Patient-Reported Changes in Communication After Computer-Based Script Training for Aphasia

Article Outline

• Abstract
• Methods
  • Subjects
  • Study Design
  • Intervention
  • Measurement of Outcomes
  • Analysis
• Results
• Discussion
  • Study Limitations
• Conclusions
• Acknowledgments
• References
• Copyright

Abstract 

Manheim LM, Halper AS, Cherney L. Patient-reported changes in communication after computer-based script training for aphasia.

Objective

To evaluate changes in patient-reported communication difficulty after a home-based, computer-delivered intervention designed to improve conversational skills in adults with aphasia.

Design

Delayed treatment design with baseline, preintervention, postintervention, and follow-up observations.

Setting

Outpatient rehabilitation.

Participants

Twenty subjects with chronic aphasia.

Interventions

Sessions with the speech-language pathologist to develop personally relevant conversational scripts, followed by 9 weeks of intensive home practice using a computer program loaded on a laptop, and weekly monitoring visits with the speech-language pathologist.

Main Outcome Measure

Communication Difficulty (CD) subscale of the Burden of Stroke Scale (BOSS).

Results

The intervention resulted in a statistically and clinically significant decrease of 6.79 points (P=.038) in the CD subscale of the BOSS during the intervention, maintained during the follow-up period.

Conclusions

The findings of this study provide positive albeit preliminary and limited support for the use of a home-based, computer-delivered language intervention program for improving patient-reported communication outcomes in adults with chronic aphasia. Additional research will be required to examine the efficacy and effectiveness of this intervention.

Key Words: Aphasia, Communication, Computer-assisted instruction, Cost, cost analysis, Rehabilitation

List of Abbreviations: BDAE, Boston Diagnostic Aphasia Examination, BOSS, Burden of Stroke Scale, CD, Communication Difficulty

IT HAS BEEN ESTIMATED that that there are more than 5 million stroke survivors in the United States,¹ 33% of whom have some impairment in language.² Thus, more than a million people are living with aphasia in the United States alone.³ Although meta-analyses⁴, ⁵ and expert opinion⁶, ⁷ indicate that individuals with aphasia benefit from treatments that improve linguistic skills, residual communication problems generally continue to have a substantial impact on their daily lives. Therefore, treatments that focus on improving their ability to participate in daily activities have received increased attention.⁸ Therapeutic interventions for aphasia are concentrated during the acute stage. Yet reacquisition of communication skills often requires long-term effort and training, generally not well supported by insurance during the chronic stage of the condition. A program that uses computer treatment as a medium for therapy and emphasizes the development of conversational skills may provide an important and low-cost means of reintegration into some of the normal activities of daily living.

Scripts can facilitate participation in personally relevant activities by guiding individual communication actions involved in a social situation.⁹, ¹⁰ Script training is designed to help speakers with aphasia use short self-chosen monologues and dialogues in natural, conversational contexts. During training, scripts are developed by the person with aphasia, with assistance from the speech-language pathologist. Cue-based massed drilling of the entire script is required to facilitate less effortful script production. This massed practice and drill can be accomplished in a low-cost manner by using computers.

There is an existing and well documented body of literature to support the positive effects of computer treatment for people confronting the long-term effects of aphasia.¹¹, ¹² Current computer programs for aphasia typically focus on tasks at the single-word level, including verbal and written word finding, single-word auditory comprehension skills, and single-word visual recognition and reading comprehension.¹³, ¹⁴, ¹⁵, ¹⁶, ¹⁷, ¹⁸ However, Petheram¹⁹ has noted that retrieving a word in response to a picture during treatment is not comparable to retrieving the same word in everyday conversation. We present a computer program that, instead of working on single-word tasks, provides a more realistic conversational context for practicing language skills.²⁰

We evaluate the impact of the program protocol in reducing patient-reported communication difficulty as measured by the CD subscale from the BOSS.²¹, ²², ²³ We then discuss the potential costs and cost savings of the computerized script training program when used in conjunction with speech and language therapy.

