Introduction

Apathy is common in neurological and psychiatric diseases [1] and is consistently linked to poor prognosis [2,3], high caregiver burden [4,5], faster cognitive and functional decline [6,7], and reduced quality of life [8]. Impulsivity is equally common, and can manifest as risky decisions, falls, excessive gambling, hyper-sexuality, inappropriate social conduct, and binge eating. Impulsive behaviours are difficult to manage, cause significant carer distress [9] and are associated with a poor prognosis [2,3]. Apathy and impulsivity are especially prevalent features of syndromes associated with frontotemporal lobar degeneration (FTLD) and frequently coexist [10–12]. Syndromes of FTLD include a variety of neurological conditions, here referencing behavioural variant frontotemporal dementia, primary progressive aphasias, progressive supranuclear palsy, and corticobasal syndrome.

Despite increased awareness of the negative impact of these complex behavioural changes, they remain under recognised and poorly treated [13–15]. There are a number of potential reasons for this, including challenges associated with current clinical rating scales [16]. Research exploring the cognitive underpinnings of apathy in terms of reward [1], effort [17], and active inference [18], and the neural underpinnings of apathy in terms of fronto-striatal circuitry [19–21], currently rely on questionnaires and clinical rating scales to validate potential mechanisms. This challenge is mitigated somewhat for impulsivity by well-established behavioural paradigms of disinhibition, such as the ‘Stop-Signal task’ [22,23]. However, self-report tools remain a simpler alternative that can be scaled to large populations, and more easily assess a wide array of impulsive behaviours. Well-designed scales also permit the inclusion of more nuance in the apathy and impulsivity literature, allowing for the consideration of separable domains that often display dissociable mechanisms [24–26]. Appropriate and meaningful scales of apathy and impulsivity are therefore crucial in advancing our understanding of the mechanisms underlying these constructs, and identifying routes to potential treatments. However, there are several limitations to the scales available.

First, there is a lack of appropriately targeted, disease-specific assessment tools to quantify apathy and impulsivity within the FTLD population. Available assessment tools for apathy and impulsivity are widely-used and well-established in other disease groups [27–29], but may not be appropriate for people affected by FTLD-syndromes, especially in view of the evidence for distinct apathy and impulsivity profiles within this population [30,31]. A disease-specific measure should be developed and validated within the intended target population, capturing concepts (symptoms and impacts) that are considered relevant and meaningful to those living with the disease [32]. Previous questionnaires have sought to do this, e.g., the Dimensional Apathy Scale [33] designed with motor dysfunction in mind. This provides an option when measuring apathy in conditions such as Parkinson’s disease [34] and ALS [35] where motivation assessments may be confounded by motor deficits. Such measures are rare in FTLD-associated syndromes, particularly with regards to impulsivity, limiting meaningful engagement with patients and hindering both assessment and treatment of these debilitating symptoms. For example, in the Barratt Impulsivity Scale [36], work-related questions such as “I change jobs” or “I plan for job security” are arguably irrelevant in this disease context as many patients are retired or otherwise unable to work. Other items, such as “I squirm at plays or lectures” contain low frequency words (“squirm”) which are difficult for some people with FTLD to understand [37,38]. Assessment tools should be accessible to the patient population and, given the increased recognition of heterogeneity within and between FTLD-associated syndromes [39], the use of a transdiagnostic scale to capture symptom commonalities across disease groups could be particularly beneficial for targeted assessment tools.

Second, apathy and impulsivity are often assessed in isolation, despite evidence for their coexistence. People with FTLD-associated syndromes can be both apathetic and impulsive [10,11,21], suggesting overlapping domains within these multifaceted behavioural constructs. Traditional dopaminergic theories which place these constructs at opposite ends of a motivational spectrum have recently been questioned in conditions associated with FTLD, due to the positive correlation between apathy and impulsivity and the limited clinical efficacy of dopaminergic medication in treating apathy [13]. Dopamine deficits and apathy are related but not synonymous [40], nor incompatible with impulsivity. Measurement tools for these symptoms would therefore benefit from sensitivity to both apathy and impulsivity.

