What you didn’t hear about the latest study of sudden and unexpected infant deaths

News accounts of a new CDC study on infant deaths have led parents to fear that babies are dying from suffocation and strangulation in their beds in skyrocketing numbers. These spins are not what the study data actually found. Giving the public only half the story has not only accentuated anxiety among parents, it has piled on perceptions that today’s parents are incompetent and at risk of harming their children without interventions from public health professionals, even down to the most intimate details of home life and parenting choices.

Many news outlets took the lead of the Washington Post, which, in the very first sentence of its coverage, laid the blame for the “skyrocketing” deaths on parents who sleep with their babies. It then proceeded to weave partial quotes from the study’s lead author that appeared to support this interpretation of the study. Reactions have been understandably heated. Let’s take a deep breath and look at what the study actually reported, as well as learn some key pieces of information that can help us avoid coming away with misunderstandings with the next health statistic we hear.

This study was published in today’s issue of Pediatrics and led by Carrie K. Shapiro-Mendoza, Ph.D., MPH, with the National Center for Chronic Disease Prevention and Health Promotion at the CDC (Centers for Disease Control and Prevention). Those who read only the abstract could easily come away with misperceptions. The abstract reported that, according to U.S. infant mortality data, since 1984, infant deaths attributed to accidental suffocation and strangulation in bed have quadrupled, going from 2.8 to 12.5 deaths/100,000 live births in 2004. The most dramatic increase occurred from 1996, it said. The conclusion stated that this study supports the need for greater prevention efforts to help parents and caregivers provide safer sleep environments. But, as we know, abstracts often don’t give the full story or reflect what a study’s data actually found.

The very first thing to question with any health statistics is how things are defined and being measured. These two things proved to be the entire story.

SUID = SIDS + ASSB + UNK

This study looked at Sudden, Unexpected Infant Deaths (SUIDs) from the CDC’s database. SUIDs include all of the different "causes of deaths" recorded by investigators at the death scene and after an autopsy and medical history are taken, whenever a baby dies suddenly and unexpectedly from natural or unnatural causes. These can include: rare inborn errors of metabolism, electrocution, hypothermia or hyperthermia, drowning, carbon monoxide poisoning, and injuries or traumas that include shaken baby syndrome and falls.

This study focused on only three injury-related categories included within SUIDs:

● Accidental Suffocation and Strangulation in Bed (ASSB)

● Sudden Infant Death Syndrome (SIDS)

● Cause Unknown (UNK)

ASSB. According to the CDC SUID investigator manual: “Asphyxia or suffocation [ASSB] is caused by the inability to breathe [which] leads to a lack of oxygen... Asphyxia can be caused by choking, constriction of the chest or abdomen, strangulation, narrowing of airway passages (severe allergic reaction or reactive airway disorders), or the inhalation of toxic gases. Common objects that are involved with asphyxia or suffocation include plastic bags, soft pillows, and soft materials such as bedding or stuffed animals.” Typical causes of ASSB can also include overlaying or accidental suffocation on a shared sleep surface; wedging or entrapment between mattress and wall, bed frame or furniture; and neck compression.

While the percentage of deaths attributed to each of these three classifications have been juggled around over the past two decades, what no one has reported to the public was that the total number of sudden, unexplained infant deaths from these three categories has dropped by nearly half (42%). Between 1984 to 2004, total SUIDs in this country went from 0.16 percent to 0.09 percent of all live births.

No only are SUIDs extremely rare, but they’ve dramatically dropped over the past twenty years.

Statisticulations

Okay, now that the panic dial has hopefully been turned down, let’s look at what’s behind these statistics. As we’ve learned, it is easy to mislead and fool people with staticulations. The main problem in trying to make sense of the specific causes of infant death within these SUID death categories, is that the definitions of the different categories and how they’re reported has changed. In other words, the specific causes for death that investigators have checked off on the forms have changed, most notably because the 1991 definition for SIDS has been increasingly more rigidly enforced.

“SIDS is a diagnosis of exclusion and is only made after all other possible causes have been ruled out,” according to the SUID investigator training manual. But according to the stricter diagnosis, any SUID that hasn’t had a thorough investigation using the new SUID reporting system could no longer be classified as SIDS. So, those deaths are now required to be classified as being Cause Unknown. The new investigation system also meant death scene investigations were more detailed, leading to more accurate classifications. So, many deaths that might once have been simply called SIDS are now classified as ASSB.

By reporting only the rise in ASSB and unknown causes, the public was left to wrongly believe that more babies were dying.

Did you catch that the most dramatic increases for ASSBs occurred after 1996? There is a reason for that. Prior to 1996, there was no uniform protocol for conducting infant death scene investigations. In 1996, the CDC released its SUID Investigation Guidelines and Reporting Form*, a uniform protocol for conducting an infant death scene investigation.

Total SUIDs dropped most dramatically during the 1980s and early 1990s, then, since 1996 they've been fairly stable, at about 0.09% of live births. As the CDC study reported, the decline in SIDS seen after 1996 was offset by an increase in ASSB and Cause Unknown deaths, “suggesting that there has been a change in the way these Sudden Unexplained Infant Deaths [SUIDs] are classified and reported.” Similarly, it stated, the use of the CDC’s 1996 SUID investigation guidelines and new reporting form by more jurisdictions across the country, “may explain why deaths are being classified as ASSB more often.”