Back to Article Outline

Methods 

Subjects 

Twenty-five subjects were recruited with aphasia subsequent to a left-hemisphere stroke confirmed by medical history and computerized tomography or magnetic resonance imaging. Subjects were at least 6 months poststroke and right-hand dominant, with no history of other premorbid neurologic or psychiatric disorders. They had completed at least 10th grade and were literate in English before their strokes. Visual acuity was no worse than 20/100 corrected in the better eye; auditory acuity was no worse than 30 dB hearing level at 500, 1000, and 2000 Hz, aided in the better ear. Subjects did not receive any other individual or group treatment while they were participating in this study. All subjects provided written informed consent under the approval of the Northwestern University Institutional Review Board.

Study Design 

A delayed treatment design was used in which subjects' outcomes were measured at 4 separate times: (1) at entry into the study (baseline), (2) approximately 6 weeks later at the start of the intervention (pretreatment), (3) at the end of the intervention (posttreatment), and (4) at a retest planned to occur at approximately 6 weeks after the end of the intervention (follow-up).

Intervention 

The intervention involved a computer-based script training program called AphasiaScripts, which was developed as part of a research study. In AphasiaScripts, an avatar that is programmed to produce natural speech with correct movements of the speech articulators serves as a virtual therapist.²⁰ Prior to treatment, the individual with aphasia and the speech-language pathologist work together to develop individualized scripts on a topic that is meaningful and relevant. After a script has been developed, it is typed into the program and recorded by the speech-language pathologist. Using AphasiaScripts, the individual with aphasia has repeated opportunities to practice the recorded conversations. Script practice has 3 phases. First, the subject listens to the entire script while it appears on the screen. Second, each sentence or conversation turn is practiced repeatedly. Third, the conversation is practiced with the virtual therapist while various forms of assistance are provided, depending on the subject's needs. These include seeing the written word, hearing the therapist's voice during choral speaking, and watching oral-motor movements of the virtual therapist. These cues are faded over time so that eventually the subject practices the conversation with the virtual therapist, without cues, as in a real conversation.²⁰

In this study, subjects worked with the speech-language pathologist to develop 3 scripts. Script development occurred over a period of 5 sessions. Then each script was practiced consecutively for 3 weeks for a total of 9 weeks of intervention. The protocol suggested that participants practice the script daily for at least 30 minutes at home on a loaned laptop. Participants kept a paper-and-pencil log of their practice times. In addition, the computer program maintained an objective measure of practice times based on log-on and log-off times as well as keystrokes made during the practice session. During the 9 weeks of practice, the subject with aphasia met once a week with a speech-language pathologist only to check status and ensure compliance. Because of differences in scheduling appointments during script development, there was variability in the length of time between pretreatment and posttreatment testing. However, the intervention was always 9 weeks.

Measurement of Outcomes 

To measure patient-reported communication difficulty, we selected the CD subscale of the BOSS. The BOSS is a comprehensive, patient-reported measure of functioning and well being.²² It is a 64-item scale made up of 12 internally consistent and unidimensional scales. The CD subscale consists of 7 items: “Because of your stroke, how difficult is it for you to (1) talk, (2) understand what people say to you, (3) understand what you read, (4) write a letter, (5) talk with a group of people, (6) be understood by others, and (7) find the words you want to say?” Responses to each item use a 5-point scale, from not at all to cannot do. We also examined 2 other subscales from the BOSS: (1) the 3-item Communication-Associated Psychological Distress subscale, intended to measure patient-reported mood, satisfaction, and normal activity restriction caused by difficulties with communicating; and (2) the Mobility subscale, which consists of 5 items that measure patient-reported mobility. All testing was conducted by a speech-language pathologist who was independent of the treating speech-language pathologist.

Analysis 

We estimated 3 fixed-effects linear regressions, which control for both within-person correlation of repeated observations and for average differences in CD scores across persons,²⁴ with the outcome scores (CD, Communication-Associated Psychological Distress, Mobility) as the dependent variables in each regression. Each subject had repeated outcome measurements at the 4 times they were asked to complete the BOSS. The regression coefficients measure the mean outcome score at the preintervention period and then the difference in scores between the preintervention period at each of the other 3 observation points. We also calculated the effect sizes based on Cohen d score, defined as the difference between the means divided by the SD of either group. Cohen equates values of .2, .5, and .8 as small, medium, and large effect sizes, respectively.²⁵

Responses on each of the BOSS individual self-reported items can take a value of 0 to 4. The individual scores were summed, and a transformed scale score was constructed, equal to [(actual score minus lowest possible score) divided by possible score range] multiplied by 100. This transformed scale can have values ranging from 0 (no difficulty) to 100 (cannot do).²³ A lower score on this and the other BOSS subscales denotes improved function.

