The River Lea originates in the suburbs north of London, winding its way southward until it reaches the city’s East End, where it empties into the Thames near Greenwich and the Isle of Dogs. In the early 1700s, the river was connected to a network of canals that supported the growing dockyards and industrial plants in the area. By the next century, the Lea had become one of the most polluted waterways in all of Britain, deployed to flush out what used to be called the city’s “stink industries.”
In June 1866, a laborer named Hedges was living with his wife on the edge of the Lea, in a neighborhood called Bromley-by-Bow. Almost nothing is known today about Hedges and his wife other than the sad facts of their demise: On June 27 of that year, both of them died of cholera.
The deaths were not in themselves notable. Cholera had haunted London since its arrival in 1832, with waves of epidemics that could kill thousands in a matter of weeks. While the disease was on the decline in recent years, a handful of cholera deaths had been reported in the preceding weeks, and it was not unheard-of for two people sharing a home to die of the disease on the same day.
But the deaths of Mr. and Mrs. Hedges turned out to be the start of a much bigger outbreak. Within a few weeks, the working-class neighborhoods surrounding the Lea were suffering one of the worst cholera epidemics in London’s history. The newspapers delivered the same sort of morbid accounting that has obsessed us all in the age of the SARS-CoV-2 coronavirus: the terrifying upward trajectory of runaway growth. Twenty cholera deaths were reported in the East End the week ending July 14. The following week’s tally was 308. By August, the weekly death toll had reached almost a thousand. London had not experienced a major outbreak of cholera for 12 years. But by the second week of August, the evidence was unmistakable: The city was under siege.
Then, as now, the first line of defense was data. Londoners were able to track the march of cholera across the East End in close to real time, thanks primarily to the work of one man: a doctor and statistician named William Farr. For most of the Victorian era, Farr oversaw the collection of public-health statistics in England and Wales. You could say without exaggeration that the news environment that surrounds us now is one that William Farr invented: a world where the latest numbers tracking the spread of a virus — how many intubations today? What’s the growth rate in hospitalizations? — have become the single most important data stream available, rendering the old metrics of stock tickers or political polls mere afterthoughts.
In 1866, Farr had become a convert to a theory of cholera first proposed by the London doctor John Snow more than a decade before — the idea, which turned out to be true, that the disease was being transmitted in drinking water. And so as the deaths began to mount in the East End, Farr immediately began investigating the water sources in the neighborhood.
By the mid-1860s, a significant portion of working-class communities were receiving their water through private companies that ran the pipes to specific addresses, much as cable companies do today. Farr decided to sort the population that had died in the recent outbreak not by residence but by the company that supplied their drinking water. The data he assembled revealed a clear pattern: An overwhelming number of people who became ill drank out of East London Waterworks Company pipes.
The company claimed that their water had been effectively filtered at their new covered reservoirs. But investigators soon tracked down the source of contamination: The water in one East London company reservoir had not been properly isolated from the nearby River Lea. Looking through the mortality reports from earlier in the summer, the investigators discovered the deaths of Mr. and Mrs. Hedges, who lived near the reservoir. An examination of their residence revealed that their toilet was expelling waste directly into the river, thereby introducing cholera bacteria into the water supply and triggering the outbreak. It was a brilliant piece of detective work, carried out with remarkable speed and efficiency. And it turned out to be a momentous one: 1866 marked the last significant cholera outbreak in the history of London.
Farr was among the first to think systematically about how data on outbreaks, their distribution in space and over time, could be used to curb them as they unfolded — and to minimize future ones. The field he helped invent has come to be called epidemiology, but in its infancy it was known by another name: vital statistics. (“Vital” as in vita, Latin for life.) The innovations in this field do not look like our traditional model of medical breakthroughs: They are not packaged in the form of miracle drugs or new imaging technologies. At their core, they are simply new ways of counting, new ways of discerning patterns.
At this stage of the coronavirus pandemic, we find ourselves in a situation not all that different from the Victorians, despite the vast gulf in scientific, technological and medical expertise that separates us from them. We lack vaccines to protect the uninfected; no drug has yet emerged to cure Covid-19, the disease caused by the virus. Our main protection right now is the one that Farr began building almost two centuries ago: the collection and analysis of data. Data lets us see where the disease is spreading and where health care systems are likely to be overrun. It allows us to calculate infection rates and map hot spots down to the level of ZIP codes.
