Identifying child victims of the South-East Asia Tsunami in Thailand

Abstract

Purpose

The purpose of this paper is to reveal difficulties associated with identifying child victims of the 2004 South-East Asia Tsunami at the Thai Tsunami Victim Identification (TTVI) operation in Phuket and explores two strategies that increased child identifications.

Design/methodology/approach

Data allowing comparison of identification proportions between adult and child (defined as ⩽16 years old) victims of six nationalities and the forensic methods used to establish identification were used in this study.

Findings

The first 100 days of the operation revealed that the proportion of adult identifications far outweighed the proportion of child identifications. Moreover, the younger the child, the longer the identification process took (p<0.0001). Children under the age of 1 year took an average of 195 days to identify compared to 130 days for children aged 16. Identification was extended, on average, 4.3 days for each year that victims younger were than 16. Identifying large numbers of child victims requires targeted protocols. Two efforts increased child identifications for the TTVI operation: using body length to distinguish post-mortem (PM) DNA samples potentially belonging to children for targeted testing, and singling out deceased parents of missing children who were previously identified by a modality other than DNA, in order to retrieve and test their PM samples as references for kinship matching. Disaster victim identification operations with similar characteristics may benefit from implementing a strategy targeting child identifications.

Originality/value

The implementation of these two strategies at the TTVI helped to overcome initial complexities, namely, the lack of ante-mortem and PM material, and increased child identifications.

Introduction

On December 26, 2004, a 9.3 magnitude earthquake erupted in the Indian Ocean off the coast of Banda Aceh in Sumatra, triggering a massive tsunami that affected 13 countries and killed over 280,000 people (INTERPOL Tsunami Evaluation Working Group, 2014; Lay et al., 2005). Thailand reported 8,195 people as either dead or missing. The effect of the disaster on local Thai families was immense, with entire villages swept away. In total, 40 other nations also suffered casualties in Thailand, including approximately 2,400 foreign tourists in Patong, Khao Lak, Krabi and Phi Phi Island (INTERPOL Tsunami Evaluation Working Group, 2014; De Valck, 2006). Approximately, 450 of the foreign victims were children, aged 16 years or younger (Figure 1). Losses often involved entire families, resulting in large numbers of related decedents. For example, the pedigree in Figure 2 illustrates a common scenario with victims across three generations, while the pedigree in Figure 3 represents an extreme example.

Disasters are often characterized by the composition of the victim population. In a disaster with a “closed population,” for example, an aviation disaster with a passenger manifest, basic information about the number and identify of the victims is available from the outset. Conversely, in a disaster with an “open population” neither the number of victims nor their identities are immediately known. The 2004 tsunami had an “open population,” therefore the Thai Tsunami Victim Identification (TTVI) did not know the victim numbers or their demographics until well into the operation.

Such a large number of related victims and a disproportionate number of deceased children had not previously been encountered on this scale in a disaster victim identification (DVI) operation. By comparison, the median age of the 2,749 World Trade Center (WTC) victims was 39 and 38 years for males and females, respectively. Moreover, only three WTC victims were aged five and under, and related victims were not common (United States Centres for Disease Control and Prevention, 2002). All but two of the 202 victims of the 2002 Bali Bombings were adults, mainly in their 20s and 30s, and most were unrelated. The two child victims were females aged 14 and 15[1]. The 2001 crash of American Airlines flight 587 resulted in numerous child fatalities, where 28 of the 251 (9 percent) passengers killed were children 16 years or under, including multiple siblings (Vidoli and Mundorff, 2012). However, AA flight 587 is not a comparable DVI operation to the tsunami because it was a closed population incident with an immediately available flight manifest and significantly fewer victims. Argentina provides a more comparable scale of missing children to the tsunami, where identifying evidence was also scarce. An estimated 500 children were disappeared by the Argentine military between 1976 and 1983—either abducted as babies and small children, or born in captivity by disappeared mothers who were later killed. Therefore, a new statistical model was developed to account for the abducted children’s missing parents. The “grandparentage index” compares genetic markers of the children’s true grandparents and has been used to confirm the identity of 117 kidnap victims (Penchaszadeh, 2015; King, 1991). DNA is also the primary identification method for the victims of the conflict that took place in the former Yugoslavia during the 1990s. This event was characterized by a large number of related boys and their fathers, with a scarcity of ante-mortem (AM) evidence. The International Commission on Missing Persons has identified over 70 percent of the 40,000 victims using DNA and anthropological methods (Parsons, 2016).

