Development and validation of the Global Post Trauma Symptom Scale-Uganda among trauma-affected adults

ABSTRACT

Background: Post-traumatic stress measurement in non-Western contexts often utilizes Western assessment tools.

Objective: The purpose of this study was to adapt and validate the Global Post-Trauma Symptom-Item Bank (GPTS-IB) among adult caregivers of youth living with HIV in Uganda.

Methods: We utilized qualitative and quantitative methods to establish validity and to develop a reduced version of the GPTS-IB.

Results: Cognitive interview results (n = 31) indicated that participants understood items of the GPTS-IB, but revisions in wording were needed for some items. A reduced version of the GPTS-IB, the Global Post Trauma Symptom Scale-Uganda (GPTSS-U) was created through item response theory analyses based on discrimination criteria and difficulty parameters. A relationship between the GPTSS-U and the PHQ-9 (r = 0.547; p < .000) scores and the PCL-5 (r = 0.9, p < .001) scores indicated acceptable discriminant and convergent validity, respectively. Test information function curve findings indicated that the GPTSS-U was most precise at moderate to severe levels of post-traumatic symptoms. Using nested linear regression models, we found incremental or predictive validity as the total scores on the GPTSS-U significantly predicted functional impairment (p < .05) beyond the ability of the Post Traumatic Stress Checklist-5.

Conclusions: Results indicated that the GPTSS-U was a valid measure of post-traumatic symptoms among caregivers of youth with HIV in Uganda. Future research should continue to assess the utility of the GPTSS-U in other non-Western contexts.

HIGHLIGHTS

• An item bank of post-traumatic stress symptoms was used to develop and ensure validity and cultural relevance among Ugandan caregivers of children living with HIV.

• Both quantitative and qualitative methods were utilized to develop the assessment tool, the Global Post-Traumatic Symptom Scale-Uganda. The scale was deemed valid and reliable and a significantly better predictor of functioning among participants than a standard post traumatic stress scale.

• Results demonstrate validity of the GPTSS-U and utility of the item-bank in contexts without validated trauma measures

1. Introduction

Exposure to traumatic events can compound daily stressors. In low- and middle-income countries (LMICs), this is concerning as the risk of trauma and mental health disorders is higher due to factors like poverty, conflict, and displacement (Fodor et al., 2014; Seedat et al., 2018). With the LMIC population projected to double by 2053 (United Nations, 2017) and increasing awareness around global mental health (Betancourt & Chambers, 2016; Murray & Jordans, 2016), understanding trauma becomes critical.

Most global trauma assessments rely heavily on Western tools, particularly those focused on posttraumatic stress Post-Traumatic Stress Disorder (PTSD) (Vermetten & Olff, 2013). Fodor et al. (2014) found that 87% of studies on traumatic stress were in high-income countries (HIC), with 51% from the U.S. Although PTSD is a recognized construct in Western trauma-affected populations, its relevance in LMICs is debated (Fodor et al., 2015), leading to conflicting views on PTSD’s global applicability (Mughal et al., 2020; Patel et al., 2022). Additionally, validating measurement scales for PTSD is complicated in settings where instability may create an increase in chronic environmental stressors (Mughal et al., 2020).

The Global Post-Trauma Symptom-Item Bank (GPTS-IB) was developed to bridge this gap (Haroz et al., 2014). This 51-item bank, based on qualitative studies, measures post-traumatic stress (PTS) and other transdiagnostic trauma-related stress and symptoms (Michalopoulos et al., 2020) which might occur in the context of ongoing trauma. The GPTS-IB aims to be adaptable and locally validated for non-Western trauma settings, providing normative scores on individual trauma symptomatology.

Caregivers of children with high-need illnesses like HIV often experience additional trauma (Bolton et al., 2014). The Diagnostic and Statistical Manual of Mental Disorders-fifth edition (DSM-5) includes illness in one’s child as a potentially traumatic event (DSM-5). Further, research has indicated that carers of children with HIV in South Africa have an increased risk of post-trauma symptoms (Bolton & Tang, 2002).

We used the GPTS-IB to assess PTS symptoms among Ugandan caregivers of children living with HIV. GPTS-IB scores were utilized to make inferences about the severity of PTS symptoms experienced both indirectly and directly among Ugandan caregivers of children living with HIV. The aim was to reduce the GPTS-IB to a relevant scale among Ugandan adults (i.e. the Global Post Trauma Symptom Scale-Uganda; GPTSS-U). Specifically, we developed and contextually validated the GPTSS-U using both qualitative (i.e. through cognitive interviews) and quantitative (i.e. through classical test theory and item response theory; IRT) methods. This is the first study utilizing the GPTS-IB to adapt and develop a reliable and valid post-trauma assessment tool in Uganda.

This study is significant for this region as services that address mental health and psychosocial functioning are increasingly delivered through task-shifting, where lay workers are trained to treat specific health disorders (Kroenke et al., 2001). As task-shifting becomes more widely utilized to address the ‘treatment gap’ in global mental health, research into its efficacy both for recipients of lay worker interventions and for their caretakers is needed (Haroz et al., 2014); this will build upon the growing body of research to show the efficacy of treatment delivery by non-specialists (Bolton et al., 2014). This is the first study of its kind to validate a PTS symptom instrument among trauma-affected caregivers of children with HIV in Uganda. Using the GPTSS-U in the caregiver context may demonstrate how mental health symptomatology varies across cultural contexts, informing tailored responses to intervention development.

2. Methods

2.1. Background and setting of study

This study was part of two larger projects in Uganda's Rakai and Masaka regions: Suubi + Adherence, a longitudinal randomized trial aimed at improving HIV treatment adherence among adolescents [NIH: 1R01HD074949-01], and Suubi Maka, a cluster trial focused on educational outcomes for orphaned children through economic empowerment [NIH: RMH081763A]. Participants were primary caregivers of youth in either study. Caregivers of children living with HIV or orphaned due to HIV were chosen as a vulnerable population at risk for multiple traumatic experiences and ongoing daily problems.

2.2. Training

Ten local staff were trained to administer psychosocial measures to caregivers in each study. They received extensive training in cognitive interviewing and quantitative assessment of psychosocial problems, with ongoing feedback from the study's principal investigator (LM) through role plays and group discussions.

2.3. Sample

Caregivers of intervention participants aged 18 or older were eligible. Exclusion criteria included active psychosis or significant developmental delay.

2.4. Consent and ethical approval

IRBs were approved at both a US and Uganda-based university (IRB AAAD2525 and AAAK3852). All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Written informed consent was obtained, and participants received UGX 10,000 (USD 3.30) for interviews and UGX 15,000 (USD 5) for the validation study.

2.5. Measures

2.5.1. Demographics

Basic demographics of age, gender, and HIV status were assessed by self-report.

2.5.2. Global Post Trauma Symptom-Item Bank (GPTS-IB)

The GPTS-IB (Haroz et al., 2014), a 51-item self-report tool, assessed post-trauma symptoms in adults. Both the GPTS-IB and the reduced Global Post Trauma Symptom Scale-Uganda (GPTSS-U) used Likert scales to measure symptom frequency over the past two weeks, with response options ranging from 0 (none of the time) to 3 (almost all of the time).

2.5.3. Functioning scale

A local measure of functional impairment, previously validated among Ugandan adults in Rakai and Masaka (Bolton & Tang, 2002), was used in the current study. It assesses tasks men (9 items) and women (9 items) typically perform for themselves, their families, and communities (e.g. farming, childcare). Participants rated the difficulty of each task over the past four weeks on a scale from 0 (no difficulty) to 4 (often cannot do).

2.5.4. Patient health questionnaire-9 (PHQ-9)

The PHQ-9 (Kroenke et al., 2001) is a nine-item self-report measure of depression symptoms over the past two weeks, with responses from 0 (not at all) to 3 (nearly every day). Widely used in non-Western contexts, including sub-Saharan Africa (Carroll et al., 2020; Makhubela & Khumalo, 2022), the PHQ-9 was used to assess discriminant validity of the GPTSS-U.

2.5.5. Post-traumatic stress-disorder checklist for DSM-5 (PCL-5)

The Post Traumatic Stress Disorder Checklist-5 (Weathers et al., 1993) is a 20-item questionnaire based on DSM-5 criteria for PTSD. Participants reported how much each symptom bothered them in the past two weeks, using response categories from 0 (none of the time) to 3 (almost all of the time) rather than the standard five-category PCL-5 scale. This adjustment maintained consistency across tools. The PCL-5 has been used among trauma-affected populations in LMICs (Bryant et al., 2017), but it has not yet been validated in Uganda.

2.5.6. Life events checklist for the DSM-5 (LEC-5)

The LEC-5 (Seedat et al., 2018; Weathers et al., 1993) is a 16-item questionnaire used to screen for lifetime exposure to potentially traumatic events linked to increased PTSD or psychosocial distress risk. Widely used in non-Western LMICs (Gray et al., 2015), the LEC-5 asked participants to report whether they had experienced or witnessed each event. Total scores were calculated for experienced events (0–16), witnessed events (0–16), and the combined total (0–32).

2.6. Procedure

2.6.1. Translation

All measures were translated into Luganda using translation-back translation, followed by group translation to ensure term equivalence, conceptual understanding, and clarity for adults with limited formal education. Two certified translators reviewed the initial translation. The research team reviewed and back-translated the instruments item by item.

2.6.2. Data analysis

To assess face and content validity of the GPTS-IB, cognitive interviews were conducted with 30 adult caregivers. Using the think-aloud method (McHenry et al., 2017), participants described each questionnaire item in their own words with examples. For example, participants were asked to describe in their own words and give examples for ‘Feeling afraid, including being afraid that a traumatic event would happen again.’ The 51 items of the GPTS-IB were read in Luganda, with responses recorded verbatim.

Responses to the think-aloud methods were assessed in terms of consistency with the intended meaning for each item. First author (LM) and a graduate assistant coded each item. When the intended meaning was inconsistent, items were revised or changed for the validation study after a full review from the research team.

2.6.3. GPTSS-U development

For the validation study, the administration included reading each item of the measures to the participants. All responses were entered into Qualtrics (2016) by local research staff in Luganda. Research staff were fluent in both Luganda and English.

Data analyses were conducted using STATA 14 (Stata Corp, 2013). The GPTS-IB was reduced to a shortened and contextually specific version (Global Post Traumatic Stress Scale-Uganda; GPTSS-U) based on item response theory analysis results. It should be noted that there was some overlap with the GPTSS-U and the PCL-5 and PHQ-9. However, any repeated question was only asked once. The PCL-5 had four items that overlapped, and the PHQ-9 had one overlapping item. We conducted an exploratory factor analysis (EFA) to meet the assumption of unidimensionality (the factor analysis resulting in a 1-factor solution or a dominant first factor, with at least 20% of the variance explained; Embretson & Reise, 2000). The assumption of local independence was tested through discrimination parameters with a slope < 4.00 for each item in the IRT model (Nguyen et al., 2014).

Based on the EFA results, a unidimensional graded response IRT model (GRM) was conducted to assess the degree to which the measure reflects the underlying latent variable (Samejima, 1997). We used the GRM as it allows for ordered response categories. IRT was used to assess the ability of each item to differentiate among individuals at different points along the latent continuum of post-traumatic stress (Wyse & De Ayla, 2010).

For each item, difficulty (or item location, b), discrimination (a), and measurement information (θ) parameters were estimated. Item difficulty parameters (b) indicate the amount of post-trauma symptoms where the probability of endorsing an item with a particular response category is 0.50. Three difficulty parameters (b₁, b₂, b₃) were estimated, corresponding with the four possible response categories of the GPTS-IB. The first difficulty parameter (b₁) indicates the level of the underlying latent trait of post-trauma symptoms, where the probability of endorsing an item with a ‘0’ instead of ‘1’ or ‘2’ is 0.50. The second difficulty parameter (b₂) is for the response of < 1. The third difficulty parameter (b₃) is for the response of < 2. Discrimination parameters indicate the ability of a particular item to distinguish between individuals with low and high amounts of the latent construct of post-trauma symptoms. Values of 0.01–0.34 are considered very low, 0.35–0.64 low, 0.65–1.34 moderate, 1.35–1.69 high, and 1.70 and above very high (Baker, 2001). Measurement information (or precision) was determined by the Test Information Curve (TIC). Measurement information represents the certainty with which the measure assesses the underlying latent trait (θ) and can vary as a function of the level of θ.

Items for the GPTSS-U were selected based on difficulty and discrimination parameters, as well as measurement information. Using item location estimates, items that spanned a range of post-trauma symptom severity were selected. Items were also selected based on high discrimination values or with estimates of 1.35 and above. After selecting items for the GPTSS-U, the TIC of the GPTS-IB to the GPTSS-U were compared. We examined whether the selected items of the GPTSS-U had a similar TIC compared to the GPTS-IB across different levels of severity. Our final step included conducting a GRM IRT model to estimate item difficulty parameters and discrimination parameters of the reduced GPTSS-U scale.

2.7. Data analysis plan – validation of the GPTSS-U

2.7.1. Internal consistency reliability

Cronbach’s alpha scores were utilized to determine the internal consistency and reliability of the GPTSS-U. A value between 0.70 and 0.79 is considered fair, 0.80 and 0.89 is considered good, and 0.90 and above is deemed excellent (Nunnally & Bernstein, 1994).

2.7.2. Construct validity

Construct validity is the degree to which a scale measures the theoretical construct it was intended to measure and is associated with other related constructs. Construct validity was determined through Spearman’s correlation coefficients (ρ) to examine the strength of the relationship between the GPTSS-U and PHQ-9, as well as the GPTSS-U and the PCL-5. The effect size was based on a coefficient between 0.10 and 0.29, representing a small association, a coefficient between 0.30 and 0.49, representing a medium association, and a coefficient of 0.50 and above, representing a significant association (Cohen et al., 2003). We hypothesized that the GPTSS-U would be positively associated with the PCL-5. We also hypothesized that the GPTSS-U would be positively correlated with the PHQ but still distinct demonstrating discriminant validity (i.e. with a correlation less than .80; Sackett & Lievens, 2008).

2.7.3. Criterion validity

Criterion validity is defined as the relationship of a scale to a criterion variable, such as a clinical diagnosis (Allen & Yen, 2001). We used components of the Bolton method (Bolton & Tang, 2002) which establishes criterion validity in LMICs of the GPTSS-U where there are no psychiatrists available to provide psychiatric diagnostic assessments. Specifically, we implemented similar methods to determine case vs non-case, but did not conduct extensive emic research with local informants at the beginning of the study. We developed cases based on a PCL-5 cutoff adjusted score of 20 or above. Non-cases were participants who had a score below 20 on the PCL-5. We conducted a two-sample t-test to examine the hypothesis that cases would have a significantly higher score on the GPTSS-U than non-cases. Experiencing a traumatic event was not a necessary precondition for being included as a member of the ‘cases’ versus ‘non-cases’ group because all participants were caregivers of a child with a serious medical condition, which has been deemed a potentially traumatic event in the DSM-5.

2.7.4. Sensitivity and specificity analyses

We conducted sensitivity and specificity analyses using receiver operating characteristic (ROC) curves to test the performance of the GPTSS-U and suggest appropriate cut-off scores for case identification. Similar to criterion validity, we used the cutoff of the PCL-5 as cases and non-cases. ROC curves plot the true positive rate (sensitivity) against the false positive rate (specificity). An area under the curve (AUC) of 0.50 indicates that the GPTSS-U shows no diagnostic utility, whereas 1.0 indicates perfect prediction. AUC values of 0.50–0.70 indicate low accuracy, 0.70–0.90 moderate accuracy, and above 0.90 high accuracy (Fischer et al., 2003). An optimal cut-off point was generated for the GPTSS-U based on maximizing sensitivity and specificity (Liu, 2012).

2.7.5. Incremental validity

Finally, we assessed incremental or predictive validity of the GPTSS-U on functional impairment using nested linear regression models. In Model 1, we examined the association of gender, age, HIV status, total LEC-5, traumatic events experienced, total PHQ-9 score, and total PCL-5 score. In Model 2, we added the GPTSS-U total score to the model. Incremental validity is if total scores on the GPTSS-U significantly predicted functional impairment (p < .05) beyond the PCL-5. This was measured by a statistically significant increase in the R² statistic when comparing Model 1 to Model 2 (Sackett & Lievens, 2008). To assess collinearity between scores on the PCL-5 and the GPTSS-U, we also examined the variance inflation factor (VIF) and tolerance values. A VIF of 5 or above indicates high collinearity, and a tolerance value of 0.2 indicates independence (Boateng et al., 2018; Craney & Surles, 2002).

3. Results

3.1. Cognitive interviews

Thirty participants (n = 30) completed the cognitive interviews (83.33% female). Average age was 49.07 years (SD = 11.18, range 33–83 years). Verbal responses were back-translated into English, compiled in Excel, and coded by main themes. Responses from the think-aloud methods were coded and reviewed by two research team members in terms of consistency with the intended meaning for each item. Most items required slight revision in wording or additional description in meaning. For example, many participants described ‘feeling detached from others’ as ‘feelings of sorrow.’ We revised the item by adding more details in the description, i.e. ‘feeling detached or cut off from others’ (i.e. alienated, isolated, lonely, withdrawn, or unable to socialize with others).

3.2. Demographics

Two hundred caregivers completed measures for the validation study (80.50% female). The average age of participants was 47.70 years (SD = 12.52, range 19–83).

3.2.1. Caregiver trauma

Caregiver scores on the revised PCL-5 ranged from 0 to 47 (out of a possible 0–60). The average score was 13.53, with a standard deviation of 8.5. Twenty-three per cent of caregivers scored above the adjusted cutoff (20). Caregiver scores on the LEC-5 ranged from 0 to 10 (out of a possible 0–10), and caregivers experienced an average of 3.75 traumatic events (SD = 2).

3.3. Creating the GPTSS-U from the GPTS-IB

3.3.1. Unidimensionality assumption

A simple factor solution was obtained using Varimax rotation. Kaiser Meyer Olkin (KMO) measure of sampling adequacy was superb with .84 (Hutcheson & Sofroniou, 1999) and a significant Bartlett’s test of sphericity (χ² ₍₁₂₇₅₎ = 4264.56, p < .001), suggesting a factor analysis was appropriate to conduct with the data. Seven factors were suggested from initial eigenvalues over 1. However, the first factor indicated over 44% of the variance, so unidimensionality was determined.

3.3.2. Local independence

The graded response IRT model indicated local independence of the GPTS-IB as all discrimination parameters were < 4.00.

3.3.3. Item selection

The graded response IRT model indicated items with discrimination parameters ranging from low to high. For example, items such as ‘headaches or migraines’ (a = 0.48), ‘problems related to a long-term condition or a physical health problem’ (a = 0.66), and ‘unable to eat or lack of appetite’ (a = 0.69) were indicated as poor in the ability to distinguish between participants with low and high amounts of post-trauma symptoms. Conversely, items such as ‘sadness, sorrow or depressed mood’ (a = 1.86), ‘emotional pain, suffering, despair or distress’ (a = 1.97), and ‘feeling confused or unable to figure out what was going on’ (a = 2.10) were indicated as high in their ability to distinguish between participants with high and low amounts of post-trauma symptoms. Discrimination parameters for the GPTS-IB are presented in Table 1. Based on these estimates and item locations across low, moderate, and severe post-trauma symptoms, the GPTS-IB was reduced to 11 items for the GPTSS-U.

Table 1.

GPTS-IB discrimination parameters.

GPTS-IB Item	Discrimination Parameter (ɑ) (Standard Error)
1. Felt afraid, including being afraid that a traumatic event or stressful event would happen again.	0.89 (0.16)
2. Felt detached or cut off from others (i.e. alienated, isolated, lonely, withdrawn, or unable to socialize with others).	0.93 (0.18)
3. Had disturbed sleep or trouble sleeping (difficulty falling or staying asleep).	0.85 (0.16)
4. Been unable to experience positive emotions (i.e. happiness, joy, excitement).	1.34 (0.20)
5. Felt apathetic or that life has become meaningless.	1.19 (0.24)
6. Had persistent negative expectations or felt hopeless.	0.82 (0.18)
7. Experienced repeated and disturbing memories about a traumatic or stressful event.	1.64 (0.21)
8. Felt agitated, irritated, frustrated, or angry.	1.03 (0.18)
9. Cried frequently.	1.30 (0.24)
10. Felt sadness, sorrow, or a depressed mood.	1.9 (0.25)
11. Experienced disturbing dreams or nightmares related to a stressful or traumatic event.	1.08 (0.19)
12. Experienced excessive worry, anxiety, or an ongoing sense of panic.	1.35 (0.21)
13. Had frequent arguments or fights with an intimate partner, family, or community member.	0.90 (0.22)
14. Had headaches/migraines.	0.48 (0.15)
15. Had trouble concentrating.	1.13 (0.20)
16. Felt forgetful or experienced memory loss.	0.87 (0.17)
17. Felt you were thinking too much.	1.55 (0.21)
18. Felt that no one understands what you have been through or are currently experiencing.	0.89 (0.17)
19. Had a lack of or reduced interest in activities that you used to enjoy.	0.89 (0.16)
20. Felt tired, fatigued, weak, or exhausted.	1.04 (0.17)
21. Experienced irritable behaviour or angry outbursts.	1.11 (0.19)
22. Engaged in reckless, self-destructive, risky, or harmful behaviour (i.e. abusing alcohol/drugs, risky sex, cutting).	0.73 (0.33)
23. Become emotionally upset when reminded of a stressful or traumatic event.	1.48 (0.22)
24. Felt a sense of loss.	1.11 (0.19)
25. Felt an ongoing sense of uncertainty, helplessness, or a lack of control.	1.11 (0.18)
26. Had thoughts of wanting to kill yourself.	1.11 (0.18)
27. Been unable to eat or a lack of appetite.	0.76 (0.40)
28. Been unable to talk about the trauma or feeling a sense of silence.	0.69 (0.17)
29. Felt that a traumatic or stressful event is actually happening again; that you are reliving it in the present (flashbacks).	1.57 (0.23)
30. Felt guilty.	1.54 (0.23)
31. Felt ashamed or damaged.	1.28 (0.21)
32. Felt jumpy or easily startled such as jumping or flinching in response to an unexpected noise.	1.45 (0.24)
33. Actively avoided external reminders of a stressful or traumatic event.	1.06 (0.18)
34. Experienced physical heart problems: heart palpitations, cardiovascular problems, or symptoms related to a heart condition.	1.26 (0.21)
35. Felt as if your surroundings or environment are not real.	1.07 (0.18)
36. Felt emotional pain (suffering), despair, or distress.	1.14 (0.23)
37. Experienced dizziness, felt confused, or disoriented.	1.97 (0.26)
38. Had an upset stomach or digestive problems.	2.10 (0.28)
39. Actively avoided thoughts, feelings, or memories related to a stressful or traumatic event.	0.72 (0.16)
40. Experienced a loss of faith and trust in the world, others, or spiritual beliefs.	1.01 (0.18)
41. Felt watchful of your surroundings or on guard.	1.20 (0.22)
42. Experienced physical trembling or shaking.	1.08 (0.19)
43. Felt chest pain or pressure in the chest.	1.12 (0.21)
44. Felt as if you have gone mad/crazy (i.e. seeing things that are not there).	0.82 (0.17)
45. Experiencing problems related to the heart (i.e. heartache, a weak heart, spoiled heart, dead heart, sick heart, tired heart, wounded heart, broken heart, heart in trouble, unclear heart, disturbed heart, falling heart, hot heart).	1.11 (0.32)
46. Experienced problems related to a long term condition or physical health problem (i.e. immune system disorders, diabetes, arthritis).	1.16 (0.20)
47. Had vision problems, eye problems, or weak/tired eyes.	0.66 (0.17)
48. Blamed yourself or others for a stressful or traumatic event.	0.84 (0.16)
49. Had an inability to remember an aspect or parts of a stressful or traumatic event.	1.15 (0.19)
50. Had a physical reaction when you were reminded of a stressful or traumatic event (i.e. sweating, racing heartbeat, nausea).	1.40 (0.21)
51. Having persistent and ongoing negative feelings such as anger, fear, guilt, horror, and sadness.	1.32 (0.21)

3.4. Descriptive statistics

Table 2 presents results from the GRM IRT model for the 11 selected items of the GPTSS-U. Discrimination parameters remained in the acceptable range for all items (a = 1.34 to a = 2.24), with item locations ranging from b₁ = −0.88 to b₃ = 3.14 across the latent construct. Test Information Curves for the GPTSS-U revealed little change from the GPTS-IB, indicating the most information at average to moderate levels of post-trauma symptoms (θ = 0 to θ = 2). Internal consistency reliability was good for the GPTSS-U with α = 0.87 and no improvement upon removal of any item. The average total score of the GPTSS-U among participants was 8.09 (SD = 5.92), with scores ranging from 0 to 29 among participants (scores could range between 0 and 33).

Table 2.

Graded Response Model Item Discrimination Parameters (a) and Standard Errors, and Item Difficulty Parameters (b₁, b₂, b₃,) and standard errors, N = 200.

GPTSS-U Item	Discrimination Parameter (ɑ) (SE)	Difficulty Parameters (b) (SE)
Experienced repeated, disturbing or unwanted thoughts or memories about a traumatic or stressful event	1.82 (.25)	b1 = -.88 (.15) b2 = .66 (.13) b3 = 1.97 (.24)
Felt sadness, sorrow or a depressed mood	2.19 (.31)	b1 = -.58 (12) b2 = 1.14 (.15) b3 = 2.18 (.25)
Experienced excessive worry, anxiety or an ongoing sense of panic	1.49 (.23)	b1 = -.07 (.13) b2 = 1.41 (.21) b3 = 2.56 (.37)
Feeling you were thinking too much	1.62 (.24)	b1 = -.61 (.14) b2 = 1.18 (.18) b3 = 2.17 (.29)
Becoming emotionally upset when reminded of a stressful or traumatic event	1.51 (.24)	b1 = .02 (.13) b2 = 2.01 (.29) b3 = 2.97 (.44)
Been unable to talk about a traumatic or stressful event or feeling a sense of silence	1.49 (.24)	b1 = .06 (.13) b2 = 2.14 (.30) b3 = 2.52 (.36)
Feeling that a stressful or traumatic event is happening again; that you are reliving it in the present (flashbacks)	1.60 (.26)	b1 = .26 (.13) b2 = 1.66 (.23) b3 = 2.44 (.34)
Feeling ashamed or damaged	1.45 (.25)	b1 = .65 (.15) b2 = 2.05 (.31) b3 = 3.14 (52)
Feeling emotional pain (suffering), despair or distress	2.02 (.30)	b1 = -.29 (.12) b2 = 1.34 (.17) b3 = 2.12 (.26)
Feeling confused or unable to figure out what was going on	2.24 (.33)	b1 = -.05 (.11) b2 = 1.35 (.16) b3 = 2.16 (.26)
Had an inability to remember parts of a stressful or traumatic event	1.34 (.22)	b1 = .10 (.14) b2 = 2.13 (.32) b3 = 2.83 (.43)

3.4.1. Construct validity

Discriminant validity was indicated through the following: (1) GPTSS-U and the PHQ-9 scores with a moderate effect size (r = 0.54 p < .001); (2) GPTSS-U scores and trauma experienced (r = 0.17), trauma witnessed (r = 0.04), and trauma experienced or witnessed (r = 0.14) all significantly positively associated with a small effect size (p < .05); and (3). GPTSS-U scores and function scale scores were also with a moderate effect size (r = .304 p < .001). Convergent validity was indicated with a strong effect size between the GPTSS-U and the PCL-5 (r = 0.91 p < .001).

Table 3 displays the polychoric correlation matrix for (1) GPTSS-U, (2) PHQ-9, (3) trauma experienced, (4) trauma witnessed, (5) trauma total, and (6) functional impairment scale.

Table 3.

Polychoric correlation matrix.

Variable	GPTSS-U	PHQ-9	Trauma experienced	Trauma witnessed	Trauma Total	PCL-5	Functioning Scale
GPTSS-U	1.00
PHQ-9	.54***	1.00
Trauma Experienced	.19**	.19**	1.00
Trauma Witnessed	.03*	−0.04	−0.04	1.00
Trauma Total	.15*	.09	.62***	.75***	1.00
PCL-5	.90***	.58***	.15*	−0.04	0.06	1.00
Functioning Scale	.31***	.20**	.61***	.70***	.95***	.26***	1.00

Note: GPTSS-U = Global Post Trauma Symptom Scale – Uganda; PHQ-P = Patient Health Questionnaire-9; PCL-5 = Post-Traumatic Stress-Disorder Checklist for DSM-5.

*p < .05; **p < .01; ***p < .001.

3.4.2. Criterion validity

Results from the two-sample t-test with unequal variances established criterion validity of the GPTSS-U scale PTS cases (n = 45) having a significantly higher mean (15.96, SD = 4.74) than non-cases (n = 155) (5.8, SD = 3.95, p < .001). In the sample (n = 200), non-cases included 154/200 (77%) participants were considered non-cases, i.e. their scores on the revised PCL-5 did not exceed the adjusted cutoff score of 20. Among the non-cases, participants experienced between 0 (n = 8) and 10 (n = 1) traumatic events, and 90% of non-cases experienced between one and six traumatic events. The average number of traumatic events experienced by non-cases was 3.56. Among the cases (n = 45, 23%), participants experienced between 0 (n = 2) and 10 (n = 1) traumatic events, and 90% of participants experienced between two and eight traumatic events. The average number of traumatic events experienced by those in the case groups was 4.33.

3.4.3. Predictive validity

The results of the incremental validity analysis are presented in Table 4. Nested regression was used to compare Models 1 and 2. Both models analyzed the effects of (1) gender, (2) age, (3) HIV status, (4) traumatic experiences, (5) symptoms of depression, and (6) symptoms of PTSD on functional impairment. Model 2 included mean GPTSS-U scores in the analysis. Standardized beta coefficients were reported for all models. Forty-three per cent of the total variance in impaired functioning was explained by Model 2, which represented a 0.13% change in R² (explained variance) with the addition of the average GPTSS-U scores. The results of Model 2 indicated that ceteris paribus, a one-unit increase in average GPTSS-U was associated with a 0.266 increase in impaired functioning. The VIF for GPTSS-U was 1.72, and the tolerance was 0.58, showing no multicollinearity. When GPTSS-U scores were included in Model 2, the association between PCL-5 and impaired functioning became non-significant. An F test of the change in R² when GPTSS-U was added to Model 2 was statistically significant (p < .05), indicating further support for predictive validity.

Table 4.

Effects of measured variables on impaired functioning presented as beta coefficients.

Model	β (S.E.)	t
Model 1
Gender	0.047 (0.89)	0.85
Age	0.114 (0.03)	1.95
HIV	0.04 (0.73)	0.7
Trauma	0.59 (0.18)	10.46**
PHQ-9	0.01 (0.11)	0.13
PCL-5	0.148 (0.05)	2.15*
Model 2
Gender	0.034 (0.89)	0.85
Age	0.116 (0.03)	2.01*
HIV	0.042 (0.72)	0.73
Trauma	0.576 (0.18)	10.24**
PHQ-9	0.002 (0.11)	0.03
PCL-5	−0.086 (0.1)	−0.66
GPTSS-U	0.266 (0.14)	2.09*

Note: GPTSS-U = Global Post Trauma Symptom Scale – Uganda; PHQ-P = Patient Health Questionnaire-9; PCL-5 = Post-Traumatic Stress-Disorder Checklist for DSM-5 (PCL-5).

*p < .05; **p < .001.

3.4.4. Clinical utility

The GPTSS-U had an AUC of 0.95 (95% CI 0.92–0.98) when post-traumatic stress symptoms were compared to no symptoms. The cut-off score determined to be ideal for the GPTSS-U was 11, with a sensitivity of 88.9% and specificity of 87.7% for post-traumatic stress symptoms compared with no symptoms.

4. Discussion

This study aimed to adapt and test the psychometric properties of the GPTSS and GPTSS-U for caregivers of youth with HIV in Uganda, using both qualitative and quantitative methods. Results showed the effectiveness of mixed methods in adapting and validating the scale for the Ugandan context. High discrimination parameters ( Table 2 ) indicated the scale's ability to distinguish between high and low levels of post-traumatic stress, while the graded response model highlighted varying difficulty parameters across items, representing expressions of the latent construct at low, moderate, and high difficulty levels. The GPTSS-U demonstrated content, criterion, and construct validity.

The GPTSS-U also predicted functioning in Ugandan caregivers better than the PCL-5. IRT allowed for precision at different symptom levels, with the GPTSS-U most accurate for moderate and severe post-traumatic stress. Test information function curve confirmed that the GPTSS-U provided the most precise measurements for moderate to severe symptoms.

The average GPTSS-U score was 8.09 (SD = 5.92), indicating most participants reported mild to moderate symptoms. The correlation with PHQ-9 and traumatic event scores supported discriminant validity. Items with low discrimination related to physical health may reflect challenges in disentangling physical and mental health problems.

Item Response Theory (IRT) and Classical Test Theory (CTT) were used to examine trauma symptoms. Cognitive interviews helped establish semantic equivalence, which is crucial for understanding post-trauma symptoms in context. The findings support conducting content validity assessments before using the GPTSS in new contexts. This study validated the GPTSS-U using qualitative and quantitative methods, consistent with other mental health validation studies in sub-Saharan Africa. For example, Michalopoulos et al. (2020) adapted the Shame Questionnaire (SQ) for Ugandan youth with HIV, and the SQ was similarly validated in Zambia (Michalopoulos et al., 2015). The International Depression Symptom Scale-General (IDSS-G), validated in LMICs, was shown to be a reliable predictor of functional impairment (Haroz et al., 2014). Our study is the first to examine post-trauma symptoms among Ugandan caregivers.

While the GPTSS is valid and reliable globally (Haroz et al., 2014), differences in item difficulty may reflect cultural factors or wording changes. This could also be a reflection of ongoing trauma. It is possible that some caregivers’ symptoms are more accurately explained by transdiagnostic trauma-related stress, which acknowledges chronic environmental stressors. In these cases, measuring symptoms associated with a ‘post’ period could be more complex. Future validation studies should explore differences by gender, age, location, and trauma type and status (i.e. chronic).

Unidimensionality was established with the EFA indicating over 44% of the variance explained by the first factor. Research supports the notion that a unidimensional solution in Exploratory Factor Analysis (EFA) can be considered adequate when the first factor explains between 40% and 50% of the total variance (Pretorius & Padmanabhanunni, 2024). As a unidimensional construct, this suggests that the GPTSS-U reflects a more universal experience of trauma symptoms, increasing the scale’s applicability across diverse cultural contexts. At the same time, the variance was on the lower end, which suggests the possibility of multidimensionality of the GPTSS-U. Future research should explore this as a multidimensional GPTSS-U would allow clinicians to identify distinct post traumatic stress symptom clusters, offering deeper insight into how trauma manifests in individuals within this context. This can support more personalized and targeted treatment strategies, especially important for caregivers facing complex or varied stressors.

A limitation is that the sample reflects only the greater Masaka region. Future studies should assess the GPTSS-U’s utility across other regions in Uganda and countries with high HIV prevalence. Engaging local health professionals, caregivers, and families will help determine appropriate language adjustments.

Although the sample size was adequate (Nguyen et al., 2014), it did not allow for disaggregation by demographic factors. Future research should be conducted with larger samples and examine factors such as gender differences in trauma symptoms. This would be important to address as existing studies have shown inconsistent results in gender-specific symptom endorsement (e.g. women reported feeling more emotionally distant than men; He et al., 2014).

Although low, research has supported the use of exploratory factor analysis with a sample size of 200 potentially producing high-quality estimates (Jung & Lee, 2011). Estimates in the literature also recommend a sample of 200–500 participants as a minimum. The minimum is contingent on the purpose of the analysis (Baker, 2001; Reeve & Fayers, 2005). As this is a preliminary study of the GPTS-IB to develop the GPTSS-U, a smaller sample size is deemed sufficient. Further, our sample size was determined to be adequate as the item difficulty parameters ranged from low to high levels across the latent construct of distress within the scale (Baker, 2001). Future research should include larger samples when possible.

Despite potential overlap, all items were retained to ensure reliability and findings support scale validity. Future studies could explore reducing redundancy to increase precision. Further, the LEC-5 did not assess the potentially traumatic event related to being a caregiver of a child living with HIV, although research has indicated this as such (Bolton et al., 2014). This may explain why 2 of the cases did not experience any traumatic events based on the items of the LEC-5, even though they had high PTS symptoms.

Given that PTSD is a recognized construct in Western trauma-affected populations and its relevance in LMICs is debated with conflicting views on PTSD’s global applicability, our purpose was to examine the presentation and expression of post-traumatic symptoms in Uganda within context. Toward that end, we developed a measure to assess post-traumatic symptoms. Because we were building off of and extending prior work in the field of PTSD in Western settings, we believed that it was important to examine our measure in relation to the PCL-5. Thus, the PCL-5 was used as a general reference of PTSD. We acknowledge that using the PCL-5 to determine case-ness of the GPTSS-U has its limitations. However, the PCL-5 was not used as a gold standard but as a referential benchmark tool given the lack of locally validated PTS tools or trained clinicians to administer diagnostic interviews. Our critique of the PCL-5 is not that it fails to measure post-traumatic stress, but that it may omit culturally specific expressions of post-traumatic stress. As such, our use of the PCL-5 was pragmatic and intended to inform, rather than define, the validity of the GPTSS-U.

This study also has many strengths. These include providing nuanced data on an underrepresented population, supporting efforts to create culturally relevant mental health tools, and raising awareness of caregiver trauma exposure in Sub-Saharan Africa.

5. Conclusions

Current findings demonstrate that the GPTSS-U is a reliable and valid instrument for assessing traumatic stress among caregivers of youth with HIV in Uganda. The utility of the GPTSS-U indicates the high importance of further examination of contextual factors that may affect the presentation and expression of post-traumatic symptoms, such as gender, socioeconomic status, social support, region, caregiving responsibilities, and additional types of trauma. Further, the mixed methods, in addition to the utilization of both CTT and IRT to develop the GPTSS-U, may be helpful in the development of relevant and culturally-validated scales in contexts such as in Sub-Saharan Africa, especially in regions with high HIV prevalence. Findings for high levels of PTS among caregivers of HIV suggest that post-traumatic stress interventions may be needed.

Authors contributions

LMM and AC were involved in the conception and design. MM, LMM, AC and EW were involved in the analysis and interpretation of the data. LMM, MM, EW, AT, DR, TN, and JSK were involved in drafting the paper. LMM, EW, DR, AT, TN, JSK, AC, and NA were involved in revising the manuscript critically for intellectual content. All of the authors provided final approval of the version to be published; and all authors agree to be accountable for all aspects of the work.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are available from the corresponding author (LM) upon reasonable request.