Abstract
Background:
Death literacy – the knowledge, skills, and capacity to navigate end-of-life care, advance care planning, dying, and death care – is shaped by cultural, religious, and linguistic factors. The Death Literacy Index (DLI) captures practical, experiential, factual, and community domains. Currently, there is limited evidence about death literacy levels among culturally and linguistically diverse communities in multicultural settings, and little is known about how translated versions of the DLI perform psychometrically in these groups or how culturally tailored, community-delivered interventions can strengthen death literacy.
Objectives:
(1) To evaluate the impact of culturally adapted, community-designed and delivered interventions on DLI subscale and total scores among Filipino, German, Italian, and Vietnamese communities in South Australia; and (2) to examine the preliminary psychometric properties (factor structure and reliability) of translated DLI instruments in these populations.
Design:
Pre–post intervention study.
Methods:
Community leaders in each group co-designed and delivered a community-specific intervention over 3–5 months using culturally familiar channels (workshops, radio/podcasts, social media, and printed/online resources). Thus, four parallel but distinct interventions were implemented, one in each cultural/language community. The DLI was professionally translated (German, Italian, Tagalog, Vietnamese) with community review. Surveys were administered in preferred languages at baseline (T1) and ~6 weeks post-intervention (T2). Psychometrics (T1, n = 242): confirmatory factor analysis (CFA) and reliability. Impact (matched pairs, n = 100): Wilcoxon signed-rank tests.
Results:
Psychometrics (T1, n = 228): Subscale CFAs supported the translated DLI across groups, while the total-scale fit was poor – reinforcing a multidimensional construct; internal consistency was good–excellent overall, with weaker reliability for Talking Support subscale (Vietnamese) and some variability in Experiential Knowledge. Intervention impact (matched pairs, n = 100): Filipino participants showed significant increases in median scores across all four DLI subscales and the total DLI score (medium–large effects); Vietnamese participants showed significant decreases in Practical Knowledge and total DLI scores, with little change on other subscales; German and Italian groups demonstrated small, mostly non-significant increases on selected DLI subscales (particularly Factual and Community Knowledge), with minimal change in total DLI scores. Across groups, DLI subscales capturing knowledge of resources and supports generally improved.
Conclusion:
Culturally tailored, community-delivered interventions can strengthen death literacy, but effects vary by cultural/linguistic context. DLI subscales function adequately across groups, whereas total-scale fit is poor. Future work should refine translations (e.g. DLI-R), address weaker domains, and adapt content and delivery to community preferences to enhance cultural responsiveness and equity in end-of-life literacy. The study demonstrates how death literacy can operate as a culturally responsive, community-level public health indicator and provides evidence to guide adaptation of both the DLI and community programs to advance equity in end-of-life literacy and care.
Plain language summary
People often find it hard to talk about death, dying, and end-of-life care. ‘Death literacy’ is the practical knowledge and confidence we need to plan, make choices, and support each other at the end of life. This study looked at how to build death literacy in multicultural communities and whether a short questionnaire (the Death Literacy Index) works well in different languages. We worked with community leaders from Filipino, German, Italian, and Vietnamese groups in South Australia. Together, we designed and delivered activities over several months-such as workshops, radio and podcast segments, social media posts, and printed resources-using preferred community channels and languages. Community members completed a survey in their chosen language before the activities and again about six weeks after. What did we find? Filipino participants showed clear improvements across all parts of death literacy. German and Italian groups had smaller, mixed improvements. Vietnamese participants did not improve overall and, in some areas, scores went down-highlighting that different communities may need different kinds of support or materials. Across groups, people generally felt more aware of resources and support available. We also checked how well the survey worked in translation. Most parts of the survey performed well across languages, but the total score did not fit as a single measure, and one section was less reliable in Vietnamese. This suggests death literacy is made up of several related skills rather than one simple trait, and that some survey wording needs refining for certain languages. Why does this matter? Culturally tailored, community-led approaches can strengthen end-of-life knowledge, but one size does not fit all. Services should partner with communities to design activities that suit their language, culture, and preferences. Future work will refine the survey (e.g., using the revised DLI-R) and adapt programs to better support every community.
Keywords
Introduction
Death literacy refers to the skills, knowledge, and capacity individuals require to navigate end-of-life and death care options, including understanding healthcare systems, engaging in advance care planning (ACP), and making informed decisions regarding dying and death care.1,2 It is a critical construct in public health and palliative care, facilitating community engagement and promoting autonomy and preparedness in end-of-life decision-making.1,2 Conceptually, death literacy aligns with broader public health frameworks, including the social determinants of health, 3 as access to information, services, and culturally safe support is shaped by structural conditions such as language, socioeconomic status, and migration history. 4 It also reflects elements of community capacity theory, 5 whereby collective knowledge, networks, and shared practices enable communities to respond to health and wellbeing needs, including end-of-life needs. Furthermore, death literacy can be considered a form of health literacy specific to dying and death, 6 with implications for health behaviour change and help-seeking within diverse cultural contexts.
Death literacy does not develop equally across populations, being shaped by structural and social inequities 7 that influence access to information, culturally safe services, and opportunities to engage with end-of-life planning. Limited English proficiency, socioeconomic disadvantage, health system navigation barriers, and experiences of marginalisation can restrict people’s ability to obtain and act on death-related knowledge, 8 contributing to inequitable palliative care access and outcomes. In this way, death literacy functions not only as an individual capability but also as an indicator of broader community understanding 9 and health equity in end-of-life care.