This research explores two efforts designed to increase child identifications in mass victim disasters: using body length to distinguish PM DNA samples potentially belonging to children for targeted testing, and to single out deceased parents of missing children who were previously identified by a modality other than DNA, in order to retrieve and test their PM samples as references for kinship matching.

Comparison of adult and child identification proportions

During the first 100 days of the TTVI operation (January 12, 2005–April 25, 2005) 1,289 victims were identified, only a fraction of which (6.4 percent), were children; 78 compared to 1,211 adults. These data comprise victim identifications from six nationalities, each with at least ten child victims, and represents the early through to late stages of the operation (Table I). Two records were excluded from analysis because the victims’ dates of birth were unknown. It is clear that in the early months of the operation a far greater proportion of adults than children were being identified. For instance, in the first 50 days, 56.3 percent (218/387) of Swedish adults were identified compared to 5.2 percent (6/115) of Swedish children. At 100 days, the disparity was still evident, with 91.5 percent (354/387) of Swedish adults identified and only 27.0 percent (31/115) of Swedish children.

Identification rates for children of specific ages were further explored. Given similar trends in the average number of days required to identify French, German, Finnish and Swedish children, their data were combined for these analyses. A negative relationship was found between age and the average number of days to identification (p<0.0001), with children under the age of 1 year requiring an average of 195 days compared to 130 days for children aged 16 years (Figure 4). An identification took 4.3 days longer for each additional year the child was under the age of 16 (95% CI =2.9–5.7 days). Lack of AM and PM evidence for child victims was likely responsible for the low proportion of child identifications.

Body length as a proxy for age

Operational policies of the TTVI dictated that “every deceased person will be treated equitably to ensure there is no discrimination based upon age, race or ethnicity” (INTERPOL Tsunami Evaluation Working Group, 2014, p. 59), and guaranteed that all victims, including children, were analyzed and reported under the one standard process. In response to the inherent limitations in trying to identify children, a number of procedures were used to augment the collection of AM evidence for children. For example, finger paintings created by missing children were requested from their schools. Conversely, increasing AM data collection also required PM data for comparison. Therefore, a decision was taken to locate PM DNA samples potentially belonging to children to expedite sample testing and the return of DNA profiles from child victims. Creating a separate subgroup of child profiles lowered the prior probability for statistical calculations of kinship matches from 1/3,000 to 1/500 (3,000 total victims vs approximately 500 children). However, there was no mechanism in place to discriminate adult DNA samples from those potentially belonging to children.

Using height as a proxy for age, the PlassData’s^TM DVI System International was queried for case files in which the victim’s body length was described as “148 cm or less.” Victim height was selected as the discerning character because changes from decomposition made it difficult to visually assess age or nationality. The decision to use 148 cm or less was taken to capture children based on their shorter stature compared to adults. This threshold measurement was intended to include as many children as possible, while excluding short-statured adult victims. The threshold also had to account for the large variation in height among adult victims, given the number of nationalities represented, so the minimum likely adult height was based on Thai standards. The average height for Thai females is 159 cm±7.11 cm (https://tall.life/height-percentile-calculator-age-country/), therefore shorter adults would be nearly 152 cm tall. While some taller children may have been excluded based on the selected threshold, especially those closer to 16 years old, a larger threshold that is adjusted depending on victim demographics may avoid the problem in future events.

Untested post-mortem (PM) samples for kinship references

Child identifications were also hindered by weak kinship matches that were unable to reach the required reporting threshold of 99.9 percent. Weak matches were often the result of a lack of DNA reference samples from close relatives. Young children usually share living quarters with family members or are often closely accompanied by a family member. As such, many parents and children died together in the tsunami. The parent was therefore also a victim, unavailable to donate DNA or provide personal items for comparison. Similar circumstances were encountered following Hurricane Katrina in Louisiana, where nearly 80 percent of the state flooded, destroying personal items and items containing DNA. In addition, a number of the Katrina victims were related, also limiting available kinship references (Donkervoort et al., 2008).

Without either additional AM DNA reference samples or additional kinship profiles to establish matches, it would have been difficult to identify many of the children. However, numerous adult victims had been identified during the early months of the operation by dental and fingerprint matching, and some of these adults were also known to be parents of missing children. Although the consequences were unforeseen at the time, PM DNA samples from victims identified by a method other than DNA had been collected, though were not sent for testing in order to prioritize resources. As a result, the availability of identified adult PM DNA samples for use as references to establish kinship matches with unidentified children was limited.

In response, a family tree for each victim was uploaded into PlassData^TM, noting family members who were also victims of the tsunami, their identification status and PM number if identified. These data were continuously updated as new identifications were established, and then used to locate and retrieve untested DNA samples from identified individuals with missing relatives, to be used as kinship references for matching. This strategy was successful because DNA samples were collected during the PM examination from every victim from the outset of the operation, even if they were not initially sent off for testing.

Outcomes of targeted strategies

The decision to use body length to locate and prioritize PM DNA samples potentially belonging to children resulted in 181 targeted bone sample submissions. The criterion to distinguish the 181 samples used height as a proxy for age to discriminate adult samples from those potentially belonging to children. By July 2005, more than half of the targeted bone samples produced DNA profiles within four weeks of being received by the laboratory. These profiles were then prioritized for matching with children. From July 2005 to December 2005, 210 of the 231 child identifications were accomplished using DNA methods, clearly illustrating that the targeted return of child PM DNA profiles facilitated child identifications. Creating a separate subgroup of child profiles also lowered the prior probability for statistical calculations of kinship matches from 1/3,000 to 1/500 (500 being the approximate number of deceased children).

The prioritization of child PM DNA would only provide identifications if sufficient AM DNA existed. In May 2005, a review of 767 case files of unidentified individuals in PlassData^TM revealed that 464 (61 percent) cases did not have sufficient AM DNA available, whether direct or kinship references. A large number of these cases involved child victims. In response, investigators targeted unidentified child AM case files, and requested family tree information that specifically included related adult victims and their identification status. From these data, previously untested bone samples from family members identified by modalities other than DNA were retrieved from the mortuary and submitted for DNA testing to be used as references samples for identifying missing relatives. PM DNA profiles from identified adults were then entered into the child’s AM file (INTERPOL AM Form, E4 page) and used for matching. Concurrent family tree data were also uploaded to the files, which was regularly updated by investigators as more adult identifications were achieved. By the end of July 2005, 1,368 AM DNA profiles (direct and kinship) were available in PlassData^TM. Of these, 347 were from untested PM samples from previously identified individuals, adding kinship data from mothers, fathers, siblings, or a combination, leading to approximately 30 identifications.

The data in Figure 5 demonstrate an increase in the percentage of child-to-adult identifications from mid-April 2005, when the two targeted strategies to facilitate child identifications were introduced, until the identification percentages almost equalized (the total data are until December 12, 2005).

In May 2005, there was an increase in children identified by DNA, which was more rapid and heightened than seen with the adults. Moreover, this increased rate of child identifications was sustained throughout the year, until child–adult identifications by DNA nearly equalized. Figures 6 and 7 illustrate the monthly increase in DNA-based child identifications compared with child identifications by all methods.

Reflections/discussion

Low numbers of DNA identifications during the early stages of a DVI operation is typical, as a number of factors must occur before DNA is fully available to establish identifications. Early identifications are often characterized as straight-forward and usually include dental and fingerprint matches. More complicated identifications, and those that rely on DNA, are more typical during the middle and later stages of an operation. While part of the increase in child identifications may be attributed to overcoming some of these inherent issues (memorandums of agreements with laboratories, established list of missing persons, collection and processing of AM samples, body recovery and mortuary operations, processing of PM samples, and more), the difference between child and adult identifications was stark. In particular, the proportion of adult identifications by DNA steadily followed this typical pattern, while children identified by DNA did not begin to increase in number until May 2005, indicating other causative factors. The measures implemented to increase child identifications may have contributed to this increase. However, the respective contributions of the two strategies are difficult to tease out. Over time, the continued identification of children conformed to the typical DVI pattern characterized by increasing percentages of identifications relying on DNA as the operation progressed (Figures 6 and 7).