Our primary hypotheses were that for the CD score, we would find the statistically significant negative (improved) change in CD scores between the preintervention and 2 postintervention measures, but no change in CD between the baseline and preintervention period. For the BOSS mobility subscale score, we expected to find no statistically significant changes because of the intervention. As a sensitivity analysis, we examined whether any of the other BOSS subscales had statistically significant changes in outcomes during the intervention period using the same fixed effects regression framework. For all tests of hypotheses, we used a 2-tailed significance test against the null hypothesis of no effect.

Back to Article Outline

Results 

The first 25 subjects entered into the study formed the sample. However, data from 5 subjects could not be analyzed because they did not have 2 test sessions prior to the intervention (baseline and preintervention), and at least 1 test session postintervention. Of the 20 subjects included in the analyses, 3 did not have the follow-up interview at 6 weeks postintervention.

The 20 subjects (13 men) were 26 to 78 years of age (mean, 54.80; SD±15.25) with an education level that ranged from 10 to 22 years (mean, 15.06; SD±3.24). The time since the onset of the stroke ranged from 10.6 to 273.7 months (mean, 53.01; SD±63.16). Severity of the aphasia, as measured by the Aphasia Quotient of the Western Aphasia Battery,²⁶ ranged from 30.5 to 85.3 (mean, 64.57; SD±15).

Mean time between surveys was 6.5 (SD±2.7) weeks from baseline to pretreatment, 13.5 (SD±2.2) weeks from pretreatment to posttreatment, and 7.0 (SD±2.8) weeks from posttreatment to follow-up. As shown in table 1, the mean value of the CD score for the 20 subjects at entry into the study (baseline) was 54.3 (SD±14.2), with a range from 32.1 to 82.1.

Table 1. Means and SDs of Outcome Variables at Baseline Entry Into Study

Abbreviation: CAPD, Communication-Associated Psychological Distress subscale.

Table 2 provides the results of the fixed effects linear regressions. The pretreatment coefficient measures the average score observed for the sample prior to the intervention. The other coefficient values measure the difference in scores from pretreatment to each of the other time points. At entry into the study (baseline), the score was just .54 units higher than the pretreatment score (not statistically significant), with an effect size of only .03. Thus, there was a negligible change in the CD score over the period of observation prior to the intervention.

Table 2. Effects of Intervention on BOSS Subscales Coefficients (SE) of Effects Using a Fixed-Effects Regression Equation

	CD	CAPD	Mobility
Pretreatment absolute score	53.74⁎(2.26)	46.49(2.93)	26.42(2.87)
Baseline score − difference from pretreatment	.54(3.19)	2.92(4.14)	−.50(4.06)
Postintervention score − difference from pretreatment	−6.79†(3.19)	−6.67(4.14)	−3.50(4.06)
Follow-up score − difference from pretreatment	−10.63⁎(3.37)	−6.64(4.38)	−2.61(4.29)
Within-subject R2	.23⁎	.13(P=.06)	.02

Abbreviation: CAPD, Communication-Associated Psychological Distress subscale.

Over the period of intervention, however, there was a statistically significant decrease of 6.79 points in the CD score (P=.038), with an effect size of .43. The average CD score decreased another 3.84 points (not statistically significant) between the end of the intervention and the follow-up some 6 weeks later, so that the change in patient-reported communication difficulty between the pretreatment and follow-up assessment was 10.63 (P=.003), with an effect size of .67.

Results for the BOSS mobility subscale are also reported in table 2. While there was a decrease of 3.50 points during the intervention in mobility impairment, it was statistically nonsignificant. The other BOSS subscale that might be expected to be sensitive to improvement in patient-reported communication difficulty is the Communication-Associated Psychological Distress subscale. As seen in table 2, there was a small (2.92 point), statistically nonsignificant increase in patient-reported Communication-Associated Psychological Distress during the period prior to the intervention (from baseline to pretreatment) and then an improvement of 6.67 points from pretreatment to posttreatment, which was not statistically significant (P=.113). However, an almost identical improvement of 6.64 points was observed between pretreatment and follow-up Communication-Associated Psychological Distress scores. To improve the power of the test, we re-estimated the regression after placing an additional constraint on the regression that postintervention and follow-up coefficients be identical. The treatment coefficient was then –6.75 (P=.07), still not statistically significant, but suggestive given the small sample size, that the intervention may have had a modest impact on patient-reported Communication-Associated Psychological Distress.