Cultural, religious, and linguistic factors deeply impact how individuals and communities engage with death, end-of-life care, and associated choices.4,10 As populations become increasingly diverse, there is growing recognition of the need for death literacy to be understood and operationalised across different cultural contexts. The Death Literacy Index (DLI), 1 developed in Australia, operationalises this construct across domains such as practical, experiential, factual, and community knowledge, and is a seminal index used across diverse cultural settings (see References 11–15). For example, it has undergone cultural adaptation and validation in multiple international contexts – including China, 16 Sweden, 17 and Turkey 18 – demonstrating its relevance across diverse populations. However, substantial variability in how death literacy manifests in culturally and linguistically diverse (CALD) groups remains an ongoing challenge for public health initiatives. While the DLI has demonstrated utility in various countries, disparities in understanding and engagement with death-related issues persist among CALD communities. These gaps are compounded by limited culturally appropriate tools and a lack of practitioner training in culturally sensitive communication.4,17
Educational interventions targeting death literacy have shown promise, particularly in increasing comfort and competence in conversations about the end of life. For example, Vivekananda et al. 19 found that online modules improved psychology students’ comfort in initiating end-of-life discussions. However, existing studies have predominantly focused on English-speaking populations,2,15,20 and those involving non-English-speaking populations have been conducted almost exclusively within their countries of origin,13,14,16–18,21 rather than examining CALD communities living in multicultural societies outside their country of origin.
Despite growing international uptake of the DLI,14,16–18,21 there remains limited understanding of its cultural responsiveness and psychometric validity within community-based multicultural contexts. Specifically, little is known about whether translated versions of the DLI adequately capture culturally specific conceptualisations of death literacy. Moreover, as studies to date have been cross-sectional, assessing death literacy at a single point in time (see, for example, References 14,16–18,21), evidence regarding whether culturally adapted interventions can meaningfully shift death literacy among CALD populations over time is non-existent. These evidence gaps limit the tool’s broader applicability and constrain public health strategies aiming to improve equity in end-of-life literacy and access. Accordingly, there is a clear need for research that both tests the DLI’s measurement properties across diverse cultural groups and evaluates culturally appropriate approaches to strengthening death literacy in community settings.
The present study addresses these gaps by evaluating community-delivered death literacy interventions among Filipino, German, Italian, and Vietnamese groups in South Australia. A convenience sampling approach was adopted. The four cultural groups – Filipino, German, Italian, and Vietnamese – were selected in partnership with the Multicultural Communities Council of South Australia, reflecting major CALD communities in the region with distinct cultural, linguistic, and migration-to-Australia profiles. Their inclusion enabled examination of death literacy across diverse cultural orientations towards dying and death, enhancing the breadth of CALD perspectives in the study and the transferability of findings to other multicultural public health contexts. It assesses both the impact of the interventions and the preliminary psychometric properties of the DLI in these populations. By doing so, the study contributes to a growing evidence base that supports culturally responsive approaches in public death literacy initiatives. This study aimed to answer the following question:
What is the impact of participating in a community-delivered intervention on the death literacy of CALD people in South Australia, and how do translated versions of the DLI perform psychometrically in these community contexts?
Methodology
The study is part of a larger 5-year program of research addressing issues in access and delivery of palliative care services to underserved populations (see The University of Adelaide 22 for further details). Comprising five separate studies, the research program rests on ongoing collaboration with community, industry, advocacy, academic, and government representatives, prioritising the voices, needs, and experiences of patients, families, and community. The philosophical orientation is that of critical realism, 23 utilising a mixed-methods, multidisciplinary approach. The study was approved by the institutional Human Research Ethics Committee (protocol ID H-2022-155).
A pre–post intervention analysis was undertaken to investigate the impact of culturally appropriate, co-designed community-based programs on the death literacy of four cultural groups: Filipino, German, Italian, and Vietnamese. The study design was developed in consultation with, and activities supported by, the Multicultural Communities Council of South Australia and Palliative Care South Australia, the peak bodies supporting and advocating for CALD communities, and promoting palliative care in South Australia, respectively. Funding was specifically allocated to support their involvement, and this, in combination with their longstanding relationships of cooperation and trust with these communities, was critical in working with community leaders and recruiting within the target communities.
Inclusion criteria for the study were people aged 18 years and over, and from one of four migrant cultural backgrounds (Filipino, Vietnamese, German, or Italian), with the aim of attracting a large sample size for each cultural group to strengthen subsequent analyses and validity of findings.
Survey translation
The original version of the DLI survey was developed in English. 1 To ensure linguistic inclusivity and cultural appropriateness, accredited professional translators forward-translated the survey from English into German, Italian, Tagalog (Filipino), and Vietnamese. Translations were shared with community leaders from each community, who provided feedback on their appropriateness, clarity and cultural relevance of translated content for their respective communities, with some minor adjustments. This feedback informed final versions prior to dissemination among each community. To support informed consent and meaningful participation, all study materials, including participant information and consent forms, were translated into the four target languages, with these and English versions made available in hardcopy and online.
Intervention
Drawing on their knowledge of their community, known and trusted leaders in each cultural group selected and implemented a package of intervention activities they considered would best engage and resonate with their communities. Community leaders were individuals with lived experience of belonging to each community, and each adopted a unique method in terms of community liaison, communication, and knowledge upskilling. Many chose to integrate new content into existing networks and familiar platforms to maximise accessibility for their target audience. For example, new content was incorporated into pre-existing podcast series, radio segments, newsletters, or social media channels. Together, these activities constituted four distinct, community-specific interventions that shared a common overarching aim – to strengthen death literacy – but being tailored to each group, differed in content, duration, and mode of delivery.
Intervention activities were implemented over a 5-month period, with some overlap in activities across groups (see Table 1). Examples included face-to-face and online activities and documentation (online or hardcopy) sharing information on death, dying, palliative care, and ACP. Community leaders led the coordination and delivery of these activities, reflecting their strong commitment to the research goals through involvement in the study and to improving community awareness and comfort around death, dying, and palliative care. The duration, number, style, method, frequency, intensity, and advertising/promotion of intervention activities varied between groups, ranging from 3 to 5 months, which may have influenced differences in resulting survey outcomes. However, all community leaders reported being motivated and committed to normalising conversations around death and dying, and increasing overall death literacy among their communities.
