Psychological Markers in Trafficking Victims, 2000-2025

KJKey Judgments

Eight judgments carry this brief. Each is an analytic conclusion, distinct from the reported facts beneath it, and each carries an explicit confidence level per §1.2. A judgment can be wrong even where every underlying fact is correct, and four of these were revised during verification.

1How to read this document

1.1 Source grading: the NATO/Admiralty scale

Every source cited in Section 12 carries a two-character Admiralty grade: a letter for the reliability of the source itself, a number for the credibility of the specific information cited. The two axes are independent, a highly reliable source can carry a specific claim that is only possibly true, and a source of unproven reliability can still carry information later confirmed elsewhere.

Source reliability:

• A: Completely reliable. No history of inaccurate reporting (a federal agency's own publication or official statistics page, speaking in its own voice).
• B: Usually reliable. Minor history of inaccurate reporting, or reliable but not a primary party to the events described (a peer-reviewed journal article, an established wire service, a credentialed investigative outlet, a research institution with transparent methodology).
• C: Fairly reliable. Some history of inaccurate reporting (a university press office relaying its own researchers' findings rather than the study itself; a regional news outlet relaying an official's statement rather than the agency's own release).
• D: Not usually reliable. Significant history of inaccurate reporting. (Not used as a load-bearing grade in this document; a source graded this low would not be cited for a factual claim.)
• E: Unreliable. Lacks a history of accuracy. (Not used in this document.)
• F: Reliability cannot be judged. (Used for a source with insufficient track record to grade, noted individually where it occurs.)

Information credibility:

• 1: Confirmed by other independent sources.
• 2: Probably true (source is reliable, information is plausible, but not independently confirmed by a separate source in this document).
• 3: Possibly true (some doubt; typically a single named source with no independent corroboration, or a convenience sample with a known funding or recruitment problem).
• 4: Doubtful (some internal contradiction, or later superseded).
• 5: Improbable (contradicted by other sources).
• 6: Credibility cannot be judged (insufficient information to assess).

The four-tier shorthand used in section subheadings (PRIMARY/OFFICIAL, NAMED INVESTIGATIVE, PEER-REVIEWED, SECONDARY/UNVERIFIED) is a quick reference only; the Admiralty grade in Section 12 is the authoritative, source-by-source grading. Rough correspondence: PRIMARY/OFFICIAL typically grades A1-A2; NAMED INVESTIGATIVE and PEER-REVIEWED typically grade B1-B2; SECONDARY/UNVERIFIED typically grades B3-C4 or lower.

1.2 Analytic confidence and the probability lexicon

This document distinguishes reported information (a fact directly stated by a source: “the study reported a sensitivity of 84.6 percent”) from analytic judgment (this document's own conclusion about what the reported information means, which requires interpretation and could be wrong even if every underlying fact is correct). A judgment is labeled “Assessment:” and carries two things: a confidence level and, where the judgment is expressed as a likelihood rather than a certainty, a calibrated probability term.

Confidence level (a function of source quality, quantity, and independence, not of how strongly the judgment is worded):

• High confidence: judgment is based on high-quality information, ideally corroborated by multiple independent sources using different methods (alternative explanations were assessed and found significantly less likely).
• Moderate confidence: judgment is credibly sourced but not independently corroborated, or the sourcing is corroborated but shares a common origin, or the analysis is sound but the sourcing is not extensive enough for High confidence.
• Low confidence: judgment rests on a single source, fragmentary information, or reasoning this document could not independently verify. A Low-confidence judgment is not a weak claim dressed up, it is this document's own honest statement that the evidence does not yet support more than that.

Calibrated probability lexicon (the words themselves, not just contextual tone, carry the meaning):

• Almost certainly: roughly 95-99 percent probability.
• Highly likely / very likely: roughly 80-95 percent.
• Likely / probably: roughly 55-80 percent.
• Roughly even chance: roughly 45-55 percent.
• Unlikely: roughly 20-45 percent.
• Highly unlikely / remote: below 20 percent.

A RECTUMmendation (an unverified assertion presented as settled fact, this project's standing term for it) is not the same failure as a properly labeled Low-confidence Assessment. The former hides its own uncertainty; the latter states it plainly. This document aims to never do the former and to do the latter whenever a judgment, rather than a directly sourced fact, is being made. On this subject the temptation toward the former is unusually strong, because the humanitarian stakes make hedging feel like indifference. It is not. A confidently wrong indicator list gets the wrong people flagged and the right people missed.

1.3 What stays a plain fact, not an assessment

Not every claim in this document needs a confidence label. A directly quoted study statistic, a specific reported sensitivity, or a directly quoted passage from a federal toolkit is reported information and is presented as such, with its Admiralty grade in Section 12 doing the epistemic work. Assessment labels and confidence levels are reserved for this document's own interpretive judgments, most concentrated in the Signals, Key Assumptions Check, and Analysis of Competing Hypotheses sections, plus the Bottom Line Up Front.

Sources rejected during research, and why, are logged in Section 8. Advocacy and mission-driven organizations, and partisan political-body factsheets, were excluded from load-bearing claims. This subject has an unusually dense advocacy literature and several of the most-circulated figures on it originate there, so the exclusion did real work in this session. Where an advocacy source was the only place a claim appeared, the claim was either dropped or re-sourced to a primary document; none ship here on advocacy sourcing alone.

1.4 Multi-model verification pass

This document was drafted by Claude (Opus 5) working from sources fetched and read in full during the drafting session. Every statistic reported below was taken from a source this session actually retrieved, not from a search-result summary; where a source could not be retrieved, that is flagged in place under VERIFICATION FLAG rather than substituted for quietly.

Verification was then dispatched to subagents running on a different model (Fable), given the completed draft and instructed to find gaps, check every figure against its primary source, and flag anything thin, wrong or outdated. The work was split five ways: the clinical prevalence studies; the screening instruments plus an independent recomputation of this document's own base-rate arithmetic; the criminological and Arizona regional sources; the federal and statutory quotations checked word by word; and a sweep for sources the drafting pass should have found and did not.

That pass earned its cost. It produced eleven corrections, of which four changed a finding rather than a detail: the UK matched cohort's matching variables (Section 5.7), which forced this document's central assessment down two confidence levels; the addition of Cole et al. 2016 (Section 5.7, Section 11.1), a matched US comparison study pointing against that same assessment; the withdrawal of a false claim that no US longitudinal post-exit data exists (Section 13); and the withdrawal of a false claim that the widely quoted health-care-contact figure had no independent replication (Section 9). It also found Arizona county-level data that the drafting pass had declared absent (Section 7.3), and established that a figure reported here as an area under the curve was in fact a correlation between two models (Section 6.2).

Corrections and additions arising from that pass are marked inline as Correction (multi-model verification pass): and Addition (multi-model verification pass): rather than silently folded into the text. A reader who wants to know how much this document changed under checking can read those markers as a list. Nothing was quietly rewritten.

One report from the verification pass was itself wrong, which is worth recording as a caution against treating any model's output as authority: it gave a depression figure from the STAR Court study as rising from 66 to 86 percent, and the source says 66 to 68 percent. The figure used here is the one retrieved directly from the paper.

Every correction and addition from the verification passes that appears in this document was independently re-verified against its primary source before being added, in keeping with the sourcing standard stated throughout, rather than taken on any model's word.

1.5 Abbreviations used in this brief

2Context: what the law counts as a psychological marker

Source line: Trafficking Victims Protection Act of 2000 and its codification at 18 U.S.C. section 1591, read as reproduced by the Cornell Legal Information Institute (source 1); definitional material from the Arizona statewide child sex trafficking study (source 30).

Almost every downstream confusion about “psychological markers” traces to the fact that the American legal definition of trafficking is itself a psychological definition. The statute does not require chains, locks, or a border crossing. It requires a state of mind produced in the victim.

18 U.S.C. section 1591 reaches anyone who knowingly recruits, entices, harbors, transports, provides, obtains, advertises, maintains, patronizes or solicits a person, “knowing, or ... in reckless disregard of the fact, that means of force, threats of force, fraud, coercion described in subsection (e)(2), or any combination of such means will be used to cause the person to engage in a commercial sex act, or that the person has not attained the age of 18 years and will be caused to engage in a commercial sex act, shall be punished as provided in subsection (b).”

Two definitions in subsection (e) do the analytic work:

• Coercion includes “(A) threats of serious harm to or physical restraint against any person; (B) any scheme, plan, or pattern intended to cause a person to believe that failure to perform an act would result in serious harm to or physical restraint against any person; or (C) the abuse or threatened abuse of law or the legal process.”
• Serious harm means “any harm, whether physical or nonphysical, including psychological, financial, or reputational harm, that is sufficiently serious, under all the surrounding circumstances, to compel a reasonable person of the same background and in the same circumstances to perform or to continue performing commercial sexual activity in order to avoid incurring that harm.”

The clause carrying the most weight for this document is “any scheme, plan, or pattern intended to cause a person to believe.” The offence is complete when a belief has been manufactured. Nothing has to be done to the body. A statute written this way is, in effect, a statute about installed psychology, and it is why the identification problem is a psychological problem rather than a forensic one: there may be no injury, no restraint, and no barrier at the door.

2.1 Terms of art that are routinely conflated, and why the distinction matters

The minor-victim rule is the single largest driver of divergent statistics in this literature. The Arizona statewide study states it plainly: “A sex trafficking victim is any child under the age of 18 exchanging sex for something of value such as money, a place to stay, drugs, clothes, or protection” (source 30). A homeless sixteen-year-old trading sex for a couch is a trafficking victim under that definition and will be counted as one in a study that applies it, and will not be counted in a study that requires a third-party trafficker. Both studies are correct. They are not measuring the same thing.

3Context: the shape and thinness of the evidence base

Source line: peer-reviewed systematic reviews and the largest located primary studies, 2010 through 2026 (sources 8, 9, 11, 12, 13, 14, 15, 16, 17, 19, 24, 25, 26, 27, 28).

Anyone writing about trafficking psychology inherits a literature that is young, small, geographically lopsided, and built almost entirely on people who have already reached a service. That is not a reason to dismiss it. It is a reason to know exactly how far each number can be carried.

3.1 The primary quantitative studies located this session

3.2 What the shape of that table does and does not support

Four structural facts about this literature bear on every number quoted later in this document.

First, sample sizes are small. Outside the two Florida administrative-data studies (sources 18 and 19) and the Mekong regional survey (source 14), the median study here has fewer than 200 participants. The dissociation study that is often summarised as showing “clinically significant dissociation comparable to dissociative disorders” has 52 participants and used purposive sampling (source 27).

