ORIGINAL ARTICLES

Unnatural sudden infant death

Roy Meadow

AbstractAimTo identify features to help paedia-tricians diVerentiate between natural andunnatural infant deaths.MethodClinical features of 81 childrenjudged by criminal and family courts tohave been killed by their parents werestudied. Health and social service records,court documents, and records from meet-ings with parents, relatives, and socialworkers were studied.ResultsInitially, 42 children had beencertified as dying from sudden infantdeath syndrome (SIDS), and 29 were givenanother cause of natural death. In 24families, more than one child died; 58 diedbefore the age of 6 months and most diedin the afternoon or evening. Seventy percent had experienced unexplained ill-nesses; over half were admitted to hospitalwithin the previous month, and 15 hadbeen discharged within 24 hours of death.The mother, father, or both were responsi-ble for death in 43, five, and two families,respectively. Most homes weredisadvantagedno regular income, re-ceiving income supportand motherssmoked. Half the perpetrators had ahistory of somatising or factitious disor-der. Death was usually by smothering and43% of children had bruises, petechiae, orblood on the face.ConclusionsAlthough certain featuresare indicative of unnatural infant death,some are also associated with SIDS.Despite the recent reduction in numbersof infants dying suddenly, inadequacies inthe assessment of their deaths exist. Untila thorough postmortem examination iscombined with evaluation of the historyand circumstances of death by an experi-enced paediatrician, most cases of covertfatal abuse will go undetected. The termSIDS requires revision or abandonment.(Arch Dis Child 1999;80:714)

Keywords: child abuse; unnatural death; smothering;sudden infant death syndrome

Infants who die suddenly and unexpectedly,and in whom a thorough postmortem examina-tion reveals no cause, are placed in the categoryof sudden infant death syndrome (SIDS). It isrecognised that among those infants there are

some who have died as a result of direct harmfrom their parents.13 Opinions about the size ofthat proportion have varied greatly.46 Thenumber of infants dying from unidentifiablenatural causes has diminished as a result of thefollowing: (1) methods for the prevention ofsevere illnesses in infancy have improved; (2)parents and doctors now recognise ill infantsearlier; and (3) diagnostic tests and postmor-tem procedures have also improved. Unlessunnatural deaths diminish at a similar rate, theproportion of babies categorised as SIDS whohave died unnatural deaths will increase andbecome more important to recognise.

Smothering is thought to be the commonestcovert reason for such unnatural sudden infantdeaths because, unlike most other forms ofphysical abuse, fatal smothering can occurwithout external signs on the face or body, andwithout incriminating evidence at postmortemexamination. Moreover, there have been sev-eral studies indicating that smothering of chil-dren by parents is not particularly rarealthough fortunately most children abused inthis way do not die.712

This study concerns the clinical features of81 young children who, although thoughtinitially to have died of natural causes,subsequently were deemed by either criminalor family courts to have been killed by theirparents.

The aim was to identify features that mighthelp paediatricians diVerentiate between natu-ral and unnatural infant deaths.

MethodsThe cases were from the past 18 years, and theinformation was assembled from my contem-poraneous notes about the children and theirfamilies, which were made when they wereencountered. Apart from a small number oflocal cases, I was involved at the request ofmedical colleagues, social services, or thepolice from diVerent areas of the UK. Duringthe same period I was requested to be involvedin a larger number of similar cases, but refusedinvolvement because of work commitments.The series of cases reported is not thought todiVer qualitatively from other cases notified tome with which I was not involved. During the18 year period I was involved with over 200other families in which children were adjudgedto have been abused, but in which no childdied. At the time of my involvement it wasusual for me to see all the documented material

Arch Dis Child 1999;80:714 7

Department ofPaediatrics and ChildHealth, St JamessUniversity Hospital,Leeds LS9 7TF, UKR Meadow

Correspondence to:Professor Meadow.

Accepted 3 September 1998

including general practitioner, hospital, healthvisitor, and social service records relating to thechild, the siblings, and the parents. In thecriminal cases, and one third of the familycourt cases, information was supplemented bytranscripts of police interviews and investiga-tions. In most cases I met separately, andjointly, the childs mother and father and keysocial workers and, sometimes, other relatives.For some of the earlier cases, a record had notbeen kept about every feature and, therefore,there is not reliable information relating to eachof the studied features for each of the 81children, or for each of the 50 families. Thenumber of children, or families, with reliablerecords is noted in the relevant part of theResults section.

