Skip main navigation

Affective Prosody and Depression After Stroke

A Pilot Study
Originally publishedhttps://doi.org/10.1161/STROKEAHA.116.013852Stroke. 2016;47:2397–2400

Abstract

Background and Purpose—

Poststroke depression (PSD) is a frequent complication of stroke with detrimental consequences in terms of quality of life and functional outcomes. In individuals with major depression, several studies have demonstrated an alteration of affective prosody. The aim of this study is to identify prosodic markers that may be predictive of PSD.

Methods—

Patient voices were recorded at baseline and 3 months after stroke. We extracted prosodic parameters, including fundamental frequency, percentage of voice breaks, and shimmer. Depression and anxiety symptoms were assessed 3 months later.

Results—

Among the 49 patients included in the study, 22.5% developed PSD 3 months after stroke. A significant decrease was observed concerning the fundamental frequency among patients who developed PSD. Discriminant analysis demonstrated that initial voice breaks coupled with shimmer are strongly predictive of subsequent PSD.

Conclusions—

Early alterations of affective prosody are associated with a higher risk of PSD 3 months after a stroke. This new physiological approach overcomes traditional barriers associated with clinical instruments and contributes to the prediction of this disorder.

Introduction

Neuropsychiatric disorders occur frequently after stroke and have a detrimental impact on functional outcomes and quality of life.1 Poststroke depression (PSD) is considered the most frequent clinical expression of these disorders and is often diagnosed at least 3 months after stroke.2 The identification of patients at risk for developing this disorder remains an important issue for clinicians because its early diagnosis is essential to improve therapeutic strategies. Although the need to identify new biomarkers3 associated with PSD is increasingly recognized, progress has been slow, despite considerable investment.4

The diagnosis of PSD is often complex because of the strong overlap between somatic symptoms related to stroke and those reported by patients during a depressive state. It is, therefore, essential that symptoms specific to PSD be identified. Impairment of affective prosody, defined as extraverbal information indicating the emotional state of the speaker, independent of verbal content,5 has been extensively reported in major depression. Prosody is defined as the set of objective parameters: the fundamental frequency of the voice signal (F0), the intensity of the sound emission, the shimmer (small variations in glottal pulse amplitude), voice breaks, and the speech rate. Prosody is mediated by the right hemisphere, which is known to influence emotional communication.6 Although several studies have demonstrated that depressed patients speak louder and at a slower rate than healthy controls,7 the findings regarding prosody have been less consistent with some studies observing reductions in F0 range, averages, or variability,5 as well as alterations of shimmer.8

The aim of this pilot study is, therefore, to evaluate the role of affective prosody in the first days after stroke in the prediction of PSD measured 3 months (M3) later. Because changes in emotional states are often coupled with modifications in laryngeal tension and subglottal pressure,9 it is hypothesized that affective prosody would be a reliable biomarker of this disorder.

Methods

Sample

Participants were patients admitted to the stroke unit for minor ischemic stroke (National Institutes of Health stroke scale <4) and confirmed by magnetic resonance imaging. Main exclusion criteria were (1) having a history of dementia, major depression, or anxiety disorders; (2) depressive and anxious symptoms during hospitalization assessed by clinical interview and Hospital Anxiety and Depression Scale (HADS)10; and (3) poststroke speech and language deficits (dysarthria and aphasia), pathological crying, or emotionalism.

Procedure and Materials

This study was approved by the institutional ethics review board, and all patients provided written consent to participate. Stroke severity was evaluated using the National Institutes of Health stroke scale. Three months after stroke, functional outcomes were assessed with the modified Rankin Scale, and the severity of clinical depression and anxiety symptoms were measured using HADS.

On the basis of HADS depression and anxiety cutoffs (≥8) and clinical symptoms based on DSM-5 (Diagnostic and Statistical Manual of Mental Disorders [Fifth Edition]), we determined groups of patients experiencing depression or anxiety at 3 months.

Prosodic Evaluation

Patient voices were recorded within 4 days after stroke onset (baseline) and M3 after stroke using 2 separate tasks: a picture description (Boston Diagnostic Aphasia Examination; BDAE11) and an oral reading story (Lion story, Barbizet). The PRAAT12 software was used by a speech therapist to process recordings.

Several voice parameters were measured: the standard deviation of the fundamental frequency, fundamental frequency range, percentage of voice breaks, and shimmer. The difference between each voice parameter measured at baseline and M3 was calculated.

Statistical Analysis

The characteristics of the depressed and nondepressed patients were compared using Mann–Whitney U tests. Chi-square tests were used to compare sex and stroke laterality. We created a model to predict group assignment through a discriminant analysis. This model is composed of a discriminant function based on linear combinations of prosodic predictor variables. The functions are generated from a sample of observations for which the assignment group is known based on clinical criteria and the HADS at 3 months. We calculated sensitivity and specificity based on predicted group and observed group membership. All statistical analyses were performed using SPSS version 20.0.

