The benefits and protective effects of behavioural treatment for dysgraphia in a case of primary progressive aphasia

Abstract

Background: Spoken and written language difficulties are the predominant symptoms in the progressive neurodegenerative disease referred to as primary progressive aphasia (PPA). There has been very little research on the effectiveness of intervention on spoken language impairments in this context and none directed specifically at progressive written language impairment.

Aims: To examine the effectiveness of behavioural intervention for dysgraphia in a case of primary progressive aphasia.

Methods & Procedures: We carried out a longitudinal single‐case study that allowed us to examine the effectiveness of a non‐intensive spell‐study‐spell intervention procedure. We did so by comparing performance on four sets of words: trained, repeated, homework, and control words at five evaluations: baseline, during intervention, after the intervention, and at 6‐ and 12‐month follow‐up.

Outcomes & Results: We find that: (1) at the end of the intervention, Trained words show a small but statistically significant improvement relative to baseline and an advantage in accuracy over Control, Homework, and Repeated word sets. (2) All word sets exhibited a decline in accuracy from the end of treatment to the 6‐month follow‐up evaluation, consistent with the degenerative nature of the illness. Nonetheless, accuracy on Trained words continued to be superior to that of Control words and not statistically different from pre‐intervention baseline levels. (3) Repeated testing and practice at home yielded modest numerical advantages relative to Control words; but these differences were, for many comparisons, not statistically significant. (4) At 12 months post‐intervention, all words sets had significantly declined relative to pre‐intervention baselines and performance on the four sets was comparable.

Conclusions: This investigation documents—for the first time—that behavioural intervention can provide both immediate and short‐term benefits for dysgraphia in the context of primary progressive aphasia.

We are grateful for the support of NIH grant DC006740; to Alexis Kruczek and Joelle Urrutia for their hard work and dedication to this project; to Ranjan Maitra (University of Iowa, Department of Statistics) for his help with statistical analyses; and for CB's friendship and example of humour and humanity throughout difficult times.

Notes

We are grateful for the support of NIH grant DC006740; to Alexis Kruczek and Joelle Urrutia for their hard work and dedication to this project; to Ranjan Maitra (University of Iowa, Department of Statistics) for his help with statistical analyses; and for CB's friendship and example of humour and humanity throughout difficult times.

1. We are grateful to Drs Hillis and Selnes at Johns Hopkins Hospital for providing us with the results of the neuropsychological diagnostic testing that they carried out with CB.

2. We had hoped to acquire similar data for the Homework words but because CB did not do the homework consistently and because she had no one at home to supervise her work and help organise the materials, we couldn't be certain of the accuracy of the homework data, nor of the conditions under which they were obtained.

3. The regression analysis can be carried out on the raw proportion differences depicted in Figure 5 or on statistically appropriate transformations of the data. For example, we also applied a logit transformation to the proportion scores obtained for each of the two kinds of word sets in each session. The logit transformation is the log‐odds or logarithm of the odds ratio of the proportion of letters correctly identified. In this case, this provides us with transformed scores that are on the real line and, consequently, greater comfort with the assumptions in the linear model. We modeled the differences in the logit of the scores in terms of a simple linear regression set‐up with time as the explanatory variable. The differences in log‐odds ratios of the trained and repeated sets varied significantly with time (coefficient  =  0.0567, p  =  .0403). (In the context of the original untransformed data, this implies that the relative proportions of the odds ratios of the scores for the repeated and trained word sets increases exponentially by 0.0567 for every unit change in time.) The model reported an R‐squared of 0.3293 and an adjusted R‐squared of 0.2684. Detailed diagnostics of the residuals indicates that the set of scores for the second observation was both outlying as well as influential. Otherwise, the homoscedasticity and normal assumption for the errors in the differences of logit‐transformed scores was reasonable. Given this, the analysis was redone with the second set of observations omitted. The main finding of a significant relationship over time (slope  =  0.0184, p  =  .0003) was strongly supported. The residuals have an autocorrelation coefficient of –0.329, so a model incorporating a first‐order autocorrelation structure was also fitted, yielding significant results comparable to those of all previous analyses (slope  =  0.08903; p < .0001). Finally, a chi‐squared test statistic evaluating the goodness of fit of the model incorporating the autocorrelation (χ²  =  2.49) indicates that the first‐order autoregressive structure for the model is not significant (p‐value  =  .115). In other words, the previous model that did not include the autocorrelation was adequate.

4. Both JRE and RSB responded very similarly to treatment except that JRE's maintenance of benefits was more short‐lived than RSB's; for the sake of brevity, we report only the results for JRE.