The causes of death of the 81 children from50 families were considered in detail duringcourt proceedings, which concluded that aparent had been responsible for the childsdeath (table 1). In 19 families, the case washeard in a criminal court, where a jurydelivered its verdict beyond all reasonabledoubt on the evidence presented to it by thediVerent parties. For 41 families, there werehearings in the family court where a judge,after considering the detailed evidence, gavejudgment based on a high degree of probability.(Ten of these cases were also proved in a crimi-nal court, the other 31 were not accompaniedby criminal proceedings.) In most cases, thefamily court proceedings took place because ofabuse of a subsequent child (particularly bysmothering), or a subsequent child being bornto a mother who was found to have had previ-ous infants who had died. In those cases, thetask of the family court, in considering the wel-fare of the subsequent child, was to decide onmatters of fact relating to the nature of thedeaths of the previous children. In 19 of the 50families, a parent confessed to harming or kill-ing the child by smothering or choking. Theconfessions occurred five months to eight yearsafter the deaths, usually in the context of courtproceedings.

ResultsThe 81 children came from 50 families. In 24families in which two or more infants died, thedeaths occurred sequentially to successiveinfants (two deaths in 18 families; three deathsin five families; and four deaths in one family).There were 46 boys and 35 girls. All except oneof the children (a step child) were the naturalchildren of the perpetrator. The age at deathranged from 1 week to 16 months and the dis-tribution of ages is shown in fig 1. In addition,there were four older children who had diedaged 112, 212, 4, and 512 years who were sib-lings of younger dead children. Those older

children had been believed to have died naturaldeaths, but were subsequently deemed to havedied unnaturally. The four older children areexcluded from those parts of the results thatdeal with the characteristics of infant deaths,although the five children aged 1215 monthsare included, because they had been catego-rised initially as SIDS.

BIRTH ORDER (50 PARENTS)Most of the dead children were born to moth-ers who had not had a previous live, healthychild. Thus, for 36 parent perpetrators therewas no living previous child, and for the 14parents who had a previous child living, 12 oftheir 17 previous children, in retrospect, wereconsidered to have been abused.

PERINATAL HISTORY (73 HISTORIES)PregnancyThirty eight pregnancies were considered to benormal. The other 35 included significantproblems, particularly alleged antepartumbleeding and abdominal pains, necessitatingadmission to hospital.

GestationThe length of gestation was < 30 weeks forseven babies, between 30 and 36 weeks for 10babies, and > 36 weeks for 56 babies.

Birth weightTwenty babies weighed less than 2500 g, 13weighed 25003000 g, and 40 had a Birthweight of > 3000 g.

Neonatal careNine infants had major neonatal problemsrequiring expert management in neonatalintensive care units. One baby had residual res-piratory problems attributed to bronchopul-monary dysplasia. The others were consideredhealthy.

MEDICAL HISTORY

Previous events (75 infants)Most of the children had had previous unusualor unexplained events reported by the perpe-trator. By far the most common reports were ofinfants stopping breathing, looking blue, ap-pearing dazed, twitching, or fitting. An acutelife threatening event was deemed to have

Table 1 Proof of parental responsibility for childrensdeaths in 50 families

Proof Number

Criminal Court alone 9Criminal Court and Family Court 10Family Court alone 31

In 19 of the families, the parent confessed.

Figure 1 The age (month of life) at which 77 youngchildren died (those in the one month column were underthe age of one month).

12

2

4

1110

8

0

6

1

3

9

7

5

16

Age (months)

Cas

es (

n)

14121110 13 1532 5 71 4 6 8 9

8 Meadow

happened if the parents story of the event wassupported by a health professional observingthe infant to be collapsed, unrouseable, orbreathless, or if on arrival at hospital there wasclear biochemical evidence (such as acidosis orhigh blood glucose) indicative of a recent majorstressful event. Forty six children had episodesof apnoea/seizures, of whom 24 had one ormore acute life threatening event, and 11 hadother unexplained illnesses. Of the rest, threehad an acute life threatening event, and ninehad another unexplained illness. Seventeen hadan unexceptional previous history.

The season of death is shown in table 2, andthe time at which the child was found dead ormoribund by the carer in table 3. These timingshave been used because of similar studies thathave been performed on children dying ofnatural causes,13 but the three time bandsdisguise what was a very clear predominance ofdeaths occurring in the afternoon and evening.Fifty five infants were found moribund or deadbetween 11:00 and 22:00.

The last occasion before death when thechild was seen to be normal, or in near normalhealth, was recorded for 65 children. Theobservers and reporters of that time intervalwere the childs carers, although in severalcases there was confirmatory evidence of thechilds previous wellbeing from relatives orfriends and, occasionally, from health visitorsor doctors. Fifty five children were observed tobe well within two hours of death, five withintwo to six hours, and five within six to 12 hoursof death.

The usual account of events was that theparent would discover the lifeless child, seekhelp from someone in the house or next door,telephone 999 for the emergency services,and attempt resuscitation. When the paramedi-cal staV arrived they too would attempt cardio-pulmonary resuscitation, although intubationwould only be performed on arrival at theemergency department of the local hospital.Several of the moribund children made an ini-tial partial response to that resuscitation beforedying.