Second, nearly every participant was recruited from a service. Post-trafficking shelters, specialty courts, mental health trusts, victim service organisations, emergency departments. People who never reach a service are systematically absent, and there is no basis in the located sources for assuming they resemble those who do. The direction of that bias is genuinely unknown: it could exclude the least symptomatic (who never needed help) or the most (who could not navigate a system).

Third, geography is lopsided. The 2016 systematic review states that “The majority of studies were conducted in low and middle-income countries with women and girls trafficked into the sex industry,” and that evidence on trafficked men and on non-sexual exploitation was “limited but emerging” (source 8). The two largest post-trafficking mental health datasets in this brief are from Southeast Asia and Moldova. Their applicability to a US southwestern border corridor is an assumption, and it is logged as one in Section 10.

Fourth, and most consequential for anyone reading a pooled figure: heterogeneity in this literature is close to total. The 2016 meta-analysis reports pooled prevalence of 50 percent for anxiety symptoms (95 percent CI 21.9-78.2), 52 percent for depression (95 percent CI 33.9-70.8) and 32 percent for PTSD (95 percent CI 8.3-54.9), with heterogeneity of I-squared = 97.0 to 98.5 percent across those estimates (source 8).

Note on a commonly conflated pair: prevalence of a symptom and prevalence of a diagnosis are not the same measurement and are freely swapped in secondary coverage of this literature. Most of the studies above screened for symptoms with self-report instruments (the Hopkins Symptom Checklist, the Harvard Trauma Questionnaire, the Brief Symptom Inventory). Abas et al. 2013 had a psychiatrist assign DSM-IV diagnoses using the Structured Clinical Interview, blind to the participants' pre- and post-trafficking histories (source 13). Its headline number, 54 percent meeting criteria for any DSM-IV disorder at an average of six months post-return, is a diagnosis figure and is not comparable to a screening figure from another study.

4Markers during exploitation: the psychology of compliance

This is the section that answers the question a reader usually means when they ask about psychological markers of trafficking: why does someone who is not locked in stay?

4.1 The coercion architecture (PEER-REVIEWED: Baldwin, Fehrenbacher and Eisenman 2015, Qualitative Health Research)

Albert Biderman, working for the US Air Force in 1957, catalogued the methods by which Korean War captors produced compliance in prisoners without leaving marks. Eight tactics: isolation, monopolisation of perception, induced debility or exhaustion, threats, occasional indulgences, demonstrating omnipotence and omniscience, degradation, and enforcing trivial demands.

Baldwin and colleagues conducted semi-structured interviews with 12 adult women trafficked into Los Angeles County from 10 countries for domestic work, sex work, or both, and analysed the accounts against Biderman's framework. They report that “Participants reported experiencing the range of nonphysical coercive tactics outlined by Biderman, including isolation, monopolization of perception, induced debility or exhaustion, threats, occasional indulgences, demonstration of omnipotence, degradation, and enforcement of trivial demands,” and that “these coercion tactics reinforced the submission of trafficked persons to their traffickers even in the absence of physical force or restraints” (source 20).

Two things about that framework are worth holding separately.

The first is that it is a description of what is done, not of what the victim is. Every item on Biderman's list is an action taken by another person. Nothing on it is a personality trait, a vulnerability, or a diagnosis. This matters because indicator lists in circulation routinely translate the effects of these tactics back into descriptors of the victim: withdrawn, evasive, inconsistent, unwilling to make eye contact, deferring to a companion. Those are the visible residue of isolation, monopolisation of perception and demonstrated omnipotence, and they are what a screening encounter actually sees.

The second is the item that carries the most explanatory weight and gets the least attention: occasional indulgences. Unpredictable intermittent reward is the mechanism that converts fear into attachment. It is the same mechanism named in the trauma-bonding literature (Section 4.2), and it is why an account of the exploitation that consists only of violence tends to read as false to survivors and to the people who work with them.

4.2 Trauma bonding and trauma-coerced attachment (PEER-REVIEWED: Casassa, Knight and Mengo 2022, Trauma, Violence, and Abuse)

Casassa and colleagues searched ten databases for literature on trauma bonding in sex trafficking, using trauma-bonding terms plus Stockholm syndrome, attachment, coercion, and manipulation, restricted to English-language work published after 2013 featuring sex trafficking victims or traffickers in a Western country. Fifteen articles met inclusion.

They identify four features of the bond itself:

1. An imbalance of power that favours the trafficker.
2. The trafficker's deliberate use of positive and negative interactions.
3. The victim's gratitude for the positive interactions and self-blame for the negative ones.
4. The victim's internalisation of the perpetrator's view.

And four related aspects:

1. Prior trauma made victims vulnerable.
2. The victim's feelings of love remained even after exiting trafficking.
3. Love is why victims do not prosecute traffickers.
4. Traffickers cultivated the bond intentionally.

The finding that should govern how much weight this construct carries is the review's own closing observation: “No article indicated how trauma bonds could be severed and replaced with healthy attachments” (source 21).

Item 3 in the second list is the one with the widest practical reach, and it is worth stating precisely rather than dramatically. In the reviewed literature, a victim's refusal to testify is described as an expression of continuing attachment rather than of intimidation alone. Nothing in the located sourcing quantifies how often each explanation applies, and this document does not claim a proportion.

Item 4 in the first list, internalisation of the perpetrator's view, is the point at which trauma bonding and the post-traumatic cognition literature (Section 5.4) describe the same thing from two directions. One calls it internalisation of the trafficker's view of the victim; the other measures it as a clinically elevated score on negative cognitions about the self. That convergence is treated as a signal in Section 9.

4.3 Non-self-identification (PEER-REVIEWED: Mahon et al. 2026, American Journal of Criminal Justice)

The single most operationally important marker in this brief is a negative one: a large fraction of trafficking victims do not describe themselves as trafficking victims, including while sitting in a programme designed for trafficking victims.

Mahon and colleagues surveyed 74 women across three sex trafficking specialty court programmes and one nonprofit in Ohio. Every participant met the legal criteria for trafficking. The paper reports: “Before court programming, 43.2% (n = 32) of participants did not see themselves as a victim or survivor of sex trafficking. This number dropped to 12.2% (n = 9) since court program participation.” The authors characterise the initial gap as a critical barrier to service engagement (source 28).

Two cautions on that study, stated rather than smoothed. It is cross-sectional and retrospective: the “before” figure is a present-day recollection of a past belief, collected after the intervention that is credited with changing it, which is a design that tends to inflate apparent change. And a specialty-court population is a population that has already been identified by a system, so the true non-identification rate in an unidentified population is unmeasured and is almost certainly higher.

The State Department's TIP Office fact sheet on trauma in child trafficking survivors, written by consultants in its Human Trafficking Expert Consultant Network, states the same phenomenon from the caregiver's side in a form worth quoting for its bluntness: “some children might have positive feelings towards their abusers and express hate that you took them away from these individuals” (source 7).

4.4 Substance use as an instrument of control (PRIMARY/OFFICIAL and REGIONAL: HHS ASPE 2010; Roe-Sepowitz, Way and Steving 2024)

Substance use in this population functions in at least three distinct ways that the sources do not always separate: as a pre-existing vulnerability that a trafficker exploits, as a dependency deliberately induced to create a withdrawal-based control lever, and as self-medication during and after exploitation. The located sources establish that the association is strong and that the direction of causation is unresolved.

HHS ASPE's 2010 review names substance-related disorders among the primary mental health conditions found in trafficking victims, alongside PTSD, anxiety and mood disorders, dissociative disorders, and complex trauma (source 4).

The Arizona statewide study gives the sharpest recent numbers. Of 309 suspected or confirmed child sex trafficking victims identified in Arizona between January 2021 and May 2023, 45 percent (n = 139) were reported to have used drugs or alcohol. Reported alcohol use rose 153.6 percent, fentanyl use 125.6 percent, and cocaine use 41.4 percent (source 30).

4.5 Synthesis: what the coercion evidence converges on

Reading Sections 4.1 through 4.4 together, three things hold across independently produced bodies of evidence.

The coercion is architectural, not incidental. Biderman's eight tactics, the four features of the trauma bond, and the statutory definition of coercion are three different institutions (an Air Force study reapplied by clinical researchers, a social-work scoping review, and the United States Code) describing a scheme intended to produce a belief. None of them describe a victim who is weak.

The bond survives the exploitation. The scoping review finds feelings of love persisting after exit; the Ohio study finds victim self-identification absent even inside a specialty court; the State Department fact sheet warns caregivers that a rescued child may direct hatred at the rescuer. Three sources, three institutions, one phenomenon.

And the effects that show up in a room are second-order. A professional does not observe monopolisation of perception. They observe a person who defers to a companion, gives an inconsistent account, and declines help. Every field indicator list in circulation is a list of these second-order effects, which is why they are non-specific (Section 6.6).

5Markers as clinical presentation: what the disorders actually are

5.1 The core symptom triad and what predicts it (PEER-REVIEWED: Abas et al. 2013, BMC Psychiatry; Hossain et al. 2010, Am J Public Health)

Across every located study, three conditions dominate: post-traumatic stress disorder, depression, and anxiety.

The best-designed prevalence figure available is from Abas and colleagues, a historical cohort of women who returned to Moldova and registered for assistance with the International Organization for Migration. 120 of 176 eligible women (68 percent) participated. At two to twelve months post-return, a psychiatrist assessed DSM-IV disorders using the Structured Clinical Interview, blind to information about the women's pre-trafficking and post-trafficking experiences. At an average of six months post-return:

The risk-factor model is the more important result. In multivariable regression:

(source 13)

Read the table with that in mind. Childhood sexual abuse and post-exit unmet needs are strong and statistically resolved. Duration is a large unresolved estimate. Social support is protective.

Hossain and colleagues, working with 204 trafficked girls and women across seven post-trafficking service settings, used the Brief Symptom Inventory and Harvard Trauma Questionnaire and adjusted for pre-trafficking abuse. They found sexual violence during trafficking associated with higher PTSD levels (AOR 5.6, 95 percent CI 1.3-25.4), more time in trafficking associated with higher depression and anxiety (AOR 2.2, 95 percent CI 1.1-4.5), and more time since trafficking associated with lower depression and anxiety but not lower PTSD (source 12).

Note the partial conflict between these two studies on duration: Hossain finds duration associated with depression and anxiety; Abas finds duration only borderline for any diagnosed disorder. They are not measuring the same outcome (symptom level versus diagnosis) and the conflict is treated as unresolved in Section 11.3 rather than papered over.

Note also the finding that both studies share, and that has the most practical bite: recovery time reduces depression and anxiety but does not reduce PTSD. Whatever the PTSD is, it is not something that fades on its own with distance from the event.

