All variables except the Overall Score on the multitasking measure were non-normally distributed (Kolmogrov-Smirnov p-values < .01), and Wilcoxon Rank Sums tests were therefore used to conduct between-group comparisons for all other variables, with Cohen’s d providing a measure of effect sizes. Results for the Multitasking test are presented in . As shown in this table, HIV+ participants demonstrated a significantly lower overall score, switched between tasks less frequently, and had significantly fewer simultaneous task attempts. Moreover, HIV+ participants exhibited a greater number of total errors, which primarily consisted of omission errors. In contrast, the groups did not differ in the number of intrusions, repetitions, sequencing errors, or tasks attempted (p values > .10). Effect sizes for the significant results were medium-to-large, ranging from 0.54 to 0.93. However, no HIV disease variables were associated with any variable from the multitasking measure.
Multitasking performance in the HIV+ and HIV− healthy comparison samples
As shown in , correlational analyses within the HIV− group revealed significant associations between the speed of information processing z score and multitasking task switches (p = .020) and simultaneous task attempts (p = .005). In addition, the executive functions z score was associated with multitasking overall score (p = .020), task switches (p = .006), and simultaneous task attempts (p = .006). Finally, the attention/working memory z score was associated with multitasking overall score (p = .007) and simultaneous task attempts (p = .048).
Spearman’s rho correlations between neuropsychological performance and Multitasking variables in the HIV− group (n = 25)
Overall score on the multitasking test was associated with the global neuropsychological summary score (i.e., GDS) at a trend level (r = −.24; p = .06) in the HIV+ group. As displayed in , within the HIV+ group, correlational analyses revealed significant relationships between the speed of information processing z score and multitasking overall score (p = .005) and simultaneous task attempts (p = .004). Significant correlations were found for the memory z score and multitasking overall score (p = .001), intrusions (p = .031), omissions (p = .005), total errors (p = .0004), and simultaneous task attempts (p = .004). The attention/working memory z score was significantly associated with multitasking overall score (p = .037), simultaneous task attempts (p = .006), total errors (p = .002), and omission errors (p = .008). Significant relationships were found between the executive functions z score and multitasking intrusions (p = .032), omissions (p = .046), and total errors (p = .008). Multitasking repetitions and task switches were not associated with any cognitive domain z score in the HIV+ group.
Spearman’s rho correlations between neuropsychological performance and Multitasking variables in the HIV+ group (n = 60)
displays the Spearman’s rho intercorrelations between the various indices of interest from the multitasking measure in the HIV+ group. Multitasking overall score was positively associated with task switches and simultaneous task attempts and negatively associated with total, omission, and sequencing errors. Multitasking total errors were strongly and positively associated with all other multitasking error types, while multitasking task switches were strongly and positively associated with simultaneous task attempts.
Intercorrelations between variables of the Multitasking test in the HIV+ group (n = 60)
Based on the responses of the 60 HIV+ participants on the IADL questionnaire, 14 participants (23.3%) met criteria for IADL-dependence, while 46 (76.7%) were deemed IADL-independent. The IADL-dependent and IADL-independent subgroups were comparable for demographic characteristics, HIV disease severity, and estimated premorbid verbal IQ (as measured with the WRAT; all p
s > .10). As might be expected from previous research (e.g., Heaton et al., 2004
), the IADL-dependent group had a significantly higher rate of NP impairment (p
= .039), endorsed greater affective distress on the BDI-II (p
< .001), and had a higher proportion of current major depression diagnoses (p
= .007), although they did not differ in proportion of individuals with lifetime substance dependence diagnoses (p
> .10). The groups also did not differ in their “best” prior level of IADL functioning across domains.
Consistent with our hypotheses, the IADL-dependent individuals displayed a lower overall score on multitasking [t
(58) = 2.35, p
= .022], but no other variables from the measure were significantly different between the groups. Given this result, a hierarchical logistic regression was conducted to predict IADL status among HIV+ participants from the overall score from the multitasking test while also accounting for the effects of neuropsychological impairment (GDS) and depression (current diagnosis). Given previous research in HIV (e.g., Heaton et al., 2004
), depression diagnosis and GDS were entered together in the first step, followed by multitasking overall score in the next step. Depression and GDS were both significantly predictive of IADL status and together resulted in a nonsignificant Hosmer and Lemeshow test [χ2
(7) = 6.46; p
= .49], which indicates adequate model fit. In the next step, adding multitasking overall score increased the fit of the model (Hosmer and Lemeshow χ2
(7) = 10.992; p
= .14), with depression (p
= .001) and multitasking overall score (p
= .019) significantly predicting IADL status, although this resulted in GDS becoming a nonsignificant contributor to the model (p
= .152). Of note, there were no significant differences on the multitasking test in HIV+ individuals with and without a current diagnosis of depression [t
(58) = 0.16, p
= .870], and the BDI-II Total Score was not significantly correlated with the overall score on the multitasking test (r
= −.039, p
An ROC curve revealed that overall score on the multitasking test was superior to chance in classifying IADL status (area under the curve [AUC] = 0.69, SE = 0.07, p = .03). A cut-point of 27 on the multitasking test was chosen as providing a reasonable balance between sensitivity and specificity for predicting IADL status. The overall hit rate for this cutoff was 65%, with excellent sensitivity (i.e., the proportion of IADL-dependent participants with overall scores on multitasking below this cutoff = 86%) and negative predictive power (i.e., the proportion of multitasking overall scores above cutoff produced by the IADL-independent sample = 88%). However, the specificity (i.e., the proportion of IADL-independent participants with multitasking overall scores above this cutoff = 57%) and positive predictive power (i.e., the proportion of multitasking overall scores below cutoff produced by the IADL-dependent sample = 38%) values were somewhat more modest. Generating odds ratios using the cut-point of 27 to indicate impairment in multitasking showed that HIV+ individuals with multitasking overall scores below this cut-point were over 8 times more likely to be classified as IADL-dependent than those who were unimpaired on this measure (OR = 8.1).
Given its utility in predicting IADL status, a post hoc
exploratory analysis was also conducted to examine potential differences between unemployed and employed individuals on the multitasking overall score. Individuals who were less than one-half time employed were considered unemployed, while individuals who were more than half time employed were considered employed (Heaton et al., 1994
). Two participants reported a work status that was ambiguous, with one classified as employed and one unemployed after further review. Individuals who identified as “retired” were not included in analyses due to classification ambiguity, leaving 56 HIV+ participants for analysis. Unemployed participants displayed lower overall scores on the multitasking test compared to employed participants [M
= 26.2, SD
= 7.2 vs. M
= 31.8, SD
= 9.4; t
(53) = −2.37, p
= .02). A nominal logistic regression was conducted that attempted to predict employment status from the overall score from the multitasking test while also including variables upon which the groups differed. The model was significant [χ2
(4) = 18.42, p
= .001], with current depression diagnosis (p
= .006), AIDS status (p
= .015), and overall score on the multitasking test (p
= .019) each providing significant, unique contributions to the prediction model.