Eventually, medicine will protect us from SARS-CoV-2, but for the time being, vital statistics are the best defense we have. In the spirit of William Farr, multiple new experiments in data gathering and analysis have sprung up during the pandemic, experiments that might save thousands of lives before the crisis is over. And they may well prevent future pandemics from developing in the first place.
William Farr in 1865. It is only “when crystallized by the intellect,” he wrote, that facts “constitute the eternal truths of science.”
Ernest Edward/National Portrait Gallery, London
Born in 1807 into a rural family of little means, William Farr was a precocious learner who attracted the support of a wealthy patron and mentors as a teenager, apprenticing with a local physician before studying medicine in Paris and at University College London. By his mid-20s, Farr had established a medical practice in London. But his true passion was for vital statistics: He was an early member of the London Statistical Society and came to believe that understanding macropatterns in mortality could become a lifesaving tool as effective as any traditional medical intervention. In fact, given the sorry state of medicine in the 1830s, data was by far the more powerful instrument. The use of data to understand patterns of life and death had been almost exclusively a commercial interest during the 18th century, a science developed largely for the mercenary aims of insurance companies. But Farr and a handful of his peers saw the potential of vital statistics as a tool for reform, a means of diagnosing the ills of society and shining light on its inequalities.
After publishing a few papers in The Lancet analyzing medical data, Farr was hired in 1837 as a “compiler of abstracts” at the General Register Office, a new government body tasked with tracking births and deaths in England and Wales. At Farr’s encouragement, the G.R.O. began recording a much wider range of data in its mortality reports, including cause of death, occupation and age.
At the G.R.O., where he worked for nearly his entire career, Farr was responsible for taking raw data and making it meaningful: discovering interesting trends in the numbers, comparing health outcomes for different subgroups in the population, inventing new forms of visualization. His statistical inquiries would at times take him to some disturbing positions. He spent years developing a bizarre theory about the connection between topographic elevation and disease, which led to some xenophobic ideas about the inferiority of lowland peoples. But the enduring legacy of Farr’s vital statistics turned out to be an egalitarian one: exposing inequalities of health outcomes, using scientific thinking to dispel the longstanding assumption, prevalent among the ruling class, of a causal link between disease and moral turpitude in low-income communities.
Counting the dead itself was not a new technique: London parish clerks had been publishing weekly “bills of mortality” since the Elizabethan era. But Farr devised new ways to make that information useful. Collecting and publishing data was not merely a matter of reporting the facts but instead a more subtle, exploratory art: testing and challenging hypotheses, building explanatory models. As Farr wrote in an essay published the year he joined the G.R.O., “Facts, however numerous, do not constitute a science. Like innumerable grains of sand on the seashore, single facts appear isolated, useless, shapeless; it is only when compared, when arranged in their natural relations, when crystallized by the intellect, that they constitute the eternal truths of science.”
The first question that Farr used statistics to answer is relevant to our present crisis as well: To what extent did urban density contribute to the death rate? Perhaps because of his own life journey — growing up in the agricultural region Shropshire, now living in the largest city on the planet — Farr decided to devote one of his first studies to the differences in health outcomes between the country and the city.
Farr was a pioneer not just in collecting data but also in devising ingenious new ways of representing it. One way of measuring the health of a society is what was called in Farr’s time a “life table”: breaking down the death rate in a given population by age. (Life tables are what allowed us to see that Covid-19’s lethality has been disproportionately concentrated among the elderly, unlike the flu pandemic of 1918, which killed an unusual number of young adults.) In one early report, Farr experimented with an ingenious way of representing those different outcomes, drawing upon data collected from three separate communities: metropolitan London, industrial Liverpool and rural Surrey. It was, in effect, a tale of two cities — and one countryside. Viewed as a triptych, the illustrations conveyed a clear message: Density was destiny.
In Surrey, the increase of mortality after birth was a gentle slope upward, like a dune rising above a waterline. The spike in the cities, by comparison, looked more like the cliffs of Dover. That steep ascent condensed thousands of individual tragedies into one vivid and scandalous image: In Liverpool, more than half of all children born were dead before their 15th birthday.
