ACTIVLIM-Hemo: A new self-reported, unidimensional and linear measure of activity limitations in persons with haemophilia

Abstract

Introduction: To assess activity limitations in people with haemophilia (PwH), the self-reported Haemophilia Activity List (HAL) is widely employed, despite several methodological limitations impacting the interpretation of categorical scores. Modern psychometric approaches avoid these limitations by using a probabilistic model, such as the Rasch model. The ACTIVLIM is a Rasch-built measurement of activity limitations previously validated in several clinical conditions like neuromuscular disorders.

Aims: This study sought to develop the ACTIVLIM-Hemo, meaning an ACTIVLIM scale version specifically adapted to assess daily activity limitations in adult PwH.

Methods: Daily activities were assessed as "impossible," "difficult" or "easy" by 114 PwH (median age of 44 years) with 63 of them reassessed after 12 days. The Rasch Rating Scale model was used to identify activities delineating a unidimensional and linear scale unbiased by demographic and clinical status. Concurrent validity was determined through correlation with the HAL sub-scores and sum score.

Results: The ACTIVLIM-Hemo included 22 pertinent activities, with difficulties independent of demographic and clinical conditions, allowing a reliable measure of activity limitations (PSI = .92) expressed on a linear and unidimensional scale in PwH (7%-100 % range, ceiling effect of 1/114) with excellent test-retest reliability (ICC = .978). Spearman rank correlations between ACTIVLIM-Hemo and HAL sub-scores ranged between .623 and .869.

Conclusions: The ACTIVLIM-Hemo is an easy-to-administer, valid and reliable alternative to HAL in assessing activity limitations in PwH. Its invariant scale can be used across conditions and time to compare the functional status of PwH over a wide measurement range.

Keywords: bleeding disorder, disability and health; haemophilia, International Classification of Functioning (ICF), outcome assessment, questionnaire, rash analysis, scale.

FIGURE 1

The activity limitation measured using ACTIVLIM‐Hemo is presented on the abscissa graduated in % range (scale measurement range, top axis) and in logits (the pass/fail probability ratio for a subject to succeed an item, bottom axis). For both units, higher values indicate a higher activity level (right side of the figure). The distribution of activity levels in our PwH sample is shown on the top panel, where most PwH lie above an activity level of 25%, excepting two subjects with a very severe activity limitation. The difficulty hierarchy of items is shown on the item map (middle panel) listed, from top to bottom, in decreasing difficulty order. The item map shows the activity level required to succeed each item, as follows. For any item, a PwH with a severe activity limitation is expected to answer “impossible” (light grey, score of 0); a PwH with a moderate activity limitation is expected to answer “difficult” (medium grey, score of 1); a PwH with a mild or no activity limitation is expected to answer “easy” (dark grey, score of 2). The item map displays the thresholds that separate consecutive response categories, being located from −8.5 to +8.5% range (−1.074 to +1.074 logits) in regard to the difficulty of each item. The item map also shows that the distance between thresholds is similar across activities, lying within 2 SD of the mean distance for all items. The conversion of raw ordinal scores to linear measures is shown on the bottom panel across the whole measurement range (from a score of 0 indicating that all items are “impossible” to a score of 44, i.e., 22*2, indicating that all items are “easy”). Given the S‐shaped relationship, a gain of 1 score point corresponds to different activity level changes across the measurement range. At the centre of the measurement, a change score from 21 to 22 results in difference of 1.4% range (.18 logits) in activity limitation, while at the measurement range's extremity, the same change score, from 1 to 2, results in an over three‐fold difference in activity of 5.2% range (.66 logits), highlighting the non‐linearity of the scores.

FIGURE 2

Differential Item Functioning (DIF) plots showing the difficulty of the 22 ACTIVILIM‐HEMO items when estimated on sub‐samples split according to sex different criteria: age split on the median of 44 years; body mass index (BMI) split on the median of 25.4; lower limb (LL) arthropathy (split on the presence of at least one LL joint with a HJHS above the median score of 2), LL surgery split on the presence of at least one surgery at the LL; pain split on the median BPI score of 1.8; time split on T1 versus T2. In each plot, more difficult items are plotted in the top/right part of the graph. Most items lie within the 95% confidence intervals (CI, solid lines) of the ideal invariance, indicating that the ACTIVLIM‐Hemo provides an invariant measure of activity limitation unbiased by either demographic or clinical conditions. Outliers are identified by their label.

FIGURE 3

Test‐retest reliability illustrated via the individual approach (left panel) following a median delay of 12 days. The activity level for each PwH (dots) is compared between T1 and T2, where higher activity levels are plotted in the top‐right part of the graph. All measures lie within the 95% confidence interval (CI) of the ideal equivalence over time. The intraclass correlation coefficient (ICC = .978) also confirms an excellent test‐retest reliability at group level. The Bland and Altman plot (right panel) shows an overall agreement at both times based on an average difference of .35% range (plain line), while most differences lie within the 95% confidence interval (CI, dashed lines) placed at 1.96 times, the SD (4.86% range) being around the mean difference.

FIGURE 4

Correlation between the linear measure of ACTIVLIM‐Hemo (abscissa), ordinal HAL sum score and sub‐scores and HAL short form (ordinates) for each PwH (dots). Spearman's rank order correlation coefficients are presented for each HAL sub‐score and for the short form.

FIGURE 5

A sample scoring form showing the evaluation of a subject with haemophilia. The online scoring form is presented to the subject in random orders to avoid any systematic effect of the item presentation sequence. The score interpretation (bottom image) shows items ordered from most (top) to least (bottom) difficult, where possible responses to each item are identified by the corresponding score (0 for “impossible,” 1 for “difficult” and 2 for “easy”) as a function of the subject's activity limitation expressed in logits. The red vertical line indicates the subject's activity limitation of −.27 logits (49.4 %range); the dashed vertical red lines indicate its 95% confidence interval between −1.16 and .61 logits (42.4–56.5 %range). The figure confirms that, given the subject's activity limitation, the expected response is observed for most items of the ACTIVLIM‐Hemo questionnaire (responses in blue squares), hence confirming the coherence of the subject's responses. The response to the item “Bicycling in the woods” is slightly unexpected (responses in yellow square) since the subject reports it as “easy” while it should be “difficult” or “impossible” given the 95% confidence interval of the subject's activity limitation. Note that the items “Skiing” and “Doing a sport more than once a week” are not answered, yet not compromising the interpretation of the subject's activity limitation.