Back to Article Outline

Discussion 

A delayed treatment design was used to measure changes in patient-reported communication outcomes prior to, during, and after the intervention. We observed no significant change in CD scores prior to the intervention, but observed a statistically significant improvement in CD scores during the intervention, and a further improvement during the postintervention follow-up. The effect size of the change from the pretreatment to the follow-up period was .67, which falls halfway between a medium and large effect size, according to Cohen.²⁵ To put these results in context, we made some comparisons between mean values in our sample of 20 and reported mean CD values among stroke survivors in a community sample.²³ The community sample had a mean CD score of 44.08 (SD±20.87) for those who were communicatively impaired and 19.32 (SD±19.99) for those communicatively unimpaired. Thus, our sample (mean baseline CD score, 54.28; SD±14.22) is more communicatively impaired and less variable than the reported community sample.²³ Participants were defined as communicatively impaired if they scored 1 to 4 on the Severity Rating Scale of the BDAE²⁷ and communicatively unimpaired if they scored a 5. Based on these data, our pretreatment score of 53.74 is only slightly less than the mean CD score of participants with aphasia with a BDAE scale point 2 (57.65).

To compare the change in self-reported communication difficulty observed in this study with clinician-tested communication improvement, we looked at which BDAE ranks would give the average CD scores we observed at the pretreatment and follow-up periods (based on the means reported in Doyle et al²³). The change we observed, from a mean CD of 53.74 at pretreatment to 43.11 at follow-up, was almost equivalent to a move from BDAE scale point 2 (average CD score of 57.7) to scale point 3 (average CD score of 42.2). BDAE scale point 2 is scored if the response is, “Conversation about familiar subjects is possible with help from the listener. There are frequent failures to convey the idea, but the patient shares the burden of communication with the examiner.” BDAE scale point 3 reflects the response, “The patient can discuss almost all everyday problems with little or no assistance. Reduction of speech and/or comprehension, however, makes conversation about certain material difficult or impossible.” While this appears to be a clinically significant change, we cannot say whether we would have found the same statistically significant change had we actually collected BDAE ranks in addition to the BOSS.

In general, it is problematic to assume that significant changes over time in patient-reported communication impairment were caused by the treatment intervention rather than some other event. To test that significant changes in CD during the intervention period were not a result of a general trend in stroke improvement over time, we used a delayed treatment design with 2 points of observation before and after the intervention. The results given this research design suggest the findings are not spurious. This conclusion is also supported by the lack of significant changes in the mobility impairment subscale of the BOSS over the same period; the BOSS mobility subscale score was also a patient-reported score, yet there was no reason to expect improvements in mobility during the treatment. The within-subject R² of .02 for the mobility regression may be contrasted to the within-subject R² of .23 for the CD regression model. Indeed, further sensitivity analyses found that no other subscales of the BOSS improved significantly over the intervention period.

The Communication-Associated Psychological Distress subscore of the BOSS, however, was the only other BOSS subscale that even approached statistical significance (all other subscales were statistically insignificant at P<.20). This finding is consistent with previous reports of a correlation of .41 between the CD and the Communication-Associated Psychological Distress subscores.²³ If the intervention operated on Communication-Associated Psychological Distress solely through its effect on CD, then one would expect the effects of the intervention on Communication-Associated Psychological Distress to be less strong and not statistically significant.

Cost considerations of using a computer program that involves home practice is worthy of discussion. The cost of the intervention described involved 5 sessions to develop the scripts and another 9 sessions of weekly follow-up. The cost of 14 sessions may be evaluated at the Medicare payment rate for the Current Procedural Terminology code 92507, which was $96 a session for the nonfacility plus facility payment in Chicago in 2008; the 14 sessions would result in a total payment of $1344. The software could be available at a cost of $350. While the intervention involved loaning laptops over the 9-week intervention, we would expect that individuals would generally have the program downloaded to their own computers, with no time limit on use.