5.2 Minors: the largest single-population figures (PEER-REVIEWED: Kiss et al. 2015, JAMA Pediatrics)

Kiss and colleagues interviewed 387 children and adolescents aged 10 to 17 in post-trafficking services in Cambodia, Thailand and Vietnam.

Pre-migration violence was reported by 22.4 percent of the sample (source 11).

Two features of that table complicate the popular picture. Depression, not PTSD, is the most common condition, and by a factor of more than two. And the gendered violence pattern inverts by type: boys in this sample were more than twice as likely to report physical violence, girls twenty-three times more likely to report sexual violence.

5.3 Complex PTSD (PEER-REVIEWED: Brewin 2020, BJPsych Advances; Ottisova, Smith and Oram 2018, Behavioral Medicine)

The ICD-11 diagnosis that best fits the described presentation is Complex PTSD. CPTSD requires all three PTSD symptom clusters (re-experiencing in the present, avoidance, and a current sense of threat) plus disturbances in self-organisation. The ICD-11 definition, as reproduced in Brewin's Box 1, gives the DSO components as “severe and persistent 1) problems in affect regulation; 2) beliefs about oneself as diminished, defeated or worthless, accompanied by feelings of shame, guilt or failure related to the traumatic event; and 3) difficulties in sustaining relationships and in feeling close to others.” All three DSO components must be present.

Brewin also records two points in his own text that bear directly on this brief. “Finally, chronic or repeated trauma is a risk factor, not a requirement, for CPTSD.” And that “studies have shown that childhood physical or sexual abuse, particularly within the family, is more strongly related to CPTSD than PTSD.” The ICD-11 definition itself gives the associated trauma types as “most commonly prolonged or repetitive events from which escape is difficult or impossible (e.g., torture, slavery, genocide campaigns, prolonged domestic violence, repeated childhood sexual or physical abuse)” (source 23).

That last clause is the seam this entire brief keeps returning to. The exposure most strongly associated with CPTSD is childhood abuse, which is also the strongest predictor of being trafficked in the first place (Section 5.7) and the strongest predictor of post-trafficking disorder (Section 5.1). Attribution of a CPTSD presentation to the trafficking specifically is therefore not straightforward.

Ottisova, Smith and Oram tested this directly. Working from the same UK cohort described in Section 5.7, they identified 51 trafficked children, 11 of whom (22 percent) carried a PTSD diagnosis, and compared CPTSD symptom counts against non-trafficked controls exposed to single or multiple trauma. Their result: “Trafficked and non-trafficked children with PTSD who had been exposed to multiple trauma showed a greater number of Complex PTSD symptoms compared to nontrafficked children with PTSD exposed to single-event traumas.” Somatic symptoms were noted in almost two thirds of trafficked children but only 10 to 11 percent of non-trafficked children (source 10).

5.4 Post-traumatic cognitions: the internal content (PEER-REVIEWED: Perry et al. 2022, Public Health Reports)

Symptom lists say what a person has. Post-traumatic cognition measures say what a person believes about themselves as a result, which is closer to what a writer or a clinician actually needs.

Perry and colleagues analysed baseline data from 110 young people aged 11 to 19 with substantiated commercial sexual exploitation or trafficking experiences, entering trauma-focused cognitive behavioural therapy in a southeastern US state between August 2013 and March 2020. Mean age 15.8. Findings: 57 of 110 (51.8 percent) met clinical criteria for post-traumatic cognitions. Increased age and a greater number of trauma categories experienced were significantly associated with meeting those criteria. Higher post-traumatic cognition scores were associated with higher post-traumatic stress symptom scores, controlling for demographics (beta = 0.95, 95 percent CI 0.64-1.26) (source 17).

Note what the significant predictor was: the number of trauma categories experienced, not any feature of the trafficking. The polyvictimisation finding again.

5.5 Where sex and labor trafficking diverge (PEER-REVIEWED: Hopper and Gonzalez 2018, Behavioral Medicine)

Hopper and Gonzalez examined psychological symptoms in 131 survivors of sex and labor trafficking, including people trafficked into and within the United States. Depression 71 percent, PTSD 61 percent. Two thirds met criteria for multiple categories of Complex PTSD, which they enumerate as affect dysregulation and impulsivity, alterations in attention and consciousness, changes in interpersonal relationships, revictimisation, somatic dysregulation, and alterations in self-perception.

The comparative findings matter more than the prevalences:

• There were no significant differences in prevalence of PTSD or depression diagnoses between sex and labor trafficking survivors.
• Sex trafficking survivors had higher rates of pre-trafficking childhood abuse and more physical and sexual violence during trafficking.
• Sex trafficking survivors reported more severe post-trauma reactions, more PTSD and CPTSD symptoms, and were more likely to meet criteria for comorbid PTSD and depression; labor trafficking survivors were more likely to meet criteria for depression alone.
• Survivors who identified as transgender endorsed more PTSD and CPTSD symptoms than male or female survivors.
• Childhood abuse exposure was linked to PTSD and CPTSD, and trafficking type predicted the number of trauma-related symptoms beyond the role of pre-trafficking child abuse (source 15).

That last clause is the one piece of located evidence pointing the other way from Section 5.3's conclusion: trafficking type retained predictive power after childhood abuse was accounted for. It is given its full weight in the ACH at Section 11.1 rather than being set aside.

Iglesias-Rios and colleagues add a gendered structure to the same picture. Using latent class analysis on 1,015 female and male survivors in post-trafficking services in Cambodia, Thailand and Vietnam, they identified two classes for each sex: for women, severe sexual and physical violence with coercion (20 percent) versus sexual violence and coercion (80 percent); for men, severe physical violence with coercion (41 percent) versus personal coercion (59 percent). Women in the severe class had roughly double the odds of anxiety (OR 2.10, 95 percent CI 1.57-2.81) and PTSD (OR 2.07, 95 percent CI 1.03-4.17) relative to the other class. For men, the difference between classes was not significant (source 14).

5.6 Dissociation and somatic presentation (PEER-REVIEWED: MacCabe et al. 2026, Journal of Trauma and Dissociation; Ottisova et al. 2018)

MacCabe and colleagues administered the DES-II to adult sex trafficking survivors using purposive sampling and analysed 52 completed surveys. They report that respondents “experienced clinically significant dissociation levels, comparable to levels associated with dissociative disorders,” that depersonalisation/derealisation and absorption/imaginative involvement were more prominent than amnesia, and that longer trafficking durations and younger age at first experience were associated with higher dissociation (source 27).

The somatic finding from the UK cohort belongs here as well: somatic symptoms in almost two thirds of trafficked children against 10 to 11 percent of non-trafficked children (source 10). Of every quantitative comparison located in this session, that is the largest gap between trafficked and non-trafficked groups. It is also the least discussed, because the presentation it describes, a young person with persistent unexplained physical complaints, is the one least likely to be read as a trafficking indicator by anyone.

5.7 The trafficked-versus-comparable-group question (PEER-REVIEWED: Ottisova et al. 2018 PLOS One; Palines et al. 2020 Child Abuse and Neglect)

Two studies put trafficked young people directly beside comparison groups, and they disagree.

Ottisova and colleagues searched the electronic health records of more than 250,000 patients in a large UK mental health trust, identified 51 trafficked children, and randomly selected a matched cohort of 191 non-trafficked children. Within the trafficked group:

The comparative results:

(source 9)

Palines and colleagues retrospectively reviewed the medical records of 143 sex trafficked children in Wisconsin and compared diagnosis rates against summarised prevalence data for three high-risk groups (runaway children, juvenile offenders, and foster care children) drawn from a scoping review. They report significantly higher rates of ADHD (52.4 percent, p < 0.0001), bipolar disorder (26.6 percent, p < 0.0001) and PTSD (19.6 percent, p < 0.05 to p < 0.0001) than all comparison groups, and higher rates of depression (45.5 percent), anxiety (19.6 percent), conduct disorder (19.6 percent), ODD (25.9 percent) and psychosis (14.0 percent) relative to some groups individually.

Their own interpretation is the part most often dropped when this study is cited: “Survivors' adaptive responses to complex trauma may lead to improper diagnosis and treatment of mental health disorders at the expense of prompt access to trauma-focused therapies” (source 16).

This is the closest thing located to a clean test: a US clinical sample, a comparison group matched on demographics rather than on diagnosis, and a comparator that is not merely high-risk but specifically sexually victimised. It found the commercially exploited group more severely symptomatic. It is a small exploited group (n = 43), its matching did not control for prior trauma load, and severity is not the same thing as a distinct signature, but it is direct evidence against the null and it is treated as such in Section 11.1.

A fourth study points the same way while narrowing what the difference consists of. Varma, Gillespie, McCracken and Greenbaum compared 27 commercially sexually exploited children against 57 age- and sex-matched child sexual abuse patients across three paediatric emergency departments and a child protection clinic, and found the groups differing on eleven variables, which were predominantly exposure-history and behavioural rather than symptom variables: violence exposure, substance use, running away, sexual history and child protective services involvement. VERIFICATION FLAG: this study's full text is paywalled and this document has read only its abstract and the summary of its variable list; the individual figures are not reproduced here and it is graded accordingly (source 40).

5.8 What predicts becoming a victim (PEER-REVIEWED: Reid et al. 2017, Am J Public Health; de Vries, Baglivio and Reid 2024, J Interpersonal Violence)

Two Florida administrative-data studies give the largest samples in this brief.

Reid and colleagues compared adverse childhood experience prevalence and cumulative ACE scores among 913 juvenile-justice-involved boys and girls for whom the Florida child abuse hotline accepted human trafficking abuse reports between 2009 and 2015 against a matched sample. ACE composite scores were higher and six ACEs indicative of child maltreatment were more prevalent in the trafficking-report group. Sexual abuse was the strongest predictor: odds of human trafficking 2.52 times greater for girls with a sexual abuse history and 8.21 times greater for boys (source 18).

De Vries, Baglivio and Reid analysed 40,531 justice-involved youth in Florida from 2011 to 2015, of whom 801 (699 female, 102 male) had trafficking investigations, using the state's risk/needs assessments matched to census-tract data.

(source 19)

Note on an important conflation: the correlates in this section are risk factors for becoming a victim. They are not markers that someone is currently a victim. A screening instrument built out of risk factors will flag the entire high-risk population, which is precisely the failure mode measured in Section 6.

6Field identification: the instruments and their measured performance

6.1 The federal position on validation (PRIMARY/OFFICIAL: HHS OTIP/NHTTAC 2018; HHS OTIP Program Instruction OTIP-IM-2024-03)

The single most useful thing about the federal screening literature is that it is candid about its own limits, in its own voice, in print.

