Printer Friendly

Age Estimation in a Colombian Modern Skeletal Sample A Test of the Transition Analysis Method.

Introduction

Skeletal age estimation is essential to forensic anthropology, as it is part of the identification process of human remains in the medico-legal setting. In the age estimation of an individual, the narrower the estimate, the more useful the skeletal analysis is for identification purposes. However, reporting a reduced age range may not only be unrealistic but also misguided given the variability in patterns of skeletal aging associated to genetics, health status, climate, cultural practices, and socioeconomic status, among others. Because aging is a highly variable process, chronological and biological age are rarely equivalent; thus age estimates are approximations based on observed similarity and concordance (Algee-Hewitt 2017).

Methods concerned with age-at-death estimation are based on the developmental changes related to growth or the degenerative features often associated with aging. Methods developed in the discipline of biological anthropology use single or multiple indicators examined using phase-based or component-based systems. Until very recently, validation studies for age estimation of adult individuals in Colombia based on skeletal methods were absent in the literature. In fact, the one and only study tests four methods based on the innominate bone (Rivera-Sandoval et al. 2018). Thus, local forensic anthropologists have been relying on foreign phase-system methods to estimate age. The most common methods used in Colombia include Todd (1920, 1921) and Suchey-Brooks (Brooks & Suchey 1990) for the pubic symphysis and Iscan et al. (1984) for the sternal rib ends. However, phase-based methods are regularly criticized because they fail to reflect biological reality (Algee-Hewitt 2017). These methods carry the assumption that the degenerative changes in various skeletal traits occur in lockstep, thereby allowing researchers to lump groups of traits into broad phases that often overlap (Dudzik & Langley 2015; Shirley & Ramirez Montes 2015). Additionally, adult phase-aging methods offer subjective interpretations of trait morphologies, which results in high error rates and have been demonstrated to be useless for the age estimation of old individuals (Boldsen et al. 2002). On the other hand, component-based methods divide all traits into individual components that are scored separately, thus offering the possibility for more objective scoring, while presuming that the features incorporated in the method undergo a degenerative progression with age (Boldsen et al. 2002; Shirley & Ramirez Montes 2015).

Transition analysis (TA) is a component-based, multifactorial aging method that uses components from different anatomical elements (cranial sutures, pubic symphysis, and auricular surface). The aim of this method is to capture the sequential aging process that occurs among individuals at different rates in order to provide more accurate (greater correspondence between chronological and estimated ages) and precise (tighter age interval lengths) age-at-death estimations. The ultimate goal of TA is to capture the complex biological changes as well as the variation in the human skeleton and provide a more holistic understanding of skeletal aging patterns (Algee-Hewitt 2017). The advantages of adopting the TA approach, though not exclusively, are twofold. First, the application of Bayesian statistics not only reduces age mimicry but also provides better age estimates for old individuals (50+ years), as age estimates are individualized and thus have their own measure of certainty (Boldsen et al. 2002). Second, these types of statistics offer important information on the uncertainty of age estimates, which responds to the traditional problem of unknown error rates (Fojas et al. 2017). The TA methodology was developed as part of "The Rockstock Manifesto for Paleodemography," which brought together biological anthropologists, demographers, and statisticians to solve theoretical and methodological problems associated with reconstructing demographic structure from skeletal samples (Hoppa & Vaupel 2002). The need for better osteological methods that applied a Bayesian focus and thus addressed the problem of accuracy and reliability of age-estimation techniques urged the development of the TA methodology, which has been tested extensively over the last few years (Bethard 2005; Gomez & Algee-Hewitt 2011; Hurst 2010; Milner & Boldsen 2012a; Shirley et al. 2010; Wilson & Algee-Hewitt 2009).

