Daily smartphone audio recordings reveal subtle changes in asthma and COPD exacerbations from day one, highlighting a promising new approach to remote monitoring that may one day enable early intervention and improve patient care.
Study: Speech as a biomarker for early exacerbation detection in asthma and COPD: the TACTICAS study. Image credit: Pixel-Shot/Shutterstock.com
Asthma and chronic obstructive pulmonary disease (COPD) are chronic, painful, and often disabling conditions that frequently worsen. Early detection of deterioration may allow timely treatment and improve outcomes. Recent research published in journals ERJ Open Research We found that voice changes can occur from the first day of deterioration, supporting the future use of voice analysis for remote monitoring and early treatment.
Speech analysis offers a new monitoring approach for asthma and COPD
Asthma and COPD exacerbations are episodes of sudden worsening of symptoms, often leading to decreased quality of life, increased drug use, hospitalization, and increased risk of death. Finding reliable ways to detect these recurrences soon after they begin is an important goal in managing both conditions, as early treatment leads to better outcomes.
Current approaches to remote monitoring rely on measures such as pulmonary function tests, blood samples, and vital signs, which can be invasive, labor-intensive, and require specialized equipment. So researchers are turning to digital speech biomarkers, as changes in the airway during an exacerbation can change the way a person speaks. Although voice analysis has shown promise in detecting a variety of other health conditions, its potential for monitoring exacerbations of asthma and COPD remains largely unexplored.
Patients had audio recordings throughout a 3-month monitoring period
The current study aimed to use specific vocal characteristics as biomarkers for the onset of exacerbation and different types of exacerbation symptoms, and to understand how voice changes reflect exacerbation of these two conditions. We also investigated how well participants could follow a voice-based app for daily monitoring.
Researchers conducted a prospective cohort study (TACTICAS (Remote Monitoring of Asthma and COPD through Speech Analysis) Study) that included 73 participants recruited at a respiratory clinic. Of these, 35 had asthma and 38 had COPD, with an average age of 62 years.
Participants used their mobile phones to record their own voices three times a day for three months. The app prompted participants to record audio in the morning, after some light exercise in the afternoon, and again in the evening alongside a symptom questionnaire. They used the TACTICAS app, co-created with asthma and COPD patients and designed for home use.
Audio capture included sustained vowels and reading a paragraph or answering a question. Participants also used a validated and accurate questionnaire to detect an exacerbation by a symptom score increase of 9 points or more from baseline for 3 days or a 12 point or more increase in score for 2 consecutive days.
Additionally, researchers contacted participants every four weeks and stored data on medication and symptom increases in an electronic case record form (eCRF).
Exacerbations were clinically verified using a combination of electronic health records (EHR) and medication use and patient-reported symptoms, and only verified events were included in the analysis. Events that were not reported but were detected by eCRF or EHR analysis were included as well.
The researchers obtained a total of 23,799 records, 2,737 of which occurred during an exacerbation. Of these, 227 were on the day of onset, recovery, or peak. During the study period, they recorded 38 exacerbations, most of which were mild to moderate. Most were captured through questionnaires.
Changes in voice appeared from the beginning of the deterioration
Analysis revealed that visible changes in speech were already present on the first day of exacerbation. Multilevel analysis identified 13 acoustic phonetic features that varied significantly at onset, including lower median pitch, shorter sustained vowels, and larger shimmers indicative of irregular vocal fold vibrations. The noise-to-harmonic ratio also increased, suggesting a decrease in overall audio quality.
These changes did not occur only at the beginning of the episode, but evolved during the exacerbation. By collecting daily voice recordings, the researchers were able to track how the voice changes from onset, peak of symptoms, and recovery, providing a detailed picture of the progress of the exacerbation.
The study also found that different vocal features appear to reflect different symptom patterns. Shimmer and noise-to-harmonic ratio were more strongly associated with chest symptoms. This may be due to decreased airflow during speech due to increased airway obstruction, whereas minimum pitch showed a strong relationship with breathlessness. The authors caution that these findings are exploratory and require confirmation in larger studies.
Interestingly, voice changes were similar in participants with asthma and COPD. The researchers suggest that this may indicate that underlying physiological changes that occur during an exacerbation are causing the speech changes, rather than a specific diagnosis. These changes can result from a combination of increased airway obstruction, sputum retention, inflammation of the vocal cords, and changes affecting the larynx. However, they also stressed that the study was not large enough to determine this with certainty.
The authors propose that these changes result from worsening airway obstruction during an exacerbation, resulting in mechanical strain on both the respiratory system and structures involved in speech production. When airway resistance increases, the respiratory muscles work harder, lung expansion becomes more restricted, and the pressure under the vocal cords decreases.
This can make it more difficult to maintain a steady voice, reduce airflow and produce shorter speaking times, increased sparkle, and a more breathy voice. As participants recovered, the pitch of their loudest voice increased. The researchers suggest that this may reflect the restoration of more normal speech mechanisms.
The study also highlights the practical benefits of audio monitoring. Participants were less likely to complete symptom questionnaires when they felt unwell, potentially hindering questionnaire-based detection of exacerbations. In contrast, recording short audio samples may provide an easier and less burdensome way to monitor respiratory health and could be an inexpensive and unobtrusive tool for future home-based monitoring.
Although the findings are promising, the authors emphasize that larger validation studies and predictive models are needed before voice-based biomarkers can be incorporated into routine clinical care.
Observational studies identify opportunities for future research
To increase confidence that each flare-up represented a genuine exacerbation, the researchers verified the events using multiple sources of information, including symptom questionnaires, electronic health records, drug use, and clinical reviews. However, they did not use the Rome criteria, a more recent framework for defining COPD exacerbations that incorporates physiological measurements such as heart rate, respiratory rate, oxygen saturation, and inflammatory markers along with symptoms.
The authors note that because these vital signs were not collected in this study, some exacerbations may be classified differently under the Rome definition. They also point out that heart rate itself can influence the characteristics of speech, making it an important factor to consider in future research.
Additionally, the findings are observational and represent associations rather than predictions. Voice also changed with exacerbation, but the study cannot determine whether these changes reliably predict an impending flare-up. This distinction is especially important because many other acute illnesses can also affect speech, such as heart failure and common respiratory infections.
Some additional limitations should also be considered. Although inhaled corticosteroids can change voice quality, the researchers did not systematically record participants’ daily inhaler use. Additionally, the study was relatively small and only included native Dutch speakers, limiting the applicability of the findings. Additionally, numerous statistical comparisons raised the possibility that some significant associations were due to chance.
Future research should validate these findings in larger multilingual populations and further refine the predictive algorithms already in development. If voice-based prediction proves to be feasible, the authors suggest that it could ultimately support treatment pathways by prompting additional diagnostic tests when early signs of exacerbation are detected.
Smartphone voice analysis could support future respiratory care
Taken together, this study shows that measurable vocal changes occur from the very beginning of an asthma or COPD exacerbation.
These results demonstrate the potential of voice as a biomarker for detecting exacerbations of chronic respiratory diseases and demonstrate the utility of home monitoring via mobile phone.
However, the authors emphasize that larger validation studies and predictive models are needed before voice-based biomarkers can be used clinically for remote exacerbation detection.
Download your PDF copy now.
