Death in Sin City: Analyzing the CDC’s mortality database

By Alex Richards and Marshall Allen
Las Vegas Sun

An ancient sage wrote that "death is the destiny of every man; the living should take this to heart."

In the United States, the Centers for Disease Control and Prevention's mortality database is the destiny of every man, woman and child, and journalists interested in health care should take this to heart.

The CDC compiles the comprehensive vital statistic on mortality – based on data collected by individual states – with information taken directly from a person's death certificate as filled out by a doctor, medical examiner or coroner. It has catalogued individual deaths by year, location, age, ethnicity, cause of death and more since the 1950s.

We were inspired to use this resource by Thomas Hargrove at Scripps Howard News Service, who has been using the database for years and gave a how-to presentation on the subject at AHCJ's 2007 conference.

For our story, we weren’t particularly interested in the leading causes of death – people die frequently from the same things no matter where they are the United States. And there are plenty of state and county reports that tally causes of death. We wanted to see if the causes of death in Nevada and Clark County, home to Las Vegas, were extremely high when compared to the national average. We write all the time about the unique ways people live in Sin City, and we were curious what the database would show about the way we die here in Las Vegas.

But it wasn’t quite that simple.

There are two options available to reporters who want to work with the CDC’s mortality database. The easier route is WONDER (Wide-ranging Online Data for Epidemiologic Research), a Web-based query system that combs through an abbreviated version of the annual data in which only critical fields are included, such as state, county, underlying cause and age. The site has a user-friendly point-and-click interface and you can export your results to a text file for further use in a spreadsheet or database manager. WONDER also provides age-adjusted rates, which is the crucial benchmark when trying to draw comparisons between counties or states.

While working on this project, we enlisted the expertise of epidemiologists at the state and University of Nevada, Las Vegas, to make sure we were on the right track. After some discussion, we decided to include causes where the county or state rate (adjusted for age) was 50 percent higher than the national average and killed at least 1 person in 100,000 between 1999 and 2004.

A reporter could put together a decent story using WONDER alone. But sometimes its general-use format precludes answering specific questions about individual communities. Richards has computer-assisted reporting expertise, so our project got a little more advanced.

To work with the multiple-cause mortality data files directly, there are a few prerequisites in the realm of software and statistics. Microsoft Access was the workhorse of the project, but it also was necessary to do some parts of the analysis in SPSS statistical analysis software – it was good for finding the medians of multiple grouped sets of numbers and finding confidence intervals for our rates of death.

The raw files are huge blocks of fixed-width ASCII text – roughly 2.5 million records and 1GB in size per year – freely available from the National Bureau of Economic Research, going back to 1959. The NBER also has detailed documentation and syntax files for popular statistics packages like SAS, SPSS and Stata, which build a framework of columns for the data before it’s imported. This can also be done manually in Access during the import process, but we don’t recommend it; there are more than 100 fields in the database.

It’s possible to work with the CDC database right away without spending the time to process the files. The National Institute for Computer-Assisted Reporting’s database library maintains the same data and makes it available specifically for journalists in ready-to-use dBase format, which can be imported or linked into most database managers and statistics programs.

The CDC’s National Bureau for Health Statistics also works closely with the U.S. Census Bureau to release timely population estimates based on race and age down to the county level. These are invaluable when trying to calculate rates for different causes of death, especially when examining a range of years or a specific age bracket. These estimate “vintages” are large files to work with, so WONDER can be used to obtain this information as well.

Something to keep in mind when looking at rates or per capita figures: there can be a dramatic difference between two groups, but that doesn’t mean it’s a statistically significant difference. It could have happened by chance. To remedy that, we found a standard error for each age-adjusted rate and then used that to calculate a 95 percent confidence interval. SPSS came in handy here to weed those causes out.

We had to make some decisions about how to group causes of death. From 1999 forward, they’re coded according to the World Health Organization’s standard – the 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10). There are about 5,000 codes, documenting everything from SARS to lightning strikes. We used the ICD-10 113 groups – a way that the NCHS commonly classifies related, specific causes – to get a better sense of how Las Vegas was unique.

We highly recommend staying humble when working with this data, and being cautious about drawing conclusions for readers. Both our stories stressed that the numbers have their limits. Most strikingly, they can tell what happened, but not why. And there can be variances based on human error and the way states and individual coroners classify causes of death.

Also, our expert epidemiologists saved us in at least one case where the data showed a surprisingly high death rate that turned out, we think, to be based on discrepancies nationwide in reporting.

In the end, we found Nevada:

• Has the 12th highest death rate among people younger than 65.

• Has a rate of deadly accidental poisoning and exposure to noxious substances, which includes drug overdoes, that’s almost twice the national rate.

• Has a rate from alcoholic liver disease 1.7 times the national rate.

• Has the highest suicide rate in the nation and a suicide rate among the elderly that’s nearly three times the national rate.

So are unhealthy people attracted to Las Vegas? Or, is there something about the city that causes some people’s lives to take a turn for the worse? It’s a chicken and egg conundrum that’s starkly illustrated by the CDC mortality database.