Results

ASCERTAINMENT

In the three year period there were over 470 000 births in the study area and 456 unexpected infant deaths, of which 363 were classified as SIDS. Subsequent checking with statutory death registrations has identified a further eight deaths, not included in the study, which met the entry criteria (ascertainment 98.3%). Of the 456 families, 30 refused an interview (93.4% consent rate), 29 other families were excluded from the analysis because of police involvement (suspected non-accidental injury), because they lived outside the study regions, or because they could not be traced.

Our study deals with 325 SIDS deaths (90% of the total SIDS deaths), 72 explained SUDI deaths (77% of the total explained SUDI deaths), and four age matched controls for each index case (1588 controls).

The major causes of death among those considered by the panel to be fully explained were infection (46%), accidental (15%), congenital anomalies (14%), and non-accidental injury (13%). The remaining deaths included metabolic disorders, bowel obstruction, bronchopulmonary dysplasia, and cardiomyopathy.

INFANTS WHO DIED OF SIDS

Clinical characteristics were divided into those identifiable at the following times: birth, after discharge from hospital, during life, and shortly before death or the reference sleep. Table 1 shows the univariate findings. The data are from medical records unless otherwise stated. The most common neonatal problems included those related to preterm delivery, respiratory problems, jaundice, intrapartum asphyxia, metabolic problems, and minor infections. Major congenital anomalies were defined as those that were potentially life threatening if not treated, or those that would lead to severe disability. The most common major congenital anomalies include positional deformities and malformations of both the genitourinary system and the cardiovascular system.

Infants who died of SIDS were significantly lighter for gestational age and had a shorter gestation. A small but significant proportion of SIDS cases were from multiple births. A greater proportion of infants who died of SIDS were unwell at or after birth (for example, needed resuscitating, were admitted to special care nursery, or had other neonatal problems). Although the rate of minor anomalies was not significantly different between cases and controls, significantly more of the infants who died of SIDS had major malformations, a third of which were cardiovascular. After discharge from hospital more infants who subsequently died of SIDS were noted by their parents to have problems with regurgitation, colic, poor weight gain, and other illnesses (the most common—found in over a third of both groups—being chest infection). Many of the infants on vitamin supplements were of short gestation, so the association of SIDS with vitamin supplementation became non-significant when controlled for gestational age.

During their short life, significantly more infants who died of SIDS experienced apparent life threatening events and more had hospital admissions. An apparent life threatening event was defined as an episode in which the infant became apparently lifeless. In the week before death, more infants who subsequently died of SIDS were diagnosed by health professionals as having mainly minor illnesses or were thought to be in poor health by their parents.

In the 24 hours preceding death, a small but significant proportion of these infants were thought by their parents to be more sweaty than usual and not feeding as well. Although most infants who died of SIDS appeared well in the final 24 hours, over a fifth scored high enough on the Baby Check to suggest they needed assessment or advice from a health care professional or continued monitoring by parents.

Table 2 shows the multivariate models obtained when these clinical characteristics are modelled together (model 1) and when background associations are taken into account (model 2).

Many of the clinical characteristics that were significant in the univariate analysis were not significant when modelled together, and of those that were, some became non-significant when controlled for other significant factors associated with SIDS. A history of an apparent life threatening event among SIDS cases remained significant in the multivariate analysis at the 5% level along with infants exhibiting birth weights below the 10th centile, multiple births, and infants with major congenital anomalies. Taking a more conservative level of significance of 1%, only prematurity and the Baby Check score remained as significant independent associations in the final model.

In the final 24 hours, significantly more of the infants who died of SIDS were showing signs and symptoms of illness to at least suggest they needed parental monitoring. This was the only variable in this time period to remain significant when controlled for other factors. Table 3 groups the scores according to the action that should be taken, as outlined previously.

Of those infants who scored more than 7, three of the 15 individual signs and symptoms occurred in a significantly greater proportion of infants who died of SIDS than controls: these included infants being drowsy most of the time when awake (14.9% v2.0% controls; odds ratio (OR) = 8.51; 95% confidence interval (CI), 1.70 to 81.56), infants wheezing (52%v 27% controls; OR = 3.00; 95% CI, 1.49 to 6.07), and infants taking less than half the normal amount of fluids in the last 24 hours (39.7% v 29.5% controls; OR = 2.57; 95% CI, 1.22 to 5.44).

These data suggest a gradient of risk for those infants who scored higher, which is reflected in the ratio of risk between the SIDS cases and controls and the unadjusted OR. However, when the age matching is taken into account the gradient of risk is, if anything, in the opposite direction, suggesting an interaction with age. The decreasing gradient is even more evident in the multivariate results. Because the median age of all infants was 100 days, table 4 examines the risk associated with the Baby Check after splitting the data into those above and below this age.