Recent medical examination or hospital admission(77 children)More than half the children had been exam-ined by doctors in the preceding week andconsidered to be healthy. These included 43children who had been patients on childrenswards in the previous four weeks, where theyhad been examined and considered to be nor-mal by paediatric staV who arranged for theirdischarge from hospital. The interval betweendischarge from hospital and death is shown intable 4.

Nearly half of the children had beendischarged from hospital within the preceding

week. Most of those had been admitted brieflybecause of an unusual or unexplained eventreported by the carer, but on investigation andobservation in hospital had been found to bewell. The usual circumstance for the 15children who had been discharged from hospi-tal within the previous 24 hours was that thechild was discharged in the morning, and thendied at home that evening.

Other unusual features of these deathsincluded seven children whose deaths wereforeseen by the carerthe parent predictingthe day on which the child would die; andanother four children who died on the anniver-sary of a previous siblings death.

EXTERNAL SIGNS AT DEATH (70 CHILDREN)Twenty seven children were reported to havebeen found with blood apparent in the mouth,nose, or on the face. In deciding upon thereport of bleeding, care was taken to establishthat the finding was of frank blood, rather thanthe common sero-sanguineous froth that canbe present in moribund children, particularlywhen subject to resuscitation. On examinationby medical staV, stale blood was seen on 20 ofthose 27 children. Ten children had eitherunusual bruises or petechiae, on the face orneck. Forty children had neither bruises,petechiae, nor report/finding of bleeding.

POSTMORTEM EXAMINATION FINDINGS

Information on the postmortem examinationsof 77 children was available, but the quality ofthat information was extremely variable. Thepostmortem examinations had been performedin diVerent parts of the country by diVerentpathologists, and extended over an 18 yearperiod during which time standards of practicehad changed greatly. In many of the earliercases, tissues had not been examined micro-scopically by newer histological techniques. Aminority of cases underwent detailed neuro-pathological examination. An eVort had beenmade previously to arrange for re-examinationof relevant tissues, but in most cases materialwas no longer available for re-examination.Therefore, this account is based on thecontemporaneous postmortem reports fromthe relevant pathologist, who was usually thelocal district hospital pathologist, although inthe last five years was more likely to have beena forensic pathologist or paediatric pathologist.In most cases it was the pathologists reportthat led to the decision of the coroner as to thecause of death and the entry on the death cer-tificate. Forty two children were categorised asSIDS or cardiopulmonary arrestSIDS, sixas not ascertainable (three of whom hadbeen found to have fractures of the skull orarm), and 29 were given a specific cause fornatural death (table 5).

Unusual features of the 42 infants catego-rised as SIDS included five infants over 12months of age; two infants who had fracturedribs, which were considered to be the result ofresuscitation; two infants under the age of 6months who had several balls of paper in thestomach, which were deemed to have beenscrewed up and eaten by the young infant.

Table 2 Season of death of 77 infants

Season Number

Winter (December, January, or February) 29Spring (March, April, or May) 12Summer (June, July, or August) 22Autumn (September, October, or November) 14

Table 3 Time of death of72 infants

Time Number

24:0008:00 1308:0016:00 2916:0024:00 30

Table 4 Interval betweenrecent hospital admissionand death

Time sincedischarge Number

< 24 hours 1517 days 22828 days 6No recent

admission 34

Unnatural sudden infant death 9

Twenty eight of the infants had had previousunexplained or unusual events in their recentmedical history.

THE PERPETRATORS

For two of the 50 families, the courts could notdecide which parent was responsible. For theothers, 43 mothers were judged to be the per-petrator and five fathers. All were white Euro-pean adults. Other features of the perpetratorsare described below.

Social classUsing the Registrar Generals classification,one family was graded social class I, one asclass II, 10 as class III, 10 as class IV, five asclass V, and 20 were unclassifiable. There was apreponderance of families in which there wasno regular wage earner. At least 35 werereceiving income support.

Maternal smoking habits (69 children)Fifty nine of the children were born to andcared for by a mother who smoked cigarettesregularly. Ten had mothers who did not smoke.

Although many of the mothers consumedalcoholic drinks, there was nothing to suggestregular overindulgence or addiction to alcohol.

Perpetrators healthThe health records of 44 perpetrating parentshad been studied for evidence of somatisation.Evidence was sought for either somatising dis-order or Munchausen syndrome (defined asfactitious disorder with predominantly physicalsigns and symptoms, plus pseudologiafantastica14). Munchausen syndrome waspresent for 10 parents (seven mothers andthree fathers) and somatising disorder waspresent in another 11 parents (defined as eithermild/moderate factitious disorder or somato-form disordersthat is, physical complaintsnot attributable to medical condition, althoughnot intentionally produced). Twenty three per-petrators had neither somatising disorder norMunchausen syndrome.