Finally, patient-reported measures may be unreliable in FTLD populations due to characteristic impairments in cognition [41,42] that in many cases lead to lack of insight, likely contributing to discrepancies observed between patient- and caregiver- reported assessments [11,43,44]. It is particularly challenging in progressive neurodegenerative conditions to determine the point at which insight is lost. People with FTLD-associated disorders may also engage in different ways with self-report leading to responses that can be characterised as ‘careless’ [45] or internally inconsistent [42]. While the patient perspective is critical in identifying important and/or bothersome symptoms and impacts of their disease, caregiver-reported measures may more reliably capture change in these symptoms over time. Assessment of these behaviours and their underlying neural correlates may therefore be more appropriately captured by carer- rather than patient-reported measures in the symptomatic stages of disease [11,46]. This is reflected in the development of other questionnaires, such as the Dimensional Apathy Scale [33], which has both a self and caregiver rated version, as well as scales which rely on structured interviews and clinical ratings [47].

In this study, we aim to address existing challenges by developing a novel, FTLD-specific, carer-rated questionnaire for apathy, impulsivity, and related behavioural change: the Cambridge Questionnaire for Apathy and Impulsivity Traits (CamQUAIT). We aim to provide a valid and reliable assessment tool capturing concepts that are relevant and meaningful to people living with FTLD. We employed Rasch analysis, and psychometric modelling, to empirically test measurement properties of this novel assessment tool within the context of FTLD.

Methods

Results

Cohort

Demographic and clinical characteristics of the development cohort are presented in Table 1 alongside details of 11 healthy controls used in analysis.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Demographics and clinical characteristics of the CamQUAIT cohort.

https://doi.org/10.1371/journal.pone.0345545.t001

Construct validity

CamQUAIT-M: motivation and support.

A Person Separation Index and Cronbach’s alpha of 0.9 indicated good separation of items along the construct and sufficient power to discriminate between 4 groups of respondents [55].

The scale showed good overall fit to the Rasch model in terms of item-person interaction statistics (item mean = 0.297, SD = 1.898; person mean = −0.155, SD = 1.274), although the item-trait interaction was significant (χ² = 96.625, p < 0.001) indicating deviation from the model expectations. Item 7 “has an intense approach to life” and item 13 “spends his/her day doing things of interest to him/her” were removed due to high fit statistics of 3.03 and 4.36 respectively (>±2.5), and highly significant Chi squared values (p < 0.001). Item 20 “is interested in having friends” showed differential item functioning by gender and was removed following an unsuccessful attempt to split the item (which can often result in the item fitting the overall item set). Item 11 “is interested in doing new things” demonstrated high residual correlation with item 19 “likes to learn new things”. After examining improvement in fit following the removal of each item in turn, item 11 was removed.

This resulted in an overall fit to the Rasch model, including good item-person fit statistics (item mean = 0.072, SD = 1.269; person mean = −0.200, SD = 1.093), a non-significant item-trait interaction (χ² = 27.5, df = 18, p > 0.05), and a high Person Separation Index of 0.9 (see Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Rasch model statistics for the two subscores of the CamQUAIT.

https://doi.org/10.1371/journal.pone.0345545.t002

Unidimensionality of the reduced item set was assessed using principal components analysis of the person residuals. The first residual accounted for 21% of the variance, once the “Rasch” factor was removed. The item set showed acceptable unidimensionality [56], with significant differences between the two item subsets at the 5% level for 9/128 (7%) persons with a lower 95% confidence interval proportion of 0.033 (<0.05), omitting test-based extremes.

CamQUAIT-C: challenging behaviours.

A Person Separation Index of 0.7 and Chronbach’s alpha of 0.8 indicated good ability to statistically differentiate between 2–3 groups.

The scale showed acceptable person fit statistics (mean = −0.164, SD = 0.950) but high item standard deviation (mean = 0.100, SD = 1.890) and a significant item-trait interaction (χ² = 75.460, df = 18, p < 0.001). No items displayed significant differential item functioning by gender. Two items showed strong misfit with high fit residuals and significant χ² values and were removed – i.e., item 18 “is unenthusiastic about his/her usual hobbies” (Fit Res = 3.22, χ² p < 0.05) and item 16 “is tearful or cries” (Fit Res = 2.94, χ² p < 0.001). Item 14 “has a poor understanding of his/her problems” was removed after theoretical examination and consideration of its disordered response structure, resulting in a significant improvement in fit (item mean = 0.027, SD = 0.992; person mean = −0.222, SD = 0.872) and a non-significant χ² of 10.6 (p = 0.560). Note that rescoring item 14 did not significantly improve fit.

The reduced item set successfully met the t-test protocol for unidimensionality, with the first residual accounting for 35% of the variance and significant differences between the two item subsets at the 5% level for 7/119 (5.9%) persons with a lower 95% confidence interval proportion of 0.020 (<0.05), omitting test-based extremes (see Table 2).