Despite those grim numbers, Farr remained hopeful that the health crisis emerging in the industrial cities could be ameliorated. “Is the excessive mortality of cities inevitable?” Farr wrote in the 1840 annual report of the G.R.O. “The first writers who established satisfactorily the high mortality of cities took a gloomy and perhaps fanatical view of the question. Cities were declared vortices of vice, misery, disease and death; they were proclaimed ‘the graves of mankind.’” And yet, he continued, “there is reason to believe that the aggregation of mankind in towns is not inevitably disastrous.”
In that same report, Farr turned his attention to another puzzling pattern in the data he had collected: what he called the laws of action of epidemics, now known to epidemiologists as Farr’s Law. Analyzing a smallpox outbreak in Liverpool, Farr divided the mortality counts into 10 separate periods. “The mortality increased up to the fourth registered period; the deaths in the first were 2,513, in the second 3,289, in the third 4,242; and it will be perceived at a glance that these numbers increased very nearly at the rate of 30 percent.” But the rate of increase, he observed, “only rises to 6 percent in the next, where it remains stationary, like a projectile at the summit of the curve which it is destined to describe.” Farr’s Law was the first attempt to describe the rise and fall of contagious diseases mathematically. All the models that have shaped so much private angst and public scrutiny — the Imperial College London models that steered Prime Minister Boris Johnson away from the initial strategy of herd immunity, the University of Washington Covid-19 projections that have heavily influenced the Trump White House — all these forecasts are descendants of the laws of action that Farr originally sketched out in 1840. When we talk about flattening the curve, the curve in question was first drawn by William Farr.
A “life table” illustration by Farr, published in “The Fifth Annual Report of the Registrar-General of Births, Deaths and Marriages in England,” 1843.
Victorian scientists would have immediately recognized many of the core categories of data assembled by epidemiologists working on Covid-19: infections, deaths, locations and so on. Today’s vital statisticians obviously have access to a wider pool of information — antibody-test results, comorbidities of victims, even different genetic strains of the virus — than Farr was able to assemble. And they have software that allows them to build models that project the epidemiological curve that Farr first identified.
But the coronavirus pandemic has also revealed some crucial holes in the way we collect data during an emerging outbreak. As unlikely as it might sound, given the existence of organizations like the C.D.C. or the W.H.O., in the early days of the coronavirus’s spread, no single data repository existed where information about all the known cases could be accessed and analyzed by public-health officials and researchers. “There really has never been a successful effort to share comprehensive open data sources during any of the modern epidemics,” says Samuel V. Scarpino, who runs the Emergent Epidemics Lab at Northeastern University. “The vast majority of public-health data during epidemics are still largely organized on pen, paper, Excel and PDFs.”
Scarpino was one of a handful of volunteers, including the Oxford research fellow Moritz Kraemer and a Ph.D. student at Tsinghua University in Beijing named Bo Xu, who formed an ad hoc organization in late January to create a 21st-century equivalent of Farr’s mortality reports: a single open-source archive of every recorded Covid-19 case anywhere in the world. By early February, the Open Covid-19 Data Working Group had assembled detailed records for 10,000 cases. Today an informal network of hundreds of volunteers has assembled records for more than a million cases in 142 countries around the world. It may well be the single most accurate portrait of the virus’s spread through the human population in existence.
Of course, the greatest value in that kind of data set lies in the clues it can give us about the future path of the disease and how that path can potentially be interrupted. But again, the work of building those models has entirely taken the form of impromptu efforts organized at a handful of academic institutions around the world. The Johns Hopkins University epidemiologist Caitlin Rivers argues that the coronavirus pandemic has made it clear that one crucial innovation we need is a new kind of institution, what Rivers called a “center for epidemic forecasting.” Rivers draws an analogy to institutions like the National Weather Service. “There were a few big storms at the turn of the century with terrible loss of life and also enormous economic consequence, so there was interest at the time in figuring how to predict the weather,” Rivers explains. With meaningful investment, Rivers believes, “we can get to the place where we are with the Weather Service, where we have reliable forecasts that inform our everyday lives as the public, and also help decision makers to understand how best to respond to these outbreaks.”