Other featuresUnusual features in the lives of the perpetratorsincluded five who had been involved with fireraising (arson), six who had been, or were,involved in litigation against doctors or hospi-tals, and seven who had sought media publicityin relation to their childs death.

DiscussionThe childhood deaths that were studied werecovert unnatural deaths which, at the time ofdeath, had been thought to be the result ofnatural causes. The series does not includeyoung children who had obvious signs at pres-entation in hospital, or at postmortem exam-ination, of major traumatic abuse, neglect, orpoisoning.

The likelihood that the court verdicts aboutparental responsibility for death were correct isvery high indeed. Even in the family courts,where the judgment is made on the balance ofprobabilities, all experts, particularly the judge,are aware of the dire implications of the verdictin terms of a parent being allowed to care for afuture child.

The circumstances of the deaths, and of thefamilies, were scrutinised carefully before thecourt hearings. Therefore, in addition to theinformation presented in this paper, there wasusually much additional incriminating evi-dence, including the recurrence of events anddeath always with the onset in the presence ofone carer, inconsistent and untruthful accountsof the events from the perpetrator, evidence ofsmothering or other abuse of a subsequentchild, and sometimes admissions by the perpe-trator. This study addresses certain generalfeatures that might be useful to paediatricianswho encounter sudden infant death, but ineach case the paediatrician must take intoaccount the additional specific features of thatindividual case.

It is necessary to consider how much thedescribed features are the result of caseselection, and how much they are characteristicof all covert unnatural infant deaths.

It is likely that the number of cases occurringin the context of Munchausen syndrome byproxy abuse, or which were caused by parentswho had Munchausen syndrome themselves ishigh because of such cases being referred pref-erentially to me.1517 Similarly, since publishingSuVocation, recurrent apnoea and suddeninfant death in 1990, many similar cases havebeen referred to me.

Because the cause of death was proved incourt, some of the features will have been usedas evidence, to persuade juries and judges topronounce that a child had been killed unnatu-rally. These features will be more characteristicof legally proven unnatural infant death thancovert unnatural infant death in general. Mul-tiple deaths, previous undiagnosed events, andmaternal somatisation are likely to be overrep-resented because of that.

In contrast, some of the described featureswill have been used in court in favour of anatural cause, and the verdict will have beenreached despite the fact that those featureswere advanced in favour of a natural cause(maternal smoking, low birth weight, neonatalproblems, and low social class are citedfrequently in court as favouring a naturalcause).13

Other features, for instance birth order, willnot have been considered in court and will nothave influenced the verdict.

Table 5 Specific pathological diagnoses (29 cases)

Diagnosis Number

Asphyxia from vomit (5 as a result of allegedseizures) 6

Asphyxia from foreign body or accident 5Other respiratory cause (bronchitis, pneumonitis,

bronchopneumonia, and bronchodysplasia) 6Cardiac (2 myocarditis, 1 patent ductus arteriosus,

and 1 cardiac asystole) 4Other causes 8

Unusual features of some of these were: two infants who hadlacerations in the throat; three with foreign bodies (coins, paperballs, and a mitten) in the upper airways or stomach, which theinfant was considered to have grasped and eaten; and four chil-dren who were deemed to have died of bronchitis, broncho-pneumonia, or patent ductus ateriosus who had taken normalfeeds and were seen to be well and free from respiratory symp-toms within two hours of death.

10 Meadow

The reason that more than half the reportedfamilies included more than one dead child islikely to be because the courts were impressedby evidence that it was highly improbable fortwo or more children to die in infancy of undi-agnosable natural causes: if there is a 1/1000chance of a child dying suddenly and unexpect-edly of natural causes in the first year of life, thechance of two children within a family so dyingis 1/1 000 000. A parent who kills only onechild is much less likely to be incriminated thanone who kills or abuses two or more.Nevertheless, the finding of 26 serial killers isworrying.

Other Australian and English studies con-firm that the youngest or only child in the fam-ily is most likely to be killed.18 Although most ofthe mothers were young, that was not the onlyreason for the finding that it was nearly alwaysthe first (or the first few) children who werekilled. It was rare for the perpetrator to havehad a previous child who was well and who hadnot been abused. This has important implica-tions for those assessing future risk and forparental treatment.

Death was much commoner in the first 7months of life. Older infants are more robustand more personable. Some of the factors thatmake it particularly stressful for a youngmother to look after a young infant might beless severe after the early months. The moder-ate excess of boys fits in with previous studies,and with reports of smothering of infants whohave not died.1012 Homicide is more likely inthe 1st year than any other year of life.19 Youngchildren are usually killed in the home, and theinfant is most likely to be killed by their ownparent in their own home.

