Methods

Around 350 children each year are admitted to the Welsh Regional Burns Unit (WRBU), which serves a population of 2.3 million. Over 50% of admissions are aged under 3 years and approximately 40 of these are from the two unitary authorities under study (Swansea and Neath Port Talbot). The total population of Swansea and Neath Port Talbot based on 2001 census data is 357 769.

This study was designed as a retrospective matched cohort study of young children (aged under 3 years) admitted with thermal injury to the WRBU, Morriston Hospital.

The initial cause of the burn was ascertained where possible as accidental, non-accidental (abusive or neglect) or undetermined. We then compared the social health outcomes of burned children with their peers up to their sixth birthday.

For the purposes of this study, the main social health outcome measures were defined as: (i) Social Services referral (defined as “in need”) and (ii) abuse or neglect, that is requiring child protection registration (CPR) or a local authority care (“looked after”) episode by age 6 years. The directors of Social Services for both unitary authorities collaborated with the study. The South West Wales Local Research Ethics Committee granted ethics approval.

Study population

Cases were identified from the WRBU database. All children aged less than 3 years who had been admitted between September 1994 and August 1997, and were resident in one of two unitary authorities (Swansea and Neath Port Talbot) at the time of admission were included. Each case was given a deprivation score obtained from the Townsend Index of Material Deprivation15 for 2001 electoral wards based on the postcode of residence at the time of injury. These cases were matched with controls for the potential confounders of sex, age and enumeration district (ie, similar socioeconomic environments).

Controls were derived from the Community Child Health Database as in previous research projects.16 Controls were selected blindly from the child health computer as the next child of the same age (within 3 months), same sex and residing in the same enumeration district as the case. Controls were excluded if they had been admitted with a burn. Seven cases could not be matched in this manner. Hence, one child of a travelling family was matched with another “traveller” control. The other six cases were matched with a control from an enumeration district with a Townsend score nearest to the case score.

Data collection and management

Data capture for all identified children continued to 31 August 2003. Electronic routine databases provided study data for Social Services referral, child protection registration or local authority care episode by age 6 years.

For all cases data were extracted from hospital medical records concerning the type of burn, treatment and outcome and whether a referral had been made to Social Services as a result of the burn.

The local Social Services databases were searched for every case and control (in case of movement between neighbouring areas), to obtain information on children who required Social Services resources, support, accommodation or child protection assessment in the years before their burn and up until their sixth birthday. This search also identified cases in whom the burn was judged as “non-accidental” after a case conference (ie, level 1 criteria used to define abuse).17

Data were treated confidentially and anonymised before being entering onto the computer database.

Statistical analysis

The sample size for statistical analysis of a categorical variable was calculated based on the primary outcome measure of Social Services referral. Using paired, individually matched cases and controls in order to detect a difference of 15% in Social Services referral between the cases and controls, 80 case/control pairs were required, based on a power of 80% with α = 0.05. As we were unsure about the completeness of the data, information on all burns cases and their matched control were retrieved for the study period (n = 145). Relevant statistical analysis was undertaken using SPSS v 12. Where possible, matched pairs were analysed using McNemar tests for categorical data. Where numbers of pairs were small, the unpaired χ² tests were run.

Results

In total, 145 cases and 145 matched control children were included in the study. One additional case who had moved abroad was lost to follow-up and therefore was not included in the analysis.

Type and cause of burn

The majority of burns occurred in the home and were water or hot liquid scalds. Eighty three (57.2%) males and 62 (42.8%) females were included in the study. At presentation the median age of the cases was 16 months (range 0–36 months), the median weight was 11.5 kg (range 3.8–17.9 kg) and the median total body surface area percentage burn was 2% (range <0.25–50%). On admission the initial cause of the burn in the cases was recorded as accidental in 130 (89.7%), non-accidental in one (0.7%), and undetermined in 14 (9.7%) (table 1).