Scale reconstruction and scoring

The final item set consisted of 15 items, 9 pertaining to the motivation and support subscore (CamQUAIT-M) and 6 to the challenging behaviours subscore (CamQUAIT-C). The final item set and accompanying scoring sheet can be found in the S1 File.

Item scoring was retained with 4 categories: Never [0], Sometimes [1], Often [2] and Always [3]. Although some items showed some disordered thresholds, a larger sample size is required to ensure efficient utilisation of all response categories. Maintaining scoring structure facilitates further data collection and the consolidation of all datasets.

Table 3 includes the score to logit to severity conversions. The Person Separation Index and Cronbach’s alpha of CamQUAIT-M indicated ability to dissociate between 4 groups, but for consistency, severity levels for both subscores were split into 3 categories based on the lower Person Separation Index and Cronbach’s alpha of CamQUAIT-C. Severity categories ranged from Mild to Moderate to Severe, with 9 levels per category for CamQUAIT-M and 6 levels for CamQUAIT-C.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. CamQUAIT conversion between section scores, logit and proposed severity rating.

https://doi.org/10.1371/journal.pone.0345545.t003

The CamQUAIT-M subscale (items 1, 3, 5, 8, 11–15) is reversed for valence. For ease of scoring the subscale of each question is also indicated on the questionnaire.

CamQUAIT performance by group and disease severity

Significant differences in performance on the CamQUAIT subscores were observed across diagnostic groups, with the motor neuron disease cohort scoring lowest on average and the bvFTD cohort receiving the highest scores (see Fig 1).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Box plot of scores from the two subscales of the CamQUAIT across disease group.

(A) Shows scores on the motivation and support subscale (CamQUAIT-M), while (B) shows scores on the impulsivity and challenging behaviours subscore (CamQUAIT-C). PSP = Progressive Supranuclear Palsy, CBS = Corticobasal Syndrome. PPA = Primary progressive Aphasia, bvFTD = behavioural variant Frontotemporal Dementia, FTD-MND = Frontotemporal Dementia with Motor Neuron Disease.

https://doi.org/10.1371/journal.pone.0345545.g001

The CamQUAIT subscores “motivation and support” (CamQUAIT-M) and “challenging behaviours” (CamQUAIT-C) were positively correlated across groups (r = 0.376, p < 0.001) but met our initial criteria of <20% shared variance (14% shared variance).

There was a positive relationship between the CamQUAIT and other widely used measures of behavioural change, such as the Revised Cambridge Behavioural Inventory (CamQUAIT-M: r = 0.663, p < 0.001; CamQUAIT-C: r = 0.595, p < 0.001) and the Apathy Evaluation Scale (CamQUAIT-M: r = 0.909, p < 0.001; CamQUAIT-C: r = 0.376, p < 0.001). This correlation was stronger when comparing the CamQUAIT-M with measures of motivation and apathy than challenging behaviours, and vice versa for the CamQUAIT-C.

Discussion

The Cambridge Questionnaire for Apathy and Impulsivity Traits (CamQUAIT) is a novel, targeted measurement tool to assess apathy, impulsivity, and related behavioural change in the context of FTLD-related disorders. The scale demonstrates good measurement properties, meeting the assumptions of unidimensionality under the Rasch model and showing good overall sensitivity to change in apathy and impulsivity across the FTLD spectrum. The CamQUAIT may therefore aid in overcoming current hurdles to assessing apathy and impulsivity within this disease group, and allow for studies to more meaningfully investigate cognitive and neural mechanisms which underpin these symptoms in FTLD.

The CamQUAIT has several potential advantages over existing measures. Firstly, it is a short and simple assessment tool which demonstrates empirically tested psychometric validity for the assessment of apathy and impulsivity in the context of FTLD-related disorders. Unlike some other clinical rating scales, such as the Lille Apathy rating scale [47], the CamQUAIT can be completed by a caregiver without the need for clinical training, or a structured interview. Most carers were able to complete the CamQUAIT easily, and the most frequently unmarked item from each subscore was removed during the Rasch processing steps. Overall, items had a low number of missing data, ranging from 0–5 unmarked items across the 132 respondents (<4%). The scale demonstrated good construct validity, correlating with other widely used measures of behavioural change such as the Cambridge Behavioural Inventory and Apathy Evaluation Scale. The CamQUAIT can therefore provide a simple option for apathy and impulsivity measurement, validated in the context of FTLD disorders, which can be readily completed by caregivers during a research visit, as seen in practise with recent clinical trials of atomoxetine in individuals with PSP [57].

