Development of the Healthcare Triggering Questionnaire in Adult Sexual Abuse Survivors

Abstract

Objective

One in 4 women and 1 in 10 men in the United States are survivors of sexual abuse (SA). For these survivors, healthcare experiences may trigger memories, thoughts, feelings or sensations related to this past abuse. Such triggering can be associated with negative responses to healthcare (e.g., anxiety, avoidance). However, to date, no healthcare triggering assessment tool exists. Therefore, the study goal was to describe the prevalence of healthcare triggering, to develop a brief Healthcare Triggering Questionnaire (HTQ), and to examine its initial validity.

Method

An initial pool of 117 items was developed based on previous research. Two-parameter logistic item response theory (IRT) models were used to develop the scales. SA survivors [male (n=233), female (n=222)] and a comparison group of non-SA individuals [male (n=114), female (n=106)] were recruited through Amazon Mechanical Turk and completed the study anonymously online.

Results

Three 10-item scales were developed: 1) The HTQ-M for males; 2) The HTQ-F for females; and, 3) The HTQ-U (unisex) for all respondents. The results supported the utility and initial validity of the gender-specific and unisex scales.

Conclusions

The HTQ scales are a psychometrically sound approach to evaluating healthcare triggering experienced by adult sexual abuse survivors. The HTQ may be considered for use by researchers interested in studying healthcare triggering, healthcare retraumatization, and healthcare adherence. The HTQ may also be of use to clinicians interested in identifying trauma survivors who are more likely to experience triggering in healthcare settings.

One in 4 women and 1 in 10 men in the United States have experienced some form of sexual abuse (SA) during their lifetimes (both contact and non-contact, Breiding, Smith, Basile, Walters, Chen, & Merrick, 2014). Such abusive experiences include rape (completed or attempted), sexual violence besides rape, sexual coercion, unwanted sexual contact, and noncontact unwanted sexual experiences (Breiding et al., 2014). For SA survivors, healthcare procedures, even those that may seem minor or routine, can share similarities with sexual abuse. Both can involve a sense of powerlessness, vulnerability, and exposure (Schachter, Stalker, Teram, Lasiuk, & Danilkewich, 2009). Due to these similarities, healthcare procedures can serve as “triggers” for sexual abuse (SA) survivors (Monahan & Forgash, 2000; Schachter et al., 2009). Triggers can be defined as, “… any event or situation in healthcare that evokes a sense of threat and results in post-traumatic stress reactions” (Dallam, 2010, p.82). For instance, prior research (Havig et al., 2008, p. 29) has identified “disrobing, lack of privacy, submissive body positioning, touch, and invasive procedures,” as common triggers of SA.

The phenomenon of healthcare triggering in SA survivors has been documented in qualitative studies across healthcare settings (Gallo-Silver & Weiner, 2006; Gesnik & Nattel, 2015; Schachter, Stalker, & Teram, 1999; Schnur & Goldsmith, 2011), empirical studies (Leeners et al., 2007; McGregor, Julich, Glover, & Gautam, 2010), and literature reviews (Havig, 2008). For example, one empirical study found that as many as 45% of SA survivors reported having abuse memories triggered during gynecologic exams (Robohm & Buttenheim, 1996). Both national (American College of Obstetricians and Gynecologists, 2011; American Medical Association, 1995) and international (Public Health Agency of Canada; Schachter et al., 2009) healthcare organizations have recognized the role triggering may play in SA survivors’ reactions to, avoidance of, or non-adherence to medical recommendations. Additionally, SA survivors themselves have expressed how important it is for healthcare providers to understand the triggering phenomenon (Schachter et al., 2009).

Empirical research indicates that SA history is significantly associated with lower rates of mammography screening (Farley, Golding, & Minkoff, 2002; Farley, Minkoff, & Barkan, 2001; Watson-Johnson, Townsend, Basile, & Richardson, 2011), as well as decreased likelihood of having had a Pap smear (Farley et al., 2002; Springs & Friedrich, 1992) and of seeking regular gynecologic care (Robohm & Buttenheim, 1996). In a survey of female SA survivors, 63.8% of participants stated that a fear of common medical examination procedures prevented them from attending regular health visits (McGregor, Julich et al., 2010).

To date, no scale has been developed to assess healthcare triggering. Without such a scale, it is impossible to measure individuals’ level of triggering, to include triggering in models of healthcare retraumatization and adherence, or to empirically assess the efficacy of interventions developed to reduce the impact of healthcare triggering in SA survivors. Given the prevalence of SA, a triggering scale is needed that could be used across healthcare settings. Therefore, the goal of this study was to develop a self-report instrument to assess healthcare triggering in adult survivors of SA. We also planned to examine associations between background characteristics (i.e., SA history and gender) and healthcare triggering. We hypothesized that SA participants would have higher levels of health care triggering than non-SA participants. The gender analyses were designed to guide decisions concerning how many forms of the instrument to create (e.g., whether to create gender-specific scales).

Methods

Participants

An anonymous, online format for data collection and recruitment was chosen based on research indicating that online communication can be preferable for stigmatized populations (Caplan & Turner, 2007), that trauma survivors feel less inhibited in responding to surveys using computer-based questionnaires (Griffin, Resick, Waldrop, & Mechanic, 2003), and that Internet-based responding gives participants the option to fill out the survey from any location at any time, which can enhance privacy. Additionally, literature suggests that the anonymity permitted by the Internet has the potential to reduce participant self-censorship, shame, embarrassment, and fear of judgment/stigmatization (East, Jackson, O’Brien, & Peters, 2008).

SA and non-SA participants were recruited through Amazon Mechanical Turk (MTurk). In brief, MTurk is an online marketplace where individuals (‘requesters’) can post work (‘Human Intelligence Tasks’ - HITs), such as surveys or translation requests, for contractors (‘Workers’) to complete. Workers log in to the MTurk site and are able to select from a list of currently available HITs, which appear on the MTurk Dashboard. Workers are able to see a brief description of the task, the time allotted, and the amount of compensation provided for completing the task (for more information on MTurk, see Mason & Suri, 2012). MTurk samples are typically more representative of the U.S. population than university samples in terms of age, gender, education levels and ethnic background (Berinsky, Huber, & Lenz, 2012; Buhrmester, Kwang, & Gosling, 2011; Litman, Robinson & Rosenzweig, 2014; Paolacci, Chandler, & Ipeirotis, 2010). Past validation studies have demonstrated that the quality of MTurk data is comparable to non-Internet based samples across a wide variety of tasks (Buhrmester et al., 2011; Paolacci et al., 2010; Sprouse, 2011; Suri & Watts, 2011), and that MTurk participants tend to be honest about self-reported information (Rand, 2011). MTurk is now extensively utilized for clinical research (see Chandler & Shapiro, 2016 for a comprehensive review).

Eligibility

To be included in the study, participants had to meet the following conditions. First, they had to reside in the United States, a criterion determined via MTurk’s built-in screening process, and implemented to reduce variability associated with medical practices/procedures in different parts of the world. Second, they had to have completed at least 100 prior assignments on MTurk with at least 90% of those assignments approved. Using experienced and reliable MTurk participants is recommended to improve data quality, to exclude automated bots or spammers, and to gather participants who take survey completion seriously (Mason & Suri, 2012). This procedure is commonly used in studies with MTurk samples (Gardner, Brown, & Boice, 2012). Although we recognize that any eligibility criterion affects the sample, overall we believe that the benefits to study validity of using the 90% criterion outweigh any potential limitations. Third, they had to report being over the age of 18. To be included in the SA subsample, participants had to additionally self-report a history of SA. Four hundred fifty-five adult SA survivors (233 male, 222 female) participated in the study, as did 220 non-SA individuals (114 male, 106 female). Data were collected during between April and June 2012.

Development of the HTQ

Initial item pool

The item development process began with a review of the extant literature on healthcare retraumatization. Based on previous qualitative and quantitative literature, we developed an initial set of 112 items. Based on our clinical experience in healthcare settings and discussions with a multidisciplinary group of healthcare providers (oncologists, nurses, therapists), we added five additional items. Together, these formed the 117 healthcare triggering items that were administered to study participants. Please see Supplementary Table 1 for source information for each item.

Measures

Background Questionnaires

SA History

Participants answered one question to determine group assignment (SA vs. non-SA). The question wording was, “Are you a survivor of sexual abuse (SA)? SA encompasses a range of offenses, including a completed nonconsensual sex act (i.e., rape), an attempted nonconsensual sex act, abusive sexual contact (i.e., unwanted touching), and non-contact SA (e.g., threatened sexual violence, exhibitionism, verbal sexual harassment)” (based on Basile & Saltzman, 2002). The response choices were “yes” or “no.”

Demographics

Participants were asked to report on their age, gender, race, ethnicity, education, region of the US where they reside (United States Census Bureau, 2001), whether they live in an urban/suburban/rural area, and whether they had received previous psychosocial support for SA.

Triggering Items

Participants responded to the initial, 117 item version of the HTQ. The instructions read, “Many individuals who have experienced sexual abuse find that aspects of the healthcare environment can trigger memories, thoughts, feelings, or sensations related to their sexual abuse. Below is a list of potential triggers likely to be encountered in the healthcare environment. We are interested to learn which of these may be a trigger for you. Please rate each item as best you can. If you are unfamiliar with an item, you can leave that question blank.” Examples of triggering items include: “breast exam,” “lying on an exam table,” and “disrobing/removing clothing.” The full set of triggering items presented to participants can be seen in Supplementary Table 1. Although participants initially responded on a 4 point scale (ranging from 0 = Not a trigger at all to 3 = A severe trigger), we collapsed the scale into a dichotomous format to reflect whether the item was a trigger (1) or not (0) for the analyses. We did this: 1) to create a simple count scale that could be easily summarized; 2) to facilitate fitting 2 parameter logistic IRT models for selecting items for the brief 10-item scales; and, 3) because endorsement of the four categories (i.e., 0–3) were often unevenly distributed. Respondents were assigned a score of 1 if their original response was 2 or 3 and a 0 if their original response was 0 or 1.

Validity

Medical Anxiety

Participants were asked to complete an eight-item scale assessing anxiety reactions to medical settings, adapted from the IDAF-4C+ (Armfield, 2010; Armfield, 2011). The scale has been shown to be reliable (α=.91) and valid. In our sample, the alpha coefficient was .94. Sample items include, “I generally avoid going to the doctor because I find the experience unpleasant or distressing,” and, “I get nervous or edgy about upcoming medical appointments.”

Healthcare Avoidance

Participants were asked “Do you avoid visiting the doctor as often as you feel you need to?” (yes/no). If they responded yes, they were asked to indicate the reason why: “lack of time/busy,” “cost/too expensive,” “do not like doctors,” “inconvenient to get to,” “fear or anxiety,” “just don’t get around to it,” and “other reason” (open-ended) (Armfield, 2010; Armfield, 2011).

Procedure

The HTQ was built as an online survey using PsychData (https://www.psychdata.com/).

Recruitment

Three MTurk sitewide postings were created to enhance recruitment of both male and female SA survivors, as well as of individuals without an SA history (see Appendix A). The male SA posting described the study purpose as, “To learn more about which aspects of medical treatments and exams are most difficult for male sexual abuse survivors.” The female SA posting described the study purpose as, “To learn more about which aspects of medical treatments and exams are most difficult for female sexual abuse survivors.” The non-SA posting described the study as a “health survey” and described its purpose as follows, “In this research project, we are specifically interested in hearing about the healthcare experiences of men and women who are not survivors of sexual abuse. In particular, we want to learn more about which aspects of medical treatments and exams are most difficult for individuals with no sexual abuse history. This will allow us to compare the experiences of non-abused individuals to abused individuals.”

All Workers (SA and non-SA) who accepted their respective HIT (n=770) were directed to the same introductory, implied consent page. This consent page described the purpose of the study, informed participants that the study had been approved by the Program for the Protection of Human Subjects at our institution, and stated that the survey was anonymous. Workers were also informed that they would be paid one dollar through MTurk for their participation. Please note, all participants (SA and non-SA) were paid the same one dollar regardless of their SA history. At the bottom of this consent page, Workers were asked to check a box signifying consent. The box was checked by 755 participants (2 individuals reported that they were not interested in participating, 13 left the consent item blank).

These 755 participants were then directed to proceed to the survey. Four-hundred fifty-five SA survivors (233 male, 222 female) and 220 individuals (114 male, 106 female) without an SA history had complete data and were used in the analyses. Given that the non-SA sample was included for a validity check, rather than for scale development, we felt this 2:1 ratio was appropriate.

At the conclusion of their study participation, all participants were presented with a list of contact information for online/phone SA survivor resources (RAINN, 1 in 6, Abuse Victim Hotline, Adult Survivors of Childhood Abuse). The provision of support resources is consistent with literature on research with traumatized populations (Becker-Blease & Freyd, 2006; Becker-Blease & Freyd, 2007).

Statistical Analyses

The initial focus in our analysis was the description of the 117 healthcare trigger items. Specifically, we summarize the response percentages associated with each healthcare trigger separately for men and women, and also separately for SA and non-SA individuals. We report and test differences using odds ratios. Subsequently, we use the data from the SA survivors to develop a set of short screening scales for men and women. We used an Item Response Theory (IRT) approach to develop these scales (Embretson & Reise, 2000). The advantage of using an IRT approach is that it allowed us to select “trigger” items across the range of SA individuals who vary from relatively low levels of healthcare triggering to those who experience high levels of triggering. In addition, IRT approaches differ from classical test theory approaches to item development in that it can create very short scales without a significant loss of validity or reliability. Thus, in developing the short screening scales, we selected items that systematically represented the range of healthcare triggering (referred to in IRT as “difficulty levels” or “thresholds”), and that had a strong ability to distinguish between people who were above and below those different thresholds (referred to in IRT as “discrimination parameters”).

Results

Sample Characteristics

Table 1 provides a description of the overall sample demographics. We did not find any meaningful statistical differences between SA and non-SA samples on any demographic features.

Table 1.

Sample Characteristics

	SA Male (n=233)	SA Female (n=222)	Non-SA Male (n=114)	Non-SA Female (n=106)
	(M, SD)	(M, SD)
Age	29.2 (9.4)	32.6 (10.8)	30.3 (10.9)	34.1 (12.6)
	n (%)	n (%)
Race
White	182 (78.1)	185 (83.3)	93 (81.6)	81 (76.4)
Other	51 (21.9)	37 (16.7)	21 (18.4)	25 (23.6)
Ethnicity
Latino/a	38 (16.3)	18 (8.1)	8 (7.0)	6 (5.7)
Not Latino/a	195 (83.7)	204 (91.9)	106 (93.0)	100 (94.3)
Education
≥College Education	101 (43.3)	105 (47.3)	46 (40.3)	58 (54.7)
< College Education	132 (56.7)	117 (52.7)	68 (59.7)	48 (45.3)
Region of Residence
New England	20 (8.6)	17 (7.7)	9 (7.9)	10 (9.5)
Middle Atlantic	47 (20.2)	41 (18.5)	17 (14.9)	14 (13.3)
South Atlantic	54 (23.2)	55 (24.8)	25 (21.9)	21 (20.0)
East South Central	11 (4.7)	19 (8.6)	5 (4.4)	5 (4.8)
East North Central	20 (8.6)	17 (7.7)	18 (15.8)	23 (21.9)
West South Central	22 (9.4)	25 (11.3)	6 (5.3)	10 (9.5)
West North Central	17 (7.3)	8 (3.6)	3 (2.6)	4 (3.8)
Mountain	9 (3.9)	14 (6.3)	4 (3.5)	2 (1.9)
Pacific	33 (14.2)	26 (11.7)	27 (23.7)	16 (15.2)
Type of Region
Urban	91 (39.1)	72 (32.4)	34 (29.8)	37 (34.9)
Suburban	100 (42.9)	99 (44.6)	59 (51.8)	53 (50.0)
Rural	42 (18.0)	51 (23.0)	21 (18.4)	16 (15.1)
Previous psychotherapy, counseling, or support groups to help with SA history
Yes	84 (36.1)	94 (42.3)	-	-
No	149 (64.0)	128 (57.7)	-	-

Note: Percentages may not exactly equal 100% due to rounding.

Impact of Sexual Abuse History on Healthcare Triggering at the Item Level

Our first hypothesis was that SA survivors will exhibit higher endorsement of healthcare triggering than non-SA individuals. Supplementary Table 2 summarizes the comparison between the SA and non-SA samples (based on odds ratios and statistical tests) on the severity of triggering (1=moderately to severely triggered; 0=slightly to no triggering) for each of the 117 healthcare trigger items. Items are rank-ordered by the size of the odds ratio: Odds ratios greater than 1 indicate a higher likelihood of triggering endorsement by the SA group compared to the non-SA group. Confirming our hypothesis, 78 of the 117 triggers (67%) had odds ratios greater than 1, indicating a greater likelihood of triggering in the SA sample. Of the 78 triggers, 44 were statistically significant (p<0.05). Of the 39 items with odds ratios less than 1, indicating greater likelihood of distress in the non-SA group, 9 were statistically significant (p<0.05).

Percentage of SA Participants Who Rated Each Item as Triggering

Supplementary Table 3 is a summary of the endorsement percentages of the 117 healthcare trigger items in SA participants. Items were rank-ordered by the percentage of participants who indicated that the item would be a trigger for memories, thoughts, feelings or sensations related to their SA. Of the 117 healthcare trigger items, one is unique for males and eight are unique for females, and these appear at the bottom of the table. All of the items were endorsed by at least one individual; indeed, the endorsement percentages per item range from 10% to 56%.

These percentages are also shown separately for males and females, and triggers that are marked with an asterisk were endorsed by a significantly different percentage of males and females (p < 0.05, two-tailed test). We recognize that the large number of statistical tests reported in Supplementary table 3 likely result in an inflated overall Type I error rate. Thus, these results should be viewed as guidelines for triggers that may be different between males and females. One consequence of adjusting for multiple comparisons would be a large increase in the type II error rate (Maxwell, 2004), and our concern here is making sure that we do not overlook potential differences rather than suggesting a difference that might not replicate. We note that thirty (26%) of the items were significantly different based on self-reported gender, supporting the decision to create three versions of the scale – a male version, a female version, and a unisex version. Interestingly, for 28 of these 30 items (93%, binomial probability < 0.0001), the endorsement percentages were higher for males.

Individual Differences in Healthcare Trigger Endorsement among the SA Sample

Supplementary Figure 1 is the stem-and-leaf plot showing the distribution of SA participants on overall healthcare trigger endorsement. Fifty-nine (13%) of the SA participants endorsed no triggers. Four SA participants (1%) endorsed all triggers. Supplementary Figure 1 also shows that the distribution of SA participants on percent of healthcare trigger items endorsed is highly positively skewed (Skewness = 1.2, SE = 0.114). The median percent of healthcare triggers endorsed is 13.8%. The median for males is 13.7% whereas the median for females is 14.1%, a difference that is not statistically significant (Mann-Whitney U Test, p = 0.515). Of the ten individuals who endorsed more than 90% of the triggers, eight were males. This may partially account for the finding that, of the triggers that were endorsed differently by males and females, the vast majority was endorsed at a higher rate by males, whereas the overall median endorsement rate was slightly higher for females.

Development of Screening Scales

A primary goal of this research was to use these data to develop short screening scales to identify patients who may be susceptible to triggering during healthcare procedures. To develop these scales, we used the data from the SA sample. From the original set of 117 items, we created three 10-item healthcare triggering scales that span a wide range of triggering response thresholds (“item difficulties”). We developed a male-specific and a female-specific scale, as well as a unisex scale. These scales were developed by fitting two-parameter logistic item response theory models and identifying items that systematically represented the range of item difficulties (b parameters) and that had high discrimination (a parameters). IRTPRO 2.1 (Cai, Thissen, & Du Toit, 2011) was used for analyses. The resulting scales (HTQ-M for males; HTQ-F for females, HTQ-U, unisex) are shown in Table 2. Each person’s score on the HTQ is the sum of items identified as triggers. We adjusted this sum for missing data by: 1) summing all the non-missing items; 2) dividing that sum by the number of non-missing items; and, 3) multiplying that result by 10.

Table 2.

Summary of Endorsement Percentages for each of the three scales for the SA group (HTQ-M, HTQ-F, HTQ-U).

HTQ-M Scale
Item	Question	Percent Severe (Total)	Discrimination Parameter (a)	Difficulty Parameter (b)
HTQ4	Genital Exam	0.57	1.31	−0.31
HTQ6	Prostate Exam	0.51	1.29	0.03
HTQ84	Provider Touches Your Buttocks	0.45	2.66	0.18
HTQ117	Feeling a Lack of Control	0.41	2.14	0.3
HTQ62	Provider whose Mannerisms are Similar to Perpetrator	0.36	2.36	0.47
HTQ72	Provider Touches Your Lips	0.30	4.19	0.58
HTQ114	Being in Close Physical Proximity to Authority Figures (e.g., Medical Provider)	0.24	2.42	0.83
HTQ12	Radiological/Imaging Tests	0.20	2.38	1.11
HTQ75	Quick Movements by Providers	0.16	2.70	1.17
HTQ94	Dark/Low Lit Exam Rooms	0.16	2.83	1.12
HTQ-F Scale
Item	Question	Percent Severe (Total)	Discrimination Parameter (a)	Difficulty Parameter (b)
HTQ117	Feeling a Lack of Control	0.52	1.06	−0.08
HTQ14	Pap smear	0.44	3.13	0.16
HTQ109	Conversation about Intimate Details	0.39	1.74	0.38
HTQ50	Hearing Others Cry Out with Anxiety or Distress	0.34	1.15	0.72
HTQ7	Mammogram	0.29	2.21	−0.73
HTQ42	Physical Touch	0.22	2.15	1.00
HTQ114	Being in Close Physical Proximity to Authority Figures (e.g., Medical Provider)	0.19	2.56	1.04
HTQ71	Provider Moving Your Head	0.13	2.06	1.44
HTQ94	Dark/Low Lit Exam Rooms	0.10	2.09	1.66
HTQ-U Scale
Item	Question	Percent Severe (Total)	Discrimination Parameter (a)	Difficulty Parameter (b)
HTQ83	Provider Touches Your Genitals	0.51	2.33	−0.02
HTQ109	Conversation about Intimate Details	0.37	2.12	0.46
HTQ108	Disrobing/Removing Clothing	0.30	2.36	0.66
HTQ42	Physical Touch	0.27	2.20	0.77
HTQ56	Being Told “Please Relax”	0.25	2.39	0.81
HTQ86	Lying on an Exam Table	0.20	4.32	0.89
HTQ76	Technician Adjusting Positioning While Lying on a Table	0.19	3.11	1.00
HTQ88	Being in a Treatment Room Alone	0.16	3.00	1.14
HTQ75	Quick Movements by Providers	0.15	3.17	1.17
HTQ94	Dark/Low Lit Exam Rooms	0.13	3.15	1.25

The difficulty parameters indicate the location of the item on the underlying latent healthcare triggering construct, such that items with lower difficulties are more likely to be endorsed by participants who have only moderate responses to healthcare triggers. As can be seen in Table 2, the difficulty parameter is, of course, strongly related to the percent of participants who endorsed the item as severe. The discrimination parameter is the slope of the logistic curve at the point over the difficulty parameter. It indicates the extent to which the item discriminates clearly between individuals above and below that difficulty parameter. Thus, it can be viewed as the degree to which the item provides clear information about where individuals are located on the underlying latent healthcare triggering scale. As indicated in Table 2, all of the items have discrimination parameters that are greater than one and most are greater than two. Not surprisingly, the items with lower discrimination parameters tend to be at the lower end of the difficulty parameters, as these items are endorsed by larger numbers of individuals. To illustrate the IRT modeling of these items, Supplementary Figure 2 shows one annotated item response curve from the HTQ-U scale. Supplementary Figure 3 shows the total item information curves for each of the three scales. These curves indicate that the scales are most informative for distinguishing between individuals who are above the midpoint of the latent healthcare triggering construct.

We evaluated the unisex screening scale for differential item functioning (DIF) between males and females. We found no evidence of DIF for any of the ten unisex items. This suggests that the unisex scale can be legitimately used with both males and females, and that differences in this scale and its correlates between males and females can be interpreted.

Response Distributions of the Three Screening Scales

Supplementary Figure 4 shows the stem-and-leaf plots of the endorsement percentages on the three 10-item screening scales: HTQ-M, HTQ-F, and HTQ-U. As shown in Supplementary Figure 4, the screening scales are all positively skewed. The median endorsement percentages for the male, female, and unisex scales are 30%, 27%, and 11%, respectively. The lower median for the unisex scale can be understood because this scale contains no gender-specific items, which, as shown in Supplementary Table 3, tend to elicit the highest endorsement percentages.

Correlations between the Screening Scales and Total Score

To assess the degree to which the screening scales are representative of the total endorsement percentage, we correlated the endorsement percentages from the screening scales with the endorsement percentage of the total set of 117 items. Because the total set of items contained the 10-item screening scales, which creates a positive bias in the correlations, we computed these correlations after removing the ten screening items from the total. For men, the correlation was 0.87, for women it was 0.88, and for the unisex scale it was 0.90 (all of these correlations are statistically significant, p < .001).

Correlations between the Scales and Criterion Validity Variables

Table 3 shows the correlations between the total endorsement percentage and: a) the three HTQ screening scales; b) scores on the measure of medical anxiety; c) a dichotomous variable assessing avoidance of doctors; and, d) among the participants who avoid doctors, a dichotomous variable assessing if they avoid doctors because of anxiety. All correlations have p-values < 0.001 except for the avoidance of doctors item. Specifically, among the HTQ 117 item version, HTQ-M, HTQ-F, and HTQ-U, all were associated with medical anxiety. However, only the HTQ-F scale was significantly correlated with the avoidance of doctors variable (p=0.032), such that among women increased triggering was associated with increased avoidance. Although avoidance of doctors was generally not strongly related to any of the scales, when we restricted the analysis to people who avoided doctors because of anxiety we found significant correlations between those who avoided doctors due to anxiety and triggering (all ps were < 0.001) (see Table 3).

Table 3.

Correlations between endorsement percentages on the HTQ scales and the criterion validity variables.

	Percent Severe (117 item version)	Percent Severe (HTQ-M)	Percent Severe (HTQ-F)	Percent Severe (HTQ-U)
Avoidance of doctors	−0.053	−0.073	−0.144	−0.082
Avoidance of doctors due to anxiety	0.269	0.289	0.290	0.287
Medical anxiety	0.495	0.483	0.508	0.468

Known Groups Validity

Table 4 shows the mean endorsement percentage of the SA and non-SA samples on the three screening scales. As can be seen in Table 4, the SA sample has a significantly higher endorsement percentage than the non-SA sample.

Table 4.

Comparisons between SA survivors and a non-SA comparison sample on the average percent endorsement on the three HTQ scales.

Outcome	Group
	SA			Non-SA
	M(%)	SD	n	M(%)	SD	n	95% Confidence Interval	t	df	p
HTQ-M	33.6	30.1	232	17.0	25.6	114	0.10–0.23	5.08	344	<0.001
HTQ-F	31.6	27.3	222	24.8	30.4	106	0.00–0.13	2.00	326	0.046
HTQ-U	25.7	29.4	452	17.0	26.8	219	0.04–0.13	3.66	669	<0.001

Discussion

The goals of the study were to develop a self-report measure of healthcare triggering for adult SA survivors and to evaluate initial validity of the scales.

Based on an original pool of 117 trigger items, we created three 10-item versions of the Healthcare Triggering Questionnaire - the HTQ-M, HTQ-F, and HTQ-U (see Supplementary Figure 5 for the final scales). Male and female SA survivors responded similarly to the unisex scale, suggesting that using one form for all participants is acceptable. However, the HTQ-M and HTQ-F contain gender-specific items (e.g., prostate exam, pap smear) that may be of particular interest to clinicians and researchers in those specialty areas.

The HTQ scales correlated highly with participants’ responses to all 117 triggers, suggesting that these 10-item scales well represent a person’s tendency to be susceptible to healthcare triggers. Results also support the criterion validity of the scale, demonstrating that higher levels of healthcare triggering were associated with higher levels of medical anxiety regardless of which form of the scale was tested. Furthermore, among participants who avoided visiting the doctor, those with higher levels of triggering were more likely to report anxiety as the reason for avoidance, consistent with past research (McGregor, Julich et al., 2010). In sum, individuals who are more likely to have memories, thoughts, feelings or sensations related to their SA triggered by healthcare are also more likely to be anxious in medical settings, and more likely to avoid doctors due to anxiety. These results are consistent with the available literature suggesting that triggers and anxiety may compromise medical attendance and adherence (Dallam, 2010; Schachter et al., 2009). Lastly, HTQ scores were significantly higher among SA survivors as compared to non-SA participants, indicating that the scale is measuring SA-related triggering as opposed to general procedure-related distress.

The HTQ scales provide researchers and clinicians with a helpful assessment tool. The HTQ scales could be used in at least five ways. First, the HTQ could be used to measure the level of healthcare triggering experienced by SA survivors (e.g., across healthcare settings in the service of determining their support needs). Second, the HTQ could be used in research seeking to identify predictors of healthcare triggering (e.g., to learn more about what aspects of the healthcare environment, provider behavior, or survivors’ cognitions make one more or less likely to be triggered by healthcare). Third, the HTQ could be used to examine the effect of triggering on other outcomes (e.g., adherence to healthcare, avoidance of healthcare, posttraumatic stress symptoms, dissociation) and to model the relationships among healthcare triggers, anxiety, and non-adherence. For example, to see if healthcare triggering might start a chain reaction such that one feels triggered, one then experiences anxiety and other posttraumatic stress symptoms, and to avoid this discomfort, one avoids or delays healthcare. Fourth, the HTQ could be used to determine the efficacy of interventions designed to reduce triggering and associated negative sequelae. For example, one could test empirically whether strategies proposed to reduce triggering (e.g., empowering patients, seeking consent, asking about treatment preferences) are efficacious. Fifth, the HTQ could be used clinically, by a wide variety of healthcare providers (e.g., psychologists, physicians, NPs, PAs, social workers, physical therapists), to identify SA survivors who are likely to experience triggering in medical environments, to learn what their triggers are, and to proactively develop a plan to avoid, minimize, or manage them. For example, if clinical staff knew in advance that a patient was likely to be triggered, and then asked for her triggers, they might: 1) learn that she tended to be triggered by a male provider, the smell of cigarettes, or someone approaching them from behind; 2) consequently work to ensure that a female, non-smoking clinical staff member was assigned to the patient; and, 3) encourage clinical staff to approach the patient from the front where possible.

As with any study, the present one has limitations. First, the study sample is largely white and non-Latino/a. Future research should validate the HTQ in more diverse samples. Second, all participants were MTurk members with at least a 90% approval rating. It is possible that these MTurk members may differ in some way from the general population. Additionally, the MTurk workers who self-selected into the study were willing to reflect on the topic of sexual abuse. It is possible that individuals who were particularly uncomfortable thinking about sexual abuse would have avoided enrolling in the study. Consequently, results here may underestimate the true extent of triggering among SA survivors. Third, the study used a dichotomous measure of SA history, not SA severity or SA characteristics (e.g., contact vs. noncontact). We took this approach because, to our knowledge, there is no literature specifically indicating that certain types of SA are more or less predictive of healthcare triggering. Given the lack of literature supporting strict inclusion/exclusion criteria, we thought it was best to include anyone who self-identifies as an SA survivor rather than to impose a particular definition of SA. We believe that this approach increases the generalizability of the study findings. However, in future research, we will seek to explore differential item functioning based on abuse characteristics and severity. Fourth, the present study was focused on SA survivors. However, healthcare may also be triggering for other groups of trauma survivors. Future research should explore the applicability of the HTQ to other trauma populations. Fifth, the avoidance measure used here was a straightforward one-item tool focused only on doctor visits. The relations between the HTQ and healthcare avoidance need to be explored further using more comprehensive measures of healthcare avoidance. Sixth, the present study included only individuals who identified as male or female, and yielded male and female-specific scales. Future research should explore the best way to use the HTQ system with transgender and gender variant individuals, including which version(s) of the HTQ should be administered. The importance of such work is highlighted by the fact that transgender individuals experience high levels of sexual abuse (Office of Justice Programs, 2014). Seventh, in the analyses of the full 117 items, we found some support for gender differences. Given the limited number of instruments administered in this study, we cannot speculate on the underlying reasons for these gender differences. Future research should work to further elucidate the similarities and differences in healthcare triggering between men and women. Eighth, we relied on participants’ self-reports of their SA status. Although it is possible that participants misrepresented their SA history, we do not believe this to be a widespread concern, given that: a) reimbursement was in no way dependent on self-reported SA history; and, b) although inaccuracies are bound to arise with any method of investigating abuse histories, self-report is not necessarily likely to introduce greater challenges than other methods of ascertaining exposure to childhood abuse (Kendall-Tackett & Becker-Blease, 2004). However, as with any study relying on self-report (online or not), we cannot confirm the presence or absence of SA history.

Overall, previous research has indicated that survivors of SA want healthcare professionals to understand how aspects of the healthcare environment can trigger memories, thoughts, feelings or sensations related to SA, and that such triggering can make medical procedures and examinations difficult to endure (McGregor, Glover, Gautam, & Julich, 2010). It is hoped that the development of the HTQ will serve as a first step towards encouraging further research on, and clinical attention to, the phenomenon of healthcare triggering.