Background

With the recognition of hoarding disorder (HD) as a distinct diagnosis in DSM-5, several measures have been developed to capture hoarding symptoms, independent of obsessive-compulsive disorder (OCD). However, there remains a dearth of research on hoarding symptoms measured cross-culturally. With respect to the U.S., we lack understanding of the manifestation of HD symptoms in Hispanic/Latino populations (H/Ls), the largest and fastest-growing minority group. Before we begin to investigate potential cultural differences in the risk and vulnerability for HD, we must first understand cultural and linguistic variability in the measurement of the HD construct. In the present study, we aimed to address this gap in the literature by testing differential item functioning (DIF) of Spanish and English versions of the Hoarding Rating Scale (HRS), a widely-used measure of hoarding that maps onto the core diagnostic criteria for HD.

Method

English or Spanish speaking participants (N=623) were recruited through Amazon’s Mechanical Turk (Mturk) system and completed respective versions of the HRS (English n = 450; 45.6% female; M age = 34.0 years; Spanish n = 173; 60.0% female; M age = 32.2 years; 100% H/L). We will use R’s ‘lordif’ package to test DIF for each item of the HRS across the two language groups.

R packages used: itemanalysis, lavaan, semTools, lordif

Run DIF Analyses

We specify results=“hide” to tell rmarkdown to include the code but not print the results. This was selected here as the results of this function will print hundreds of iterations of these models, which are used to determine if DIF exists.

HRSdif <- lordif(HRSdat[c(1:5)], HRSdat[,6], criterion = "Chisqr",alpha=.01, minCell = 5)

Summarize Results of DIF Analyses

datatable(HRSdif$stats, extensions = c('FixedColumns','FixedHeader'), options = list(pageLength=10, dom = 'tip', scrollX = TRUE, scrollY=TRUE, fixedHeader=TRUE, fixedColumns = list(leftColumns = 2),
  initComplete = JS("
    function(settings, json) {
      $(this.api().table().header()).css({
        'background-color': '#000',
        'color': '#fff'
      });
    }")
))

The significant chi square test for models 1 and 3 indicates omnibus (i.e., overall) DIF for items 4 and 5, which represent distress and impairment, respectively. The significant chi square tests for Models 1 and 2 indicates that there is uniform DIF for items 4 and 5. There was non-uniform DIF for item 4 only, as indicated by the significant chi square test for Models 1 and 3 for this item.

Notably, all items had the number of reponse options reduced from 9 to 7 due to sparseness.

Visualize Results of DIF Analyses

plot(HRSdif, labels = c("English", "Spanish"))
Trait distributions indicate that although there is broad overlap in self-reported hoarding in English and Spanish respondents, English respondents generally reported lower HRS scores.

Trait distributions indicate that although there is broad overlap in self-reported hoarding in English and Spanish respondents, English respondents generally reported lower HRS scores.

Item 4, which measures distress due to clutter, difficulty discarding, and excessive acquiring, exhibits non-uniform DIF. Upper left: ICCs for English and Spanish groups. Upper right: absolute difference in ICCs for English and Spanish groups. Bottom left: item response functions for each response category by language group. Bottom right: difference in ICCs weighted by score distribution for focal group (Spanish), indicating some impact just above average scores.

Item 4, which measures distress due to clutter, difficulty discarding, and excessive acquiring, exhibits non-uniform DIF. Upper left: ICCs for English and Spanish groups. Upper right: absolute difference in ICCs for English and Spanish groups. Bottom left: item response functions for each response category by language group. Bottom right: difference in ICCs weighted by score distribution for focal group (Spanish), indicating some impact just above average scores.

Item 5, which measures impairment due to clutter, difficulty discarding, and excessive acquiring, exhibits uniform DIF. Upper left: ICCs for English and Spanish groups. Upper right: absolute difference in ICCs for English and Spanish groups. Bottom left: item response functions for each response category by language group. Bottom right: difference in ICCs weighted by score distribution for focal group (Spanish), indicating that the impact of DIF is mainly at higher levels of hoarding.

Item 5, which measures impairment due to clutter, difficulty discarding, and excessive acquiring, exhibits uniform DIF. Upper left: ICCs for English and Spanish groups. Upper right: absolute difference in ICCs for English and Spanish groups. Bottom left: item response functions for each response category by language group. Bottom right: difference in ICCs weighted by score distribution for focal group (Spanish), indicating that the impact of DIF is mainly at higher levels of hoarding.

Test characteristic curve (TCC) plots for all items (left) and DIF items (right). On DIF items, Spanish-speaking adults would score slightly lower if language group-specific item parameter estimates were used to score HRS levels.

Test characteristic curve (TCC) plots for all items (left) and DIF items (right). On DIF items, Spanish-speaking adults would score slightly lower if language group-specific item parameter estimates were used to score HRS levels.

Individual-level DIF impact. Left: in most cases, accounting for DIF led to marginally higher scores, although the overlal impact appears to wash out when viewed as an average Right: Accounting for DIF led to slightly lower scores for Spanish speaking participants and slightly higher scores for English participants. These groupwise differences were relatively consistent across levels of theta (i.e., levels of hoarding severity).

Individual-level DIF impact. Left: in most cases, accounting for DIF led to marginally higher scores, although the overlal impact appears to wash out when viewed as an average Right: Accounting for DIF led to slightly lower scores for Spanish speaking participants and slightly higher scores for English participants. These groupwise differences were relatively consistent across levels of theta (i.e., levels of hoarding severity).