In response to the necessity of establishing scientific validity of applied methods and techniques and thus meet evidentiary standards required by the Daubert ruling (Lesciotto 2015; Rivera-Sandoval et al. 2018), this study examined the ability of the TA method to return accurate and precise age estimates for a sample of individual modern Colombian skeletons. Additional objectives of this study resemble a subset of the goals set by Milner and Boldsen (2012a) and included the following: (1) establish which anatomical structure(s) worked best for estimating age; (2) determine interval bias for different age cohorts, and (3) determine where the greatest uncertainty was with respect to age intervals obtained. Finally, interobserver reliability was evaluated between two examiners in an effort to evaluate the repeatability of the method in an approach similar to Fojas et al. (2017).

Methods

A total of 58 individuals drawn from the Antioquia Modern Skeletal Collection curated at the National School of Criminology and Forensic Sciences located in Medellin, Colombia, were used in this study. Specimens from this collection include identified bodies from a municipal cemetery that, following the expiration of time allotted in the cemetery, as assigned by the government, were not claimed by their families. Thus, the disposition of the remains was arranged by the cemetery, which donated them to the National School of Criminology for scientific purposes. All specimens have demographic data and cause-of-death information associated.

Individuals were selected on the basis of availability of recorded sex and age. No specimens with trauma or pathological conditions were included in the analysis. The sample consisted of 43 males and 15 females with ages between 16 and 83 years (mean age = 47.28 years, SD = 21.93). Young adults (ages 21-40) and old individuals (ages 61-83) dominated the sample (Fig. 1). Other age groups included late adolescents (ages 16-20) and middle-aged individuals (ages 41-60). From the total sample, 31 individuals from all age categories had fragmentary and incomplete crania from which only one trait could be scored. Additionally, six specimens from categories classified as young adults and old individuals had damaged and incomplete pelvic structures from which 25-50% of the features could not be scored. Forty-six percent of the total sample had all features available to score. Anatomical structures were scored by the observers following the descriptions contained in the transition analysis age estimation-skeletal scoring manuals (Milner & Boldsen 2013, 2016). Researchers blindly applied the method to chronological age, and raw scores were input into the ADBOU software package (Milner & Boldsen 2012b). Ancestry was marked as unknown, because although place of birth was known, the "black" and "white" categories available in the program do not fit the admixture characteristic of the Colombian population. Best point estimates as well as 95% confidence intervals were obtained using the forensic prior distribution, which is based on a 1996 report on U.S. homicide deaths (Boldsen et al. 2002). The forensic distribution was selected over the archaeological prior derived from 17th-century Danish parish records, as the Colombian skeletal sample has a higher frequency of young adult males whose cause of death is associated to gunshot trauma, thus sharing similar demographics with the forensic prior distribution.

With the aim of evaluating the ability of the method to provide accurate age estimates, Pearson correlation coefficient was calculated between chronological and estimated ages. Although ADBOU reports point estimates as well as 95% confidence intervals for each anatomical structure, the age estimates used for the correlation test were those obtained through the analysis of features from the three anatomical structures combined. Point estimates from all structures were also used to calculate inaccuracy and bias. Inaccuracy or absolute error is calculated as the absolute difference between the estimated and the documented age (|estimated age-documented age|) divided by the sample size (n), while bias indicates under- or overestimation of the individual's documented age. Bias is estimated similarly but takes into consideration the [+ or -] sign. Both estimates are given in years. Displays of raw data were built to provide a sense of accuracy and precision of the TA method and whether estimated intervals encompassed chronological ages. Statistical tests were calculated using SPSS, version 23 (IBM Corp. 2015).