Although the numbers are small in each cell the data suggest that the pattern of risk is different for the two age groups. For infants less than 101 days old, the increased risk appears to be among those infants with scores of 8–12, whereas for infants more than 100 days old, the risk increased with the score. A similar multivariate analysis to the one described above (table 2, model 2) for the two different age groups showed that the Baby Check scores were only significant for the younger infants scoring between 8 and 12 (multivariate OR = 20.50; 95% CI, 4.03 to 104.30). If in the same age specific models we use a simple dichotomous variable representing the Baby Check score (healthyv a score > 7), this does not reach significance among the older infants (OR = 2.49; 95% CI, 0.73 to 8.52) but remains significant among the younger infants (OR = 10.77; 95% CI, 2.98 to 38.92).

INFANTS WHO DIED OF EXPLAINED SUDI

In the univariate analysis, all the clinical features and health markers that were significant in the SIDS cases were in the same direction in the explained SUDI cases. Table 5 shows the results of conducting a similar multivariate analysis (but excluding factors related to the sleeping environment) for the infants who died of explained SUDI.

Unlike the infants who died of SIDS, neonatal problems and hospital admissions were significantly more common among the explained SUDI cases in the multivariate analysis, although hospital admissions became non-significant when controlled for other factors. The significance of neonatal problems is perhaps not surprising because 14% of these deaths were later explained by congenital anomalies. In both SIDS and explained SUDI deaths episodes of apparent life threatening events and a high Baby Check score in the 24 hours preceding death/reference sleep were much more frequent than in the controls.

Table 6 shows Baby Check scores for explained SUDI cases in relation to controls.

There were relatively few cases in some of the categories, so although there appears to be a progressive and substantial rise in the risk ratio for death with increasing Baby Check scores, the confidence intervals of the age adjusted odds ratio are wide. Only those scores greater than or equal to 20 are significantly greater than unity.

SIDS CASES COMPARED WITH EXPLAINED SUDI CASES

In the multivariate analysis, a history of apparent life threatening disease or illness in the 24 hours preceding death measured by Baby Check remained significant for both groups of death. Figure 1shows the number of episodes of apparent life threatening disease in the two index groups and the control group.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Frequency of episodes of lifelessness.

Figure 2 compares the Baby Check scores between the two index groups and the controls. Of the control infants a small proportion (7.5%) had a Baby Check score higher than 7 in the 24 hours before the reference sleep; among the SIDS cases this proportion was three times higher (21.4%), and among the explained SUDI cases the proportion was more than six times higher than the controls (49.2%).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Frequency of Baby Check scores.

For those infants who scored between 8 and 12, 34% of the parents of infants who died of SIDS and 47% of control parents thought their infant's health was good in the last 24 hours; among the three explained SUDI cases, none of the parents thought the infant was well. Of those infants who scored higher than 12, 19% of both SIDS and control parents thought their infant was healthy, whereas 42% SIDS and 38% control parents did not contact a health professional, despite concerns about the infant's health. A much greater proportion of infants who subsequently died of explained SUDI were identified as more severely ill (score > 19) compared with the other two groups; consequently, of those infants who scored more than 12, only 11% of parents thought their infant was healthy and the same proportion failed to contact a health professional.

Discussion

Our data demonstrate poorer health and a higher frequency of symptoms among infants dying suddenly and unexpectedly when compared with controls, both during life and shortly before death. This was in addition to the range of disadvantageous clinical characteristics identified antenatally, at birth, and after discharge from hospital. The data emphasise the importance of apparent life threatening diseases in this regard, and highlight the potential of the Baby Check score in helping to prevent at least some of these deaths.

The validity of data on reported health and illness depends heavily on accurate recall, and comparisons might be affected by differential recall among cases and controls. How confident are we that our results were not the result of recall bias? First, we attempted to build up a confidential and intimate relationship with the parents, contacting them through their own health visitor, and utilising experienced health care professionals to interview the parents in their own homes and facilitate bereavement care. Second, we ensured that all data were collected very soon after the death (for cases) and the reference sleep (for controls). Third, we sought not just parental opinion of infant health at birth or during life, but specific details from the parents of any medical contacts or hospitalisation to verify the severity of illness. Fourth, to measure infant health in the 24 hours before death/interview we used the Baby Check, which requires specific signs and symptoms and should therefore be less prone to recall bias. We cannot completely exclude the possibility of some recall bias, but the willingness and honesty with which parents answered some of the more intimate questions, the consistency of the findings over a range of clinical characteristics, and the proportional similarity of medical contacts and hospitalisation between the cases and controls suggest the steps we have taken has minimised the impact.

A history of an apparent life threatening disease was associated with deaths in both index groups, independently of poor infant health in the 24 hours preceding the death, when controlling for many other factors in the multivariate analysis. This emphasises the importance of apparent life threatening diseases as a marker of risk. About one in nine infants dying of SIDS, and one in seven explained deaths, experienced an episode of lifelessness compared with around one in 30 control infants. Of those babies with an apparent life threatening disease, 66% of SIDS cases, 67% of explained SUDI cases, and 67% of controls were brought to medical attention, but relatively more of the index infants (47% SIDS, 44% explained SUDIv 33% controls) were taken to hospital. If the rate of apparent life threatening diseases in controls is representative of the general population, as we believe, it suggests that in a population of 250 000 with around 3000 deliveries a year there would be 100 infants a year experiencing an apparent life threatening diseases, of which 30 would arrive at hospital, but 30 would not be seen by a doctor at all.

Our results are consistent with our previous report, which highlighted that among the epidemiological characteristics only the age distribution, frequency of congenital malformations, and higher rates of smoking among mothers of SIDS cases were significantly different between infants who died of explained SUDI and those who died of SIDS.13 Several studies have looked at signs of illness preceding death and found specific symptoms among infants who died of SIDS, including respiratory problems, irritability, rash, change of cry, fever,19 gastrointestinal illness,19-22and maternal reports of infants becoming sleepy, listless, and droopy.18 19 Collectively, signs of illness before death appear to be more prevalent among infants who died of SIDS, although there appears to be no difference in major signs of illness10 or unreported severe illness.9 23Maternal descriptions of previous episodes of lifelessness (“held their breath”, “stopped breathing”) reported in the prospective Tasmanian study6 suggested an apparent life threatening event in the postnatal period. Similarly, events described by parents that were found to be significant in other studies include cyanotic attacks,24 convulsions,22 and tachypnoea,20 although this last finding disappears when controlled for gestational age and growth.

Our data are therefore consistent with other published work. However, the size of our study, which included virtually all the sudden deaths from a population of about half a million births, and the inclusion of deaths which, while unexpected, were subsequently explained, makes the findings more robust. We have been able to quantify, with much greater precision than previous studies, for both SIDS and explained SUDI, the characteristics of vulnerability arising in the neonatal period, and markers of the severity of illness in the 24 hours preceding death.

Although many of the epidemiological risk factors for SIDS (young maternal age, high parity, socioeconomic circumstance, etc) could be ascertained objectively from existing medical records, none was sufficiently specific to allow any particular intervention to be targeted at those infants most at risk of sudden death. In contrast, the Baby Check score is a means of quantifying acute illness, can be used by parents to help them decide whether to seek medical attention, and can be used by health professionals as a triage tool to discriminate more reliably between those babies who should be assessed in hospital and those who need not.15 It measures the severity of the present illness and therefore can trigger interventions that could be life saving. In our study, we were obviously unable to accrue a prospective Baby Check score on those infants destined to die and therefore relied on recall (with the attendant problem of possible bias). We also needed to modify the score to reflect the circumstances of the data collection because we could not collect all possible dimensions of the Baby Check. Because the maximum possible modified score was 96, rather than 111, our retrospective score would tend to underestimate the baby's severity of illness. Although we used parental recognition of febrile illness as a proxy measure for a temperature reading, recent studies suggest that touch might overestimate a fever but would not fail to recognise it.25On this measure, the score will therefore have overestimated the severity of illness for some infants, but the proportional difference between SIDS and control cases remained the same (Baby Check score > 7: fever included: 21.4% SIDS v 7.6% controls; fever excluded: 18.2% SIDS v5.9% controls). Given the modifications to Baby Check made for our study and the possibility of recall bias it is difficult to derive a measure of specificity from the control infants. Concurrent assessment of the Baby Check26 suggests that our retrospective scoring system might have slightly overestimated the degree of illness, although the socioeconomic status of the 104 Cambridge mothers used in this field study was higher than the population norm. An accurate assessment of both the sensitivity and specificity of the Baby Check can only be made with a large scale population based intervention study.

According to the retrospective Baby Check score, over a fifth of infants who died of SIDS and nearly a half of the infants who died of explained SUDI were identifiably unwell in the 24 hours preceding death. Most of the infants who died of SIDS, especially the younger ones, displayed minor signs and symptoms, whereas for the explained deaths the signs and symptoms were potentially more serious. For both groups the evidence suggests that for a small proportion of infants showing signs of being severely unwell, appropriate action by parents or health care professionals might have prevented the death. This finding has important implications. Some studies have shown that a proportion of explained sudden unexpected deaths of infants, particularly those related to infections or to trauma, might be preventable.17 21 27-29 It is well known from the series of CESDI reports that failure to recognise severity of illness is a recurring theme, and that this failure of recognition might not be only by the parents but by health professionals, who may give inappropriate reassurance that the baby is well. Although there are clear general messages for the necessity of better professional education and training, the use of the Baby Check score has the potential to guide less confident or experienced workers to “do the right thing”. Concerns that the widespread use of Baby Check might result in a large extra burden on general practice appear to be unfounded.30 31

We are concerned at the level of discordance between the Baby Check score and the parental recognition of ill health in their baby, and the fact that this discordance occurred quite commonly in the control group as well as the index cases. Although many parents are good at recognising illness in their baby, some are less able, and others might not realise the importance of potentially serious signs such as subcostal recession. Many disadvantaged parents, particularly if young and of low educational achievement, might not be able to communicate their concerns effectively to health care professionals in a way that will convey the possible seriousness of the infant's condition. Baby Check, which has been carefully validated for use by doctors and parents, might allow parents to refine their judgement of an infant's degree of illness. Thus, there is the potential both to prevent “unnecessary” visits to the doctor, and to give parents more confidence in presenting their child when it is appropriate. Parental access to the Baby Check could be universal if it was routinely included in the “personal child health record”.