The methods used in the Indiana SIDS Report were simple, straight forward, and easy statistical techniques that anyone could duplicate with the right software, namely Word for Windows, Excel, and Epi-Info. And the only essential of those is Epi-Info. Another spreadsheet would do just as well for the tables. Harvard Graphics might do even better for the graphs than did Word.
Measuring the impact of the Back to Sleep campaign was done using 2x2 tables with statcalc in Epi-Info. Although it was repetitive and thus a little boring to be testing virtually every conceivable relationship, it was easy because Epi-Info readily provides p-values, relative risk measures, and confidence intervals for every 2x2 table done.
The report did have another purpose though, beyond measuring the impact of the Back to Sleep campaign. We wanted to create a statistical snapshot of the SIDS situation in Indiana. Doing so was even easier than testing for the campaign’s impact. It was where we started.
The easiest way to begin any statistical analysis is with graphs. Using the graph portion of Word for Windows, virtually every conceivable variable from the death certificates was graphed against SIDS deaths and/or death rates. Most of those graphs appear here. If cohort data had been available, it might have been helpful to graph SIDS rates against weight gain and/or prenatal care from the birth certificates, but in the absence of timely cohort data, that was not an option.
Two graphs which were not published were (1.) a pie chart of "Percentage of Indiana SIDS Deaths, by Type of Region, 1991-1995 Average," and (2.) a pie chart of "Indiana SIDS Deaths per 1000 Live Births, by Type of Region, 1991-1995 Average." The first was a simple effort showing a roughly equal split between urban and suburban areas for SIDS deaths, with 45% and 42% of the state’s SIDS deaths, respectively. Only 13% of the state’s SIDS deaths were in rural areas. But of course this is somewhat misleading since the actual number of live births varies greatly from one type of region to another. In an effort to standardize the data, the first pie chart morphed into the second showing urban areas to have 38% of Indiana’s SIDS deaths; suburban areas, 32%; and rural areas, 30%. The desire that all of the report be accessible to the general public made us decide not to include this chart, since it had the potential to confuse some people without adding anything to the understanding of the statistically knowledgeable.
The other type of analysis that was done but not published was to examine the annual SIDS rates by counties. With this it quickly became obvious that SIDS deaths are thankfully so rare that not even our largest counties had stable rates all years. As it happens, SIDS is so rare that even when counties were consolidated into regions, not one of Indiana’s 14 regions had a stable SIDS rate all five years. Of course when rates are not stable, the information they provide is not reliable and is likely to be misleading. Therefore, in order to get a reliable picture of the SIDS situation in Indiana it was necessary to use five-year averages. Even so, the vast majority of county rates are unstable and hence unreliable. (See the "Sudden Infant Death Syndrome (SIDS) in Indiana, by County, 1991-1995" Table.) For that reason, regional five-year averages also appear. (See the "Sudden Infant Death Syndrome (SIDS) in Indiana, by Region, 1991-1995" Table.) Again, SIDS deaths are thankfully so rare that 3 of the state’s 14 regions do not have a stable rate even for the five-year average.
The other question which may arise about methods is ‘why was 1994 excluded from consideration when testing the impact of the Back to Sleep campaign?’ Or alternatively ‘why was the break between "before" and "after" not made at 1992 when the original recommendation was made?’
Although the American Academy of Pediatrics made the original recommendation to place healthy infants to sleep on their backs in 1992, that recommendation did not show any impact in Indiana until 1994 when the National Institutes of Health launched a concerted Back to Sleep campaign. Nationally that campaign began in March. In Indiana, it began in August when public service announcements about the recommendations aired on cable TV throughout the state. The data suggest that those public service announcements had immediate impact. The number of SIDS deaths in 1994 was 111, noticeably below the 128 per year average of the previous five years but still well above the numbers for 1995 and 1996. This also meant that, strictly speaking, 1994 was neither a "before" year nor an "after" year, since some months fell in each category. For that reason it could not safely be included in either category.
Bruce Meyer