Intervention activities by community group.
‘Train-the-trainer’ workshops
The intervention cycle began with a 5-h ‘Train-the-trainer’ workshop delivered in partnership with Palliative Care South Australia. Approximately three leaders from each participating CALD community attended. These individuals, considered research collaborators rather than participants, were fully informed of their role of disseminating information and initiating conversations about palliative care, death, dying, and grief within their communities. All were trusted figures within their communities and played a critical role in ensuring intervention activities were culturally appropriate and linguistically accessible. They were mindful of cultural sensitivities, such as avoiding activities during significant religious or cultural periods, and were also well-positioned to connect community members with relevant supports and services if required.
The workshop was designed to equip community leaders with the knowledge, skills, and resources necessary to confidently engage their communities on these sensitive topics. Following the workshop, community leaders, with ongoing support provided by Multicultural Communities Council of South Australia (MCCSA), Palliative Care South Australia (PCSA), and the research team, co-developed and delivered culturally relevant resources and activities tailored to their communities’ preferences and needs.
Recruitment and consent processes
A designated MCCSA staff member (F.D.) served as a key liaison between the MCCSA, PCSA, researchers, and community representatives. Community leaders played a central role in inviting and recruiting participants through their established local networks and trusted relationships. Recruitment strategies were tailored to accommodate diversity in communication practices, guided by community leaders, ensuring information was clear, respectful, and accessible. Strategies included face-to-face invitations, phone calls, newsletters, social media, and other culturally relevant channels.
During the first community activity, an interactive discussion on palliative care, death, dying, and grief, leaders provided information about study aims and outlined the process for completing the survey (online or hardcopy) using culturally appropriate and accessible language. At this meeting, potential participants received participant information and consent forms in their preferred language and had the opportunity to reflect on participation. It was clearly communicated, in invitations and during the meeting, that attendance did not require study involvement. Following consent, participants were provided with the DLI.
Data collection
Self-administered surveys (and online options) were available in participants’ preferred languages (including English). Completed hardcopy surveys for T1 could be returned via a pre-paid envelope or during community face-to-face events (such as workshops). During these events, community leaders assisted participants to complete surveys where needed, particularly in cases where language or literacy barriers were present. Following collection during the initial workshop, a 6-month period for intervention activities began, followed by re-administration of the DLI survey (T2) approximately 6 weeks after the intervention. Participants were invited but not required to provide a unique identifier that enabled matching of T1 and T2 data. On completion/submission of each survey, a $25 gift card was provided as a token of appreciation for participants’ time and contribution.
Data confidentiality and storage
No identifying information was collected; identifiers provided by participants for data matching (T1–T2) were removed from the dataset following linkage. Consent forms and completed hardcopy surveys were stored securely in a locked filing cabinet on university premises, with survey data transferred by university researchers to Qualtrics (Provo, UT, USA), within a secure password and user-protected server. All online survey responses were similarly entered into Qualtrics and ensuing datasets stored on password- and user-protected university servers. Participants submitting online were directed to an independent webpage where they provided addresses for delivery of gift vouchers. All contact details were destroyed following the postage of these. During data analysis, responses were aggregated to ensure individual anonymity.
Measures
Sociodemographic information: Sociodemographic information collected included participants’ gender), marital status, living arrangements, education, and cultural group.
DLI: The DLI is a multidimensional, 29-item instrument designed to assess individuals’ knowledge and skills related to end-of-life and death care options. Developed in Australia through extensive qualitative research with carers and community networks, the DLI operationalises the construct of death literacy as comprising four domains: Practical Knowledge, Experiential Knowledge, Factual Knowledge, and Community Capacity (the latter two domains are subdivided into two subscales. 1 The DLI has been psychometrically validated in several cultural contexts, demonstrating robust reliability and construct validity. In Australia, the original scale showed excellent model fit (comparative fit index (CFI) = 0.955, root mean square error of approximation (RMSEA) = 0.049) and high internal consistency (α = 0.82–0.95). 24 In the United Kingdom, confirmatory factor analysis (CFA) replicated the higher-order structure with strong internal consistency (α = 0.79–0.90), good convergent and discriminant validity, and minimal socio-demographic variability in scores. 15 Similarly, the Turkish and Chinese validations confirmed the scale’s internal structure and cross-cultural applicability, with acceptable reliability (Cronbach’s α = 0.68–0.97) and fit indices (CFI > 0.93; RMSEA < 0.055). 18 A Swedish-language adaptation (DLI-S) employed a multi-stage translation, cognitive interviewing, and expert review process. The final Swedish version demonstrated excellent reliability (overall α = 0.94; subscale α = 0.81–0.92) and test–retest stability (ICC = 0.66–0.85), supporting its validity within that national context. 17 Collectively, these studies affirm the DLI as a rigorous tool for evaluating death literacy across diverse settings.
Data analysis
Initially, a multi-phase analysis was employed. Two distinct datasets informed the analyses:
Baseline (T1) data, used exclusively for psychometric testing of the DLI, and
Matched-pairs (T1–T2) data, used for assessing within-group change in DLI scores over time.
Psychometric testing was conducted, consisting of CFAs and reliability analyses. These were conducted to establish the construct validity and internal consistency of the DLI and its subscales within the target populations. Matched-pairs analyses were then conducted to assess within-group changes in death literacy knowledge over time across Filipino, German, Italian, and Vietnamese language groups. All matched samples met or exceeded the minimum required sample size of 15 participants per group as determined by a priori power analysis using G*Power 3.1.9.7, assuming a large effect size (dz = 0.80), alpha = 0.05, and power = 0.80. R-studio was used for all analyses.
