What does the Dementia Rating Scale-2 measure? The relationship of neuropsychological measures to DRS-2 total and subscale scores in non-demented individuals with Parkinson’s disease

Abstract

Objective:

The Dementia Rating Scale-2 (DRS-2) is recommended for assessing global cognition in Parkinson’s disease (PD) by the Movement Disorder Society. However, empirical evidence is limited regarding the degree to which the DRS-2 corresponds to traditional neurocognitive domains (i.e., construct validity) in PD. Thus, this study aims to determine the construct validity of the DRS-2 in a non-demented sample of PD patients.

Method:

Patients with PD (n = 359; mean age = 64.50 ± 8.53, education = 14.97 ± 2.73, disease duration = 8.48 ± 4.87, UPDRS Part III motor scale scores = 25.23 ± 10.17) completed the DRS-2 as part of a comprehensive neuropsychological assessment consisting of attention/working memory, executive function, language, delayed recall, and visuoperceptual-spatial skills. Bootstrapped bias-corrected Spearman rho’s correlations andhierarchical linear regressions were performed to examine construct validity of DRS-2 total and subscale scores.

Results:

Speeded measures of set-shifting, rapid word generation to letter and semantic cues, and simple visuoperceptual skills largely accounted for variance in DRS-2 total scores. Most DRS-2 subscale scores showed weak relationships with theoretically related neuropsychological measures.

Conclusions:

DRS-2 total scores reflect impairment across a range of cognitive domains (i.e., executive, language, and visuoperception), while DRS-2 subscale scores have limited construct validity. Together, the DRS-2 does not appear to have utility beyond screening for global cognition in PD.

Introduction

Parkinson’s disease (PD) is a progressive movement disorder that presents with an array of non-motor symptoms, including cognitive impairment. Prevalence estimates of cognitive impairment in the early stages of PD (without dementia) range from 19–30% (Elgh et al., 2009; Muslimović et al., 2005). The most common pattern of cognitive impairment in PD includes deficits in executive function (Woods &Tröster, 2003), visuospatial skills (McKinlay et al., 2009), and learning and memory (Higginson et al., 2003; Massman et al., 1990). As the disease progresses, up to 80% of patients will be diagnosed with Parkinson’s disease dementia (PDD; Aarsland et al., 2003a; Hely et al., 2008). Cognitive impairment in PD is associated with a host of clinical outcomes including increases in healthcare costs and caregiver burden as well as diminished quality of life and functional independence, and increased rates of mortality in patients (Aarsland et al., 2000; Hughes et al., 2004).

Assessment of neurocognitive status is important for identifying individuals at greatest risk for subsequent cognitive decline and dementia. Due to demands on neuropsychological services, the use of cognitive screeners becomes a viable first line option in this effort. Ideal cognitive screeners have short administration times, cover multiple cognitive domains, and help identify individuals at risk for cognitive decline. Among the most common cognitive screeners are the Montreal Cognitive Assessment (MoCA) (Nasreddine et al., 2005), the Mini Mental State Examination (MMSE) (Folstein et al., 1975), and the Dementia Rating Scale (DRS) (Mattis, 1988). Of these three, the DRS is the most comprehensive, though it takes the longest to administer (i.e., around 20–25 minutes). The DRS includes 24 subtests that are used to generate five subscales of Attention, Initiation/Perseveration, Construction, Conceptualization, and Memory, which summed together provide an overall index score of mental status/global cognition (Mattis, 1988). The DRS was later revised to include updated age- and education-corrected normative data and made minor refinements in administration and scoring (i.e., referred to as the DRS-2; Jurica et al., 2001). An alternate DRS form was developed to enable serial assessments and monitor global cognition over time (Schmidt, 2004).

Although the original purpose of the scale was to assess cognition in individuals with ‘organic mental syndromes,’ the DRS is commonly used to screen for cognitive status (e.g., normal cognition, mild cognitive impairment, dementia). In particular, the DRS total score is purported to adequately detect and distinguish changes in cognitive status within and between several cognitively impaired populations with various etiologies (Emery et al., 1996; Salmon et al., 1989; Paolo et al., 1995), including PD (Aarsland et al., 2003b, Noe et al., 2004; Bezdicek et al., 2015; Matteau et al., 2012; Pirogovsky et al., 2014; Villeneuve et al., 2011). In fact, the DRS is among the “recommended” tools by the Movement Disorders Society Task Force for cognitive screening in PD (Litvan et al., 2012; Skorvanek et al., 2018) due to its adequate internal consistency and excellent test-retest reliability and sensitivity to change.

Several studies have examined how traditional neuropsychological measures map onto the DRS total and subscale scores (i.e., construct validity). In other words, how well do DRS total and subscale scores represent the hypothesized cognitive construct (i.e., global cognition, memory, etc.)? In non-PD populations, several studies support the construct validity of the DRS total score with other cognitive screeners (Strauss et al., 2006), estimates of intellectual functioning (Chase et al., 1984; Smith et al., 1994), and cognitive composite performance (Knox et al., 2003). Indeed, some studies suggest adequate construct validity for most, but not all, DRS subscale scores (Marson et al., 1997; Smith et al., 1994), while other studies do not (Bobholz & Brandt, 1993; Kessler et al., 1994; Woodard et al., 1996).