When Social Services investigated cases, non-accidental injury was proven in four (2.8%) following a case conference. One was reassigned as undetermined and 129 (89%) as accidental (table 1). Of the four non-accidental cases, one child was placed in local authority care and three on the child protection register (CPR) for physical injury after statutory case conference.

Thirty two cases (22%) were referred to Social Services at the time of the burn for support following concerns by hospital staff. Most of these had no further Social Services input (22/32, 68.8%). Of the 10 that had further Social Services input, four of these children later came into local authority care or were placed on the CPR.

Five cases previously known to Social Services were not referred at the time of their burn. However, none of these children were subsequently placed in local authority care or on the CPR.

None of the cases sustained any further burns during the study period.

Comparison of cases and controls up to age 6 years

Baseline comparisons of Social Services referrals

Ten cases (6.8%) had already been referred to Social Services prior to the burn date, the majority for suspected abuse, compared with four controls (2.8%) (table 2A). There was no statistically significant difference between the case and control groups in terms of numbers that had been referred to Social Services prior to the burn date (95% CI −0.005 to 0.088, p = 0.15).

Comparisons of Social Services referrals at age <3 years

The total number of cases referred to Social Services aged less than 3 years was 25 (excluding those only referred at the time of the injury) (25/145, 17.2%). In the control group 11 (7.6%) children who were under 3 years of age were referred to Social Services (table 2B). There was a statistically significant difference between the cases and controls (95% CI 0.027 to 0.166, p = 0.013). The figure for referrals for all Swansea children aged less than 3 years for 2003 was 674. Based on the total number of children (7148), this equates to a 9.4% referral rate.

Comparisons of Social Services referrals at age 6 years

Twenty six (18%) controls and 66 (46%) of the burned cases had been referred to Social Services by their sixth birthday. This was statistically significant (95% CI 0.176 to 0.376, p = 0.000). Thirty four cases (23%) had more than one referral. Twenty cases (14%) had a single referral at the time of the burn and no contact thereafter. If the latter were excluded, then 46 cases (32%) had been referred to Social Services versus 26 controls (18%) (table 2B). Significantly more cases than controls had Social Services referral by the age of 6 years (95% CI 0.047 to 0.23, p = 0.006).

By their sixth birthday 14 cases were on the CPR or in local authority care (eight abuse and six neglect). There were significantly more children “in care” or on the CPR in the case group (14) compared to the control group (two) (95% CI 0.030 to 0.13, p = 0.004).

Nine of these 14 case children had been referred to Social Services at the time of their burn because of staff concerns around non-accidental injury. Eight of these children were subsequently identified as abused and one as neglected (table 3).

At the time of the burn, four (2.8%) cases were immediately identified as being non-accidentally injured. When these were removed from the comparison, it was found that of the remaining cases, 10 (6.9%) were recorded as being either accidental or undetermined and subsequently went on to be in local authority care or on the CPR by age 6 years (four abuse and six neglect). This was significant when compared with the two controls in local authority care or on the CPR by age 6 (95% CI 0.007 to 0.10, p = 0.039). There were five (3.4%) children who were not referred by hospital staff at the time of the burn and these were subsequently neglected.

Discussion

In our study most children (89.7%) were considered to be accidentally injured at presentation and no burns were attributed to neglect. We identified that 2.8% of young children admitted with a burn were non-accidentally injured. This was a robust diagnosis based on the child being placed in local authority care or on the child protection register following a case conference. Meeting these criteria provides the highest level of confidence that abuse has taken place.17 However, abuse was not excluded with the same rigour as no explicit criteria for referral were in use in the burns unit at the time of the study and clinical investigations were not standardised. Referral to Social Services was based on the history, clinical features, presentation of the burn injury and relevant family factors. Cases were referred on account of maternal substance misuse, observed poor parenting skills, parental hostility towards a child and inappropriate refusal of treatment. These factors have been linked with abusive injuries in the published literature.3 7 18

US studies describe a high proportion of non-accidental burns in children (3–26%)8 10 19 20 compared with UK series (1–16%).3 11 12 13 14 This could be due to a true difference between countries, better identification of cases or different referral practices. Selection bias or different definitions of abuse make direct comparison between studies and different countries difficult. Our study has shown that non-accidental injury as the cause of a burn was proven in only a small proportion of the case population at presentation (just 2.8%), which supports the findings of other UK studies.11 12 14

Children from socially disadvantaged families sustain more burns than other children,1 3 6 7 21 which can be partly explained by accidents resulting from overcrowding and unsafe heating practices.3 21 22 23 They are also more likely to be burned non-accidentally – either deliberately or through neglect. The boundaries between accidents, inadequate supervision and neglect are rarely clear cut; a pattern of parental supervision, which is persistently inadequate, compromises a child’s safety and becomes neglect.24 25 26 We looked at Social Services referral as a measure of social disadvantage and, like several other studies, found that some children admitted with burns were known to welfare services already.8 18 Indeed, 6.8% of our burn cases had been referred to Social Services before they attended with a burn. There was no significant difference in the number of matched controls (ie, children living in similar socioeconomic environments) referred to Social Services at a comparable age.

Almost one third of the cases in our study had been referred to Social Services because of concerns about their welfare (children “in need”) by their sixth birthday. In total, 9.7% had been abused or neglected by age 6 years (as defined by registration on the child protection register or placement in local authority care) and this included the cases that were identified as having a non-accidental burn at presentation. Significantly more cases than controls were in need or had been non-accidentally injured (abused or neglected) by their sixth birthday. The results do not appear to be biased by children being over scrutinised as a result of referral at the time of the burn as the majority of children known to Social Services in later years had not been referred at the time of the burn. Furthermore, some children known to Social Services before the burn and most cases referred by hospital staff as a consequence of the burn injury had no further Social Services input. It is not possible to elicit from our study whether referral was unnecessary or the support given at this time prevented further problems.

We had planned to examine the likelihood of labelling non-accidental burns as accidental at initial presentation but the limitations of a retrospective study and relatively small numbers made this difficult. However, we were able to ascertain that 7.0% of cases were sent home from hospital with an “accidental” burn and were subsequently abused or neglected by age 6 years.

We have shown that burned children aged less than 3 years at admission sustain more abuse or neglect at the hands of their carers over the next 3 years. Also those children with burns appear to be at higher risk of further neglect/abuse and welfare concerns (“in need”) than matched controls. We therefore postulate that health professionals and agencies involved in protecting children are under-diagnosing non-accidental burn injuries (predominantly neglect) at presentation. How can we get it right? How can we identify the correct cause of burns in young children and so give support to families that need it, and more importantly safeguard children? We cannot allow children to “return to jeopardy”, a term coined by Hultman8 who also demonstrated that some burned children went home to suffer further neglect and abuse.

We further hypothesise that a burn that requires hospitalisation in a young child increases the likelihood of safeguarding concerns or even significant harm. So we propose that a burn injury in a young child could be used as a surrogate marker indicating a need for closer supervision and follow-up by professionals. Undoubtedly it would be good practice for all children aged under 3 years admitted following a burn to undergo careful evaluation. Where concerns are substantiated, children should be monitored carefully in the years that follow.

We have highlighted a need for specific training in relation to burn injuries in young children – especially neglect cases – since these were poorly identified in our study. The recent study by Chester et al13 also highlighted this issue. Better inter-agency communication between professionals and improved staff training would facilitate the recognition of non-accidental burn injuries in young children. Robust initial injury documentation and assessment would also aid in the evaluation of subsequent injuries.

The challenge for future research is to determine criteria that could be used at the time of the initial burn, which would trigger a more in-depth evaluation from a safeguarding perspective and prompt appropriate referral.27 Prospective follow-up studies would be invaluable in assessing outcomes and the effectiveness of using such measures.