Psychometric testing: Subscale structure was tested separately within each cultural group using CFA in R (lavaan v0.6-x). For a given subscale, only respondents with complete data on all constituent items were retained (list-wise deletion). Models were estimated with maximum-likelihood and missing = ‘listwise’, yielding identical sample sizes for CFA and reliability analyses. Model fit was evaluated with the CFI, Tucker-Lewis Index (TLI), RMSEA, and standardised root mean square residual (SRMR). Standardised factor loadings and their range are reported. Internal consistency for the same complete-case datasets was assessed with Cronbach’s α (psych v2.3-x). Corrected item-total correlations (CITCs), the range of inter-item correlations (IICs), and the average inter-item correlations (AICs) were also calculated. Rows with any missing item values were excluded, ensuring the sample size (n) matched that used in the corresponding CFA.
Mixed-methods analysis: Matched-pairs analyses were conducted to assess within-group changes in death literacy knowledge over time across the four language groups. Shapiro–Wilk tests conducted on T1 scores for each DLI subscale and total score indicated that many distributions deviated significantly from normality, particularly within the Filipino, German, and Vietnamese groups (p < 0.05 across multiple domains). In addition, matched-pairs subgroup sizes within each cultural group were modest (ns = 16–30; see Table 4), reducing the robustness of parametric repeated-measures tests and increasing sensitivity to outliers. The DLI items also employ ordinal Likert-type response formats, for which non-parametric tests are generally recommended when distributional assumptions are not met. Accordingly, the Wilcoxon signed-rank test was employed to assess within-group differences over time.
Findings
The findings section is structured into three parts: (1) Participant Characteristics, providing demographic and background information; (2) Psychometric Properties of the DLI, detailing the factor structure, subscale reliability, and item-level performance across cultural groups; and (3) Matched-Pairs Analysis, presenting pre–post changes in DLI scores and intervention effects within each group.
Participant characteristics
Table 2 presents participant characteristics at baseline (T1: psychometric testing sample) and in the matched-pairs sample (participants with both T1 and T2 data). A total of 228 participants provided baseline (T1) data and 100 provided matched T1–T2 data (Table 2). Percentages reflect the proportion of respondents who selected each option for a given question. Living arrangements and education were multiple-response items, therefore percentages exceed 100% within these categories.
Participant characteristics at baseline (T1; n = 228) and for the matched-pairs sample with both T1 and T2 data (n = 100).
Psychometric properties of the DLI
The psychometric properties of the DLI, including CFA fit indices and internal consistency estimates, disaggregated by cultural group and subscale are provided in Table 3. Specifically, standardised factor loadings (λ) for each subscale and cultural group are summarised within the table.
Psychometric properties and standardised factor loadings (λ) of the DLI subscales by cultural group.
AIC: average inter-item correlation; CFI: Comparative Fit Index; CITC: corrected item-total correlation; IIC: inter-item correlation; RMSEA: root mean square error of approximation; SRMR: standardised root mean square residual; TLI: Tucker-Lewis Index; α: Cronbach’s alpha (internal consistency); λ: standardised factor loadings.
Confirmatory factor analysis
To evaluate the structural validity of the DLI across cultural groups, we conducted CFAs for each subscale within each group. Model fit was assessed using four commonly reported indices: the CFI, TLI, RMSEA, and SRMR. Following conventional cut-offs recommended by Hu and Bentler, 25 good model fit was indicated by CFI and TLI values ⩾0.95, RMSEA ⩽0.06, and SRMR ⩽0.08. Values approaching these cut-offs were considered to reflect acceptable fit.
Among Filipino participants, model fit was generally acceptable to excellent. All subscales had CFI values ⩾0.922 and TLI values between 0.742 and 1.002. RMSEA values were more variable, with some exceeding ideal thresholds (e.g. Hands-on Support = 0.346), while SRMR values remained consistently low (all <0.08). Experiential Knowledge (CFI = 0.982; TLI = 0.963; RMSEA = 0.148; SRMR = 0.033) and Community Support (CFI = 0.974; TLI = 0.922; RMSEA = 0.209; SRMR = 0.037) demonstrated strong comparative fit but with some elevation in RMSEA, which may reflect model complexity or sample characteristics.
The Vietnamese sample showed good fit for several subscales, such as Hands-on Support (CFI = 1.000; TLI = 1.026; RMSEA = 0.000; SRMR = 0.015). However, Talking Support presented concerns, with a low Cronbach’s alpha (0.533), lower TLI (0.598), and SRMR above 0.06 (0.068), suggesting potential issues with internal coherence and structural validity. Experiential Knowledge also showed less-than-ideal fit (CFI = 0.772; TLI = 0.543; RMSEA = 0.362), indicating the need for further refinement or cultural adaptation of this subscale.
CFA results for German participants revealed acceptable model fit across most subscales, with CFI values ⩾0.841 and SRMR values <0.08. However, some subscales such as Accessing Help (RMSEA = 0.272) and Experiential Knowledge (RMSEA = 0.271) exceeded the recommended threshold, while others – like Talking Support (CFI = 1.000; TLI = 1.051; RMSEA = 0.000; SRMR = 0.030) – fit the model exceptionally well.
The Italian data supported strong model fit across most domains, with several subscales meeting all thresholds (e.g. Accessing Help: CFI = 1.000; TLI = 1.019; RMSEA = 0.000; SRMR = 0.016). However, Talking Support (TLI = 0.570; RMSEA = 0.354) and Factual Knowledge (TLI = 0.718; RMSEA = 0.257) showed weaker fit, despite acceptable CFI and SRMR values, suggesting the need for further exploration of item functioning in these domains.