In stark contrast, fewer studies have examined the construct validity of the DRS in PD. To our knowledge, only one study has examined the construct validity of the DRS total and subscale scores in a sample of non-demented PD. Brown et al. (1999) used the original version of the scale and found total scores were significantly correlated with global cognitive rank scores, which were derived from a combination of neuropsychological performance and clinical judgement. In a small subset of the sample, neuropsychological measures (i.e., Digit Span Forward, Wisconsin Card Sort Test Perseverative Responses, WAIS-R Similarities, WAIS-R Block Design WMS-Logical Memory, WMS-Visual Reproduction) were correlated with DRS subscales to test for construct validity. The authors reported adequate convergent validity for four of the five DRS subscales, including DRS Attention with Digit Span Forward, DRS Initiation/Perseveration with Wisconsin Card Sort Test Perseverative Responses, DRS Conceptualization with WAIS-R Similarities, and DRS Memory with WMS-Logical Memory Immediate Recall.

While findings from this early validation study by Brown et al. (1999) are interesting, additional psychometric investigation is needed because the sample size was small (n = 14), few neuropsychological measures were used to determine construct validity, and the DRS and neuropsychological measures draw from different normative samples. Equally important, the possible influence of psychological symptoms (i.e., depression, anxiety, and apathy) on DRS total and subscale scores or neuropsychological measures was not included. This is important given the high prevalence of psychological disorders in this population (den Brok et al., 2015, Scott et al., 2020; Dissanayaka et al., 2010; Reijnders et al., 2008).

As such, the aims of the current study were two-fold and involved a large cohort of 359 non-demented individuals with idiopathic PD who had completed the DRS-2 and a comprehensive neuropsychological evaluation. First, we examined the construct validity of DRS-2 total and subscale scores against neuropsychological measures from five cognitive domains (e.g., executive function, delayed recall, visuoperceptual-spatial, language, attention/working memory). Second, we wanted to learn which neuropsychological tests correlated most strongly with DRS-2 total and subscale scores, while controlling for demographic and disease-related variables as well as mood and motivation symptoms (e.g., depression, anxiety, and apathy).

We predicted that neuropsychological measures of executive function would be most strongly associated with DRS-2 total scores based on observations that executive dysfunction is common in non-demented individuals with PD (Kudlicka et al., 2011) and more than half the items contributing to DRS-2 total scores are from DRS-2 executive function subscales (e.g., Initiation/Perseveration and Conceptualization). Given that neuropsychological measures are process impure or comprise tasks that overlap with more than one cognitive construct, we expected DRS-2 subscales to correlate with multiple neuropsychological measures to some degree. However, when considering the tasks that contribute to the bulk of a DRS-2 subscale score, we also expected a given DRS-2 subscale to correlate most strongly with neuropsychological measures of the intended cognitive construct. Specifically, we predicted (1) attention/working memory measures would correlate most strongly with DRS-2 Attention subscale score, (2) executive function measures would correlate most strongly with DRS-2 subscales of Initiation/Perseveration and Conceptualization, (3) visuoperceptual-spatial measures would correlate most strongly with DRS-2 Construction subscale, (4) delayed recall measures would correlate most strongly with DRS-2 Memory subscale.

Method

Design

A retrospective chart review was completed on individuals with PD at the University of Florida (UF) Norman Fixel Institute for Neurological Diseases. Data included participants’ demographics, disease-related characteristics, neuropsychological assessment, and mood/motivation measures. At the time of initial assessment, all patients provided written, informed consent for their personal health information to be included in a large clinical research database maintained (CJ) at the UF Normal Fixel Institute for Neurological Disorders (INFORM). Informed consent was obtained according to the UF Institutional Review Board guidelines, and the study was conducted in accordance with the Helsinki Declaration.

Participants

Participants included a convenience sampleof individuals with idiopathic PD from a large ongoing clinical research database of movement disorder patients seen at the UF Norman Fixel Institute for Neurological Disorders (INFORM). All participants received a comprehensive neuropsychological evaluation as a part of their standard clinical care. At times, evaluations were shortened due to time availability, severity of cognitive impairment, or some combination. For the current study, inclusion criteria were: (1) evaluation between 2002 and 2019, (2) a diagnosis of idiopathic PD made by a fellowship-trained and board-certified movement disorders specialist, (3) completion of the DRS-2, (4) availability all neuropsychological measuresused in the current study and (5) disease-related measuresused in the current study.

Exclusion criteria entailed a) any current major psychiatric disturbance (e.g., active psychosis or current major depressive episode without treatment within the past two months); b) a comorbid essential tremor diagnosis; c) previous brain surgery (e.g., deep brain stimulation, pallidotomy); d) history of other neurological conditions (e.g., epilepsy, stroke, or brain injury); e) evidence of significant cognitive impairment based on scores below 125 on the DRS-2. We selected 125 as our cutoff score to exclude cases with possible dementia as scores below 125 correspond to the 6–10th percentile and lower (Jurica et al., 2001; Lucas et al., 1998).