The Adult Human Trafficking Screening Tool and Guide was published in January 2018, funded by HHS/ACF/OTIP under contract HHSP233201500071I and produced by NHTTAC, which is managed by ICF. Its primary authors were Wendy Macias-Konstantopoulos of Massachusetts General Hospital and Harvard Medical School, and Julie Owens.

What it says about itself:

• “This Toolkit provides a screening tool to use in identifying adults who you suspect may have experienced sex or labor trafficking. While this tool is not yet validated, it has been developed based on the latest research and best practices in screening.”
• “The AHTST and Toolkit were created to combine literature and promising practices for screening tools for a variety of interpersonal crimes and other health concerns, including domestic violence, sexual assault, human trafficking, and HIV. It has not yet been validated or evaluated in the field.”

What it says about the field as a whole:

• “No validated tools exist for the purpose of screening clients/patients for trafficking across all public health settings (public health, health care, behavioral health, and social services), and very few validated trafficking screening tools exist at all. In fact, only an extremely small number of screening tools of any kind have been rigorously investigated and evaluated.”
• “No research data or evaluation criteria are available to compare the effectiveness of different methods of screening.”
• “None of the existing screening tools are validated across multiple public health settings.”
• “validation studies were not publicly available for many of the screeners online at the time of this review.”

And on red-flag checklists specifically:

• “A 'red flag' checklist is not a screening tool, but it can be useful alongside screening tools. Red flag checklists serve as quick guides that help professionals recognize a cluster of symptoms. They are not formal and typically are not validated.” (An earlier draft elided the middle two sentences. They are restored here because the toolkit's own position is that checklists are a useful complement to the AHTST, and quoting only the negative half would misrepresent it.)
• “There appears to be no consensus of opinion in trafficking research literature or among experts about whether red flag checklists are useful instruments.” The toolkit records that some experts avoid them “because they can be too easily rushed through by busy professionals,” and that others use them because “no better options exist.”
• “Reviewed tools contain an average of 40 questions and require 60” minutes or more, which the toolkit gives as a reason many are unusable in time-sensitive settings.

(source 2)

The position has not changed as of the most recent federal guidance located. OTIP Program Instruction OTIP-IM-2024-03, issued 12 December 2024, states that OTIP “highly encourages the use of validated screening tools, when available and possible, but inclusion of any particular tool in this PI is not meant as an endorsement,” and notes that “neither of these approaches requires disclosure. It is more important to identify and understand the person's unique needs than to secure a disclosure of human trafficking” (source 3).

6.2 The Vera Trafficking Victim Identification Tool (PRIMARY/OFFICIAL: Simich et al. 2014, NIJ, NCJ 246712)

The most-cited validated US instrument was developed by the Vera Institute of Justice under NIJ award 2011-MU-MU-0066, with Laura Simich as principal investigator, published June 2014.

Method: working with 11 victim service providers across California, Colorado, New York, Texas and Washington, Vera collected data on more than 230 cases. Of the 180 individuals who responded to the screening questions, 96 (53 percent) were trafficking victims and 84 (47 percent) were victims of other crimes such as domestic violence, smuggling, prostitution or labor exploitation. Of the trafficking victims, 38 (40 percent) were sex trafficking victims and 58 (60 percent) labor trafficking victims.

Results: 87 percent of the questions significantly predicted trafficking victimisation in general, 71 percent predicted labor trafficking specifically, 81 percent predicted sex trafficking. A short version of 16 questions was found to predict victimisation accurately for both types.

Reported model performance, general trafficking model:

(source 5)

6.3 The paediatric child sex trafficking screens (PEER-REVIEWED: Greenbaum et al. 2018, J Adolescent Health; Hurst et al. 2021, Pediatrics)

Greenbaum and colleagues evaluated a short child sex trafficking screening tool across 16 US sites: 5 paediatric emergency departments, 6 child advocacy centres and 5 teen clinics, with 810 participants. Overall CST prevalence in the sample was 11.1 percent (13.2 percent in emergency departments, 6.3 percent in child advocacy centres, 16.4 percent in teen clinics). Reported performance: “sensitivity, specificity, and positive likelihood ratio of 84.44% (75.28, 91.23), 57.50% (53.80, 61.11), and 1.99% [sic] (1.76, 2.25), respectively” (source 25). The percent sign on the likelihood ratio is the published abstract's own, verified against the publisher-deposited text; a likelihood ratio is a ratio and not a percentage, and the value should be read as 1.99. The “[sic]” is inserted here so a reader does not assume this document mis-transcribed it.

Hurst and colleagues administered a 6-item confidential screening tool by electronic tablet in a single urban paediatric emergency department at a Level 1 trauma centre with roughly 80,000 annual visits. 212 participants analysed, 72.6 percent female, median age 15. A positive screen required two or more affirmative responses. Results: 109 positive screens, 26 confirmed trafficked, prevalence 12.3 percent.

4 false negatives (15.4 percent of trafficked patients) and 87 false positives (41.0 percent of the entire sample) (source 24).

6.4 The adult emergency department screen (PEER-REVIEWED: Chisolm-Straker et al. 2021, JACEP Open)

The Rapid Appraisal for Trafficking (RAFT) is a 4-item instrument and, of the located instruments, the only one with a genuine external validation sample.

Setting: six emergency departments, five in New York City and one in Fort Worth, Texas. 4,127 patients enrolled: 3,292 in the NYC derivation group and 835 in the Fort Worth validation group.

Trafficking prevalence: 1.1 percent in NYC (36 cases: 20 labor, 16 sex) and 1.4 percent in Fort Worth (12 cases: 8 labor, 4 sex).

Performance: in the derivation group “RAFT was 89% sensitive (95% confidence interval [CI], 79%-99%) and 74% specific (95% CI, 73%-76%).” In external validation “RAFT was 100% sensitive (95% CI, 100%-100%) and 61% specific (95% CI, 56%-65%).” Combined, affirmation to any one of the four items gave 92 percent sensitivity and 72 percent specificity.

Stated limitations: the exclusion criteria “may have decreased the prevalence of trafficking identified”; “patients that were ineligible for participation may have been at higher risk for having a trafficking experience,” the authors naming patients unable to consent (including those presenting with intoxication, substance use disorder or mental illness complications), those who could not speak with the interviewer alone, and those presenting and being dispositioned in the middle of the night; a participation decline rate of 48 percent in New York City and 50 percent in Fort Worth, of unknown directional bias; a data-collection pause from March to August 2020; and the authors' own recommendation that “RAFT should be validated in other EDs and other kinds of areas, including rural settings, reservations, and free-standing EDs.” The authors also state that “implications of a false screen on RAFT are fairly benign” (source 26).

6.5 The base-rate arithmetic

The reported statistics above are enough to work out what actually happens when these instruments are used, and the answer is not what the sensitivity figures imply.

Applying the RAFT NYC figures (1.1 percent prevalence, 89 percent sensitivity, 74 percent specificity) to a notional 10,000 emergency department patients:

Running the same arithmetic on the Fort Worth figures (1.4 percent prevalence, 100 percent sensitivity, 61 percent specificity) gives a positive predictive value of approximately 3.5 percent.

Note the contrast with the paediatric settings. Hurst reports a measured positive predictive value of 20.2 percent at 12.3 percent prevalence, and the same arithmetic applied to Greenbaum's multi-site figures (11.1 percent prevalence, 84.44 percent sensitivity, 57.5 percent specificity) gives approximately 20 percent. Two independent paediatric studies, one measured and one derived, converge on the same answer: in a high-prevalence adolescent setting, about one positive screen in five is a real case. That is roughly five times better than the general adult emergency department and still means four in five positives are not trafficked.

6.6 Red-flag checklists, and the branding indicator

The federal toolkit's own assessment of red-flag checklists is quoted in Section 6.1: not screening tools, not formal, typically not validated, and with no expert consensus on whether they are useful at all (source 2).

The specific indicator most often cited in public-facing awareness material is trafficker branding: a tattoo, usually a name, symbol or barcode, marking ownership.

Cross-reference: a companion brief in this project, Human Trafficking Victim Marking and Identification, 2015-2025, takes the marking question as its whole subject at 80 graded sources and reaches a compatible conclusion by a different route, separating the evidentiary value of a tattoo for a person already identified from its value for picking someone out of a crowd. Its sources are not re-graded here and none of them is used as a source in this document. A reader who needs the marking question in depth should go there rather than treat this subsection as the last word on it; what follows is this brief's own narrower finding, reached from the sources it did retrieve.

The general problem is structural rather than specific to tattoos. Section 5.8 established that the strongest measured correlates of juvenile trafficking victimisation are repeated running away and high adverse-childhood-experience scores. Those are also the defining features of the runaway and child-welfare populations from which trafficking victims are drawn.

7Regional focus: Arizona, and the southeastern corridor

7.1 Jurisdictional overview and statewide identification (REGIONAL: Roe-Sepowitz, Way and Steving 2024, ASU STIR and Mercy Care, hosted by the Arizona Governor's Office of Youth, Faith and Family)

Arizona is the sixth largest state in the United States, with 15 counties, 141 law enforcement agencies and 14,591 sworn officers. The state's anti-trafficking research is unusually well developed and is concentrated in one place: the Office of Sex Trafficking Intervention Research at the Arizona State University School of Social Work, directed by Dominique Roe-Sepowitz, working in partnership with Mercy Care, the Phoenix Police Department, and the Governor's Office of Youth, Faith and Family.

Mercy Care has coordinated the Trafficked Youth Collaborative since 2017, originally in Maricopa County and now statewide. The collaborative's process routes an identified child through crisis stabilisation, medical clearance and mental health evaluation before placement, using a 23-hour crisis assessment, a model adopted because child victims were previously being taken straight to group homes.

The most recent statewide figures, covering January 2021 through May 2023:

By year, for the three counties this document's regional focus covers: Pima 21 (2021), 37 (2022), 8 (through May 2023); Cochise 0, 6, 1; Santa Cruz 0, 0, 0. Tucson Police Department appears on the report's closing list of referring agencies.

Youth of colour exceeded 50 percent of identified victims, which the authors describe as a disproportionality “not uncommon in sex trafficking victimization research.”

The report's closing operational list includes, without elaboration, the line: “No standardized screening in most settings” (source 30).

The predecessor Maricopa County study covering three years to December 2020 documented 291 child sex trafficking cases and recorded a steady decline in average age at detection from 16.6 years in 2017 to 15.3 years in 2020, and a shift in guardianship from more than 90 percent Arizona DCS in 2017 to 59 percent by year three, with parental guardianship rising to 38 percent (sources 30 and 33).