Forecasts are only as good as the underlying data that support them, and in the case of disease outbreaks, most of the data collection — even in comprehensive archives like the one assembled by the Open Covid-19 Data group — suffers from a crucial liability: The information is captured too late. Numbers like hospitalizations or deaths are vital statistics to be sure, but they are tracking the end stages in the path of a disease. In the case of Covid-19, by the time the average person makes it to the hospital, around 10 days have passed since their initial contact with the virus. “Public-health reporting is usually very late,” says the epidemiologist Larry Brilliant, who helped eradicate smallpox in the 1970s. “It’s just shortly before the peak of an outbreak, historically, because as people get more alarmed, they go to their doctor, and their doctor goes to the public-health official and they report it.”
With a disease like Covid-19, where presymptomatic and asymptomatic carriers are capable of spreading the virus, the lag in reporting can make the difference between a runaway outbreak and effective containment. A typical case of Covid-19 that ends in a death follows this timeline, which can stretch to 30 days or more:
Infection -> Incubation -> Presymptomatic spread -> Symptoms and spread -> Doctor’s visit -> Hospitalization -> Intensive care -> Death
In the standard regime, even in the best-case scenario, data collection doesn’t begin until Day 10, during the doctor’s visit. Covid-19 has prompted an inspiring scramble of experiments designed to move the data gathering earlier on the timeline. Some of them involve what is called “sentinel surveillance” — widespread, early-stage testing in critical populations that may be at risk. “There’s testing for the individual who needs to understand do they have this disease, do they need to isolate or seek care,” says Lorna Thorpe, director of the epidemiology division at New York University’s medical school. “But to manage the outbreak, you need to know where it is, you need to be ahead of it.” Much like the outbreak of 1866, Covid-19 has hit hardest in low-income communities, which generally have reduced access to the health care system, where most data is collected. “Oftentimes the communities that need our attention during the outbreak, that are most likely to be hit early, may also be the ones that we understand the least about,” Scarpino says.
In part because of the limited supply of tests, the first few months of data about Covid-19 were almost entirely oriented toward people experiencing severe symptoms, who would show up at hospitals. But a sentinel-surveillance program could have targeted communities — like nursing homes or low-income neighborhoods — that had not yet experienced symptomatic infections, potentially detecting those outbreaks before they became unstoppable. Thorpe points to the success of the Seattle Flu Study, an initiative that began in 2019, which set up testing kiosks, analyzed samples from hospitals and distributed home nasal swabs to a broad section of the city’s population, asking them to send in samples if they developed symptoms of respiratory infection. Tellingly, the program was the first to detect community transmission of SARS-CoV-2 in the United States.
The Seattle Flu Study was a variation on another emerging technique that has already played an important role in the fight against Covid-19: “syndromic surveillance.” The idea is simple: Supplement the official data from patients entering the health care system with data tracking the appearances of disease symptoms before they get to a doctor or a hospital. One influential early project that drew on this approach was a program called Google Flu Trends, introduced in 2008 as a collaboration between Google and the C.D.C. The service didn’t track symptoms directly but instead analyzed patterns in Google search queries associated with influenza: “My child has a fever,” say, or “aches and pains.” By mapping those queries geographically, the service aimed to identify influenza hot spots days or weeks before they showed up on the radar of the C.D.C. Then, in 2011, an epidemiologist at Boston Children’s Hospital named John Brownstein helped create a website called Flu Near You that relied on user-submitted data tracking fever and other flu symptoms directly through a small but statistically representative group of volunteers. In the early days of the SARS-CoV-2 outbreak, Brownstein spun off a new version called Covid Near You. “Most people with Covid have mild illness and are unlikely to interact with any health system,” Brownstein says. “Data from self-reported symptoms can help fill in gaps, especially in light of limited testing.” A visitor to the site answers a few simple questions: What’s your ZIP code? How are you feeling? If you’re not feeling well, what are your symptoms? The data collected allows the service to map emerging hot spots before they show up in the clinics or in the official county health reports, effectively shifting the data-collection timeline five days to the left. In late March, when much of the focus was on the explosion of cases in New York City, Covid Near You was already picking up a surge in Covid-19 symptoms in less densely settled areas. “Despite the urban hot spots,” Brownstein says, they saw signs of outbreaks in rural communities, “especially in locations where people may have second homes.”
