Infant Deaths

Article excerpt

Mapping project logs decade of mortality for poorest residents, polluted areas

On the map of infant death, Tuli Hughes lives in a red zone.

Within yards of her house in San Francisco's most impoverished corner, six babies have died in four years. Hughes has suffered five doomed pregnancies.

Yet the Fijian-American mother resides in a city that studies have ranked the lowest in the nation in infant mortality. Within a few miles of her home, survival rates for babies are among the best in the United States, but the infant mortality rate in her ZIP code compares with that in Bulgaria or Jamaica.

Twenty years after U.S. health officials vowed to solve the glaring health disparities leaving babies in certain neighborhoods at far greater risk of death than the general population, two of us at the San Francisco Chronicle decided to examine why they had failed.

My colleague Reynolds Holding and I started our project with a map logging 10 years of infant mortality rates in ZIP codes up and down California. That map wound up serving as a guide in a yearlong investigation that took us from the most impoverished corners of the San Francisco Bay Area to the pesticide-soaked fields of the Central Valley and then into the polished corridors of the booming neonatal intensive care industry.

The research resulted in a five-part series showing that while concentrating national efforts on medicine and technology, health care leaders have overlooked evidence that pollution and the stress of inner-city life may be a threat to many newborn babies.

It also illustrated how some of the medical marvels aiding the survival of the smallest and sickest babies - infants born as many as 16 weeks early - fail to reach more than 1,000 of the infants who die in California each year because of a breakdown in the state's health care delivery system.

Along the way, we got a primer in the mushrooming complexities of working with mortality data and the incredible amount of legwork needed to bring such statistics to life.

Statistical problems

The Chronicle's, series began with the typical paranoid musings of any new parent. As I lay awake at night listening to the breathing of my new baby, I wondered: How often do babies really die? And, are some babies much more likely to die than others?

These ponderings eventually led us to build a map charting where infant deaths were occurring in California. Building this map, first constructed with an old batch of California death and birth data sitting around the paper, turned out to be far more complex than it appeared. First, we used Microsoft Access to sort deaths of infants less than 1 year of age out of the huge database of California deaths and total them by ZIP code of residence.

But just knowing how many deaths were occurring for babies who lived in a given ZIP was of little use without knowing how many infants lived there. So, we had to use birth records to create similar ZIP-code-by-ZIP-code birth totals. With this, we could calculate infant mortality rates, or the number of infant deaths per thousand births, for each ZIP code.

These totals quickly highlighted a statistical problem that was to haunt us throughout the project: the difficulty of obtaining statistically significant results for areas with very small populations. Because we were looking at areas as small as ZIP codes, which often had only a couple of hundred births each year, it was easy for anomalies and annual fluctuations to skew the data.

Interviews and consultations with infant mortality researchers at Stanford University and around the country helped us develop a methodology for handling this problem. We decided to look at ZIP codes' infant mortality rates for an entire 10-year period and to eliminate any ZIP codes that were so sparsely populated they didn't have at least 1,000 births over a decade. (We would refine this standard even further before publication by performing statistical significance checks, but this was the methodology we used to build our initial map. …