Nearly all the perpetrators were the childsmother. This is likely to be a correct conclusionfor these sorts of death. Fathers may beresponsible for an infants death, but previouswork suggests they are less likely than mothersto kill children under 1 year of age and they aremore likely to kill the child by shaking or bydirect physical force, which usually cause signsthat are unlikely to be mistaken for a naturaldeath. It is interesting to reflect that the fatherscause death by methods that are moreindicative of impetuous violence, whereas themethods used by mothers suggests a clearintent to kill or cause grievous bodily harm.18

Deaths were commonest in disadvantagedfamilies, receiving income support. During thestudy period, the proportion of adult womensmokers in the UK has been in the range2530%.20 Most of the perpetrating motherswere regular smokers, and few if any desistedduring pregnancy. Both these factors have beencited as risk factors for SIDS, smoking in preg-nancy being particularly important.21 22

The perinatal events and early histories ofthe children are relevant. The pregnanciescontained an excess of problems. Part of thatexcess was due to some mothers who, in retro-spect, were considered by the obstetricians tohave had factitious obstetric disorder.

About a quarter of the infants were low birthweight, compared with the general UK pro-portion of 7% (and a proportion of 1012% in

disadvantaged families). About three quartersof the infants had unusual previous medicalhistories. The story of recurrent episodes ofapnoea and seizures is one that has beenreported before in children who are smothered,and particularly in those who are incurringMunchausen syndrome by proxy abuse.7 10 12

About half the deaths occurred in what seemedto be the context of Munchausen syndrome byproxy,23 which is a similar proportion to thatfound for children who survived suVocation inthe 199294 British Isles survey.11 Some of theother episodes are less easy to understand, butthe findings accord with experience of otherfatal child abuse in which there has been anexcess of preceding unusual medical events orhospital admissions,24 and it is also a findingthat has been observed in the UK in babiessubsequently labelled as cot death.25

Most of the deaths are likely to have beencaused by smothering. Many of the precedinghistories were typical of smothering. The 19 (of50) perpetrators who confessed to killing theirchild described the smothering. One was amother who had also pushed balls of paperdown the infants throat because she thoughtthat would choke the child (it did nottheballs of paper went into the stomach). Severalof the more recent cases have been the subjectof meetings involving all the medical experts ofthe diVerent parties; and the usual conclusionhas been that the mode of death has beensmothering. In three cases the neuropathologi-cal changes, and the external marks on thebody, raised the likelihood of shaking as thecause, or as a contributory factor.

No cases of poisoning were identified. Aboutone third of the cases, mainly the recent ones,had undergone detailed toxicological tests thatwere negative. The possibility that a few of theearlier deaths were attributable to poisoningcannot be excluded, but the contemporaneousinformation does not make it likely.

Paediatricians are familiar with the story ofnatural SIDS death when a previously healthy4 month old baby is found dead in his cot in themorning as his parents awaken, having seemednormal or merely snuZy the previous evening.Quite often those bereaved parents say they lastsaw their child to be well eight or 10 hours ear-lier. These deaths were diVerent. Nearly all thechildren were seen to be well shortly beforethey died, and most of them died in thedaytime or early evening. This should not besurprisingobviously, when asleep, parents donot kill their children. Moreover, it is likely thatmany natural deaths occur more slowly, thechild becoming gradually iller during the night,unobserved. Nevertheless, near miss cot deathsshould always be reviewed carefully: the chanceof a parent finding a child dying a natural deatha few seconds before stopping breathing is verysmall. If the parent is on the scene at that timeit must raise the possibility that the parent hasbeen responsible for the child not breathing.Natural cot deaths are more likely to be associ-ated with a child being found dead, rather thanmoribund. Most of the deaths occurred whenthere was just one parent in the house,although a significant proportion happened

Unnatural sudden infant death 11

when both parents were at home, with one par-ent being alone with the child for long enoughto kill.

Nearly half of the children had had briefadmissions to hospital within the precedingweek. Fifteen children had been dischargedfrom hospital childrens wards in the last 24hours. All paediatricians will feel that thoseinfants were discharged inappropriately andtoo early. It is neither a shortage of hospitalbeds nor of facilities that causes paediatriciansto discharge children so rapidly, it is part of anevolutionary culture which correctly believeshome to be a safer and happier place than hos-pital for most children, and which seeks toadmit ever fewer children to hospital for evershorter periods. But now that the averagelength of stay in the UK for an acute paediatricadmission is under two days, it is time torecognise that too many risks are being taken. Ifa young child is admitted to hospital as a resultof a young, harassed mother from a poor homerecounting a startling or unusual event, and thebaby is found to be completely normal, it doesnot mean all is well. When next morning thepathology reports and the x ray reports comeback normal, it would be safer to discharge thehospital notes and normal investigation reportsout of the door than the baby. That mother hasbrought the child to hospital for a reason thatwe have not understood. There is ampleevidence that children suVering recurrentphysical abuse, as well as Munchausen syn-drome by proxy abuse, have many warning sig-nals and previous hospital encounters beforethe final event that maims or kills.