Measures

Cognitive screener

The DRS-2 includes five subscales: Attention (e.g., digit span, detecting A’s; 37 points possible), Initiation/Perseveration (e.g., performing alternating movements, copying related patterns, semantic fluency; 37 points possible), Construction (e.g., copying designs; 6 points possible), Conceptualization (e.g., similarities; 39 points possible), and Memory (e.g., sentence recall, orientation, word recognition, design recognition; 25 points possible). An overall total score is derived from all five subscales (144 points possible).

Neuropsychological measures

Standard neuropsychological measures included the domains of: a) attention/working memory, b) executive function; c) delayed recall; d) language, and e) visuoperceptual-spatial skills. Table 1 depicts the specific measures that were used for each neurocognitive domain in the current study. The assignment of tasks to “domains” was rationally based and consistent with previous approaches (Jones et al., 2014; Kenney et al., 2020; Ratajska et al., 2021).

Table 1.

Neuropsychological tests and cognitive composites.

Domain	Neuropsychological Tests	Raw Scores Used
Executive Function	TMT-B	Completion Time
	Stroop Test (Interference Trial)	Total Number of Correct Items
	Letter Fluency (COWAT-FAS)	Total Number of Words (all 3 Trials)
Attention/Working Memory	WAIS-III Digit Span Forward	Total Number of Points
	WAIS-III Digit Span Backward	Total Number of Points
Language	BNT	Total Correct Spontaneous Responses
	Semantic Fluency (Animals)	Total Number of Words
Delayed Recall	WMS-III Logical Memory	Delayed Total Recall
	HVLT-R	Delayed Total Recall
Visuoperceptual-Spatial	Benton JOLO	Total Number of Correct Items
	Benton FRT	Total Number of Correct Items

Notes: HVLT-R = Hopkins Verbal Learning Test – Revised (Brandt & Benedict, 2001); WMS-III = Wechsler Memory Scale – Third Edition (Wechsler, 1997); COWAT-FAS = Controlled Oral Word Association Test (Tombaugh et al., 1999); Stroop Test is the Golden version (Golden, 1978); TMT-B = Trail Making Test Part B (Reitan, 1992); Benton JOLO = Benton Judgment of Line Orientation (Benton et al., 1994); Benton FRT = Benton Facial Recognition Test (Benton & Van Allen, 1968); BNT = Boston Naming Test (Kaplan et al., 1983); Semantic Fluency (Animals) Test (Tombaugh et al., 1999).

Motivation and mood measures

Participants completed standardized rating scales of motivation and mood, including the Apathy Scale (AS; Starkstein et al., 1992), Beck Depression Inventory–II (BDI-II; Beck et al., 1996), and State Trait Anxiety Inventory (STAI; Spielberger et al., 1970). The AS (Starkstein et al., 1992) assess symptoms of apathy over the past four weeks. Total scores range from 0–42, with higher scores representing greater apathetic symptoms, and a suggested clinical cutoff of 14 or greater (Starkstein et al., 1992; Leentjens et al., 2008). The BDI-II (Beck et al., 1996) measures symptoms of depression over the past two weeks. Total scores range from 0–63, with higher scores representing greater depressive symptoms, and a suggested clinical cutoff of 14 or greater (Leentjens et al., 2000; Torbey et al., 2015). The STAI (Spielberger et al., 1970) is a measure of anxiety that includes a scale of “state” anxiety and a scale of more longstanding “trait” anxiety. Total scores for each of the two STAI scales range from 20–80, with higher scores representing greater levels of anxiety (Dissanayaka et al., 2015).

Clinical measures of Parkinson’s disease motor symptom severity

Participants received standard measures for staging the severity of their motor symptoms and disease course. These included the Unified Parkinson’s Disease Rating Scale (UPDRS; Fahn & Elton, 1987) and the Hoehn and Yahr Scale (H&Y; Hoehn & Yahr, 1998). Part III of the UPDRS is a motor examination that quantifies the type, number, and severity of motor symptoms common in PD. Higher scores reflect a greater amount of and/or more severe motor symptoms. The Hoehn-Yahr Scale is a clinician-rated scale of disease-related disability and staging that ranges from 0 (“No visible symptoms of PD”) to 5 (“PD symptoms on both sides of the body and unable to walk”). Higher scores reflect more advanced disease stage.

Statistical analyses

To examine construct validity of DRS-2 total and subscale scores, we conducted a series of bootstrapped bias-corrected Spearman’s rho correlations due to non-normality (Shapiro Wilks and Kolmogorov-Smirnov ps < .001 except for Stroop, Letter Fluency, and WMS-III Logical Memory Delayed Recall). We used the following to determine adequate convergent validity: significant correlation coefficients that were greater than 0.5 were interpreted as strong, coefficients of 0.3 to 0.5 were interpreted as moderate, and coefficients less than 0.3 were interpreted as weak (Cohen, 1988; Lopez et al., 2018). To examine the relationships between DRS-2 total and subscale scores and neuropsychological tests, we conducted separate hierarchical linear regression analyses with DRS-2 total and subscale scores serving as the dependent variable, and neuropsychological measures that were significant in the correlational analyses serving as independent variables. Age, education, sex, and UPDRS Part III motor scores were entered into Block 1 to control for demographic and disease-related variables, and BDI-II, AS, and STAI-State scores were entered into Block 2 to account for mood and motivation symptoms in all regression analyses. Owing to non-normality, all regression analyses were bootstrapped bias-corrected using 1000 samples. Since the DRS-2 and neuropsychological tasks were based on different normative samples and not all neuropsychological tasks adjusted for education, we converted raw DRS-2 total and subscale scores as well as neuropsychological tasks to z-scores using sample-specific means and standard deviations, and we controlled for demographic variables (age, education, sex) and motor severity (UPDRS Part III motor scores) in the analyses. All statistical analyses were conducted using Statistical Package for the Social Sciences (SPSS) Version 27.