In order to assess interobserver reliability and the proportion of agreement between two examiners, a subsample of 16 individuals was randomly selected. Both observers have between five and ten years of experience with traditional age estimation methods, but Observer 1 was trained in the TA method directly from the developers, whereas Observer 2 was trained secondhand. A number of coefficients can be applied in the calculation of interobserver reliability. Cohen's kappa is perhaps the most popular in the medical and forensic fields (Shirley & Ramirez Montes 2015; Wongpakaran et al. 2013), but Cohen's kappa indices pose two paradoxes: (1) a low kappa can occur despite the high percentage of agreement, and (2) unbalanced marginal distributions generate higher values of kappa than more balanced marginal distributions (Wongpakaran et al. 2013). Thus, kappa varies considerably with prevalence and bias at equal agreement rate. Alternatively, Gwet's AC1 coefficient overcomes the paradoxes in Cohen's kappa, as AC1 does not punish examiners who produce similar ratings or marginal distributions. Chance agreement is adjusted such that the AC1 between two or more examiners is expressed as the conditional probability that two randomly selected examiners will agree, given that no agreement will occur by chance (Gwet 2014). Finally, bias is minimized in the definition of chance agreement for AC1 by using average marginal distributions as its estimates of proportions (Xie [n.d.]). Thus, AC1 is considered a more robust measure of reliability. Gwet's AC1 coefficients were calculated for the combined sex subsample with 95% confidence intervals. Due to the small sample size of individual sex subsamples (11 males and 5 females), AC1 values were not calculated for males and females separately. Ten traits were scored on the pubic symphysis (5 on each side) and 18 on the auricular surface (9 on each side), for a total of 28 traits. Cranial features were not included, since most individuals had fragmentary and incomplete skulls whose total traits could not be scored. AC1 values range from 0 to 1, where 1 denotes greater reliability, that is, an almost perfect match, and values close to 0 indicate lesser reliability. Specific guidelines for interpretation proposed by Fleiss et al. (2003) are the following: <.40 = poor agreement, .41-75 = fair to good agreement, and .75-1= very good agreement. Gwet's AC1 values were calculated using AgreeStat 2015.6 (Advanced Analytics, LLC 2015).

According to Gwet's (2010, 2013) approach to calculate sample size for agreement coefficient, the sample size depends on the desired error margin. The following is the formula for calculating minimum necessary sample size (n):

[mathematical expression not reproducible]

N is the number of subjects in the entire population; r is the relative error (the difference between the true value obtained from the population and the estimated value obtained from the sample, effect size); [pi].sub.A] is the overall agreement probability; and [pi].sub.E] is the chance-agreement probability (Gwet 2014). Given that some parameters are unknown at the design stage, Gwet (2010) proposes the rule of thumb of assuming a best-case scenario where chance-agreement probability is 0, and [pi].sub.A] is assigned an anticipated value to obtain the absolute minimum sample size. Thus, for an error margin of 25% for the interobserver reliability test, the required sample size was 16 individuals.

Results

The Pearson correlation test shows a high correlation index (0.9) (p < 0.001) between estimated and chronological age and a coefficient of determination ([R.sup.2]) of 0.81, indicating that there is a strong correlation between estimated age and age at death. Bias and inaccuracy indices (Table 1) and displays of raw data (Fig. 2) provide abetter sense of the method's capacity to encompass chronological ages for each age cohort. These figures represent age estimates based on traits from each anatomical structure and features from all structures combined in which individuals were arranged from young to old ages. Plots illustrate correspondence between chronological ages, point estimates, and 95% confidence intervals, so accuracy and precision can be inferred by looking at this correspondence.

Of the three anatomical structures, the pubic symphysis provided better estimates, especially for late adolescents and individuals in the second decade of life. However, it tended to overestimate the ages of the oldest individuals, as can be seen by the correspondence between the bars and the chronological ages (Fig. 2b) and bias values of 10.35 and 22.6 years. Additionally, the pubic symphysis provided age intervals that were particularly wide for individuals in the third decade as well as for middle-aged and old individuals. The same trends were observed for the auricular surface (Fig. 2c), which tended to overestimate the ages of the old individuals by 16 and 18.8 years. Both the pubic symphysis and the auricular surface provided relatively accurate estimations for late adolescents and individuals between 21 and 30 years of age, as inaccuracy indices ranged from 2.1 to 6.3 years.