7.2 Tucson and Pima County (REGIONAL: Roe-Sepowitz, Bracy and Hogan 2019, ASU STIR Youth Experiences Survey Year Six)

The closest thing to Tucson-specific psychological data located in this session comes from the sixth year of the Youth Experiences Survey, conducted in July 2019 with homeless young adults aged 18 to 25 in Phoenix and Tucson, in partnership with Our Family Services of Tucson, Native American Connections, UMOM and one-n-ten of Phoenix. A seven-page paper survey was distributed over two weeks in transitional housing, drop-in centres and on the streets. N = 167, average age 20.9. 70.1 percent were raised in Arizona.

Baseline psychological picture for the whole sample:

Trafficking prevalence within that sample:

The three most common reasons participants gave for their sex trafficking victimisation were a place to stay (50 percent), money (44.6 percent) and food (33.8 percent).

The trafficked versus non-trafficked comparison within this sample is the part with the most analytic value. The sex trafficked group was significantly more likely to identify as LGBTQ; engage in self-harming behaviours; report mental health diagnoses and more than one diagnosis, specifically depression, PTSD, schizophrenia and borderline personality disorder; report a history of suicide attempts; report a medical issue, specifically asthma; witness domestic violence in the household in both directions; experience abuse in a domestic violence relationship and be the abuser in one; experience each of the ten adverse childhood experiences and four or more ACEs; experience emotional childhood abuse, physical abuse, and sexual abuse by a parent or guardian both as a child and as an adolescent; and experience bullying and harassment by school peers (source 31).

Note also that the trafficked group differed on asthma. That is a useful corrective for anyone tempted to read the significant-differences list as a psychological profile: it is a list of everything that reached significance, and it includes a respiratory condition.

7.3 Cochise and Santa Cruz counties

What the data shows for the two border counties: Cochise County recorded 7 identified child sex trafficking victims across the study period, 2.3 percent of the state total, distributed 0 in 2021, 6 in 2022 and 1 through May 2023. Santa Cruz County recorded zero in every period. Pima County recorded 66, 21.4 percent of the state total and second only to Maricopa (source 30).

The 2017 statewide incidence report is explicit about its own geographic limits in a way that bears directly on how these zeros should be read: “This incidence number represents only areas including Pinal County, Pima County, Maricopa County, and Yavapai County.” Its contributing-agency roster runs to 31 organisations, including Tucson Police Department, Tucson Gospel Rescue Mission and Our Family Services, and contains no Cochise or Santa Cruz County agency at all (source 32).

Cross-reference: a companion brief in this project, Origin Regions and Northbound Trafficking Routes, 2015-2025, works the criminal-offence side of the same two counties from Arizona state offence data and the FBI Uniform Crime Reporting human trafficking file, and reaches a compatible conclusion from independent sources: a real Cochise County trafficking-offence cluster concentrated in Douglas, Willcox and Tombstone from 2022, and a Santa Cruz County zero that is a reported nil in the state system but a non-submission in the federal one. That material is not re-graded here and none of it is used as a source in this document; it is named so a reader who needs the offence picture knows where it is. What remains genuinely absent, in both briefs, is any victim-psychology or victim-identification data specific to either county.

7.4 The Super Bowl claim (RESEARCH INSTITUTION: Roe-Sepowitz et al. 2015, McCain Institute)

Arizona is the site of the most-repeated single myth in American trafficking discourse, and also of the research that tested it.

The study scanned online sex advertisements on Backpage.com using a Sex Trafficking Matrix instrument and placed decoy advertisements to measure buyer response, comparing the ten days before and including Super Bowl Sunday in 2014 against 2015. Phoenix findings: a 30.3 percent increase in advertisement volume, a 22.1 percent increase in decoy responses (950 contacts in 2014 against 1,160 in 2015), and 73.3 percent of contacts originating from the three local area codes. Twenty-three Phoenix advertisements were flagged as possibly involving minors and reported to the National Center for Missing and Exploited Children, down from 34 the previous year. The FBI reported that of 27 juveniles recovered in the year preceding the Super Bowl, 18 were found in the three weeks before it.

The two discrepancies between the report and the press announcement, stated rather than reconciled: the announcement gives 21 flagged advertisements against the report's 23, and rounds 73.3 percent to “some 70 percent.” The report is the primary document and its figures are used above.

The report states in its own voice: “This does not indicate that the Super Bowl caused more sex trafficking.” Roe-Sepowitz is quoted in the announcement as saying that “sex trafficking is big business every day of the year, not just during the Super Bowl” (sources 34 and 35).

7.5 Material located and excluded as too dated to be load-bearing

The 2017 ASU STIR report “Incidence of Identified Sex Trafficking Victims in Arizona: 2015 and 2016” is a genuine primary regional source with an unusually complete list of contributing agencies, including Tucson Police Department, Tucson Gospel Rescue Mission, Our Family Services, the Bureau of Indian Affairs Arizona office, Arizona DCS, Arizona DPS and Arizona Juvenile Probation (source 32). Its incidence counts are not used as current figures in this document because they predate the statewide collaborative's expansion beyond Maricopa County, and identification counts in this field track detection capacity (Section 7.1). It is cited for its agency roster and its methodology, not for its totals.

8Sources rejected this session, and why

Notes on sources whose status moved during research:

The Adult Human Trafficking Screening Tool was first located through Freedom Network USA, an excluded advocacy organisation. Rather than cite it there, the document uses the copy at the Pennsylvania Office of Rural Health (a university-hosted mirror of the HHS document) for content and cites the HHS/NHTTAC original as the source, with the December 2024 OTIP program instruction (retrieved directly from acf.gov) as independent confirmation that the federal position is current. No claim rests on the advocacy host.

Lederer and Wetzel 2014 (source 29) was retained but downgraded. It is a peer-reviewed law journal article and is the origin of several of the most-circulated US figures in this field. It is also authored by the president of Global Centurion, an advocacy organisation, and its acknowledgments name Abolition International, the Charlotte Lozier Institute and the Greenbaum Foundation among its funders. Its sample of 106 was recruited through service-provider focus groups, and its psychological figures are self-reported diagnoses rather than instrument-based assessments. It is graded B3, its funding is disclosed in the citation, and no claim in this document rests on it alone.

The Vera Institute of Justice (source 5) was retained as a research institution rather than excluded as advocacy. The decision rests on the specific document: the cited work is a final report to the National Institute of Justice under a named federal award, published with its full methodology, sample limitations and regression tables, and NIJ is the publisher of record. Its methodological weaknesses are stated in Section 6.2 on their own merits, not on the basis of who wrote it.

One excluded source turned out to be right on the underlying claim. Advocacy material asserting that traffickers deliberately induce drug dependency as a control mechanism is prevalent and, as a description, matches the coercion architecture in Section 4.1. Primary confirmation of the specific claim was not found in this session, so the claim is not made in this document (Section 4.4), and this note records the excluded sources' agreement without promoting it to a citation.

9Signals: genuine cross-source correlation

The standard applied here: a correlated signal requires two or more independently produced sources, from different institutions, using different methods, arriving at the same finding without one citing or reprinting the other. Multiple outlets covering the same study is not correlation. It is one data point reported several times, and it counts as one. Two research groups analysing the same underlying dataset is a real but weaker correlation than two independent bodies of evidence, and is rated as such rather than collapsed into the stronger category.

Signal 1 (moderate): Coercion in trafficking is an architecture of manufactured belief, and the same architecture appears in a clinical framework, a social-work review, and federal statute.

Three independently produced sources describe the same mechanism. Baldwin and colleagues, applying a 1957 US Air Force framework to interviews with 12 trafficked women in Los Angeles County, report the full range of Biderman's non-physical tactics reinforcing submission “even in the absence of physical force or restraints” (source 20). Casassa and colleagues, reviewing fifteen articles across ten databases, identify deliberate alternation of positive and negative interaction, victim self-blame for the negative, and internalisation of the perpetrator's view (source 21). And 18 U.S.C. section 1591(e)(2) defines coercion to include “any scheme, plan, or pattern intended to cause a person to believe that failure to perform an act would result in serious harm” (source 1).

Why this now rates moderate rather than strong: the empirical base is a single qualitative study of twelve women. The convergence with Biderman's 1957 framework and with the statutory definition of coercion is real and is why the signal is retained rather than dropped, but it is conceptual convergence, not independent replication. A reader should treat the coercion architecture as a well-motivated organising framework with thin direct evidence behind it, not as a corroborated finding.

Signal 2 (strong): Childhood sexual abuse predicts both entry into trafficking and post-trafficking psychiatric outcome, across four independent datasets on three continents.

Four studies, four institutions, four methods, four populations:

• Abas et al. (Moldova, 120 women, psychiatrist-administered SCID blind to exposure): childhood sexual abuse AOR 4.68 (95 percent CI 1.04-20.92) for any DSM-IV disorder six months post-return (source 13).
• Reid et al. (Florida, 913 justice-involved youth with accepted trafficking abuse reports, administrative records): sexual abuse the strongest predictor of trafficking, OR 2.52 for girls and 8.21 for boys (source 18).
• Hopper and Gonzalez (US, 131 survivors, symptom instruments): childhood abuse exposure linked to PTSD and CPTSD; sex trafficking survivors had higher rates of pre-trafficking childhood abuse than labor trafficking survivors (source 15).
• Roe-Sepowitz et al. (Arizona, 167 homeless young adults, self-report survey): the sex trafficked group significantly more likely to have experienced sexual abuse by a parent or guardian both as a child and as an adolescent (source 31).

Add to those Brewin's account of the ICD-11 literature, which records childhood physical or sexual abuse as “more strongly related to CPTSD than PTSD” (source 23), and the convergence extends to the diagnostic literature that was not studying trafficking at all.

This is the strongest correlation in the brief. Nothing in it requires trafficking to be the causal agent, and that is the point: the association runs both before and after the trafficking, which is what makes attributing the resulting psychology to the trafficking itself so difficult.

Signal 3 (strong): Trafficking screening instruments are highly sensitive and poorly specific, measured independently in adult and paediatric settings by different research groups.

Four independent validation exercises, three research groups, two age bands, ten sites across at least seven states:

Sensitivity in a band from 84 to 100 percent, specificity in a band from 53 to 74 percent, with no overlap between the two bands.

Corroborating this from an entirely different direction: HHS's own 2018 toolkit states plainly that “No validated tools exist for the purpose of screening clients/patients for trafficking across all public health settings” and that its own instrument “has not yet been validated or evaluated in the field” (source 2), and the December 2024 program instruction, while now naming six setting-specific validated tools, still offers no evidence comparing their effectiveness (source 3). Federal policy and independent measurement agree on the shape of the problem.

Signal 4 (moderate, downgraded from strong in the multi-model verification pass): Complex trauma presentation in trafficked young people is driven by multiple trauma exposure rather than by trafficking as such.

Three independent findings, three research groups:

• Ottisova, Smith and Oram: trafficked and non-trafficked children with PTSD who had experienced multiple trauma showed more CPTSD symptoms than non-trafficked children with PTSD exposed to single-event trauma. The grouping fell along the multiple-versus-single trauma line, not the trafficked-versus-not line (source 10).
• Perry et al.: in 110 young people with substantiated CSE/T, the significant predictor of meeting clinical criteria for post-traumatic cognitions was the number of trauma categories experienced (source 17).
• The State Department TIP Office fact sheet, written independently of both, opens by naming polyvictimisation as the frame: children who are trafficking victims “often experience multiple forms of trauma, referred to as polyvictimization... This intersection of multiple traumas leads to complex trauma, making children more vulnerable to exploitation” (source 7).

Different institutions (a UK mental health trust, a US university research team, the Department of State), different methods (records cohort, therapy-intake survey, expert consensus fact sheet), same finding.

Signal 5 (moderate): The trafficking bond persists after physical separation from the trafficker.

• Casassa et al.: across the reviewed literature, “victim's feelings of love remained even after exiting trafficking,” and no reviewed article described how such a bond could be severed (source 21).
• Mahon et al.: 43.2 percent of women in Ohio specialty courts, all legally trafficking victims, did not identify as victims before programming, and 12.2 percent still did not afterward (source 28).
• US State Department TIP Office: caregivers are advised that “some children might have positive feelings towards their abusers and express hate that you took them away from these individuals” (source 7).

Three institutions, three methods. Rated moderate rather than strong because the three describe closely related but not identical constructs (continuing affective attachment, refusal of the victim label, and hostility toward a rescuer), and none of them measures the same variable as the others.

Signal 6 (narrow): Depression, not PTSD, is the most prevalent condition in trafficked minors.

• Kiss et al., 387 trafficked children in three Southeast Asian countries: depression 56.3 percent, anxiety 32.6 percent, PTSD 25.5 percent (source 11).
• Palines et al., 143 trafficked children in Wisconsin: depression 45.5 percent, PTSD 19.6 percent (source 16).
• Hopper and Gonzalez, 131 mixed-age survivors: depression 71 percent, PTSD 61 percent (source 15).

Three studies on two continents with different instruments and different age ranges, all placing depression above PTSD. Rated narrow because they are all clinic- or service-recruited samples, and because two of the three are studies of minors while the third is not, so the populations are not equivalent.

This signal matters because it runs against the popular framing, which treats PTSD as the signature condition. The located evidence does not support that framing. Note the tension with Laird's meta-analysis (Section 11.1), where PTSD carries by far the largest pooled odds ratio relative to non-exploited peers. Those two findings are compatible: PTSD can be the condition most elevated by exploitation while depression is the condition most common among the exploited.

Signal 7 (weak, tested and reported as weak): Duration of trafficking as a driver of severity.

This was tested against three sources and does not hold consistently.

• Abas et al.: duration of trafficking AOR 1.12 per month over a 2 to 31 month range, 95 percent CI 0.98-1.29, p = 0.089. Large point estimate, not statistically resolved at n = 120 (source 13).
• Hossain et al.: more time in trafficking associated with higher depression and anxiety, AOR 2.2, 95 percent CI 1.1-4.5. Significant (source 12).
• MacCabe et al.: longer trafficking durations associated with higher DES-II dissociation scores, in a 52-person purposive sample with no comparison group (source 27).

Named specifically as NOT correlation

Each of the following rests on a single source. They are not weaker facts for it. They must not be read as corroborated.

• The 87.8 percent health care contact figure. That 87.8 percent of trafficking victims had contact with a health care provider while being trafficked, with 63.3 percent treated at a hospital or emergency room, comes from Lederer and Wetzel's convenience sample, of whom 98 answered the health care contact questions (source 29). It is the most-repeated statistic in the trafficking-and-health-care literature and it is one study, service-recruited, partly advocacy-funded.

• The 43.2 percent non-self-identification figure. One study, 74 women, three courts, one state, retrospective self-report of a prior belief state (source 28).
• The Arizona homeless young-adult prevalence figures (38.9 percent sex trafficking, 53.3 percent any trafficking). One survey, one state, 167 participants, self-report, drawn from four partner agencies (source 31).
• The somatic-symptom gap (two thirds of trafficked children versus 10-11 percent of non-trafficked). One records study, 51 trafficked children, one UK mental health trust (source 10). This is the largest trafficked-versus-comparison difference in the brief and it is single-sourced.
• The dissociation profile (depersonalisation and absorption above amnesia). One study, 52 purposively sampled adults, no comparison group (source 27).
• The Vera tool's 87 percent of questions predicting victimisation. One study, one sample of 180, no external validation set (source 5).
• The 70 percent local-area-code finding from the Super Bowl decoy study. One university press announcement describing one unlocated study report (source 34).

10Key Assumptions Check

A Key Assumptions Check names the load-bearing premises that the analysis above rests on and that were not verified in this session. These are flagged not because they are doubted, which is a different thing, but because they are unverified, and because if any of them is wrong, specific conclusions above move.

Assumption 1: That findings from post-trafficking service populations generalise to trafficked people who never reach a service.

Why unverified: by construction. Every prevalence, symptom and comparison figure in Sections 4 and 5 comes from people already in a shelter, court, clinic or mental health service. No located study sampled a general population and then identified trafficking within it, except the two Florida administrative studies, which sampled a justice-involved population instead.

What changes if wrong: nearly every prevalence figure in the brief. The direction of the bias is genuinely unknown and could run either way. If service populations over-represent the most symptomatic, the figures are inflated. If the most controlled and most bonded victims are precisely the ones who never reach a service, the figures are deflated and the trauma-bonding findings in particular understate the phenomenon.

Assumption 2: That the Southeast Asian and Eastern European post-trafficking cohorts describe psychology applicable to a US southwestern border corridor.

Why unverified: the two largest and best-designed mental health datasets in this brief are from Cambodia, Thailand and Vietnam (sources 11, 14) and from Moldova (source 13). Trafficking type, cultural framing of mental distress, the instruments' validation populations, and the post-exit service environment all differ materially from Arizona.

What changes if wrong: Sections 5.1 and 5.2 lose their strongest numbers, and the brief falls back on much smaller US samples. Note that the correlation in Signal 2 survives this, because it is attested in the US studies independently.

Assumption 3: That self-reported trafficking status in survey research corresponds to the legal category.

Why unverified: the Arizona YES figures (source 31) rest on participants' own reports of having been “sex trafficked” or “labor exploited.” Section 4.3 establishes that a large fraction of legally trafficked people do not apply the label to themselves. If that under-labelling operates in the YES sample, the 38.9 percent figure is a floor and not an estimate.

What changes if wrong: Section 7.2's prevalence figures move, and the trafficked-versus-non-trafficked comparison in that study is contaminated because some of the comparison group are misclassified cases.

Assumption 4: That the screening-instrument prevalence figures measure the true rate of trafficking in those settings.

Why unverified: the RAFT study's own limitation section states that “Patients that were ineligible for participation may have been at higher risk,” naming intoxication, mental illness, inability to consent alone, and nighttime presentation, and reports a 48 to 50 percent participation decline rate with unknown directional bias (source 26). The excluded and declining groups are plausibly enriched for trafficking.

What changes if wrong: the base-rate arithmetic in Section 6.5. If the true emergency department prevalence is meaningfully above 1.1 percent, positive predictive value rises and the “96 out of 100 positives are not trafficked” conclusion weakens proportionally. At 3 percent prevalence the same sensitivity and specificity give a positive predictive value near 10 percent, which is still low but a materially different picture. The conclusion is directionally robust but its magnitude is not.

Assumption 5: That the psychological literature and the identification literature are describing the same population.

Why unverified: the clinical studies overwhelmingly describe people identified as trafficking victims and receiving post-exit services. The screening studies describe people being sorted in real time, most of whom are still in the situation. These are different points in a trajectory and the brief treats findings from one as informing the other.

What changes if wrong: Section 4.5's claim that field indicators are second-order effects of coercion tactics. If the psychology of someone still under a trafficker's control differs materially from that of someone months post-exit, then the symptom profiles in Section 5 are a poor guide to what a screener will encounter, and the whole identification enterprise is even harder than Section 6 makes it look.

Assumption 6: That the low Cochise County count and the Santa Cruz County zero reflect absent identification infrastructure rather than absent trafficking.

Why unverified: the inference that these figures measure detection rather than incidence rests on the Arizona report's own general warning about identification counts, on the 2017 report's explicit exclusion of both counties from its scope, and on the referral-based design of the collaborative. No source located this session tests incidence in either county by an independent method.

What changes if wrong: Section 7.3, and any regional picture built on it. If the border counties genuinely have proportionally less child sex trafficking than the Phoenix and Tucson metropolitan areas, then a corridor-focused regional account overstates the local phenomenon, and the Santa Cruz zero is a finding rather than an artefact. Note that the companion routes brief reaches the opposite conclusion for Cochise County from criminal-offence data, which is a different instrument measuring a different thing, and that neither brief resolves the incidence question.

Assumption 7 (added in the multi-model verification pass): that this document's peer-reviewed sources are institutionally independent of one another.

Why unverified, and it largely is not: Cathy Zimmerman is an author on sources 8, 11, 12 and 14. Sian Oram is an author on sources 8, 9, 10 and 13. Sources 9 and 10 are the same 51-child cohort analysed twice. Seven of this document's peer-reviewed sources trace to a single London School of Hygiene and Tropical Medicine and King's College London research network. The drafting pass treated several of these as independent corroboration in Section 9 and did not flag the overlap anywhere.

What changes if wrong, which is to say if that network shares a methodological bias: Signals 2, 4 and 7 all weaken, because each counts sources from that network toward its independence requirement. The correction has been applied inside those signals individually. The two findings least exposed to this are the screening-instrument performance band (entirely separate research groups) and the Laird meta-analysis (source 41, Deakin University, outside the network).

Assumption 8 (added in the multi-model verification pass): that a non-significant result in a small sample is evidence of no difference.