Results

Sample characteristics

The initial sample size was 583 and was reduced to 359 due to inclusion/exclusion criteria: (a) missing data for all neuropsychological measures (n = 163) or (b) disease-related measures (n = 38) and (b) evidence of significant cognitive impairment based on scores below 125 on the DRS-2 (n = 23). As a group, the sample was in their mid-60s, well-educated (around 2 years college), predominately male (71.58%), and mostly non-Hispanic white (95%). Participants were generally in the early-mid stages of disease severity based on ratings using the H&Y and motor scale of UPDRS (i.e., mean ‘on medication’ score of 25.23). On average, scores on indices of apathy, depression, and anxiety were below clinical cutoffs. See Table 2 for study sample characteristics.

Table 2.

Sample characteristics (n = 359).

	M (SD)	Range
Demographics
Age (Years)	64.50 (8.53)	39–87
Education (Years)	14.97 (2.73)	12–21
Sex (M/F)	257/102	-
Race (% White)	95.0	-
Disease-Related Characteristics
Disease Duration (Years)	8.48 (4.87)	1–39
Unified Parkinson’s disease Rating Scale Part III (Motor Scale; On Medication)	25.23 (10.17)	4–65
Hoehn and Yahr Stage (%)
1.0	1.9	-
1.5	1.3	-
2.0	60.2	-
2.5	18.1	-
3.0	14.9	-
4.0	3.6	-
Dementia Rating Scale-2 (Raw Scores)
Attention	35.86 (1.21)	30–37
Initiation/Perseveration	35.87 (2.06)	26–37
Construction	5.82 (.474)	3–6
Conceptualization	36.5 (2.21)	28–39
Memory	23.01 (1.87)	16–25
Total Score	137.08 (4.32)	125–144
Mood and Motivation (Raw Scores)
Beck Depression Inventory-II Total Score	9.94 (6.65)	0–39
State Trait Anxiety Inventory-State Total Score	38.13 (11.34)	20–74
State Trait Anxiety Inventory-Trait Total Score	36.47 (10.21)	20–65
Apathy Scale Total Score	11.22 (6.40)	0–30
Neuropsychological Measures (Raw Internal Z Scores)
Trail Making Test-Part B	0.00 (1.00)	−1.28–2.33
Stroop Test (Interference Trial)	0.00 (1.00)	−2.54–3.30
Letter Fluency (FAS)	0.00 (1.00)	−2.54–2.61
Digit Span Forward	−0.00 (1.00)	−2.87–2.59
Digit Span Backward	0.00 (1.00)	−2.87–2.59
Boston Naming Test	0.00 (1.00)	−2.22–2.73
Semantic Fluency (Animals)	0.00 (1.00)	−2.68–3.58
Logical Memory Delayed Recall	0.00 (0.99)	−2.39–2.56
Hopkins Verbal Learning Test-Revised Delayed Recall	0.00 (0.99)	−2.14–1.47
Judgement of Line Orientation	0.00 (0.99)	−2.92–1.53
Facial Recognition Test	0.00 (0.99)	−3.32–1.92

Construct validity of DRS-2 total scores

Spearman’s rho correlations revealed moderate correlations between DRS-2 total score and measures of executive function (Trail Making Test B, Stroop, Letter Fluency). There were moderate correlations between DRS-2 total score and measures of delayed recall (Logical Memory Delayed Recall, Hopkins Verbal Learning Test Revised Delayed Recall) and language (Boston Naming Test, Semantic Fluency). The remaining correlations with attention/working memory (Digit Span Forward, Digit Span Backward) and visuoperceptual-spatial (Facial Recognition Test, Judgement of Line Orientation) measures were weak. See Table 3.

Table 3.

Bootstrapped Spearman correlations between DRS total score and neuropsychological measures with 95% confidence intervals (n = 359).

	Spearman’s rho	95% Confidence Interval
		Lower	Upper
Executive Function
Trail Making Test Part B	−.457a	−.550	−.369
Stroop (Interference Trial)	.422a	.340	.502
Letter Fluency (FAS)	.428a	.329	.525
Attention/Working Memory
Digit Span Forward	.235a	.124	.331
Digit Span Backward	.245a	.142	.346
Language
Boston Naming Test	.364a	.268	.454
Semantic Fluency (Animals)	.436a	.345	.518
Delayed Recall
Logical Memory Delayed Recall	.357a	.258	.447
Hopkins Verbal Learning Test-Revised Delayed Recall	.383a	.293	.464
Visuoperceptual-Spatial
Judgement of Line Orientation	.218a	.107	.319
Facial Recognition Test	.307a	.213	.397