Results from cranial sutures should be carefully considered, because only 27 individuals from the pooled sample had complete cranial features that could be scored, which clearly affects the interpretation of the results. Cranial sutures overestimated the ages of late adolescents (Fig. 2a) by 4.9 years and underestimated the ages of individuals between 21 and 30 years of age, middle-aged, and old individuals by 5.6 to 29.3 years. Age intervals provided by cranial sutures were extremely wide for middle-aged and old individuals, and the inaccuracy index for individuals aged 30 and beyond increased dramatically in comparison to results obtained from other anatomical structures. In contrast, when features from all structures were combined (Fig. 2d), age estimates improved in terms of accuracy and precision, especially for late adolescents and young adults, as inaccuracy indexes ranged from 1.75 to 8.05 years compared to higher values provided by individual structures. Additionally, interval lengths were narrower for these age groups. Estimated ages for middle-aged individuals fell within intervals of 31 to 41 years, but the ages of individuals representing this age group were underestimated by 11.5 years. Thus, the greatest uncertainty in age estimation for this test sample was for middle-aged individuals. Although there was slight under-and overestimation of the ages of old individuals (-1.76 years to 2.37), when all traits were combined, estimates for this age group improved in accuracy and precision in comparison to results obtained by features from individual structures.

Gwet's AC1 coefficients calculated to evaluate interob-server reliability (Tables 2 and 3) indicate that for the pooled subsample there was fair to good agreement in most of the traits and very good agreement in eight traits, with the superior protuberance of the right side of the pubic symphysis as the trait with the greatest agreement. However, there were five traits in which there was poor agreement, such as the dorsal symphyseal texture of the right side of the pubic symphysis and the middle surface morphology of both sides of the auricular surface, among others. Of the randomly selected subsample for the interobserver error analysis, 69% of the cases were correctly estimated into confidence intervals by the examiner with less experience using the method (Observer 2), while 94% were correctly estimated by the examiner with more experience (Observer 1).

Discussion

The primary goal of this study was to evaluate if age estimates provided by the ADBOU software program involving features from the three anatomical portions are accurate and precise in relation to chronological age at death in a Colombian population subsample. Results from the Pearson correlation test indicate a strong association between point estimates and chronological ages, as 81% of the variability in age estimations in this test sample can be explained by the variability in chronological age. However, bias and inaccuracy values were particularly useful to examine how TA performed on each age cohort, as the correlation coefficient alone does not indicate where the method might have performed poorly. The analysis of the performance of the method when incorporating multiple anatomical features to age estimates will be discussed later in this section.

Although TA was designed to incorporate information from various anatomical regions from the skeleton, in recent years researchers have also evaluated the performance of the method in single anatomical areas (Hurst 2010; Milner & Boldsen 2012a). This study indicates that in the final contribution to age estimation, the pubic symphysis performed slightly better than the auricular surface. The pubic symphysis provided narrow age intervals for late adolescents, similar to phase-based methods that use the same anatomical structure. Furthermore, this element offered fairly close point estimates to chronological ages for late adolescents. Age intervals were also promising for individuals between 21 and 30 years of age; however, for middle-aged and old individuals, age intervals broadened so substantially that neither the pubic symphysis nor the auricular surface provided significantly useful information about age. Other scholars have reported age underestimation of middle-aged and old individuals by the auricular surface (Hens et al. 2008; Milner & Boldsen 2012a), including the study by Rivera-Sandoval et al. (2018), which found that Lovejoy etal. (1985), Buckberry and Chamberlain (2002), and Rouge-Maillart et al. (2009) underestimate individuals older than 65 years. Conversely, the analysis presented here suggests the opposite for a Colombian population: both the auricular surface and the pubic symphysis overestimated the ages of old individuals. This finding indicates that the forensic prior distribution may have influenced the results, particularly in older age where the effects of this distribution are most evident. The forensic prior distribution was calculated from the 1996 homicide data, which has a higher frequency of young males as opposed to old individuals (Hurst 2010).

