Telemonitoring to Reduce COPD Exacerbations: A Work in Progress

COPD exacerbations are highly consequential to patients and society. They contribute to disease progression^1–3 and mortality^4,5 while incurring high health care costs. In a recent analysis, national medical costs attributable to COPD were projected to top $60.5 billion in 2029.⁶ An estimated 45–70% of the total cost is ascribed to management of exacerbations.⁷ To that end, early prediction of COPD exacerbations and timely intervention for prevention of a full-blown event are highly attractive propositions.

Telemonitoring refers to remote patient monitoring of digitally transmitted health-related data with the purpose of detecting changes in patient condition.⁸ In the context of COPD, telemonitoring has been utilized to serve multiple purposes including facilitating self-management, assessing adherence, and detecting early disease decompensation.⁹ Despite the massive interest in the topic, high-quality clinical studies are few and far between, likely due to the cost and the implementation challenges. In the only randomized, controlled trial, to this author’s knowledge, testing the impact of telemonitoring on COPD exacerbations, Andersen and colleagues demonstrated a lower number of hospitalizations caused by exacerbations of COPD without any change in the risk of having at least 1 hospitalization or time to first hospitalization compared with usual care.¹⁰

In this issue of Respiratory Care, Kronborg et al report the results of their single blinded randomized controlled study that tested the effect of a telemonitoring prediction algorithm on preventing COPD exacerbations.¹¹ The algorithm integrated measurements of oxygen saturation, blood pressure, pulse rate, and weight, alerting COPD nurses to act when high risk for exacerbation was identified. The data were also available to the clinicians who assessed risk of exacerbation in the control group but without the risk prediction algorithm. The methodology aimed to isolate the effectiveness algorithm support over that of clinical assessment in preventing COPD exacerbations. The groups were not well matched with slightly higher quality of life scores, larger proportion of male gender, higher educational status, and presence of comorbidities in the intervention group. Notably, patients in the intervention group had higher health utilization at baseline with history of more acute hospitalizations and out-patient visits per patient compared with the control group. After 6 months of telemonitoring, the between-group difference in acute hospitalizations, the primary outcome, was not significant in unadjusted and adjusted analyses.

This study had significant limitations, some of which reflect the challenges commonly encountered in implementation science in telemonitoring. First, the study did not recruit the number of subjects as required in the sample size calculation (137 vs 200 subjects) and attrition was greater than expected during the follow-up. This is somewhat surprising since the general attitude of patients with COPD toward telemonitoring is reported to be positive.^12,13 Second, nearly 40% of the follow-up data were missing and required imputation. Third, the clinical interventions for patients deemed at high risk for exacerbation in both intervention and control groups were not standardized. Accurate prediction will provide value only if there is an effective preventive intervention delivered in a standard way. Fourth, the single blind design meant COPD health professionals knew group assignments and may have treated their respective groups under the influence of personal views toward the intervention, introducing a performance bias. Fifth, as having had history of exacerbation is the most important predictor of future exacerbations, an imbalance between the groups is an important limiting factor in interpreting results.

What are the learnings from the study by Kronborg et al for COPD telemonitoring clinical trialists? The large proportion of missing data may imply that telemonitoring procedures should be further simplified to promote patient adherence. Instead of scheduled data transmission that relies on patient collection, continuous and noninvasive parameters obtained via wearable sensors may be preferable. Patient surveys could be made more accessible (eg, via voice interactive tools), to augment response rates. In order to ensure balanced groups, leading risk factors for exacerbations such as severe lung disease and previous history of exacerbations can be used to stratify the randomization procedure. Importantly, use of telemonitoring is likely to have the highest impact among patients with high health care utilization, which can serve as an inclusion criterion for future studies. How the intervention is structured is of paramount importance. In today’s financial climate, a majority of health care systems will not be able to afford routine employment of a separate clinician team in ambulatory patient monitoring. Furthermore, the prediction tool utilized in the trial was previously reported to have outstanding discriminatory performance (area under the curve 0.95, ie, ability to correctly classify exacerbation 95 out of 100 clinical scenarios).¹⁴ One might argue that such a powerful tool eliminates the need for further adjudication by humans. Once a potential exacerbation is identified, a standard management pathway (implemented by the patient’s own primary care physician or office) can be followed. Since telemonitoring programs will also be judged on cost-effectiveness, study designs with minimal human interference will have the highest chance of adoption. If clinician involvement is planned, alert thresholds can be adjusted to improve precision and recall, preventing alert fatigue and negative bias toward the prediction tool. In the case of this trial, the prior excellent performance of the tool¹⁴ could have been introduced to the clinical nurses at the outset to instill higher confidence in the warnings.

Kronborg and colleagues have developed a powerful prediction tool for COPD exacerbations, which was integrated into their telehealth system. They designed a clinical study to determine effectiveness of their predictive tool and shared their negative findings. Their thoughtful approach to research and willingness to share the work are commendable. In doing so, they provide important lessons for those who are invested in making telemonitoring for COPD successful. As Bill Gates said, it is fine to celebrate success, but it is more important to heed the lessons of failure.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

No funding was received for this article.