Notes:

^ap < .001

A bootstrapped hierarchical linear regression analysis was performed to determine which neuropsychological measures accounted for the most variance in the DRS-2 total score, after controlling for demographic and disease-related variables (Block 1; age, education, sex, and UPDRS Part III motor scores), and mood and motivation symptoms (Block 2: BDI-II, AS, STAI-State scores). All raw scores from the neuropsychological exam were entered into Block 3. Results indicated that the overall model accounted for 38.6% of the variance in DRS-2 total scores, [F(18,341) = 14.36, p < .001, R² = .386]. There was no significant multicollinearity among predictors (Tolerance > .8 and Variation Inflation Factor < 1.2). Inspection of bootstrapped coefficients revealed Trail Making Test Part B (B = −.157), Facial Recognition Test (B = .127), Semantic Fluency (B = .123), and Letter Fluency (B = .122) were weakly associated with DRS-2 total scores [ΔR² = .244, p < .001]. Sex (B = .291) was also associated with DRS-2 total scores. See Table 4 for the final model with bootstrapped coefficients.

Table 4.

Final model with bootstrapped coefficients: DRS-2 total scores regressed on neuropsychological tests controlling for age, education, sex, and UPDRS Part III motor scores (n = 359).

	B	p
Age	−.001	.899
Education	.027	.137
Sex	.291	.011
UPDRS Part III Motor Scores	−.003	.540
Beck Depression Inventory-II Total Score	.012	.206
Apathy Scale Total Score	−.016	.070
State Trait Anxiety Inventory-State Total Score	.001	.898
Trail Making Test Part B	−.157	.009
Stroop (Interference Trial)	.079	.177
Letter Fluency (FAS)	.122	.024
Digit Span Forward	.047	.363
Digit Span Backward	.032	.522
Boston Naming Test	.084	.088
Semantic Fluency (Animals)	.123	.029
Logical Memory Delayed Recall	.017	.744
Hopkins Verbal Learning Test Revised Delayed Recall	.077	.155
Judgement of Line Orientation	.040	.444
Facial Recognition Test	.127	.008

Construct validity of DRS-2 subscale scores

Spearman’s rho correlations revealed weak correlations between DRS-2 Attention subscale score with Trail Making Test B, Stroop, Letter Fluency, Digit Span Forward, Digit Span Backward, Semantic Fluency, Logical Memory Delayed Recall, Hopkins Verbal Learning Test Revised Delayed Recall, and Facial Recognition Test. DRS-2 Initiation/Perseveration subscale score was moderately correlated with Trail Making Test B and Stroop, while weakly correlated with Letter Fluency, Digit Span Backward, Semantic Fluency, Boston Naming Test, Logical Memory Delayed Recall, Hopkins Verbal Learning Test Revised Delayed Recall, Judgement of Line Orientation, and Facial Recognition Test. DRS-2 Construction subscale score was weakly associated with Trail Making Test B, Stroop, Semantic Fluency, Hopkins Verbal Learning Test Revised Delayed Recall, and Judgement of Line Orientation. DRS-2 Conceptualization subscale score was correlated with all measures from the neuropsychological exam. The magnitude of all relationships was weak except for a moderate relationship emerging between DRS-2 Conceptualization and Letter Fluency. DRS-2 Memory subscale score was weakly associated with Trail Making Test B, Stroop, Letter Fluency, Digit Span Forward, Boston Naming Test, Semantic Fluency, Hopkins Verbal Learning Test Revised Delayed Recall, and Facial Recognition Test, while moderately correlated with Logical Memory Delayed Recall. See Table 5.

Table 5.

Bootstrapped Spearman correlations between DRS subscale scores and neuropsychological measures with bias corrected 95% confidence intervals (n = 359).