Results

Among the 49 patients included in the study, 22.5% developed PSD M3 after stroke. The sociodemographic and clinical characteristics are presented in the Table. None of the patients received thrombolysis in the acute phase; 54% had a right brain lesion.

Table. Comparison of Demographic and Clinical Data

Total (n=49)PSD (n=11)No PSD (n=38)P Value
Baseline
 Age59.26 (±16.26)60.27 (±11.92)59.14 (±17.47)0.883
 % Women22.0044.4418.42<0.001*
 NIHSS0.81 (±1.18)0.55 (±1.21)0.89 (±1.17)0.288
 Lesion laterality (% right)54.0036.3652.630.237
M3
 Rankin0.10 (±0.37)0.09 (±0.30)0.14 (±0.42)0.497
 MOCA27.20 (±2.50)26.45 (±3.04)27.42 (±2.32)0.506
 HADS_A5.24 (±3.68)9.82 (±4.35)6.81 (±4.26)0.046*
 HADS_D7.49 (±4.42)10.64 (±2.46)3.68 (±2.19)<0.001*

HADS indicates Hospital Anxiety and Depression Scale; M3, 3 months; MOCA, Montreal Cognitive Assessment; NIHSS, National Institutes of Health stroke scale; and PSD, poststroke depression.

*P<0.001.

Affective Prosody

Among the prosodic variables assessed by both tasks, only fundamental frequency was associated with patient sex (P<0.001) but not with lesion lateralization. Concerning the reading task, the analyses showed that baseline voice break percentages were significantly higher in patients with PSD (P=0.047) and decreased significantly in this group between baseline and M3 (P=0.002). Shimmer was significantly lower in the PSD group at baseline (P=0.002) and at M3 (P=0.033). Changes in F0 standard deviation were significantly different between the PSD and the non-PSD groups for both the reading and the description tasks (P=0.020 and P=0.012; online-only Data Supplement). There was a significant inverse association between F0 standard deviation and severity of depressive symptoms when adjusting for sex (P=0.015). There was no difference concerning prosody for patients with or without anxiety.

Discriminant Analysis

Discriminant analysis of the baseline voice break percentages and baseline shimmer in the reading task demonstrated discrimination between PSD and non-PSD groups (Wilks λ=0.074; χ2=10.262; P=0.002). Patients with a higher discriminant baseline score based on the prosodic parameters were significantly more depressed at M3 (P=0.003); 77.8% of the original observations were correctly classified (Figure). Sensitivity and specificity based on discriminant scores were, respectively, 0.88 and 0.75 (online-only Data Supplement).

Figure.

Figure. Baseline prosodic discriminant score linked to 3-month depressive symptoms. M3 indicates 3 months; and PSD, poststroke depression. ***P<0.001.

Discussion

The main results of this pilot study are that a combination of prosodic markers measured early after stroke permits the identification of patients at risk of subsequent PSD. Furthermore, lower shimmer at M3 is observed in patients with PSD, and changes in prosodic characteristics between baseline and M3 are observed in individuals with this disorder.

The observation of modified prosodic parameters in patients with PSD is in line with previous reports in major depression,5 reinforcing the hypothesis that both disorders may share similar pathophysiological mechanisms. This result suggests that the investigation of affective prosody could help clinicians to improve assessments. In accordance with previous studies, we observed that fundamental frequency is influenced by sex.13 Because women were over-represented in our study, we adjusted our analysis by sex, and our results were replicated when examining men only. Furthermore, although several studies have suggested the potential influence of left brain lesions on the risk of PSD,14 recent reviews do not support this association,1,4 and there were no significant differences between PSD and non-PSD groups concerning stroke lateralization in our study. We observed significantly higher levels of anxiety symptoms among PSD patients, thus raising concerns about the potential influence of anxiety on prosodic parameters. However, we did not observe an association between prosodic parameters and the severity of anxiety.

The main result of this study is the strong discriminant ability of measurements of baseline prosodic parameters in the identification of patients at risk of PSD at M3. A recent study demonstrated that hearing one’s own voice influences one’s own mood,15 suggesting that early deterioration of affective prosody could, therefore, negatively influence patients’ mood. Fundamental frequency variation is known to be more useful for tracking within-patient changes than interindividual differences.5 For this study, decreases in F0 variability between baseline and M3 in PSD patients were observed, and although this marker is not a predictor at baseline, these findings suggest the potential interest for tracking within-patient changes.

Despite the well-known association between right hemispheric lesions and affective prosody,6 we did not observe differences for stroke location and PSD. Nonetheless, the results of this study should be interpreted cautiously because of several limitations. We applied strict exclusion criteria resulting in a small sample size of patients, and therefore, these results cannot be generalized. We also cannot exclude the potential role of subtle language deficits after stroke, although this risk was likely reduced by the exclusion criteria applied. In conclusion, these results indicate that prosodic characteristics are modified in stroke patients with PSD M3 after stroke and that baseline evaluation of this new biomarker could improve our ability to detect patients at higher risk of PSD and, therefore, improving its management.

Footnotes

The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.116.013852/-/DC1.

Correspondence to Charlotte Cosin, MS, 146 rue Léo Saignat, 33076 Bordeaux, Cedex, France. E-mail

References

  • 1. Hackett ML, Köhler S, O’Brien JT, Mead GE. Neuropsychiatric outcomes of stroke.Lancet Neurol. 2014; 13:525–534. doi: 10.1016/S1474-4422(14)70016-X.CrossrefMedlineGoogle Scholar
  • 2. Hackett ML, Anderson CS. Predictors of depression after stroke: a systematic review of observational studies.Stroke. 2005; 36:2296–2301. doi: 10.1161/01.STR.0000183622.75135.a4.LinkGoogle Scholar
  • 3. Cosin C, Sibon I, Poli M, Allard M, Debruxelles S, Renou P, et al. Circadian sleep/wake rhythm abnormalities as a risk factor of a poststroke apathy.Int J Stroke. 2015; 10:710–715. doi: 10.1111/ijs.12433.CrossrefMedlineGoogle Scholar
  • 4. Ferro JM, Caeiro L, Figueira ML. Neuropsychiatric sequelae of stroke.Nat Rev Neurol. 2016; 12:269–280. doi: 10.1038/nrneurol.2016.46.CrossrefMedlineGoogle Scholar
  • 5. Cummins N, Scherer S, Krajewski J, Schnieder S, Epps J, Quatieri TF. A review of depression and suicide risk assessment using speech analysis.Speech Commun. 2015; 71:10–49. doi: 10.1016/j.specom.2015.03.004.CrossrefGoogle Scholar
  • 6. Heilman KM, Leon SA, Burtis DB, Ashizawa T, Subramony SH. Affective communication deficits associated with cerebellar degeneration.Neurocase. 2014; 20:18–26. doi: 10.1080/13554794.2012.713496.CrossrefMedlineGoogle Scholar
  • 7. Hardy P, Jouvent R, Widlöcher D. Speech pause time and the retardation rating scale for depression (ERD). Towards a reciprocal validation.J Affect Disord. 1984; 6:123–127. doi: 10.1016/0165-0327(84)90014-4.CrossrefMedlineGoogle Scholar
  • 8. Hönig F, Batliner A, Nöth E, Schnieder S, Krajewski J. Automatic modelling of depressed speech: relevant features and relevance of gender.Presented at the Fifteenth Annual Conference of the International Speech Communication Association, Singapore, September 14–18, 2014. https://www5.informatik.uni-erlangen.de/Forschung/Publikationen/2014/Hoenig14-AMO.pdf. Accessed February 2, 2016.Google Scholar
  • 9. Schirmer A, Kotz SA. Beyond the right hemisphere: brain mechanisms mediating vocal emotional processing.Trends Cogn Sci. 2006; 10:24–30. doi: 10.1016/j.tics.2005.11.009.CrossrefMedlineGoogle Scholar
  • 10. Zigmond AS, Snaith RP. The hospital anxiety and depression scale.Acta Psychiatr Scand. 1983; 67:361–370.CrossrefMedlineGoogle Scholar
  • 11. Goodglass HK, Kaplan E. Assessment of Aphasia and Related Disorders. Philadelphia, PA: Lea & Febiger; 1972.Google Scholar
  • 12. Boersma P. Praat, a system for doing phonetics by computer.Glot Int. 2002; 5:341–345.Google Scholar
  • 13. Titze IR. Physiologic and acoustic differences between male and female voices.J Acoust Soc Am. 1989; 85:1699–1707. doi: 10.1121/1.397959.CrossrefMedlineGoogle Scholar
  • 14. Rajashekaran P, Pai K, Thunga R, Unnikrishnan B. Post-stroke depression and lesion location: a hospital based cross-sectional study.Indian J Psychiatry. 2013; 55:343–348. doi: 10.4103/0019-5545.120546.CrossrefMedlineGoogle Scholar
  • 15. Aucouturier JJ, Johansson P, Hall L, Segnini R, Mercadié L, Watanabe K. Covert digital manipulation of vocal emotion alter speakers’ emotional states in a congruent direction.Proc Natl Acad Sci U S A. 2016; 113:948–953. doi: 10.1073/pnas.1506552113.CrossrefMedlineGoogle Scholar

eLetters(0)

eLetters should relate to an article recently published in the journal and are not a forum for providing unpublished data. Comments are reviewed for appropriate use of tone and language. Comments are not peer-reviewed. Acceptable comments are posted to the journal website only. Comments are not published in an issue and are not indexed in PubMed. Comments should be no longer than 500 words and will only be posted online. References are limited to 10. Authors of the article cited in the comment will be invited to reply, as appropriate.

Comments and feedback on AHA/ASA Scientific Statements and Guidelines should be directed to the AHA/ASA Manuscript Oversight Committee via its Correspondence page.