A lengthening expiratory phase can signal a COPD exacerbation before the patient recognises a meaningful change. The physiological rationale for acoustic monitoring is straightforward, and understanding it is important when assessing or challenging the technology. Airway changes that come before a severe exacerbation start well before symptoms become obvious. They also alter the sounds of breathing. Appropriate sensors and signal processing can detect and quantify those changes remotely.
This article covers the physiology, shows what early exacerbation sounds like acoustically, and sets out where the signal performs well and where it is less dependable.
The Physiology Behind Early Exacerbation
Viral upper respiratory tract infection is the most common trigger for a COPD exacerbation. Bacterial infection and exposure to environmental irritants are also important causes. Early inflammation in the airways increases mucus production, mucosal oedema, and smooth muscle bronchoconstriction. These processes also occur in stable COPD, but during an exacerbation they develop more quickly and with greater intensity.
The consequence is progressively restricted airflow. At first, the effect is greatest during forced expiration, which is why FEV1 remains the standard diagnostic marker. However, the change also appears during tidal breathing. Dynamic airway collapse, excess secretions, and narrowed airways together alter the sounds produced through each breath cycle. The expiratory phase becomes longer compared with inspiration, high-frequency crackles become denser in the early morning, and intermittent wheeze can emerge at frequencies absent from the patient's baseline acoustic profile.
These acoustic changes generally come 48 to 96 hours before the patient feels breathlessness. People with COPD often compensate for increasing obstruction by changing their activity and breathing pattern before they consciously identify deterioration. They slow down, pause more on stairs, and sit more often without necessarily viewing this as a change in their condition. That compensation does not obscure the acoustic signal to the same extent.
Acoustic Features Seen in Pilot Data
In our pilot, we measure four main acoustic features for each patient: inspiratory-to-expiratory ratio (I:E ratio), crackle density score (CDS), wheeze index (WI), and overnight respiratory rate (ORR). A personal baseline for each feature is set during the first four to six weeks of monitoring. Alerts assess departures from that baseline, rather than a single absolute threshold.
The I:E ratio is the most consistent early indicator in our pilot data. At rest, a normal I:E ratio is approximately 1:2. In early COPD exacerbation, it moves towards 1:3 or longer as airway emptying takes more effort. In patients who later need clinical intervention, we see this change earlier on average than changes in the other features.
The next most consistent early finding is higher crackle density. COPD crackles are usually coarse, reflecting secretion movement and peripheral airways snapping open. More crackles in the lower lobes during quiet tidal breathing suggest increasing secretion retention, a recognised precursor of infective exacerbation.
"Pilot data" needs a clear definition here. Our enrolled cohort is small, the findings have not undergone independent peer review, and we are reporting patterns rather than statistically validated thresholds. Those are the limits of what we can responsibly claim today.
When the Signal Can Be Trusted
The acoustic signal is clearest when COPD is the main pathology and monitoring has continued long enough to establish a meaningful individual baseline. In such patients, drift in the I:E ratio together with rising crackle density gives a clear directional indication several days before symptomatic crisis.
Reliability falls in several situations. Patients with substantial cardiac comorbidity, especially heart failure, may have acoustic profiles resembling early exacerbation. Fine bibasal crackles, tachypnoea, and a reduced I:E ratio can arise from either condition. The system does not distinguish between these causes, and clinicians should not expect it to. For a patient with known cardiac disease, an alert should lead to clinical assessment covering cardiovascular and respiratory causes.
Current smokers and people who smoke heavily in enclosed domestic spaces generate more ambient noise in overnight recordings, lowering confidence in feature extraction. Monitoring can still have value for this group, but the signal-to-noise ratio is lower. Alert thresholds may therefore require different calibration to limit false positives.
What Follows an Alert
An alert is not a diagnosis. It signals the need for a proportionate clinical response. In our pilot, the monitoring coordinator makes a same-day telephone call after an alert. A structured question set covers current breathlessness score, recent sputum changes, medication adherence, and any rescue medication taken in the past 48 hours. The call takes approximately eight minutes.
The call assessment guides what happens next. In most pilot alerts, it does not lead to emergency escalation. More often, the response is a rescue course of steroids or antibiotics, an earlier review, or referral to community physiotherapy. The benefit is not mainly finding patients who need immediate emergency care. It is finding deterioration that, without treatment, could require emergency care in 48 to 72 hours.
That distinction is central. We are not replacing emergency response. We are working to make it less necessary.
The Evidence Gap We Aim to Close
The clinical rationale for acoustic early warning is sound. The key issue is whether deployment in real homes, with real patients and clinical workflows, improves outcomes beyond current telephone follow-up and patient self-reporting. A properly designed clinical study is needed, and we are working towards one. This is not simply a disclaimer. Clinicians adopting early-stage technology should know precisely what is being claimed and the basis for it. Detecting an early signal is not the same as proving better outcomes. We are working towards that evidence and do not claim to have reached it.