The family trees uploaded into PlassData’s^TM helped reveal the extent of missing related victims. Typically, kinship matching software searches for matches using an AM-to-PM algorithm, while a second kinship matching algorithm aggregates PM samples potentially belonging to related family members. A similar PM-to-PM match approach was used during the World Trade Center Human Identification project, to aggregate unidentified fragments with the same DNA profile, even if that profile had not yet been associated with a particular victim. When the profile was eventually matched to a victim, the entire group of fragments was linked to the identification (Shaler, 2005). The PM-to-PM approach taken at the TTVI identified kinship relationships among unidentified remains, some reaching the reporting threshold of 99.9 percent. Familial combinations included parent–child pairs, two parents and one child, and one parent and two children. In a number of instances, at least one adult family member was identified through this process. As such, the PM sample from the newly identified adult could then be used to identify related children. Once one individual from a family tree unit was identified, the other related-deceased were putatively identified, followed by additional targeted comparisons that progressively enabled identification using DNA kinship analysis.

Conclusion

The DVI operation in Phuket, Thailand following the South-East Asia Tsunami highlights some of the challenges associated with a large number of child victims, particularly when related family members are also missing. Initially, children proved more difficult to identify than adults because the methods used to identify the adults (fingerprints and dental comparisons) were not as successful in identifying children. Therefore, two child targeted procedures were implemented which proved successful: using body length to distinguish PM DNA samples potentially belonging to children for targeted testing, and to single out deceased parents of missing children who were previously identified by a modality other than DNA, in order to retrieve and test their PM samples as references for kinship matching. In combination, these strategies contributed to improvements in the number and rate of child identifications. From July 2005 to December 2005, 210 children were identified using DNA methods, from a total of 231 children identified using all methods. This investigation indicates that understanding the different identification challenges for adults and children can augment DVI efforts, and may necessitate enhanced AM and PM collection strategies, as well as changes to operational policies, communication policies for families and the media, and resource allocation. Based on the success of the child targeted strategies discussed here, we recommend implementing targeted strategies that are be based on the unique demographic composition of the victims.

Note

1.www.indo.com/bali121002_bekup/mourning4.html (accessed April 2, 2015).

Figure 1

Age and sex distribution of the 442 non-Thai child victims

Figure 2

A family with victims across three generations

Figure 3

A local Thai family that lost twenty-two members in the tsunami. ‘X’ represents relatives who were deceased prior to the tsunami. Black shapes represent tsunami victims; white shapes represent living relatives

Figure 4

Average time in days required to identify children using all methods as a function of age

Figure 5

Increases in child identifications, by country, following implementation of targeted child identification strategies at approximately 100 days into the operation

Figure 6

Percentage of total adult and child identifications established using DNA throughout the first year of the TTVI operation (all nationalities)

Figure 7

The percentage of total child identifications established each month by each identification method (all nationalities)

Table I

Adults and children identified in the first 50 and 100 days of the TTVI operation, by country

Country	Adult/child	Victims identifieda	% IDb at 50 days	% IDb at 100 days
Finland	Adult	109	67.0	89.0
	Child	51	5.9	29.4
France	Adult	49	28.6	53.1
	Child	18	5.6	33.3
Germany	Adult	443	28.5	84.4
	Child	35	5.7	28.6
Norway	Adult	55	90.7	90.7
	Child	22	17.4	65.2
Sweden	Adult	387	56.3	91.5
	Child	115	5.2	27.0
Thailand	Adult	893	0.5	3.7
	Child	155	0.0	0.6

Notes: ^aThe number of identifications achieved by December 12, 2005; ^bidentified