The 20 subjects in this study reported spending an average of 44.0 (SD±30.3) hours practicing scripts on their laptop during the 9 weeks of the intervention. Practice ranged from 11.3 hours to 66.1 hours. The protocol's suggested minimum requirement of 30 minutes of practice a day would have resulted in 31.5 hours of practice. Evidence of perceived treatment value is supported by 65% of subjects spending more than 31.5 hours with the intervention.

The cost of an additional 44 hours of speech therapy at the Medicare payment level would be $4224 at an outpatient facility and $2992 at a private office (using the 2008 Medicare office and outpatient facility payment rates in Illinois). Neither Medicare nor other insurers are likely to pay for these additional hours of speech and language therapy that allow beneficiaries with aphasia to meet with a speech-language pathologist for intensive practice of their scripts. Intensive practice is a key component for reducing speech and language impairments.⁵, ²⁸, ²⁹ Thus, the use of the software program provides a low-cost method of incorporating intensive practice into a more limited speech and language therapy intervention.

Study Limitations 

There are a number of limitations to the study. The sample in this study may not be representative of a general population of individuals with aphasia and communication impairment. The delayed treatment design is not comparable to a randomized controlled trial. Although there was no change in patient-reported communication difficulty prior to the intervention and no change for BOSS subscale measures unrelated to the intervention, it is still possible that the associations between the intervention and patient-reported improvements in communication are not causative. Further, the improvement in patient-reported communication does not necessarily indicate an improvement in actual communication impairment, and other clinician-tested indicators of change would complement the findings of this study.

Finally, the findings do not imply that the mix of the speech-language pathologist's time and independent computer practice time by the person with aphasia is the most cost-effective use of resources. The results, however, do provide reason to explore further these questions and others related to dosage and subject characteristics that are best suited to treatment.

One advantage of a computer program is that individuals with aphasia would have it available beyond the intervention period and could learn how to alter scripts to fit new situations. Generalizability of these findings to home settings awaits further evaluation. The enthusiasm and adherence to guidelines by study participants who practiced more than the requested hours suggested that the program was considered rewarding by most subjects. However, we do not know how difficult it will be to duplicate the intervention using the computer software in routine clinical practice with individuals with aphasia. Evaluation of a larger, community practice–based demonstration of the protocol and computer program, either by rigidly following a research protocol or by allowing community speech-language pathologists and individuals with aphasia more latitude, is warranted.

Back to Article Outline

Conclusions 

The findings of this study provide positive, albeit preliminary and limited, support for the use of a home-based computer-delivered language intervention program for improving patient-reported communication outcomes in adults with chronic aphasia. Additional research will be required to examine the efficacy and effectiveness of this intervention.

Back to Article Outline

Acknowledgments 

We thank Rosalind Hurwitz, MS, for assistance with data collection, Allen Heinemann, PhD, for helpful comments and suggestions about the article, and Audrey Holland, PhD, a consultant on this project, for her advice throughout the duration of the grant.