Why unverified: this document repeatedly relies on null findings from samples of 41 to 120 people. A compulsory-admission odds ratio of 0.27 with an interval of 0.06 to 1.25 describes a possible 73 percent reduction that the study could not resolve, not an absence of effect. The same applies to the trafficked-versus-multiple-trauma comparison at n = 11 against 21, and applied to the Abas duration coefficient until the verification pass caught it.

What changes if wrong: it is wrong, and the corrections are applied in place at Sections 5.1, 5.3, 5.7, 9 and 11.3. This assumption is recorded here because a reader should apply the same scepticism to any remaining null in this document that this document failed to apply to itself.

11Analysis of Competing Hypotheses

Analysis of Competing Hypotheses is a discipline against anchoring: the failure mode where an analyst adopts the first plausible explanation and then collects only evidence that confirms it. The method requires laying out the rival explanations for a contested question side by side and testing each against the same body of evidence, including evidence that fits none of them.

Three questions in this brief are genuinely contested by the located sourcing.

11.1 Is there a psychological profile specific to trafficking, distinguishable from that of other severe chronic trauma?

What the in-scope sourcing supports: trafficked people show high rates of PTSD, depression, anxiety, dissociation, self-harm, negative self-concept and somatic symptoms. Studies that place them beside comparison groups return mixed results depending on which comparison group is used.

H1: There is a distinct trafficking signature. Evidence for: Cole et al. propensity-matched 43 commercially exploited youth against 172 sexually abused non-exploited youth in a US clinical dataset and found significant differences on standardized PTSD and behaviour-checklist measures, avoidance, hyperarousal, dissociation and several behavioural indicators, concluding that the exploited group carried “more severe symptomatology and behavior problems” (source 36). Palines et al. found significantly higher rates of ADHD, bipolar disorder and PTSD in 143 trafficked Wisconsin children than in all three high-risk comparison groups (source 16). Hopper and Gonzalez found trafficking type predictive of symptom count beyond the role of pre-trafficking child abuse (source 15). The Arizona YES survey found the sex trafficked group significantly different from non-trafficked homeless young adults on self-harm, multiple diagnoses, suicide attempts and every ACE (source 31). The somatic finding in the UK cohort, two thirds versus 10-11 percent, is a very large gap (source 10).

The authors state that “the findings are correlational and cannot imply causation of CSE.”

Two things about this source matter for the hypothesis under test. It is from Deakin University in Australia, entirely outside the LSHTM and King's College London network that supplies much of the rest of this document's evidence, so it is genuinely independent. And its comparator is non-exploited peers, not other trauma-exposed youth, so it establishes that exploited youth differ sharply from the general adolescent population without establishing that they differ from comparably abused adolescents. It is decisive against a strawman version of H2 and not decisive against the real one. Note also that its running-away odds ratio of 2.28 is well below the 4.36 de Vries found in Florida, which is a useful reminder that these pooled figures average over very different populations.

H2: There is no trafficking-specific signature; the profile is that of severe chronic interpersonal trauma, and the apparent differences are differences in trauma dose. Evidence for: the UK matched cohort found no significant difference between trafficked and non-trafficked children in the same service on pathway into care, compulsory admission, inpatient stay or change in functioning; the only significant difference was duration of service contact (source 9). Ottisova, Smith and Oram found the CPTSD grouping fell along the multiple-versus-single-trauma line rather than the trafficked-versus-not line (source 10). Perry et al. found number of trauma categories, not any trafficking feature, predicted post-traumatic cognitions (source 17). Abas et al. found childhood sexual abuse the dominant predictor of post-trafficking disorder, with trafficking duration only borderline (source 13). Brewin records childhood abuse as the exposure most strongly associated with CPTSD generally (source 23).

H3: The question is unanswerable as posed, because “trafficking” is a legal category rather than an exposure type, and it bundles heterogeneous experiences. Evidence for: Section 2's terms of art. A sixteen-year-old trading sex for a couch and a woman held in domestic servitude for a decade are both trafficking victims in law and share almost nothing as an exposure. Hopper and Gonzalez found sex and labor trafficking survivors differing systematically from each other (source 15). Iglesias-Rios et al. found two distinct violence-and-coercion classes within each sex, with different mental health associations (source 14).

Weighing: H2 has the better-controlled design behind it, but not by as much as an earlier draft of this document claimed.

What survives that correction, and it is still substantial: the CPTSD analysis from the same research group (source 10) grouped trafficked children with non-trafficked multiple-trauma children and separated both from single-trauma children, with trafficked versus multiple-trauma non-significant at z = -1.58, p = 0.11. Perry's finding that number of trauma categories, not any trafficking feature, predicted post-traumatic cognitions is from an independent US sample (source 17). Abas's finding that childhood sexual abuse dominates the risk model is from a third research network in a third country (source 13). H2 rests on three legs, and only one of them was damaged.

Against that, H1's strongest study, Palines, compared against pooled literature figures for other populations rather than a matched local sample, and its authors themselves attribute the excess diagnoses to misdiagnosis of complex trauma rather than real excess pathology. The Arizona YES finding is real but is a comparison at the extreme tail of an already extreme distribution, and its list of significant differences includes asthma, which suggests it is detecting general severity rather than a psychological signature.

The evidence for H1 that this document takes most seriously is Hopper and Gonzalez's result that trafficking type predicted the number of trauma-related symptoms beyond the role of pre-trafficking child abuse (source 15). That is the one located finding designed to isolate exactly the effect H2 denies, and it found one. It is reported from the abstract only; the paper's full text is paywalled and this document could not retrieve it, so the analysis method and effect size behind that sentence are unverified (Section 13).

H3 is not a dodge. It explains why H1 and H2 both find support: they are averaging over experiences that do not belong in the same category. The two studies that disaggregated (sources 14 and 15) both found structure inside the “trafficking” label that the aggregate studies cannot see.

What holds: the symptom types documented in trafficking victims are the symptom types of severe chronic interpersonal trauma, and no located source describes a symptom that occurs in trafficking and not elsewhere. Trauma dose and prior childhood abuse are powerful predictors across every study that measured them.

What no longer holds: that matched comparisons find no difference. Two of the three located comparison studies capable of detecting a difference found one, and the two that found differences found them in the direction of greater severity among the trafficked or exploited group. The study most often quoted for the null could not have found a diagnostic difference by design.

The synthesis the evidence actually supports, and the reason this sits at Low rather than resolving toward H1: what distinguishes commercially exploited youth from comparably abused youth in the located studies is largely severity and exposure history rather than symptom kind. Varma's discriminating variables are running away, substance use, violence exposure and prior child protective services involvement. Cole's are severity scores on standard instruments plus behavioural problems. Neither describes a symptom a clinician could see in a room and attribute to trafficking. So the practical conclusion for identification survives intact even as the theoretical claim weakens: there is probably no marker that identifies a trafficking victim as such, but this document was wrong to present that as an established finding about the underlying psychology rather than as an open question with evidence on both sides. Section 13 carries the specific study that would settle it.

11.2 Is trauma bonding a distinct construct, or a relabelling of coercive control plus ordinary attachment?

What the in-scope sourcing supports: the phenomenon is consistently described. Its measurement base is weak. The most famous name for it is not a diagnosis.

H1: Trauma bonding (or trauma-coerced attachment) is a distinct psychological construct with its own mechanism, namely intermittent reinforcement under conditions of captivity. Evidence for: Casassa et al.'s four features are consistent across fifteen articles and specify a mechanism, the deliberate alternation of positive and negative interaction, that is not part of the definition of coercive control alone (source 21). Biderman's “occasional indulgences” tactic independently names the same mechanism from a source that had never heard of trafficking (source 20).

H2: It is coercive control plus normal attachment processes, and the special label adds nothing. Evidence for: every feature Casassa et al. identify (power imbalance, alternating treatment, self-blame, internalisation of the perpetrator's view) is describable in the vocabulary of coercive control and the ICD-11 CPTSD criterion of negative self-concept (source 23). The scoping review found only fifteen qualifying articles, and reported that not one described how the bond could be severed, which is what one would expect if the construct has not yet been operationalised well enough to be intervened upon (source 21).

H3: It is the same folk concept as “Stockholm syndrome” with a more clinical name, and inherits that concept's lack of empirical grounding. Evidence for: Namnyak et al.'s systematic review of Stockholm syndrome across PubMed, EMBASE, PsycINFO and CINAHL found twelve papers, mostly case reports, “ambiguity in the use of the term,” “no validated diagnostic criteria,” and that the diagnosis “is not described in any international classification system”; they conclude the apparent similarities between well-publicised cases “may be due to reporting and publication bias” (source 22). Casassa et al.'s own search strategy included Stockholm syndrome as a search term for trauma bonding, which is itself evidence the two concepts are not cleanly separated in the literature.

Weighing: H3 is the weakest of the three but is not empty, and it is the reason this section exists. The Stockholm syndrome literature is a cautionary case of a phenomenon that is vividly described, universally recognised, and empirically hollow. H1 has the better mechanism: intermittent reinforcement is a well-established behavioural principle independent of this literature, and its appearance in Biderman's 1957 catalogue and in a 2022 review of trafficking cases is not a coincidence that H2 explains well. H2 is the most parsimonious and is probably right that the label adds less than its popularity implies.

11.3 Does the duration of trafficking drive the severity of the psychological outcome?

What the in-scope sourcing supports: three studies tested this and did not agree.

H1: Longer exploitation produces worse outcomes, dose-response. Evidence for: Hossain et al., more time in trafficking associated with higher depression and anxiety, AOR 2.2, 95 percent CI 1.1-4.5 (source 12). MacCabe et al., longer duration associated with higher dissociation (source 27). Intuitively expected.

H2: Duration does not independently drive outcome once prior history and post-exit conditions are controlled. Evidence for: Abas et al., in the only study using blind structured clinical diagnosis, reported a duration association that did not reach significance (AOR 1.12, 95 percent CI 0.98-1.29, p = 0.089) against childhood sexual abuse at 4.68 and unmet post-trafficking needs at 1.80 (source 13).

H3: Duration matters for some outcomes and not others, and the studies disagree because they measured different outcomes. Evidence for: Hossain measured symptom levels of depression and anxiety on self-report inventories; Abas measured diagnosed DSM-IV disorder by structured interview; MacCabe measured dissociation on the DES-II. Note that Hossain's own result already contains this pattern internally: more time since trafficking lowered depression and anxiety but did not lower PTSD, meaning duration-related effects behaved differently by outcome within a single study (source 12).

Weighing, revised in the multi-model verification pass: with the unit error corrected, all three studies now have positive point estimates for duration. Hossain finds it significant for depression and anxiety (AOR 2.2, CI 1.1-4.5), MacCabe finds it associated with dissociation, and Abas finds a large per-month estimate that missed significance in a small sample. What looked in the drafting pass like a genuine three-way disagreement is closer to consistency plus one underpowered study, and the earlier draft's remark that Hossain and Abas sharing a research network made their disagreement “more informative than their agreement would have been” was reasoning built on a disagreement that largely was not there. H3 remains attractive because the three studies do measure different outcomes, but it is no longer needed to reconcile a conflict.

12Sources Cited

Grades are Admiralty two-axis (Section 1.1). All URLs were retrieved during the compilation session on 2026-09-09 unless otherwise noted.

Note on grading, added in the multi-model verification pass: the credibility digits on six entries were corrected after an audit against the scale in Section 1.1 found them inconsistent with it. A “1” means confirmed by other independent sources, and it had been applied to several sources this document elsewhere describes as single-sourced. Sources 5, 9, 22, 30 and 31 were regraded downward on the credibility axis for that reason. The audit also noted that source 29 was downgraded to B3 partly for advocacy funding and service-provider recruitment while source 31 shares both properties, which is recorded here as an inconsistency this document has narrowed but not fully resolved; source 31's funders are named in its entry so a reader can weigh it.

Primary / official (US federal and statute)

Peer-reviewed

Research institutions and regional sources

Added in the multi-model verification pass

Sources located but excluded

See the table and notes in Section 8. Nothing in that category is cited for a factual claim anywhere in this document.

13Open items for a follow-up research pass

Resolved during this session

RESOLVED: whether a validated US trafficking screening instrument exists. It does, in narrow settings. RAFT (source 26) has genuine external validation in adult emergency departments, and the Greenbaum tool (source 25) has multi-site paediatric evaluation. What does not exist, per HHS's own statement, is a tool validated across public health settings generally (source 2). Landed in Section 6.

RESOLVED: whether the trafficked-versus-comparable-group question has been tested directly. It has, twice, with conflicting results, by different designs (sources 9 and 16). Landed in Section 5.7 and Section 11.1.

RESOLVED: whether Stockholm syndrome has a clinical evidence base. It does not. Namnyak et al. 2008 is definitive on the point and no later systematic review contradicting it was located (source 22). Landed in Section 11.2.

RESOLVED, partially: the Arizona regional picture. Statewide and Tucson-metropolitan data was located and is strong (sources 30, 31, 32, 33). County-level data for Cochise and Santa Cruz was not, and remains open below.

Still open

1. Cochise and Santa Cruz County victim-psychology data. Narrowed in the multi-model verification pass. Identification counts for both counties were found and are now in Sections 7.1 and 7.3, so that part is closed. What remains absent is any psychological or victim-identification data specific to either county. The closest located item is the SAATURN final evaluation (Stevens and Black, University of Arizona SIROW, April 2019), a task force whose stated service area was Pima, Cochise and Santa Cruz counties but whose published numbers are effectively Pima-only, with Cochise and Santa Cruz appearing solely as outreach activity. That report was not retrieved in full this session and is not cited as a source. Next step: retrieve the SAATURN 2019 and 2016 evaluations in full, and request Arizona Human Trafficking Council annual reports for 2021 and 2023 to 2025 directly (goyff.az.gov returned 403 to automated fetches; a browser may work).
2. The HHS ASPE substance use coercion policy brief. Downloaded but not readable; PDF content streams did not extract across two attempts. Next step: re-fetch through a different route (the ASPE report landing page rather than the direct /private/pdf/264166/ path) or retrieve the HTML summary. Flagged in place at Section 4.4.
3. The canonical acf.gov URL for the Adult Human Trafficking Screening Tool. The path /otip/resource/adult-human-trafficking-screening-tool-and-guide returned 404 during this session after redirecting from www.acf.hhs.gov to acf.gov. The document itself was read from a Penn State mirror and its current federal status was confirmed independently via the December 2024 program instruction (source 3). Next step: locate the current canonical acf.gov or nhttac.acf.hhs.gov path so the citation points at the government copy.
4. The National Human Trafficking Hotline data page. acf.gov/otip/research-policy/data/nhth-data returned HTTP 403 on attempt. No hotline figure is used anywhere in this document. Next step: retry, or use the OPRE evaluation reports of the hotline (acf.gov/opre) which are federal evaluations rather than operator-published statistics and would not carry the advocacy-source problem.
5. RESOLVED, partially: the branding-tattoo indicator. The verification pass established that no study measuring branding prevalence, sensitivity or specificity in an actual victim population exists in the peer-reviewed literature, searched across Europe PMC and including the Journal of Forensic Nursing. The only quantitative figure is a survey of 40 survivor-serving organisations estimating roughly 47 percent of the survivors they serve have been branded, which is a provider estimate rather than a victim count. Section 6.6 is rewritten on that. Separately, a companion brief in this project now treats marking and identification as its whole subject at 80 graded sources, and a reader wanting depth should use it. What remains open here is narrow and probably unanswerable from the published record: no victim-level prevalence or discriminative figure exists, and this document could not establish that one is forthcoming.
6. RESOLVED: the underlying ASU Super Bowl study report. Located and retrieved in full at the McCain Institute during the verification pass, added as source 35, and Section 7.4 rewritten on it. Two figures moved: 23 flagged advertisements rather than the press announcement's 21, and 73.3 percent local area codes rather than “some 70 percent.” The report also states in its own voice that “This does not indicate that the Super Bowl caused more sex trafficking.”
7. RESOLVED, and the earlier claim was wrong. US longitudinal post-exit data. An earlier draft asserted that “no located study tracks US survivors across years.” That assertion was false and is withdrawn. The verification pass located several, and one was retrieved and read in full: Bath and colleagues' exhaustive court-file review of 364 commercially sexually exploited youth in the Los Angeles STAR Court, 2012 to 2017, with a mean 494 days of supervision and paired pre-versus-post outcomes (runaway episodes 2.2 to 1.8; out-of-home placements 4.3 to 1.8; prostitution citations 0.97 to 0.17, all p < .001; depression recorded at 66 percent baseline against 68 percent post-supervision, traumatic stress disorder 30 to 37 percent) (source 38). Others located but not retrieved in full, and therefore not cited as sources here: an NIJ-funded quasi-experimental evaluation of the My Life My Choice survivor-mentor programme (Rothman, Bair-Merritt and Farrell, NCJ 253459, 2019, data at ICPSR 37599); Twis, Cimino and Plunk's double pre/post study of 95 youth survivors with six-month follow-up (PLOS One 19(1):e0291207, 2024); and two analyses of commercial sexual exploitation in the Add Health national cohort (Barnert et al., Public Health Reports 2022; Godoy et al., Int J Environ Res Public Health 2025). What remains genuinely open, stated precisely this time: no located US study follows identified trafficking survivors prospectively with repeated standardized symptom measures. The US longitudinal evidence is programme pre/post administrative data and general-population cohort analyses, not a survivor symptom cohort. Next step: retrieve the four studies named above in full and determine whether any carries repeated standardized instrument scores rather than file-derived diagnoses.
8. The study that would settle Section 11.1. Following from the addition of Cole et al., the specific missing design is now nameable: a comparison of trafficked or commercially exploited people against a comparison group matched on trauma load and prior victimisation rather than on demographics or on diagnosis, using the same standardized instruments in both arms. Cole matched on demographics and found differences; the UK cohort matched on diagnosis and could not have found them; nobody located has matched on prior trauma exposure, which is the variable every other finding in this brief points at. Next step: search the polyvictimisation literature specifically for trafficking-versus-matched-trauma-load designs.
9. Doychak and Raghavan on trauma-coerced attachment. Their 2020 Journal of Human Trafficking article is the most-cited primary study on the construct and is behind a Cloudflare challenge on the publisher's site that could not be cleared this session. Section 4.2 rests on the Casassa scoping review instead, which is a review rather than a primary study. Next step: retrieve through an institutional repository or the John Jay College faculty repository.
10. The Hopper 2017 polyvictimisation paper. “Polyvictimization and Developmental Trauma Adaptations in Sex Trafficked Youth,” Journal of Child and Adolescent Trauma, was located but is behind a Springer authentication redirect. It bears directly on Section 11.1 and would strengthen or weaken H2 depending on what it found. Next step: retrieve via PMC or an author repository.
11. Whether the somatic-symptom finding replicates. Source 10's two-thirds versus 10-11 percent gap is the largest trafficked-versus-comparison difference in this brief and is single-sourced from 51 children. Next step: search the paediatric and psychosomatic literature specifically for somatisation in trafficked versus non-trafficked youth. If it replicates, it is the strongest candidate for anything resembling a trafficking-specific marker and Section 11.1's assessment moves.

14Note on the eventual World-Building Document

This document draws no line from any real finding to any fictional character, faction, organisation or location. It names no Vampires of Tucson character. It does not represent that any fictional entity operates the way any real trafficking network operates, and it does not represent that any fictional person's psychology corresponds to any real survivor's.

Where the eventual World-Building Document draws on this material, that is a separate, later step taken under clearly acknowledged fictional licence. It should be treated as such rather than presented as continuous with the sourcing standard used here. A brief graded source by source is not a warrant for the fiction built afterward, and the fiction cannot inherit the brief's epistemic standing by being adjacent to it.

Three specific cautions apply to any fictional use of this material.

The first is the finding in Section 11.1, stated at the confidence it ended at rather than the confidence it started at. If the fiction depicts a character who is identifiable as a trafficking victim by psychological presentation alone, that is not supported by anything in this brief. The symptom types are those of severe chronic interpersonal trauma, and what separates exploited youth from comparably abused youth in the located studies is severity and exposure history, not a symptom a clinician could see in a room and attribute to trafficking. Note the shape of that claim carefully, because this document had to correct itself on it: the evidence does not establish that trafficked people are psychologically indistinguishable from other severely traumatised people. Two matched comparisons found real differences. What it establishes is that no marker identifies a trafficking victim as such.

The second is the base-rate arithmetic in Section 6.5. Fiction that depicts identification as a matter of noticing the right sign is depicting something the measured performance of every located instrument contradicts. What identification actually looks like, on the evidence, is a low-yield screen with a good negative predictive value, followed by a relationship built slowly enough that a person who does not consider themselves a victim might eventually say something.

The third is Section 4.3. The most consistently attested psychological fact in this brief is that victims commonly do not identify as victims and may direct hostility at the people who remove them from the situation. A narrative that assumes gratitude is not describing what the sources describe.

This brief is not authoritative outside the purpose stated in its header, is not clinical guidance, is not a diagnostic instrument, and should not be used to assess any real person. Every figure in it belongs to the specific population the cited study sampled, and none of them is a general truth about trafficking victims.