New technology has also made syndromic surveillance more feasible. The San Francisco-based start-up Kinsa has been selling an internet-connected thermometer since 2014. According to Inder Singh, Kinsa’s chief executive and founder, who formerly oversaw the Clinton Foundation’s work on infectious disease, the original vision was for the company to detect these early patterns of illness without forcing people to change their usual routines. “The idea was: Let’s take an existing behavior, the only thing that people do in the home when illness strikes,” Singh explains. “They grab the thermometer.” From the consumer’s point of view, the interaction with Kinsa’s thermometer is straightforward enough, but behind the scenes the device sends anonymous, geolocated information about the results to Kinsa’s servers. That new data stream enables the company to maintain what it calls health weather maps for the entire country, with real-time data on atypical fevers reported down to the level of individual counties.
Starting on March 4, 2020, Kinsa’s charts began tracking a statistically meaningful increase in the number of fevers in New York, 19 days before the city went into a full lockdown. (The first case in the city was reported on March 1.) By March 10, the number of people registering an elevated temperature in Brooklyn was 50 percent higher than normal, suggesting that the virus was already rampant throughout the five boroughs, even though the official case load was still less than 200.
One limitation of our current data has to do with geography rather than time. As Marc Gourevitch, chair of the department of population health at N.Y.U.’s medical school, observes, most of our tools for mapping outbreaks aren’t granular enough. “In many cities and urban neighborhoods,” Gourevitch says, “there can be great variation within a couple of blocks, or a fraction of a mile, in terms of the conditions that really drive health. So if you want to look at variations in health and risk and outcomes, you need to take a granular view of the geography that you’re talking about if you want to be able to think about strategies of protecting at these small scales. It’s really the scale at which health is fundamentally determined: whether it’s by crowding, access to good schools, to air quality — all kinds of drivers that vary on a small scale.” By default, our health care data is generally organized geographically by county. But in a city like New York, where a single county contains millions of people, that scale is all wrong for tracking a fast-moving virus.
In many cases, that wide-angle view has been established deliberately as a privacy protection. A few years ago, Gourevitch helped organize an online resource called City Health Dashboard, which presents community life-expectancy averages by census tract, showcasing the broad inequalities in health outcomes in communities living just a few blocks from one another. But even that resource was controversial. “It took years and pressure to get state authorities and the C.D.C. to contribute to the estimates of life expectancy at the census-tract level,” Gourevitch says. “That was a multiyear effort because of legitimate concerns about the privacy issue.”
One potential solution that Gourevitch sees is a kind of geographic blurring for outbreak data. In John Snow’s famous map of the 1854 cholera outbreak — the one that ultimately led to the understanding that the disease was caused by contaminated water — he documented deaths at the level of individual street addresses, revealing a cluster of deaths around a widely used drinking well. But in the middle of an outbreak like Covid-19, you don’t need to be zoomed in that far to get a meaningful sense of where the outbreak is spreading. Instead of a pushpin on the map denoting an infection at a specific address, Gourevitch suggests deliberately making the targeting less precise: perhaps a city block, not a specific address. That level of granularity would be tight enough to detect the spread of the outbreak through microcommunities in the city, but not so tight that individual identities can be discerned in public data.
A Farr visualization charting temperature and mortality rates in London, printed in “Report on the Mortality of Cholera in England, 1848-49.”
British Library/Science Photo Library
While all these forms of disease surveillance offer improvements on the basic model that Farr and Snow helped invent in the middle of the 19th century, they share one key characteristic: They are based on data assembled from human beings, as they pass either through the health care system or through some self-reporting mechanism. Shifting the timeline even further to the left may require new sources of data that are not anchored in individual cases.
In the early 1990s, a Dutch microbiologist named Gertjan Medema was conducting experiments with triathletes racing in the Rhine delta. Medema and his colleagues were interested in the health impact of open-river swimming, and so as part of their experiment they collected river water, which they subsequently analyzed for the presence of a whole host of pathogens: bacteria, fecal pathogens, enteric viruses and other dangerous microbes. In those days, testing a sample for the presence of these organisms took weeks. While Medema and his team were still waiting for their results, the news broke of an unusual outbreak of polio in Streefkerk, a town six miles downriver from the testing site. Medema analyzed the river water they collected three weeks before and discovered that poliovirus was clearly detectable in the samples. The river held clues pointing to an emerging outbreak weeks before the health authorities did. “It was a lucky coincidence — if ‘luck’ is the right word to use,” Medema says now.
Back in 1992, those clues were effectively useless from a public-health perspective because they took too long to decipher. But today’s tools allow scientists like Medema to detect a virus based on its precise genetic sequence in a matter of hours. Because many dangerous pathogens are expelled in human waste, sewage samples are the most direct way of surveying viral or bacterial activity in a given community — short of testing people directly. “When I saw SARS-CoV-2 hit in China,” Medema says, “I was looking for reports of fecal shedding of the virus.” Before long, evidence began circulating that some sufferers of Covid-19 experienced diarrhea. “And that’s when I said, it may be that this virus will come to our country, so we’d better prepare sewage surveillance. Not because we think it’s an important risk for transmission, but because you could use sewage to monitor the circulation of the virus in the population.”
On Feb. 6, Medema and his colleagues gathered sewage samples from six points in the Netherlands, including a waste-treatment plant near Amsterdam’s Schiphol Airport, on the premise that the virus could potentially first arrive via air travel. The results came back negative. But a month later, when the outbreak was still in its earliest stages in the Netherlands, they returned to the same locations to collect samples. This time, they found evidence of the virus in several of the locations. “If we compare our prior sewage reporting with the number of reported cases,” Medema says, “it looks likely we can pick up the signal of the virus if we are at about one in 100,000 people reported infected.” (A preliminary study of a sewage-treatment plant in New Haven, Conn., this spring showed that presence of the virus in wastewater peaked seven days before reported Covid-19 cases.) In Farr’s era, sewage was a primary cause of epidemics. But in the 21st century, sewage might well offer us important data to contain their spread.
Not all pathogens are expelled in human excrement, which means that Medema’s approach has some limitations as a defense against future outbreaks. But sewage surveillance has a critical advantage over syndromic surveillance with a virus like SARS-CoV-2, which has an unusually high concentration of carriers who show no symptoms whatsoever. “The difficulty for this kind of virus is that containment doesn’t work because there is a lot of silent transmission,” Medema says. “But we can use sewage surveillance with these sorts of viruses — to pick them up and understand the virus circulation better. There’s a projection that we may see waves and waves of this virus. Maybe sewage surveillance can be an early warning to see if there’s another wave coming.”
The most radical technique for shifting the data-collection timeline to the left — but the one that might offer the most significant protection against future epidemics — involves cutting people out of the equation altogether. The underlying data that allowed William Farr to draw the first epidemic curve back in 1840 was, understandably, limited to patterns of life and death in the human population. Syndromic or sewage surveillance allows us to pick up signals earlier in the cycle by detecting symptoms or fecal shedding before people make contact with the health system. But for many of the most terrifying diseases that have emerged in the past few decades, the initial human cases showed up in the middle of a much longer timeline. “Covid, SARS, MERS, swine flu, bird flu, Ebola, H.I.V., Zika all were at one point animal diseases,” Larry Brilliant says. “Instead of syndromic surveillance, going two steps to the left is surveillance of animal diseases. You move it where it belongs into the realm of the zoonotic diseases, about 50 of which have jumped species from animals to humans in the last three decades.”
The promise of applying Farr’s vital statistics to the realm of animal diseases is a simple one: You can stop an emerging zoonotic disease before it makes the jump from animal to human. Animal surveillance could ward off the potential pandemic that experts have historically worried about the most: an influenza outbreak along the lines of the 1918 avian flu. “When you have 20 of your chickens die off, and your whole livelihood depends upon them,” Brilliant says, “if you have a hotline as they do in Cambodia, you can call the government and say ‘I have 20 dead chickens,’ and they’ll come and bring you 30 live ones and clean up your place. That’s a phenomenal bi-direction system that cleans up the virus for you, puts you back into business, and the epidemic is aborted. Being able to survey bats, pigs, birds — that’s going way beyond syndromic surveillance. That’s what we’re going to have to do in the age of pandemics.”
Public-health data began with that most elemental form of accounting: how many people died on this day in this place. The insights that arose from the collection of that information helped turn cities from the “graves of mankind” into communities that today enjoy some of the longest life expectancies anywhere on the planet. But during an epidemic, from the perspective of vital statistics, a human death tells the story of an infection that happened in the past. A hundred dead chickens, on the other hand, could tell the story of a future infection — and maybe even stop it from emerging at all.
Steven Johnson is the author of twelve books, including his account of the 1854 cholera epidemic, “The Ghost Map,” and most recently, “Enemy of All Mankind: A True Story of Piracy, Power, and History’s First Global Manhunt.”