Back to Article Outline

References 

1. American Heart Association. 2008 Heart disease and stroke statistics update. Dallas: American Heart Association; 2008;
   • View In Article
2. Pedersen PM, Jorgensen HS, Nakayama H, Raaschou H, Olsen TS. Aphasia in acute stroke: incidence, determinants, and recovery. Ann Neurol. 1995;38:659–666
   • View In Article
   • MEDLINE
   • CrossRef
3. NIH Pub No. 97-4257 October, 2008 (National Institute on Deafness and Other Communication Disorders, Fact Sheet: Aphasia). http://www.nidcd.nih.gov/health/voice/aphasia.aspAccessed January 15, 2009
   • View In Article
4. Robey RR. The efficacy of treatment for aphasic persons: a meta-analysis. Brain Lang. 1994;47:582–608
   • View In Article
   • MEDLINE
   • CrossRef
5. Robey RR. A meta-analysis of clinical outcomes in the treatment of aphasia. J Speech Lang Hear Res. 1998;41:172–187
   • View In Article
   • MEDLINE
6. Albert ML. Treatment of aphasia. Arch Neurol. 1998;55:1417–1419
   • View In Article
   • MEDLINE
   • CrossRef
7. Holland AL, Fromm DS, DeRuyter F, Stein M. Treatment efficacy: aphasia. J Speech Hear Res. 1996;9:S27–S36
   • View In Article
8. Cherney LR, Robey RR. Aphasia treatment: recovery prognosis and clinical effectiveness. In:  Chapey R editors. Language intervention strategies in aphasia and related neurogenic communication disorders. Philadelphia: Lippincott, Williams & Wilkins; 2001;p. 148–172
   • View In Article
9. Sharkey NE, Mitchell DC. Word recognition in a functional context: the use of scripts in reading. J Mem Lang. 1985;24:253–270
   • View In Article
10. Walker CH, Yekovich FR. Script-based inferences: effects of text and knowledge variables on recognition memory. J Verbal Learn Verbal Behav. 1984;23:357–370
    • View In Article
11. Katz RC. Computer applications in aphasia treatment. In:  Chapey R editors. Language intervention strategies in aphasia and related neurogenic communication disorders. Philadelphia: Lippincott, Williams & Wilkins; 2001;p. 718–741
    • View In Article
12. Loverso FL, Prescott TE, Selinger M. Microcomputer treatment applications in aphasiology. Aphasiology. 1992;6:155–162
    • View In Article
13. Fink RB, Brecher A, Schwartz MF. A computer-implemented protocol for treatment of naming disorders: evaluation of clinician-guided and partially self-guided instruction. Aphasiology. 2002;16:1061–1086
    • View In Article
14. Katz RC, Wertz RT. The efficacy of computer-provided reading treatment for chronic aphasic adults. J Speech Lang Hear Res. 1997;40:493–507
    • View In Article
    • MEDLINE
15. Mortley J, Enderby P, Petheram B. Using a computer to improve functional writing in a patient with severe dysgraphia. Aphasiology. 2001;15:443–461
    • View In Article
16. Pedersen PM, Vinter K, Olsen TS. Improvement of oral naming by unsupervised computerized rehabilitation. Aphasiology. 2001;15:151–169
    • View In Article
17. Petheram B. Exploring the home-based use of microcomputers in aphasia therapy. Aphasiology. 1996;10:267–282
    • View In Article
18. Raymer AM, Konen FP, Saffell D. Computerized training for impairments of word comprehension and retrieval in aphasia. Aphasiology. 2006;20:257–268
    • View In Article
19. Petheram B. Computers and aphasia: a means of delivery and a delivery of means. Aphasiology. 2004;18:187–191
    • View In Article
20. Cherney LR, Halper AS, Holland AL, Cole R. Computerized script training for aphasia: preliminary results. Am J Speech Lang Pathol. 2008;17:19–34
    • View In Article
    • CrossRef
21. Doyle PJ. Measuring health outcomes in stroke survivors. Arch Phys Med Rehabil. 2002;83:S39–S43
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (87 KB)
    • CrossRef
22. Doyle PM, McNeil MR, Mikolic JM, et al. The burden of stroke scale (BOSS) provides valid and reliable score estimates of functioning and well-being in stroke survivors with and without communication disorders. J Clin Epidemiol. 2004;57:997–1007
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (250 KB)
    • CrossRef
23. Doyle PJ, McNeil MR, Hula WD. The burden of stroke scale (BOSS): validating patient-reported communication difficulty and associated psychological distress in stroke survivors. Aphasiology. 2003;17:291–304
    • View In Article
24. Greene WH. Econometric analysis. New York: MacMillan Publishing Co; 1990;p 480-5
    • View In Article
25. Cohen J. A power primer. Psychol Bull. 1992;112:155–159
    • View In Article
    • CrossRef
26. Kertesz A. Western aphasia battery. New York: Harcourt Brace Jovanovich; 1982;
    • View In Article
27. Goodglass H, Kaplan E, Baressi B. The assessment of aphasia and related disorders. 3rd ed.. Philadelphia: Lippincott, Williams & Wilkins; 2001;
    • View In Article
28. Bhogal SK, Teasell R, Speechley M. Intensity of aphasia therapy: impact on recovery. Stroke. 2003;3:987–993
    • View In Article
29. Cherney LR, Patterson J, Raymer A, Frymark T, Schooling T. Evidence-based systematic review: effects of intensity of treatment and constraint-induced language therapy for individuals with stroke-induced aphasia. J Speech Lang Hear Res. 2008;51:1282–1299
    • View In Article
    • CrossRef