This study did not include reliable data onthe status of the young children in terms ofprevious entry on the Child Protection Regis-ter. A minority are known to have been onChild Protection Registers; most were not.Even though a recent study from NorthAmerica suggested that screening child protec-tion agency records for previous referrals wasan ineVective way of detecting infanticidesmisdiagnosed as SIDS,26 checking the ChildProtection Register is a sensible step whendealing with young infants who present withunusual or recurrent events.

Currently, many paediatric units are failingto heed warning signs and failing to protectsome very vulnerable children. Many of themothers who killed their child in the afternoon/evening, after their being discharged in themorning, had been resident in hospital withtheir child. One wonders how often a doctor ornurse sat down with that mother beforedischarge to find out whether she was confi-dent and happy to return home with her infant,or discussed the family with their health visitoror general practitioner.

Because these were cases considered origi-nally to be natural deaths, major external signswould not be expected. The association ofbleeding with smothering has been reportedbefore.10 12 It is important, nevertheless, to dis-tinguish between the pinkish brown sero-sanguineous mucus that may be on the face ofa dead baby who has been subject to resuscita-tion, and the bleeding or crusted red blood that

may be found after smothering and physicalabuse. Although smothering was believed tohave been the cause of death for most of thechildren, neither purpura nor petechial haem-orrhages were common. More than half thechildren had neither bruises, petechiae, norsigns of recent bleeding. The corpse appearednormal. Similarly, fatal smothering might notreveal any specific abnormality during post-mortem examination.

The circumstances of the study did not allowthe sort of pathological re-evaluation that wasneeded. Reliance had to be placed on the notestaken from postmortem reports that had beenperformed by a variety of investigators indiVerent parts of the country, over a longperiod of time. Those pathologists would havebeen commissioned by the local coroner whohas to be informed of sudden unexpecteddeaths. That coroner has an assistant, the coro-ners oYcer, who explores the background tothe death. Unless there is something very unu-sual or suspicious about the circumstances, thecoroner invites the local hospital pathologist todo a postmortem examination. That patholo-gist might not have particular experience ofinfants and children, and is rarely a paediatricpathologist.

Both the coroner and the pathologist areunder pressure to provide a diagnosis fast sothat a death certificate can be issued and thebody released to the family for burial. Essen-tially, they are having to provide a quickdecision about a natural or unnatural death,and the pathologist has to do that beforedetailed histological or other test results areavailable. If the pathologist does not provide anatural cause for the coroner, the coroner hasto arrange for a formal inquest, and there isconsiderable pressure to avoid such events.Inevitably, there is a tendency for all parties toseek a natural cause rather than an unnaturalone. It is unusual for either the coroner or thecoroners oYcer to have a medical background.If the coroners oYcer has identified obvioussuspicious features, the coroner commissionsan approved forensic pathologist. Some, butnot all, forensic pathologists have considerableexperience of infants, but their work will berelated to forensic examinations and they arenot involved regularly with postmortem exami-nations of ordinary infant deaths.

The postmortem examination conclusionsdemonstrate the inadequacies of the past prac-tice. SIDS has been used at times as apathological diagnosis to evade awkwardtruths. In this series it was used for infants overthe age of 12 months, for infants with fractures,and in bizarre circumstances, such as whencoins, balls of paper, and mittens were presentin the airways or intestines of the children.Such findings were explained away on the basisthat the infant had grasped and eaten the object(even though their developmental age madethat impossible). Postmortem examination hadrevealed unusual foreign bodies in the stom-achs of five children, and in the airways of twochildren. At the time of postmortem examina-tion, these had been considered to be the resultof accidental, coincidental ingestion. Yet, those

12 Meadow

dealing with accidental ingestion know howrare it is for children below the age of 1 year toswallow foreign bodies. Moreover, the chanceof such a young child accidentally ingesting aforeign body, and then dying of natural causeswithin the next four hours while the foreignbody is still in the stomach, must be infinitesi-mally small. The finding of a foreign body inthe alimentary canal of an infant who isthought to have died of natural circumstancesshould arouse considerable suspicion.

Another distressing feature to a paediatricianis the way that some respiratory and cardiacconditions that make children ill were diag-nosed on the basis of a few inflammatory cellsin the relevant organ, and no account was takenof the history; the problem being that thepathologist probably only had a brief storyfrom the coroners oYcer, and had no adequateprevious medical history, and no advice or helpfrom a paediatrician. For the pathologist it wasnot customary to ask the local paediatricianwhether it was feasible for a 3 month baby topick up a small coin and eat it, or to die frombronchopneumonia half an hour after beingobserved by two parents to be awake, happy,and in normal health.

There continues to be much interest in thedevelopment of more specific tests (such asintra-alveolar haemosiderin)27 that might iden-tify smothering and other forms of childabuse.28 The UK Royal College of Pathologistshas issued guidelines on minimum prerequi-sites for a thorough postmortem examinationof infants.29 Despite the considerable resourceimplications of this thorough assessment andinvestigation, it is likely that most infants whodie suddenly and unexpectedly in the UK arenow having that thorough postmortem exam-ination. Nevertheless, unless it is combinedwith a careful evaluation of the childs previoushistory, by an experienced paediatrician, mis-takes will continue to be made. The price ofthose mistakes is high because other siblingsborn to the perpetrator are at risk. Thosecountries and USA States that have an experi-enced child death review team are much lesslikely to make dangerous mistakes, particularlyif a death scene investigation is included.3033

The UK Confidential Enquiry into Still-birthsand Deaths in Infancy (CESDI) continues toperform useful research and studies of selectedissues,34 35 but it is not a substitute for carefulmultidisciplinary investigation of all unex-pected infant deaths.

As the number of infant deaths continues tofall, it becomes feasible to organise a multidis-ciplinary confidential enquiry into every unex-pected infant death. The time for suchcomprehensive investigation has already ar-rived. Even though the number of infantscategorised as SIDS in the UK has fallen inrecent years to below 400/year, it is a nationalscandal that we accept a situation in which somany young children die of unknown causes. Ifone out of every thousand 21 year olds diedsuddenly and unexpectedly without an identi-fiable cause there would be a national outcry.

It is sad that the term SIDS has become abarrier to the sensible and sensitive investiga-

tion of infant deaths. Despite the compassion-ate and constructive work of so many lay andmedical people concerned with SIDS organisa-tions in diVerent countries, the label hasbecome counterproductive. For many of thecases in this series, SIDS deaths were presentedto court as if they represented a single disorder,and information given regardless of the factthat nearly all the earlier epidemiologicalsurveys of SIDS will have included many chil-dren who died of covert unnatural deaths.Many social workers, and sometimes healthprofessionals, had been misled at case discus-sions and conferences, and in the assessment ofmany of these families, because of the way thatthe label SIDS had been given to a previouschildWe dont have to worry about thatdeath, it was a natural one, the child died ofSIDS.

If the term SIDS is to continue to be usedthen it should be limited in the way suggestedby the expert panel of the US NationalInstitute of Child Health and Development36 tothe sudden death of an infant under 1 year ofage which remains unexplained after a thor-ough case investigation, including performanceof a complete autopsy, examination of thedeath scene and review of the clinical history.

However, it would probably be in the bestinterests of children to abandon the term andacknowledge that when a previously well childdies suddenly, and a thorough assessment andpostmortem examination reveals no cause, wesay unexplained or undetermined ratherthan pretending that it is a natural death fromsome yet to be identified agent. Nobody doubtsthat there are many diVerent causes of deathfor children categorised as SIDS, and yet wecontinue to use the term as a diagnosticcategory, and to discuss it as a single diseaseentity. We should be honest and admit that wedo not know and, above all, we should be angryand intent on preventing so many youngchildren dying suddenly and unexpectedly soearly in life.

Key messages+ Covert unnatural deaths are most com-

mon in children younger than 8 months

+ The childs mother is the usual perpetra-tor

+ Many of the infants killed have hadprevious unusual illness episodes orrecent admission to hospital

+ Infants who die in the afternoon orevening, after recent admission to hospi-tal, or who come from a family in whicha previous child has died unexpectedlyshould raise suspicion

+ Over a third of the infants believed tohave been smothered had either pe-techiae or frank blood about the face,although many do not disclose other sig-nificant abnormality at postmortem ex-amination

Unnatural sudden infant death 13

Warning to meta-analystsSome of the cases reported in this paper willhave been included in other papers originatingfrom the UKnotably references 10, 12, and17. It would be unwise to include another UKpaper in any meta-analysis alongside thispaper, because of the risk of overlap.

I am grateful to the many colleagues who have involved me withthese families, and who have been generous in supplyingadditional information for the study. I thank J Darling and DSmith for their thoughtful comments and M Jones for her eY-cient work with so many drafts.

1 Oliver JE. Dead children from problem families in NE Wilt-shire. BMJ 1983;286:11517.

2 Taylor EM, Emery JL. Two year study of the causes of post-perinatal deaths classified in terms of preventability. ArchDis Child 1982;57:66873.

3 Committee on Child Abuse and NeglectAmerican Acad-emy (1994). Distinguishing sudden infant death syndromefrom child abuse fatalities. Pediatrics 1994;94:1246.

4 Asch SS. Crib deaths: their possible relationship topost-partum depression and infanticide. J Mt Sinai HospNY 1968;35:21420.

5 Kukull WA, Peterson DR. Sudden infant death and infanti-cide. Am J Epidemiol 1977;106:4856.

6 Emery JL. Infanticide, filicide and cot death. Arch Dis Child1985;60:5057.

7 Rosen CL, Frost JD, Bricker T, et al. Two siblings withrecurrent cardiorespiratory arrest: Munchausen syndromeby proxy or child abuse? Pediatrics 1983;71:71520.

8 Meadow R. Fictitious epilepsy. Lancet 1984;2:258.9 Rosen CL, Frost JD, Glaze DG. Child abuse and recurrent

infant apnea. J Pediatr 1986;109:10657.10 Meadow R. SuVocation, recurrent apnoea, and sudden

infant death. J Pediatr 1990;117:3517.11 McClure RJ, Davis PM, Meadow SR, Sibert JR. Epidemiol-

ogy of Munchausen syndrome by proxy, non-accidentalpoisoning, and non-accidental suVocation. Arch Dis Child1996;75:5761.

12 Southall DP, Plunkett MCB, Banks MW, Falkov AF, Samu-els MP. Covert video recordings of life-threatening childabuse: lessons for child protection. Pediatrics 1997;100:73560.

13 Golding J, Limerick S, Macfarlane A. Sudden infant death:patterns, puzzles and problems. Shepton Mallet: Open BooksPublishing Ltd, 1985.

14 Diagnostic and statistical manual of mental disorders. 4thed. Washington DC: American Psychiatric Association,1994.

15 Meadow R. Munchausen syndrome by proxy. The hinter-land of child abuse. Lancet 1977;ii:3435.

16 Meadow R. Munchausen syndrome by proxy. Arch Dis Child1982;57:928.

17 Meadow R. Munchausen syndrome by proxy abuseperpetrated by men. Arch Dis Child 1998;78:21016.

18 Wilczynski A. Child homicide. London: Greenwich MedicalMedia Ltd, 1997.

19 Creighton SJ. Fatal child abusehow preventable is it?Child Abuse Review 1995;4:31828.

20 Bower H. Smoking in Britain. BMJ 1998;316:727.21 Blair PS, Fleming PJ, Bensley D, et al. Smoking and the

sudden infant death syndrome: results from 19935case-control study for confidential enquiry into stillbirthsand deaths in infancy. BMJ 1996;313:1958.

22 Alm B, Milerad J, Wennergren G, et al. A case-control studyof smoking and sudden infant death syndrome in the Scan-dinavian countries, 19921995. Arch Dis Child 1998;78:32934.

23 Meadow R. What is, and what is not, Munchausensyndrome by proxy? Arch Dis Child 1995;6:5348.

24 Korbin JE. Fatal maltreatment by mothers: a proposedframework. Child Abuse Negl 1989;13:4819.

25 Stanton AN, Oakley JR. Pattern of illnesses before cotdeaths. Arch Dis Child 1983;58:87881.

26 OHalloran RL, Ferratta F, Harris M, Ilbeigi P, Rom CD.Child abuse reports in families with sudden infant deathsyndrome. Am J Forensic Med Pathol 1998;19:5762.

27 Beecroft DM, Lockett BK. Intra-alveolar pulmonarysiderophages in sudden infant death: a marker for previousimposed suVocation. Pathology 1997;29:603.

28 Berry PJ. Pathological findings in SIDS. J Clin Pathol 1992;45:1116.

29 Royal College of Pathologists. Minimum guidelines forpost-mortem investigation of post-neonatal infant deaths or sud-den unexpected deaths in infancy. Guidelines for post-mortemreports. London: Royal College of Pathologists, 1993:appendix 1, 7.10.

30 Reece RM. Fatal child abuse and sudden infant deathsyndrome: a critical diagnostic decision. Pediatrics 1993;91:4239.

31 Granik LA, Durfee M, Wells SJ. Child death review teams: amanual for design and implementation. American Bar Associ-ation Order Fullfillment 549, 750 Lake Shore Dr, Chicago,IL 60611.

32 Bass M, Kravath RE, Glass L. Death-scene investigation insudden infant death. N Engl J Med 1986;315:1005.

33 Perrot LJ, Nawojszyk S. Non-natural death masquerading asSIDS (sudden infant death syndrome). Am J Forensic MedPathol 1988;9:10511.

34 CESDI Confidential Enquiry into Still-births and Deaths inInfancy. Fifth annual report. London. Maternal and ChildHealth Research Consortium, 1998.

35 Fleming PJ, Blair PS, Bacon CJ, et al. Environment ofinfants during sleep and risk of the sudden infant deathsyndrome: results of 19935 case-control study forconfidential enquiry into stillbirths and deaths in infancy.BMJ 1996;313:1915.

36 Willinger M, James LS, Catz C. Defining the sudden infantdeath syndrome (SIDS): deliberations of an expert panelconvened by the National Institute of Child Health andHuman Development. Pediatr Pathol 1991;11:67784.

14 Meadow