	DRS-2 Subscale
	Attention	Initiation/Perseveration	Construction	Conceptualization	Memory
Neuropsychological Measures
Executive Function
Trail Making Test Part B
Spearman’s rho	−.232c	−.325c	−.172c	−.245c	−.222c
95% Confidence Interval: Lower	−.333	−.423	−.282	−.348	−.324
95% Confidence Interval: Upper	−.123	−.233	−.066	−.144	−.125
Stroop Test (Interference Trial)
Spearman’s rho	.157c	.334c	.084c	.205c	.239c
95% Confidence Interval: Lower	.064	.231	−.024	.110	.141
95% Confidence Interval: Upper	.252	.434	.188	.304	.334
Letter Fluency (FAS)
Spearman’s rho	.169b	.229c	.027	.331c	.195c
95% Confidence Interval: Lower	.060	.132	−.077	.224	.093
95% Confidence Interval: Upper	.268	.326	.125	.434	.292
Attention Working/Memory
Digit Span Forward
Spearman’s rho	.163b	.101	−.022	.116a	.196c
95% Confidence Interval: Lower	.059	−.002	−.119	.004	.089
95% Confidence Interval: Upper	.262	.195	.071	.227	.292
Digit Span Backward
Spearman’s rho	.230c	.113a	.089	.130a	.096
95% Confidence Interval: Lower	.134	.000	−.012	.020	.006
95% Confidence Interval: Upper	.330	.217	.195	.247	.192
Language
Boston Naming Test
Spearman’s rho	.097	.274c	.078	.219c	.270c
95% Confidence Interval: Lower	−.014	.176	−.024	.119	.171
95% Confidence Interval: Upper	.212	.373	.184	.323	.368
Semantic Fluency (Animals)
Spearman’s rho	.145b	.348c	.145b	.255c	.225c
95% Confidence Interval: Lower	.044	.259	.047	.145	.127
95% Confidence Interval: Upper	.250	.434	.245	.355	.325
Delayed Recall
Logical Memory Delayed Recall
Spearman’s rho	.160b	.178c	.096	.172c	.320c
95% Confidence Interval: Lower	.065	.074	−.017	.074	.217
95% Confidence Interval: Upper	.250	.284	.203	.269	.416
Hopkins Verbal Learning Test-Revised Delayed Recall
Spearman’s rho	.123b	.230c	.163b	.244c	.212c
95% Confidence Interval: Lower	.022	.133	.065	.134	.110
95% Confidence Interval: Upper	.216	.324	.254	.352	.309
Visuoperceptual-Spatial
Judgement of Line Orientation
Spearman’s rho	.072	.172c	.204c	.132a	.096
95% Confidence Interval: Lower	−.023	.064	.110	.027	−.004
95% Confidence Interval: Upper	.178	.282	.298	.232	.195
Facial Recognition Test
Spearman’s rho	.177c	.198c	.067	.153b	.173c
95% Confidence Interval: Lower	.079	.098	−.045	.058	.063
95% Confidence Interval: Upper	.274	.296	.173	.246	.276

Notes:

^ap < .05,

^bp < .006,

^cp ≤ .001

Separate bootstrapped hierarchical linear regression analyses were performed to determine which neuropsychological measures accounted for the most variance in the DRS-2 subscale scores, after controlling for demographic and disease-related variables (Block 1: age, education, sex, and UPDRS Part III motor scores), and mood and motivation symptoms (Block 2: BDI-II, AS, STAI-State scores). Only scores from the neuropsychological exam that were significantly associated with a given DRS-2 subscale score were entered in Block 3. See Table 6 for all DRS-2 subscale score final models with bootstrapped coefficients.

Table 6.

Final models with bootstrapped coefficients: DRS-2 subscale scores regressed on neuropsychological tests controlling for age, education, sex, and UPDRS Part III motor scores (n = 359).

A. DRS-2 Attention Subscale Score	B	p
Age	−.009	.182
Education	.017	.448
Sex	.259	.047
UPDRS Part III Motor Scores	.008	.147
Beck Depression Inventory-II Total Score	.000	.975
Apathy Scale Total Score	−.007	.528
State Trait Anxiety Inventory-State Total Score	−.001	.910
Trail Making Test Part B	−.113	.113
Stroop (Interference Trial)	.019	.790
Letter Fluency (FAS)	−.020	.767
Digit Span Forward	.009	.893
Digit Span Backward	.199	.001
Semantic Fluency (Animals)	.001	.983
Logical Memory Delayed Recall	−.037	.566
Hopkins Verbal Learning Test Revised Delayed Recall	.016	.813
Facial Recognition Test	.043	.463
B. DRS-2 Initiation/Perseveration Subscale Score	B	p
Age	−.004	.512
Education	.000	.981
Sex	.153	.201
UPDRS Part III Motor Scores	−.004	.446
Beck Depression Inventory-II Total Score	.009	.362
Apathy Scale Total Score	−.014	.129
State Trait Anxiety Inventory-State Total Score	−.005	.346
Trail Making Test Part B	−.165	.009
Stroop (Interference Trial)	.104	.042
Letter Fluency (FAS)	.099	.047
Digit Span Backward	.004	.943
Boston Naming Test	.079	.128
Semantic Fluency (Animals)	.107	.039
Logical Memory Delayed Recall	−.105	.064
Hopkins Verbal Learning Test Revised Delayed Recall	.024	.672
Judgement of Line Orientation	−.013	.815
Facial Recognition Test	.142	.005
C. DRS-2 Construction Subscale Score	B	p
Age	−.003	.714
Education	.000	.982
Sex	.136	.328
UPDRS Part III Motor Scores	.001	.838
Beck Depression Inventory-II Total Score	−.009	.410
Apathy Scale Total Score	−.007	.494
State Trait Anxiety Inventory-State Total Score	.000	.972
Trail Making Test Part B	−.028	.691
Semantic Fluency (Animals)	.033	.598
Hopkins Verbal Learning Test Revised Delayed Recall	.106	.095
Judgement of Line Orientation	.218	.001
D. DRS-2 Conceptualization Subscale Score	B	p
Age	.003	.637
Education	.054	.012
Sex	.131	.333
UPDRS Part III Motor Scores	.001	.899
Beck Depression Inventory-II Total Score	.001	.933
Apathy Scale Total Score	.005	.633
State Trait Anxiety Inventory-State Total Score	.010	.081
Trail Making Test Part B	−.048	.496
Stroop (Interference Test)	.048	.498
Letter Fluency (FAS)	.223	.001
Digit Span Forward	.023	.710
Digit Span Backward	−.023	.702
Boston Naming Test	−.003	.963
Semantic Fluency (Animals)	.069	.301
Logical Memory Delayed Recall	−.086	.180
Hopkins Verbal Learning Test Revised Delayed Recall	.115	.077
Judgement of Line Orientation	.029	.644
Facial Recognition Test	.065	.255
E. DRS-2 Memory Subscale Score	B	p
Age	.001	.814
Education	.007	.815
Sex	.196	.047
UPDRS Part III Motor Scores	−.006	.264
Beck Depression Inventory-II Total Score	.014	.167
Apathy Scale Total Score	−.012	.214
State Trait Anxiety Inventory-State Total Score	−.006	.256
Trail Making Test Part B	−.040	.568
Stroop (Interference Trial)	.042	.510
Letter Fluency (FAS)	.003	.960
Digit Span Forward	.093	.080
Boston Naming Test	.082	.155
Semantic Fluency (Animals)	.061	.347
Logical Memory Delayed Recall	.255	.001
Hopkins Verbal Learning Test Revised Delayed Recall	−.002	.985
Facial Recognition Test	.014	.817

DRS-2 attention.

Significant correlates of DRS-2 Attention were entered into Block 3 (Trail Making Test B, Stroop, Letter Fluency, Digit Span Forward, Digit Span Backward, Semantic Fluency, Logical Memory Delayed Recall, Hopkins Verbal Learning Test Revised Delayed Recall, Facial Recognition Test). The final model accounted for 10.3% of the variance, [F(16,343) = 2.323, p < .003, R² = .103]. Sex (B = .259) and Digit Span Forward (B = .199) had a weak effect on this subscale (ΔR² = .061, p < .001). Multicollinearity statistics were within the acceptable limits (Tolerance > .8 and Variation Inflation Factor < 1.2). See Table 6A.

DRS-2 initiation/perseveration.

Significant correlates of DRS-2 Initiation/Perseveration were entered into Block 3 (Trail Making Test B, Stroop, Letter Fluency, Digit Span Backward, Semantic Fluency, Boston Naming Test, Logical Memory Delayed Recall, Hopkins Verbal Learning Test Revised Delayed Recall, Judgement of Line Orientation, Facial Recognition Test). The final model accounted for 23.9% of the variance, [F(17,352) = 4.173, p < .001, R² = .239]. Trail Making Test Part B (B = −.165), Facial Recognition Test (B = .142), Semantic Fluency (B = .107), Stroop Interference Trial (B = .104), and Letter Fluency (B = .099) had weak effects on this subscale (ΔR² = .142, p < .001). Multicollinearity statistics were within the acceptable limits (Tolerance>.8 and Variation Inflation Factor < 1.2). See Table 6B.

DRS-2 construction.

Significant correlates of DRS-2 Construction were entered into Block 3 (Trail Making Test B, Stroop, Semantic Fluency, Hopkins Verbal Learning Test Revised Delayed Recall, Judgement of Line Orientation). The final model accounted for 9.1% of the variance in DRS-2 Construction, [F(11,348) = 2.879, p < .001, R² = .091]. Judgement of Line Orientation (B = .218) had the only, albeit weak, effecton this subscale (ΔR² = .066, p < .001). Multicollinearity statistics were within the acceptable limits (Tolerance > .8 and Variation Inflation Factor < 1.2). See Table 6C.

DRS-2 conceptualization.

Significant correlates of DRS-2 Conceptualization were entered into Block 3 (Trail Making Test B, Stroop, Letter Fluency, Digit Span Forward, Digit Span Backward, Boston Naming Test, Semantic Fluency, Logical Memory Delayed Recall, Hopkins Verbal Learning Test Revised Delayed Recall, Judgement of Line Orientation, Facial Recognition Test). The final model accounted for 16.9% of the variance, [F(18,341) = 3.617, p < .001, R² = .169]. Education (B = .054) and Letter Fluency (B = .223) had weak effects on this subscale (ΔR² = .111, p < .001). Multicollinearity statistics were within the acceptable limits (Tolerance > .8 and Variation Inflation Factor < 1.2). See Table 6D.

DRS-2 memory.

Significant correlates of DRS-2 Memory were entered into Block 3 (Trail Making Test B, Stroop, Letter Fluency, Digit Span Forward, Boston Naming Test, Semantic Fluency, Logical Memory Delayed Recall, Hopkins Verbal Learning Test Revised Delayed Recall, Facial Recognition Test). The final model accounted for 20.4% of the variance, [F(16,343) = 5.144, p < .001, R² = .204]. Sex (B = .196) and Logical Memory Delayed Recall (B = .255) had weak effects on this subscale (ΔR² = .134, p < .001). Multicollinearity statistics were within the acceptable limits (Tolerance > .8 and Variation Inflation Factor < 1.3). See Table 6E.

Discussion

The overall goal of the current study was to examine the construct validity of the DRS-2 total and subscale scores in a sample of non-demented patients with PD. Measures of executive function, followed by language and visuoperceptual skills, largely accounted for variance in DRS-2 total scores. In addition, neuropsychological measures from the same hypothesized cognitive domain largely accounted for variance in DRS-2 subscale scores. The magnitude of these relationships was weak across DRS-2 total and subscale scores, despite statistical significance, raising questions about the construct validity of both DRS-2 total and subscale scores in non-demented samples of individuals with PD.