Contributions of cranial sutures, independently from the other structures examined, could not be fully considered in this study due to the small number of crania with all features scored. Nevertheless, the lack of cranial sutures does not affect the performance of the procedure, as in forensic casework skeletons are often found without crania or with incomplete crania, as exemplified by the sample of skeletons examined in this study. In the application of the method, however, it is not recommended that TA be used to estimate the age of individuals with only one anatomical element present, particularly if that element is the cranium.

As expected based on the design of the original method, age estimates improved when traits from all anatomical units were combined. For late adolescents (ages 16-20) and young adults between 21 and 30, estimates were accurate, while for individuals between 31 and 40 and old individuals (ages 61-83), estimates improved in precision, although intervals extended to 50 years in length in the latter group. Age calculations for middle-aged individuals (ages 41-60) tended to be underestimated, though this may ultimately be an effect of a small sample size. As pointed out by Milner and Boldsen (2012a), the long-standing problem of accurate and precise age estimations for middle-aged individuals persists. No pelvic traits seem to be particularly characteristic of middle age, and thus the issue seems to be biological, as signs of variability increase with skeletal aging at this age period. On the other hand, TA age estimates for old individuals are particularly promising considering this is one of the first validation studies for age estimation using a modern Colombian skeletal sample. Interval lengths decrease in old individuals due in part to selective mortality (Milner and Boldsen 2012a); that is, only the fittest individuals at truly advanced years will survive, thus reducing the variation in skeletal traits related to aging, a trend that TA seems to capture well. Results presented here demonstrate good utility because they provide smaller age intervals not only for old individuals but also for young adults in comparison to the foreign phase-based methods of the pubic symphysis that are currently applied in Colombia. Furthermore, results obtained by this study offer the possibility of avoiding open-ended terminal age intervals in the forensic analysis of individuals of truly advanced age.

Gwet's AC1 coefficients for the analysis of interobserver reliability indicate that the two observers were in fair to good agreement in the scores given to the majority of the traits, with poor agreement in 5 of the 28 traits scored. Shirley and Ramirez Montes (2015) report that in a component system, disagreement between observers is more likely to occur when there are more than three coding possibilities for each trait. However, results from this study indicate that discrepancies between examiners are likely due to other variables. First, although both examiners have applied phase-based methods for age estimation in the past eight years, Observer 1 has more experience with the TA method. Second, problems with translation and/or descriptions could have played a part given that the method is written in English and some of the descriptions do not translate well into Spanish (although the analysis was complemented with the use of photographs). For instance, words like "rampart" or "coarse grained/packed fine sand" do not adequately translate into Spanish in the context of skeletal analysis, thus affecting the possibility of applying the method in a consistent manner. In addition, although both observers are native Spanish speakers, Observer 1 is fluent in English, which might have affected the application of the method. Lastly, since TA is a relatively new method, many Colombian forensic anthropologists are not yet very familiar with its use.

Conclusions

Overall, results indicate that TA did not perform equally well for all age cohorts in this test sample. For late adolescents estimates were both accurate and precise, and for young adults between 21 and 30 years estimates were still accurate but decreased in precision. For individuals in the third decade of life the method appeared to perform relatively well, although precision decreased in comparison to age estimates of those in the second decade. Middle-aged individuals were significantly underestimated, although the sample size analyzed in this study was likely too small to provide reliable results. Estimations for old individuals were far from ideal--particularly in terms of precision--but better than estimates provided by phase-based systems that arbitrarily force the examiner to fit the complex variation observed in the aging process into one or more phases.

As confirmed by the analysis of the contribution of each skeletal structure to the method's performance, the sole focus on cranial sutures and pelvic joints affects final age estimates, as features from all three anatomical units combined do not fully encompass the skeletal variation related to age progression. Hence, additional age-informative features from various portions of the skeleton need to be taken into consideration in an effort to design and validate TA age-estimation methods (Getz et al. 2017; Milner et al. 2016). Meanwhile, results from this article indicate that the method works better for young and old individuals, thus demonstrating its utility in forensic applications. This method may be applied to middle-aged individuals; however, practitioners should be aware of the methodological biases outlined above. Moreover, additional work that examines a larger sample of modern Colombian skeletons is necessary to determine if results found in this study are consistent. Finally, this article underlines the necessity of translating the TA method into Spanish and disseminating it in the Colombian forensic anthropology community.