In addition to subscale-level CFA, we examined model fit for the total scale within each cultural group. CFA model fit for the full DLI scale was poor across all groups relative to established thresholds. CFI values ranged from 0.214 (Vietnamese) to 0.441 (Filipino), and TLI values ranged from 0.285 to 0.433. RMSEA values were elevated across groups (range: 0.188–0.389), and SRMR values exceeded acceptable limits in all cases (range: 0.141–0.167). Furthermore, some groups showed problematic factor loadings, including negative or near-zero values in the Vietnamese sample (range: −0.201 to 0.854). These results suggest that a unidimensional or single-factor model for the total DLI scale may not adequately capture the structural complexity of the instrument, reinforcing the appropriateness of treating the DLI as a multidimensional construct. This finding aligns with the original DLI validation framework, which conceptualised death literacy as a composite of distinct but related domains.
From an interpretive perspective, these total-scale findings indicate that, although the 29 items cohere sufficiently to produce highly reliable composite scores, they do not conform to a strictly unidimensional measurement structure in these CALD samples. Thus, in the present study, interpretation at the subscale level is most reliable, and total DLI scores are treated as pragmatic indicators of overall death literacy rather than as evidence of a single underlying latent factor. As such, poor model fit for the total-scale CFA does not undermine the interpretability of our findings; instead, it reinforces the importance of respecting the multidimensional nature of the DLI when drawing conclusions. Overall, these CFA results support the structural validity of the DLI at the subscale level, while the consistently poor fit for the total scale indicates that a unidimensional single-factor model does not adequately represent this instrument within these CALD samples.
Internal consistency
We evaluated the internal consistency of the DLI across four cultural groups – Filipino, German, Italian, and Vietnamese – using both subscale-level and total-scale reliability and item statistics. Internal consistency was evaluated using Cronbach’s alpha (α), with higher coefficients indicating stronger reliability (≥0.70 acceptable, ≥0.80 good, and ≥0.90 excellent). 26 We also examined CITC, IICs, and AICs. CITC values >0.30 are considered desirable as they indicate that individual items contribute meaningfully to the overall scale. 27 IICs are expected to fall within 0.15–0.50 for broader constructs, 28 with higher values suggesting tighter item cohesion but potentially indicating redundancy if they exceed 0.80. Similarly, AICs in the range of 0.15–0.50 are typically considered indicative of adequate internal consistency, with values above this range suggesting strong item coherence but also the potential for overlapping content.
Total scale analyses supported the psychometric adequacy of the DLI across all cultural groups. The total scale Cronbach’s alpha values ranged from 0.894 (Vietnamese) to 0.958 (Filipino), indicating excellent reliability across the board. CITC values were acceptable to strong for most groups: Tagalog (0.499–0.752), German (0.225–0.717), and Italian (0.258–0.774). Vietnamese participants showed a broader range (−0.060 to 0.764), including some items below the recommended threshold, likely driven by the weaker reliability of the Talking Support subscale. AIC values for the total scale were well within or above the acceptable range: 0.441 for Tagalog, 0.214 for Vietnamese, 0.291 for German, and 0.283 for Italian. These findings support the coherence of the full DLI instrument and its internal structure, although item-level refinement may be warranted for specific subpopulations.
At the subscale level, Filipino participants showed high to excellent reliability, with Cronbach’s alpha ranging from 0.863 (Talking Support) to 0.951 (Experiential Knowledge). AIC values were high (0.613–0.793), exceeding the typical range and indicating very strong item cohesion. While this reflects excellent reliability, the upper-end values suggest potential conceptual overlap among items. CITC values were consistently strong (0.633–0.939), and IICs ranged from 0.478 to 0.958.
In the Vietnamese sample, reliability was good for most subscales (α = 0.834–0.903), with AICs between 0.536 and 0.664 – again exceeding the ideal range, but indicative of internal consistency. However, the Talking Support subscale showed poor internal consistency (α = 0.533), a very low AIC of 0.228, and a CITC range from 0.178 to 0.475, including negative IICs (as low as −0.383). These findings suggest problematic item coherence, possibly due to linguistic or cultural misalignment in how the ‘talking support’ domain is perceived.
German participants demonstrated moderate to strong reliability across subscales (α = 0.739–0.910), with AICs between 0.419 and 0.665. Talking Support and Hands-on Support had AICs near the lower bound (0.446 and 0.419, respectively) and CITC values close to the threshold, suggesting some variability in item performance. IICs were wide-ranging (0.198–0.841), reinforcing heterogeneity in item responses.
Italian participants demonstrated consistently high subscale reliability (α = 0.836–0.931), with AICs from 0.558 to 0.731. These values, although above the recommended range, indicate strong internal consistency. CITC values were uniformly strong (0.532–0.874), and IICs ranged from 0.326 to 0.869, supporting scale integrity.
Overall, CITC, IIC, and AIC values reinforced the pattern of strong psychometric performance for both the total scale and subscales across all cultural groups. Subscales such as Accessing Help and Community Support consistently demonstrated strong internal consistency and coherent item functioning. Talking Support, however, showed notably lower reliability in the Vietnamese and, to a lesser extent, German groups, indicating a need for potential refinement or cultural adaptation of this subscale to ensure broader applicability and construct clarity.
Matched-pairs findings
Table 4 displays descriptive and inferential statistics from the matched-pairs analysis conducted for each cultural group and DLI subscale. Medians and distributional characteristics are shown for T1 and T2, alongside Wilcoxon signed-rank test results, and effect sizes, clarifying the magnitude and significance of within-group changes.
Descriptive statistics and Wilcoxon signed-rank effect sizes (r) for DLI domains by cultural group.
Within-group changes were analysed using the Wilcoxon signed-rank test (two-tailed). Medians (median), interquartile ranges (IQR), and minimum and maximum values (Min, Max) are reported for T1 and T2 scores.
diffMed: difference in medians (T2–T1); p: p value; r: effect size (r = Z/√n) where n is the matched-pairs sample size; W: Wilcoxon test statistic; Z: standardised test statistic.
Within-group changes in death literacy knowledge
Among Filipino participants, statistically significant improvements were observed across all four domains of death literacy and the total knowledge score. Large effect sizes were found for Factual Knowledge (r = 0.728, p = 0.004) and Community Knowledge (r = 0.740, p = 0.002), while medium effect sizes were evident for Experiential Knowledge (r = 0.578, p = 0.022), Practical Knowledge (r = 0.485, p = 0.048), and Total Knowledge (r = 0.568, p = 0.025). These findings reflect consistent and substantial gains following the intervention, particularly in domains involving factual understanding and community-based literacy. The combination of statistical significance and practical effect magnitudes suggests that the intervention was highly effective for this group.
In contrast, the Vietnamese group demonstrated a more mixed pattern. A statistically significant decline was observed in Practical Knowledge (r = 0.321, p = 0.036), while the Total Knowledge score also showed a significant reduction (r = 0.476, p = 0.023). Other domains such as Factual Knowledge (r = 0.139, p = 0.055) and Community Knowledge (r = 0.289, p = 0.339) did not show significant change, although the former approached significance. Experiential Knowledge remained stable (p = 0.239, r = 0.084). These findings suggest that the intervention may not have been as effective for the Vietnamese group, with some indications of decreased scores over time, warranting further investigation into cultural or implementation factors influencing these results.
The German group exhibited largely stable scores across most domains, with non-significant results observed for Experiential (r = 0.267, p = 0.541), Practical (r = 0.212, p = 0.891), and Total Knowledge (r = 0.044, p = 0.842). However, Factual Knowledge approached statistical significance (p = 0.052) with a medium effect size (r = 0.404), and Community Knowledge also showed a medium effect (r = 0.416, p = 0.090). These results suggest possible improvements in specific knowledge areas that may not have reached conventional thresholds for statistical significance, potentially due to sample variability.
Among Italian participants, most subscales showed no significant change over time. Community Knowledge exhibited the largest shift with a medium effect size (r = 0.590, p = 0.073), approaching statistical significance. Other domains, including Experiential (r = 0.300, p = 0.248), Factual (r = 0.452, p = 0.443), and Practical Knowledge (r = 0.293, p = 0.629), demonstrated small or negligible effects. The Total Knowledge score remained unchanged (r = 0.168, p = 0.798). These patterns suggest minimal change in death literacy over time, though modest gains in community-related knowledge may warrant future exploration.
Narrative synthesis of cross-cultural differences
Taken together, the psychometric and matched-pairs analyses illustrate how cultural context shapes both the measurement and development of death literacy. Filipino participants demonstrated the most consistent and substantial improvements across all domains, suggesting that the intervention was highly resonant for this group. The high reliability and cohesive factor structures observed among Filipino participants further suggest that the DLI concepts aligned closely with existing cultural knowledge frameworks.
In contrast, Vietnamese participants demonstrated a more mixed pattern of change across domains, including declines in practical and total knowledge scores. These outcomes should be interpreted with caution, as they may reflect measurement limitations rather than true reductions in death literacy. In particular, lower internal consistency and weaker model fit for the Talking Support and Experiential Knowledge subscales suggest that some items may not have functioned optimally within this sample. This may indicate a need for linguistic refinement, clearer contextual examples, or additional cognitive testing to ensure that DLI constructs are interpreted consistently. Implementation differences (e.g. facilitator familiarity with language needs or community engagement styles) may also have contributed to variability in outcomes. Together, these findings highlight the importance of validating and adapting death literacy measures prior to broader comparative use in diverse CALD communities.
The German and Italian groups displayed modest or domain-specific changes, reflecting more stable death literacy levels at baseline and possibly narrower growth potential over a short intervention period. Psychometric indicators also suggested some variability in item responsiveness across these groups, which may have reduced the sensitivity of the measure to detect small shifts in knowledge or confidence over time. Improvements in community-oriented domains (e.g. Community Knowledge in Italians) may therefore reflect increased awareness of support pathways rather than shifts in deeper cultural attitudes.
Across all four groups, the DLI functioned most reliably at the subscale level. This supports the conceptualisation of death literacy as a multidimensional construct shaped by cultural norms around communication, knowledge sharing, and caregiving roles. The variability in matched-pairs outcomes demonstrates the importance of cultural tailoring, both in how death literacy is measured and in how interventions are delivered. Future initiatives will benefit from identifying which aspects of death literacy are most salient, and which require culturally grounded adaptation within specific community contexts.
Taken together, these findings underscore the utility of matched-pairs analysis for detecting within-group change. The Filipino group demonstrated the most robust gains, with consistent, statistically significant improvements and medium-to-large effect sizes across all domains. The Vietnamese group showed unexpected declines in some areas, particularly practical and total knowledge. German and Italian participants exhibited more nuanced patterns, with some moderate improvements that did not consistently reach significance. These variations highlight the importance of culturally tailored approaches to enhancing death literacy and suggest that intervention strategies may require adaptation to better address the needs of different language and cultural groups.
Discussion
Main results
This study employed a multi-phase, mixed-methods approach to examine the psychometric properties and effectiveness of a culturally tailored interventions designed to enhance death literacy among four language groups in South Australia: Filipino, German, Italian, and Vietnamese. Guided by frameworks of health literacy, 10 community capacity, 5 and the social determinants of health, 3 this analysis positions death literacy as a specialised but inherently multidimensional form of health literacy that is distributed across individuals, families, communities, and systems rather than residing solely in individual knowledge or attitudes. CFA supported structural validity of the DLI at the subscale level, although the total scale consistently demonstrated poor model fit across all groups. This reinforces the conceptualisation of death literacy as a multidimensional construct rather than a unidimensional trait. Internal consistency was strong to excellent across most subscales and language groups, with the exception of the Talking Support and Experimental Knowledge subscales in the Vietnamese group, which showed low reliability and item coherence, suggesting potential issues in cultural or linguistic alignment.
Importantly, variation in model fit and reliability also mapped onto intervention outcomes: groups where subscales performed more strongly (particularly the Filipino cohort) demonstrated clearer gains in corresponding domains, whereas subscales with weak performance (e.g. Vietnamese Talking Support and Experiential Knowledge) aligned with smaller or negative pre–post shifts. This interplay highlights how psychometric responsiveness can shape the extent to which community-level changes in death literacy are detectable.
Matched-pairs analyses revealed that the Filipino group experienced the most robust and consistent improvements in death literacy, with statistically significant pre–post gains and medium-to-large effect sizes observed across all domains. Interpreted through health literacy and community capacity lenses, these findings suggest that the intervention worked not only at the level of individual skills (e.g. factual understanding) but also by activating existing social networks and trusted community leaders as ‘literacy mediators’ who helped participants interpret, discuss, and apply information about death, dying, palliative care, and ACP.4,5,8 In community empowerment terms, the Filipino activities, integrated into familiar spaces such as community meetings, radio, and social media, appear to have strengthened collective efficacy and shared norms around talking about death, thereby supporting more interactive and critical forms of death literacy rather than merely functional knowledge. These findings suggest that the intervention was highly effective for this group, particularly in enhancing factual and community-based knowledge.
In contrast, the Vietnamese group showed a more mixed pattern, including statistically significant declines in practical and total death literacy knowledge. These findings raise questions about the cultural fit and salience of the intervention, as well as the appropriateness of the DLI as a measurement tool for this cohort. From a social determinants and culturally competent care perspective,3,4,8 these patterns may reflect broader structural inequities that shape how Vietnamese participants access, interpret, and act on information about end-of-life care – for example, language barriers, differential familiarity with health and social care systems, and cultural norms that discourage open discussion of death. Weaker reliability and poorer model fit for Talking Support and Experiential Knowledge in this group further suggest that some items may not align with locally meaningful ways of learning from or talking about death, underscoring the importance of culturally adapted language, examples, and modes of delivery in both intervention design and measurement.4,8,13,16,17 It is possible that the DLI did not adequately capture culturally specific understandings or expressions of death literacy among Vietnamese participants, which may have contributed to lower reliability in some subscales and unexpected shifts in post-intervention scores. However, it is also possible that variations in the assistance provided to participants by community workers to understand, respond to, and complete the DLI may have varied across language groups and impacted on responses recorded at T1, but not T2.
The German and Italian groups exhibited minimal change overall, with some domains – particularly community knowledge, showing small-to-moderate effect sizes that approached but did not reach statistical significance. These more modest or domain-specific shifts are consistent with the idea that baseline death literacy, settlement history, and existing access to services (key social determinants of health3,7) may have limited observable gains over a relatively short intervention period, while still allowing for targeted improvements in community-facing domains such as knowledge of resources and support pathways.5,7,9
Taken together, our psychometric and intervention findings affirm the cross-cultural utility of the DLI subscales while emphasising that death literacy is best understood as a composite of related capacities, practical, experiential, factual, and community, rather than a single latent trait. In line with community capacity theory, 5 different communities may display distinct understandings across these domains (e.g. strong experiential and community knowledge but weaker factual or system navigation knowledge), which has direct implications for tailoring interventions. These findings affirm the cross-cultural utility of the DLI subscales while also highlighting the need for ongoing refinement to enhance cultural responsiveness and scale performance. The variation in intervention outcomes underscores the importance of adapting both content and delivery methods to better align with the cultural, linguistic, and contextual needs of specific communities, particularly where participants may require assistance to fully understand and respond to the DLI items.
Interpretations of main results
Viewed through a social determinants and human rights lens,3,23 lower death literacy among some CALD groups should not be interpreted solely as an individual deficit, but as a marker of structural inequities in access to information, culturally safe services, and perhaps opportunities to participate meaningfully in end-of-life decision-making.4,7,8 In this sense, death literacy functions as both a community-level and system-level indicator of equity in end-of-life care. The co-designed, community-led model tested here, grounded in partnerships with community organisations and supported by a train-the-trainer approach, aligns with rights-based, community empowerment, and culturally competent care frameworks that emphasise participation, self-determination, and shared control over how topics such as death and dying are discussed and acted upon.4,8,9,23 These results also point to the value of localised, community-informed strategies in improving public understanding and engagement with end-of-life issues. We emphasise that a shared language or birth country does not necessarily equate to shared cultural norms or beliefs, or indeed a shared sense of community, highlighting the need to work closely with each self-identified community to ensure that the study has meaning and relevance for participants. Leonard et al. 24 reported that people speaking English at home had higher DLI scores compared to people speaking other languages. It is thus timely to investigate death literacy among those from diverse cultural and linguistic backgrounds in Australia, to determine whether gaps exist in death literacy levels across and within communities.
Implications for research and practice
The findings from this study support the positioning of death literacy as a community health indicator, reflecting not only individual knowledge and preparedness but also the collective capability and structural conditions that enable communities to engage confidently with end-of-life care. Subscale-level insights, therefore, have dual value: they guide intervention targeting and also highlight where system-level barriers may drive lower literacy, signalling areas for public health strengthening.
There was considerable overlap across the four CALD groups in terms of intervention activities selected and actioned. However, we acknowledge that each community’s experiences varied, being tailored to meet their needs and preferences. Whilst representing a fundamental design element allowing for culturally tailored approaches (representing a strength of this study and aligned with the study philosophy), this may have introduced variability and outcome differences in the survey results. This may have stemmed in part from cultural responsiveness, in addition to variability in the implementation strategies and capacities for each community leader. However, it is not the authors’ intention that the engagement with and interventions within communities in this study are exactly replicable, but that they illustrate the feasibility of working with existing strengths and capacities of underserved communities to achieve valued outcomes, here, to enhance death literacy within each community.
From a theoretical standpoint, this flexibility is consistent with community capacity and empowerment approaches, which prioritise responsiveness to local priorities. It also reinforces the need to interpret effect sizes cautiously and to pair quantitative results with qualitative insights about how different communities experience and enact death literacy over time.5,9,23
Strengths and limitations of this study
This study has several strengths, including a co-designed, community-led intervention model across four CALD communities, use of professionally translated consent documents and measures, and rigorous psychometric assessment of the DLI alongside pre–post evaluation. However, it is important to highlight that sample sizes for matched pairs within cultural groups were small and uneven (ns = 16–30), which may have reduced power to detect small-to-moderate effects and limited the precision of subgroup comparisons. Consequently, findings should be interpreted as indicative of changes within the participating communities rather than generalisable to all Filipino, German, Italian, or Vietnamese populations.
Although there was considerable overlap across the four CALD groups in terms of intervention activities these varied, being tailored to meet their needs and preferences. Although this flexibility is consistent with community capacity and empowerment approaches, it may have introduced variability in survey results that complicates direct comparison across groups. Finally, total-scale CFA fit was consistently poor across groups, indicating that the DLI is best interpreted as a multidimensional measure at the subscale level rather than a unidimensional scale, which also constrains how total scores should be interpreted.
Notwithstanding professional forward translation of the DLI survey by accredited translators, individual and group feedback from several community leaders indicated that many participants found the various content and concepts unfamiliar, taboo, or challenging to understand, even in their preferred languages, over and above accuracy of translations. Some leaders reported that community members were sometimes initially reluctant to engage in discussions relating to death and dying, describing the content as uncomfortable or emotionally distressing. However, they further indicated that they believed the interventions enhanced confidence in the topic, helping to normalise such conversations within their communities; this may not have been captured by the re-administration of the DLI survey.
The authors acknowledge that opting to utilise a forward-translation method for the DLI tool (i.e. from English to German, Italian, Filipino, or Vietnamese) may have generated less valid translations, impacting study results. Future studies should include strategies to improve the validity of the translation, such as forward–backward translation and further co-creation with community leaders. These strategies may have also minimised what our community leaders reported as unfamiliarity of the concepts and language within the DLI. Partly to clarify these, and partly as some community members struggled to read the survey (due to either literacy or health challenges), some community leaders reported helping participants to complete the survey. Other studies using the DLI have adopted similar approaches, particularly with respect to illiterate participants, as Semerci Çakmak et al. 7 outlined in their paper reporting on Turkish adults. Although the extent of assistance across T1 and T2 is unknown, it is plausible that more assistance was provided at T1 (as this was consistently completed during an initial community workshop). In addition, such assistance likely varied across time and groups, all of which may have influenced results; future researchers could document such procedural issues more closely.
For future research, the use of the DLI-R 20 is recommended, as it incorporates simplified language aimed at improving clarity and comprehension for a lay audience; as such, it may lead to less confusion among prospective participants, especially among those for whom English proficiency is a concern.
Conclusion
To the authors’ knowledge, this study is the first to investigate the impact of community-delivered and designed interventions on death literacy among CALD populations. While variations were observed across groups, participants generally exhibited gains in knowledge of community resources and support systems related to end-of-life care within each community. The findings reinforce that resourcing and partnering with communities is critical for meaningful engagement, particularly when addressing sensitive topics such as death and dying. More broadly, this work highlights the need to adapt interventions as well as mode of delivery and content of measurement tools to ensure cultural responsiveness and accessibility. Importantly, by linking subscale-level psychometric performance with observed intervention effects, the study demonstrates how reliable domains of the DLI can sensitively track community change and inform the cultural refinement of weaker domains. These findings clarify the DLI’s value as both a research and evaluation tool capable of guiding targeted improvements in programmes and monitoring community-level readiness and equity in end-of-life care. These results further demonstrate the potential for community-driven approaches to strengthen death literacy within diverse populations in Western contexts, where ageing societies and increasing cultural diversity highlight the importance of ensuring end-of-life information is provided in ways (where, how, and by whom) that are preferred by and accessible to all. At a policy level, embedding death literacy education and DLI-based indicators within multicultural and national palliative care frameworks may support earlier engagement with services and advance equity in end-of-life outcomes. These findings contribute to the growing body of death literacy research in Australia and internationally, adding an important cultural lens to this work.
Footnotes
Acknowledgements
We acknowledge the valuable contribution of the dedicated community leaders representing four community groups (Filipino, German, Italian, and Vietnamese), including Miriam Cocking, Dina Solis, Virginia Cossid-Short, Antje Richter, Dianne Bradey, Sharon Subera, Rosalia Panepinto, Trang Do, and Nga Dang. Their commitment, cultural insight, and sustained engagement were fundamental to the success of this project. We also gratefully acknowledge the support provided by Shyla Mills, CEO of PCSA.
Ethical considerations
The study was approved by the institutional University Human Research Ethics Committee (protocol ID H-2022-155).
Consent to participate
Participants gave written consent to take part in this study and for their aggregated data to be reported in publications.
Author contributions
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by funding from The Hospital Research Foundation.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data availability statement
In line with participant consent, data were accessible solely to the research team. Requests for further information or access may be directed to Associate Professor Ali Lakhani at