Most prior studies used estimates of intellectual functioning or ranks of cognitive impairment to assess construct validity of the DRS total score. While interesting, this is a different conceptual question posed by the current study. Only one prior study used cognitive composites derived from a comprehensive neuropsychological exam to assess construct validity of the DRS-2 total and subscale scores in a large non-PD sample (Knox et al., 2003). The DRS-2 total score was strongly associated with cognitive composites of attention, executive function, visuospatial skills, language, and immediate memory, but not delayed memory. In the present study, we used individual neuropsychological measures to examine construct validity of the DRS-2 total and subscale scores instead of cognitive composites. By doing so, we were able to learn which neuropsychological measures accounted for the most variance in the DRS-2 total score. Our findings suggest deficits in speeded measures of set-shifting, rapid word generation to letter and semantic cues, and simple visuoperceptual skills are uniquely associated with DRS-2 total scores in a non-demented sample of patients with PD. From a psychometric perspective, several DRS-2 subtests are conceptually related to different aspects of executive function. At the same time, cognitive tests are not ‘process pure,’ or rarely assess a domain of cognition in isolation. For instance, semantic fluency is largely considered a language-related task (Whiteside et al., 2016) with a frontal-executive component in semantic-lexical retrieval (Aita et al., 2019). By the same token, most neuropsychological measures that emerged as unique correlates of the DRS-2 total score have a processing speed component (e.g., Trail Making Test Part B, Letter Fluency-FAS, Semantic Fluency-Animals), apart from Facial Recognition Test. Although UPDRS Part III motor scores did not emerge as a significant correlate of DRS-2 total scores, this begs the question as to whether our findings reflect a slowed ‘speed of processing’ in our sample. From a neurobiological perspective, nigrostriatal dopamine disruption (Siepel et al., 2014), the key neuropathological feature of the disease, as well as prominent grey-white matter changes in the frontal regions (Biundo et al., 2011; Gallagher et al., 2013) are linked with worse executive dysfunction in PD. Future work should investigate construct validity of the DRS-2 total score as compared with other groups (e.g., normal aging, Alzheimer’s disease) using a similar approach as the current study to determine whether our findings reflect cognitive deficits associated with the sample population, or the utility of the DRS-2 total score.

Overall, four out of the five DRS-2 subscales were uniquely associated with conceptually related measures (DRS-2 Attention with Digit Span Backward, DRS-2 Conceptualization with Letter Fluency, Construction with Judgement of Line Orientation, DRS-2 Memory with Logical Memory Delayed Recall), while Initiation/Perseveration was associated with both conceptually related (DRS-2 Initiation/Perseveration with Letter Fluency, Semantic Fluency, Trail Making Test Part B, Stroop Interference Test) and unrelated measures (Facial Recognition Test). One possible interpretation of our findings is DRS-2 subscale scores correlated with conceptually related neuropsychological measures as evidenced by statistical significance. Most of these correlations were weak in magnitude, falling below recommended values (e.g., r ≥ .50; Carlson & Herdman, 2012), which directly challenges this interpretation. Together, construct validity of the DRS-2 subscales was below expectation, consistent with previous studies in PD (Brown et al., 1999) and non-PD populations (Bobholz & Brandt, 1993; Kessler et al., 1994; Woodard et al., 1996). At the conceptual level, the DRS-2 subscale scores were developed using face validity, which is a less rigorous method, and may, in part, account for the heterogeneity of findings in the literature. Thus, clinicians and researchers should interpret cognitive performance using DRS-2 subscale scores with caution.

There are limitations to this study. First, we did not examine ethnoracial identity as a covariate in the current study as most of our sample was non-Hispanic white, well-educated, males who fell within the early-mid stages of the disease. The demographic composition of our sample highlights larger issues in research and clinical recruitment and retention, namely individuals with idiopathic PD from historically marginalized backgrounds are underrepresented in the PD research and clinic samples. Therefore, replication in larger samples differing in demographic and disease-related characteristics is needed. Second, there remains a significant amount of variability left unexplained in DRS-2 total and subscale scores which may, in part, be due to the process impurity. Although domains of cognition are thought to be distinct and separate, we acknowledge that the domains themselves, nor the measures developed to assess them, are process pure. For the purposes of the current study, we viewed domains of cognition as mostly orthogonal, and performed separate regression analyses for DRS-2 subscales, which may be viewed as a potential limitation. Additionally, we converted raw DRS-2 total and subscale scores, as well as raw neuropsychological measures, to z-scores using internal sample means and standard deviations to examine relationships on a similar metric, which may be viewed as another potential limitation. We did not use the standardized z-scores for DRS-2 and neuropsychological measures because these standardized scores are drawn from different normative sampling data with varying demographic corrections, which would have introduced another limitation. Therefore, the use of raw scores with demographic corrections in the current study should be considered a strength.

In conclusion, the DRS-2 total score is disproportionately impacted by executive, semantic-lexical retrieval, and visuoperceptual difficulties. Further, DRS-2 subscale scores show weak construct validity. Together, findings suggest the DRS-2 not appear to have utility beyond a screening tool for global cognition in PD.