In contrast to some alternative assessment tools which are widely used – but developed principally for psychiatric or healthy populations – the CamQUAIT contains items that are relevant at face value to people with advancing cognitive and motor disability. As with the Dimensional Apathy Scale [33], this should reduce the influence of confounding deficits when assessing apathy and impulsivity, allowing for a measure which is both more accessible and has higher face validity in studies focussing on individuals with syndromes associated with FTLD. The logit scoring system provides a measurement scaling system for assessment, examining the trade-off between respondent “ability” (behaviour, personality traits, etc.) and item difficulty, placing items and people along the same continuum by converting raw data into equal interval logit scores. This approach has gained traction in some groups, scoring individuals based on their ability level, while items are scored based on their difficulty, according to performance of the whole sample. By meeting the assumptions of the Rasch model, one can be confident that the items in a given questionnaire assess the same latent trait within the chosen population.

The CamQUAIT item set was derived from data collected from FTLD patients’ informants only, making it a targeted assessment. Carers of people with bvFTD consistently reported the highest scores while carers of people with motor neuron disease report the lowest. CamQUAIT-M revealed strong endorsement of motivational deficits relating to everyday activities and interactions with friends across diagnostic groups, in line with previous studies [11]. In contrast, the CamQUAIT-C revealed predominance of challenging behaviours in bvFTD and PSP groups, consistent with previous findings and the broader literature in both bvFTD [41,58,59] and PSP [60–62], for which this category of symptoms also form part of the diagnostic criteria. Indeed, the latest diagnostic criteria for PSP recognised impulsivity as a core feature [50]. The scale also effectively captured the full spectrum of severity from normal to severe, indicated by good item-person fit of the scale subscores. Moreover, the scale remains applicable transdiagnostically as all groups scored higher on average than controls, indicating endorsement of these behaviours across groups to varying extents. This suggests that the CamQUAIT can be used in research exploring neuropsychiatric symptoms of specific disease cohorts, as well as studies which seek to collapse across diagnostic boundaries. This is useful given the increasing influence of transdiagnostic practises, particularly in FTLD where borders between diagnostic groups are frequently blurred [39] and collapsing across groups may be necessary to power clinical trials [63].

Finally, although the CamQUAIT-M and CamQUAIT-C subscores were positively correlated, the correlation was sufficiently low (<15% shared variance) to justify two subscores and supported by high multidimensionality when including all items in a single Rasch analysis. This places the CamQUAIT in a position to facilitate the investigation of both apathy and impulsivity arising from FTLD and related disorders [11,19,24], potentially driving forward new approaches which prioritise explaining these symptoms as two sides of a motivational coin, rather than two distinct and separate entities. This may also drive the discovery of new treatment options which are able to target apathy and impulsivity concurrently, with methylphenidate currently showing promise in the treatment of apathy [64], whilst already being approved for the modulation of disinhibition in ADHD [65]. Depending on the research question, the CamQUAIT may therefore supplement dimensional or patient-reported measures of apathy and impulsivity, or stand alone as a validated alternative to other scales. In particular, the CamQUAIT may help to identify common underlying mechanisms behind the co-occurrence of apathy and impulsivity in clinical cohorts.

This study has several limitations. A rescoring process was not repeated to identify and resolve residual misfit. In view of the relatively small sample size, particularly at a diagnostic group level, it was considered that there was insufficient data to warrant further item rescoring due to the risk of categories working ineffectively. A larger sample may result in appropriate endorsement of all response categories, and we therefore prioritised the maintenance of a simple and consistent scoring structure. In addition, not all psychometric properties were evaluated in this study, nor was direct input from patients and caregivers collected in item development, for example on ‘meaningfulness’ or ‘impact’ of items and their underlying behaviours. The latter could help to confirm the meaningfulness of the CamQUAIT for those living with FTLD-related syndromes and be a useful future step.

In summary, studies assessing the benefit of symptomatic treatments for apathy and impulsivity in FTLD-associated syndromes should use appropriate tools with established measurement properties for the population in question. In this paper, we demonstrate that the novel 9-item and 6-item CamQUAIT subscores meet the Rasch model expectations and demonstrate good validity for the assessment of apathy and impulsivity in these syndromes. We hope that the CamQUAIT will be a useful addition for studies evaluating the benefit of potential new symptomatic treatments for apathy and impulsivity in FTLD-associated disorders.

Supporting information

Acknowledgments

The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care. For the purpose of open access, the authors have applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission.