauded for their mixture of context and compassion, as well as their data-driven objectivity. They provoked heated discussion on a topic so many people prefer to avoid. Over five different days, the editorial page published numerous letters to the editor in response to the stories. The Globe’s comment board drew robust responses.
One reader thanked Globe reporters, saying, “It makes a big crack in the concrete wall of ignorance and denial in our society of the disparities and inequalities that our ‘culture’ has long tolerated.”
Government officials also responded. The series spurred state officials to release more data on the number of COVID cases and deaths at nursing homes, as they faced repeated pressure from the Globe to exhibit more transparency. Officials also publicly vowed to do better in overseeing safety at nursing homes, where Massachusetts had one of the highest death rates in the nation. One in seven long-term care residents died in the pandemic.
The state attorney general also began a probe into Medicaid nursing home discrimination, acting on information from the Globe’s undercover operation of bias against the poor at these facilities.
Overall, the powerful data analysis of the series and its deeply human stories served as a catalyst for discussions about death, dying and disparities, subjects so relevant in this pandemic year and beyond.
We used a variety of tools and techniques, including:
● Linear and multiple regression to identify factors correlated with higher mortality rates at nursing homes. (First, we had to painstakingly gather data from a variety of state and federal databases and merge them together.)
● Geocoding. We examined 21 years of detailed death certificate data to look for patterns in deaths (such as where people die by race and income) and identify deaths potentially related to COVID-19. But because the data did not include people’s income, we geocoded the data based on the residential address of the decedent, looked up the median income for that Census tract, and then classified the person into an income bucket we devised partly based on Pew Research Center definitions.
● Survey research. We commissioned Suffolk University to conduct a poll of Massachusetts residents and their views on deaths. We also surveyed the families of 450 people who died in 2016 by mail (using contact information from the state’s death data). The latter survey also included open-ended questions that gave reporters ideas for stories and names of potential interview subjects.
● Audits. To identify Medicaid discrimination at nursing homes, we mimicked “pair testing” approaches often used by investigators to identify potential discrimination. We sent two emails to hundreds of nursing homes expressing interest in potentially placing an elderly parent in the facility. One set of emails indicated the person would have to rely entirely on Medicaid. The other set of emails indicated the person was affluent enough to afford to pay the bill themselves..
● Content analysis. We conducted a content analysis of transcripts for the Massachusetts governor’s press conferences in the first month of the pandemic, finding that the governor and his staff rarely mentioned nursing homes at the briefings compared to hospitals.
What was the hardest part of this project?
We faced several challenges in putting together this series. First, we had to sue the Massachusetts Department of Public Health twice to obtain the death data that was the backbone of the series, spending more than $16,000 in legal fees. Second, the data from the state and federal government was often incomplete and missing unique identifiers to easily merge the databases. (We wound up painstakingly joining databases by the names of nursing homes and manually fixing errors where the names of nursing homes were listed differently in different databases). In some cases, the state also withheld the pricese number of deaths at nursing homes, forcing us to contact the nursing homes themselves to find out the actual numbers and fill in that missing information in our database.
In order to successfully survey 450 families who had someone die in the same year, we needed to send out more than 5,000 letters – since most people don’t respond to mail surveys. (The state death data doesn’t contain phone numbers or email addresses for family members – just the name and address of the person who provided information to the state about the deceased.)
Of course, one of the biggest challenges was the pandemic. We were getting close to publishing a series about disparities in death and discrimination in nursing homes when the coronavirus pandemic erupted in the state – forcing us to pause publication of the series and then re-report many part of it.
What can others learn from this project?
The series powerfully demonstrates the importance and utility of death data – and it’s worth it to even sue to get it. We obtained detailed data for 1.2 million deaths in Massachusetts, giving us incredibly detailed information for everyone who died in the state over more than two decades. That data could be used in any number of stories.
We also demonstrated how journalists can borrow some techniques from academia — including multiple regression, paired testing, and content analysis — to do their own studies. Multiple regression is a great technique to help figure out what factors are significantly related to deaths or something else reporters are studying. Paired testing is a useful way to to identify potential discrimination. Content analysis is a great way to measure what words or editorial choices people are making in their speeches, underscoring their priorities. We also showed two different ways of using scientific surveys – conducting your own or hiring a professional pollster. But the bottom line is journalists can take advantage of all these techniques themselves.
We think the series also showed how journalists can investigate a topic, then be forced to pivot when another major story breaks — like the pandemic. When that thappened to us, we didn’t panic — we actually saw an opportunity to incorporate new pandemc death data and see if our earlier findings about inequities still held up. They did and the result was a highly challenging and illuminating series to produce.