A major caution when interpreting national trends about specific causes for infant deaths, according to the CDC, is that “many SUID cases are not investigated, and when they are, cause-of-death data are not collected and reported consistently.” Inaccurate classifications for causes of death make it impossible to accurately sort out national trends.

“Autopsy findings alone often cannot differentiate between ASSB and SIDS because pathophysiological findings that can distinguish an ASSB death from a SIDS death are not always evident,” the CDC authors said. That means, the specific category a death is attributed to is a judgment call.

“The impressive decline in SIDS during the 1990s has been credited to national efforts promoting a supine sleep position for infants,” wrote the CDC authors. But, as they noted, this is another claim that cannot be credibly made from the available data:

Yet since 1998, safe-sleep prevention efforts may have had little effect in reducing total SUID deaths overall. Instead, the relatively stagnant total SUID rate together with the observation that declining SIDS rates are being offset by increasing ASSB and unknown cause rates, suggests that the way these deaths are classified has changed. This change in classification and reporting has been observed in recent studies.

This CDC study also involved a lot of calculations to estimate the proportion of deaths in each category, using mortality data from the Compressed Mortality File (accessed via CDC WONDER). This added another complication because the codes** for cause of death in the database changed beginning in 1999. As they explained:

Codes for cause of death were defined according to the International Classification of Diseases, Ninth Revision (ICD-9) for 1984–1998, and the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) for 1999–2004. The change from ICD-9 to ICD-10 in 1999 resulted in comparability problems for many causes of death, including those in our analysis (eg, both ASSB and SIDS were more likely to be selected as the underlying cause of death under ICD-10 than under ICD-9). These comparability problems result in discontinuities in the trends for these causes. We accounted for this comparability issue in our calculations of the annual percent change by creating a dummy variable in our regression models to control for the discontinuity caused by changing revisions.

In fact, looking closely at the CDC study, there is a lot of missing data, negating the ability to soundly support much of the claims and conclusions being made in the media. Specific characteristics associated with ASSB deaths were not even reported for any of the years, except for 2003-2004, based on reading the text of 931 death certificates. The authors stated that while this information provided some insights, it was limited by “incomplete textual information in nearly half of the reports and lack of knowledge about certifier reporting practices.” Among those breakdowns, no potential mechanism was listed on the death certificates for more than a third of the deaths (35.1%). Among the cases with a mechanism noted, suffocation attributed to overlaying or rolling on top of or against infant while sleeping was listed in 33.8% of the deaths — but only 27.5% of deaths were known to have occurred in a bed, 6.8% in a crib, 10% on a sofa or recliner. The location for more than half (54%) of deaths were unknown or could not be determined. Co-sleeping or bed sharing was reported in 51.2% of those cases, but 42.6% of cases didn’t record it at all or it couldn’t be determined. Only 2.6% of the babies who died had been laying face down or were in an obstructed position.

Whether or not any of these correlations mean these are the contributing causes, or that there is a trend of greater or fewer numbers of babies’ deaths associated with these risk factors, cannot be concluded from this single observational snapshot. Remember, too, a lot more deaths were being lumped into the ASSB category that hadn’t been before.

Nearly 7 times as many babies in the ASSB category during 2002-2004 died on Mondays as compared to Thursdays or Fridays — whether that means anything or is another statistical fluke isn’t known, either. As the authors cautioned: “because ours is a descriptive study, there is no comparison group, and we do not assess risk factors related to ASSB.”

The bottom line is that twenty years of infant deaths show that far fewer babies overall are dying from these accidental injuries. There is no evidence that parents today are being more careless or are less capable of caring for their babies, or that there is a major epidemic of babies being smothered to death.

While such perceptions of an epidemic might be used to support the need for more public health programs to oversee parents to ensure they’re providing “safer sleep environment,” as the authors noted, “public health officials must first know what constitutes a dangerous environment” and whether these unknown deaths are potentially preventable. They don’t know that.

© 2009 Sandy Szwarc

* Incidentally, this 1996 form proved cumbersome and not user-friendly, and national evaluations found that it was not consistently being used. So, the CDC revised it and released a new 8-page form on March 1, 2006, as well as launched a new, even more detailed, SUID registry database “to allow the CDC and its partners to monitor death certificates and trends and to identify risk factors.” This is our heads up that we’ll be hearing lots more about correlations gathered from this database.

** As readers will remember, the new ICD-10 codes are the attempts of the government and third-party payers to monitor the minutiae of health care, and have not been readily adopted yet. The new system includes 155,000 codes, including an increase in the number of diagnoses to 68,000 from the current 13,500. To document medical procedures, doctors will have to check the correct code from among 87,000 options, compared to the 3,000 in ICD-9. Instead of just five codes for a sprained ankle, there are now 45! Just two weeks ago, in response to complaints, the government Centers for Medicare and Medicaid Services, which maintains the codes for billing and diagnosis, extended the deadline for using ICD-10. It announced medical providers will now have until October 1, 2013 to comply.

This is another heads up for us to view health statistic trends we hear over upcoming years especially skeptically — expect all sorts of new “epidemics” as diagnoses are newly recorded.

Opening picture: from The Party Shoppe.