Acknowledgments

The authors would like to thank John Fredy Ramirez Santana, who kindly provided access to the Antioquia Modern Skeletal Collection in Medellin, Colombia.

References

Advanced Analytics, LLC. Agree-Stat2015.6-Mac, Gaithersburg, MD, 2015. http://www.agreestat.com/. Accessed December 1, 2017.

Algee-Hewitt BFB. Appendix B: Age estimation in modern forensic anthropology. In: Langley NR, Tersigni-Tarrant MA, eds. Forensic Anthropology: A Comprehensive Introduction. 2nd ed. Boca Raton, FL: CRC Press; 2017:381-419.

Bethard J. A Test of the Transition Analysis Method for Estimation of Age-at-Death in Adult Human Skeletal Remains [master's thesis]. Knoxville: University of Tennessee; 2005.

Boldsen JL, Milner GR, Konigsberg LW, Wood JW. Transition analysis: A new method for estimating age from skeletons. In: Hoppa RD, Vaupel JW, eds. Paleodemography: Age Distributions from Skeletal Samples. Cambridge: Cambridge University Press; 2002:73-106.

Brooks S, Suchey JM. Skeletal age determination based on the os pubis: A comparison of the Acsadi-Nemeskeri and Suchey-Brooks methods. Human Evolution 1990;5(3):227-238.

Buckberry JL, Chamberlain AT. Age estimation from the auricular surface of the ilium: A revised method. American Journal of Physical Anthropology 2002;119(3):231-239.

Dudzik B, Langley NR. Estimating age from the pubic symphysis: A new component-based system. Forensic Science International 2015;257:98-105.

Fleiss JL, Levin B, Paik MC. Statistical Methods for Rates and Proportions. 3rd ed. New York: Wiley-Interscience; 2003.

Fojas CL, Kim J, Minsky-Rowland JD, Algee-Hewitt BFB. Testing inter-observer reliability of the Transition Analysis aging method on the William M. Bass forensic skeletal collection. American Journal of Physical Anthropology 2017;165(1): 183-193.

Getz SM, Milner GR, Boldsen JL. Rethinking adult skeletal age estimation: An expanded approach to transition analysis using binary traits. In: Proceedings of the 69th Annual Meeting of the American Academy of Forensic Sciences, February 13-18, 2017; New Orleans, LA.

Gomez JM, Algee-Hewitt BFB. Does practice make perfect? A test of the effect of experience on age-at-death estimation using transition analysis. In: Proceedings of the 80th Annual Meeting of the American Association of Physical Anthropologists, April 13-16, 2011; Minneapolis, MN.

Gwet KL. Handbook of Inter-Rater Reliability: The Definitive Guide to Measuring the Extent of Agreement Among Raters. 4th ed. Gaithersburg, MD: Advanced Analytics, LLC; 2014.

Gwet KL. Inter-rater reliability discussion corner. Sample size determination. June 28, 2010. https://agreestat.com/blog_irr/sample_size_determination.html. Accessed May 5,2018.

Gwet KL. Sample size calculations for agreement coefficients. May 6, 2013. https://agreestat.com/blog_irr/articles/blog2013_05may06.pdf. Accessed May 5, 2018.

Hens SM, Rastelli E, Belcastro G. Age estimation from the human os coxa: A test on a documented Italian collection. Journal of Forensic Sciences 2008;53(5):1040-1043.

Hoppa RD, Vaupel JW. The Rockstock Manifesto for palodemog-raphy: The way from stage to age. In: Hoppa RD, Vaupel JW, eds. Paleodemography: Age Distributions from Skeletal Samples. Cambridge: Cambridge University Press; 2002:1-8.

Hurst C V A test of the forensic application of transition analysis with the pubic symphysis. In: Latham KE, Finnegan M, eds. Age Estimation of the Human Skeleton. Springfield, IL: Charles C. Thomas; 2010:262-272.

IBM Corp. IBM SPSS Statistics for Macintosh, Version 23.0, Armonk, NY, 2015. http://www-01.ibm.com/support/docview.wss?uid=swg24038592. Accessed February 2, 2018.

Iscan MY, Loth SR, Wright RK. Metamorphosis at the sternal rib end: A new method to estimate age at death in white males. American Journal of Physical Anthropology 1984;65(2): 147-156.

Lesciotto KM. The impact of Daubert on the admissibility of forensic anthropology expert testimony. Journal of Forensic Sciences 2015;60(3):549-555.

Lovejoy CO, Meindl RS, Pryzbeck TR, Mensforth RP Chronological metamorphosis of the auricular surface of the ilium: A new method for the determination of adult skeletal age at death. American Journal of Physical Anthropology 1985;68(1): 15-28.

Milner GR, Boldsen JL. Transition analysis: A validation study with known-age modern American skeletons. American Journal of Physical Anthropology 2012a;148(l):98-110.

Miner GR, Boldsen JL. Transition analysis program. Fordisc Version 1.0, 2012b. http://anth.la.psu.edu/research/research-labs/bioarch/docs/ADBOU2Setup.zip/view/Accessed December 15,2015.

Miner GR, Boldsen JL. Transition analysis age estimation: Skeletal scoring manual. Fordisc Version 1.0, 2013. http://anth.la.psu.edu/research/research-labs/bioarch/docs/TAManual2013June.pdf/view/Accessed December 15, 2015.

Miner GR, Boldsen JL. Transition analysis age estimation: Skeletal scoring manual. Fordisc Version 1.02, 2016. http://math.mercyhurst.edu/~sousley/Software/Accessed December 20,2016.

Milner GR, Boldsen JL, Ousley SD, Getz SM, Weise S, Tarp P. Estimating age from adult skeletons: New directions in transition analysis using a wide array of traits. In: Proceedings of the 85th Annual Meeting of the American Association of Physical Anthropologists, April 12-16, 2016; Atlanta, GA.

Rivera-Sandoval J, Monsalve T, Cattaneo C. A test of four innominate bone age assessment methods in a modern skeletal collection from Medellin, Colombia. Forensic Science International 2018;282:232.el-232.e8.

Rouge-Maillart C, Vielle B, Jousset N, Chappard D, Telmon N, Cunha E. Development of a method to estimate skeletal age at death in adults using the acetabulum and the auricular surface on a Portuguese population. Forensic Science International 2009;188(l-3):91-95.

Shirley NR, Algee-Hewitt BFB, Wilson RJ. The reality of aging virtually: a test of transition analysis on pelvic laser scans. In: Proceedings of the 79th Annual Meeting of the American Association of Physical Anthropologists, April 9-12, 2010; Albuquerque, NM.

Shirley NR, Ramirez Montes PA. Age estimation in forensic anthropology: Quantification of observer error in phase versus component-based methods. Journal of Forensic Sciences 2015;60(1):107-111.

Todd TW. Age changes in the pubic bone I: The male white pubis. American Journal of Physical Anthropology 1920;3(3):285-334.

Todd TW. Age changes in the pubic bone II: The pubis of the male Negro-White hybrid; III: The pubis of the White female; IV: The pubis of the female Negro-White hybrid. American Journal of Physical Anthropology 1921;4(l):l-70.

Wilson RJ, Algee-Hewitt BFB. [Interfacing age: A test of the ADBOU age estimation software in a forensic context. In: Proceedings of the 78th Annual Meeting of the American Association of Physical Anthropologists, March 31-April 4, 2009; Chicago, IL.

Wongpakaran N, Wongpakaran T, Wedding D, Gwet KL. A comparison of Cohen's Kappa and Gwet's AC1 when calculating inter-rater reliability coefficients: A study conducted with personality disorder samples. BMC Medical Research Methodology 2013;13(l):1-7.

Xie Q. Agree or disagree? A demonstration of an alternative statistic to Cohen's Kappa for measuring the extent and reliability of agreement between observers. MacroSy s, LLC. (n.d.) https://s3.amazonaws.com/sitesusa/wp-content/uploads/sites/242/2014/05/J4_Xie_2013FCSM.pdf/Accessed December 5, 2017

Maria Alexandra Lopez-Cerquera (a*) Diego Alejandro Casallas (b)

(a) Department of Natural and Behavioral Sciences, Pellissippi State Community College, 7201 Strawberry Plains Pike, Knoxville, TN 37914, USA

(b) Human Identification Working Group, Fiscalia General de la Nacion, Diagonal 22B No. 52-01, Bogota, Colombia

(*) Correspondence to: Maria Alexandra Lopez-Cerquera, Department of Natural and Behavioral Sciences, Pellissippi State Community College, 7201 Strawberry Plains Pike, Knoxville, TN 37914, USA

E-mail: malopezcerquera@pstcc.edu

Submitted 29 January 2018; Revised 22 May 2018; Accepted 20 June 2018

DOI: 10.5744/fa.2018.1030
TABLE 1--Bias and inaccuracy of age estimates by anatomical structure.
Age groups 41-60 and 71-83 were not subdivided into additional age
ranges given the small number of individuals that represented them.

       Pubic  Symphysis   Auricular Surface  Cranial Sutures
Ages   Bias   Inaccuracy  Bias   Inaccuracy  Bias    Inaccuracy

16-20  -1.05   2.15        2.68   6.28         4.96   5.36
21-30  -0.17   5.31       -3.11   6.11        -5.60   8.85
31-40   3.50  10.10        0.52  11.27         4.16  24.13
41-60  -4.84  16.57       -6.88  11.60       -29.35  29.35
61-70  10.35  19.80       16.09  25.00       -15.08  31.88
71-83  22.60  22.60       18.88  18.88       -15.74  22.46

       All Structures
Ages   Bias    Inaccuracy

16-20   -1.11   1.75
21-30   -1.14   5.10
31-40   -0.54   8.05
41-60  -11.56  12.50
61-70   -1.76  13.76
71-83    2.37   5.80

TABLE 2--Gwet's ACI values of pubic symphysis traits for the combined
sex subsample. Values in bold indicate poor agreement.

Pubic Symphysis            Right  Left
Symphyseal relief          .784   .704
Dorsal symphyseal texture  .378   .572
Superior protuberance      .893   .782
Ventral symphyseal margin  .512   .689
Dorsal symphyseal margin   .804   .593

TABLE 3--Gwet'sACI values of auricular surface traits for the
combined sex subsample. Values in bold indicate poor agreement.

Auricular Surface                   Right  Left

Superior demiface topography        .421   .766
Inferior demiface topography        .530   .653
Superior surface characteristics    .699   .480
Middle surface characteristcs       .381   .351
Inferior surface characteristcs     .386   .670
Inferior surface texture            .778   .843
Superior posterior iliac exostoses  .790   .649
Inferior posterior iliac exostoses  .355   .500
Posterior exostoses                 .593   .637
COPYRIGHT 2019 University Press of Florida
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2019 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:TECHNICAL NOTE
Author:Lopez-Cerquera, Maria Alexandra; Casallas, Diego Alejandro
Publication:Forensic Anthropology
Article Type:Report
Geographic Code:3COLO
Date:Jan 1, 2019
Words:5290
Previous Article:Success Rates of Forensic Surface Search for Osseous Remains in a New England, USA, Environment.
Next Article:Standardization of Postcranial Nonmetric Traits and Their Utility in Ancestry Analysis.
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |