Early detection of decompensation of chronic heart failure using a non-contact monitor of nocturnal respiratory patterns

Abstract

Heart failure affects 1-2% of the adult population in the United Kingdom and accounts for the majority of hospitalisations in patients with cardiovascular disease. The financial implications are enormous as it consumes 1-2% of the national health care budget with 70% of these costs relating to hospitalisation expenses. Prevention of these admissions may be possible by detecting early signs of decompensation in patients with chronic heart failure (CHF) and instituting interventions that may steer the course of disease back to stability without the need for a hospital inpatient stay. Further, Sleep Disordered Breathing (SDB) and in particular Central Sleep Apnoea (CSA) is found in patients with CHF and at any symptomatic stage of the condition. This may be associated with Cheyne-Stokes Respiration (CSR), which has been shown to be an independent predictor of mortality. In the first study of this thesis, I investigated the accuracy of the SleepMinderTM (SM) device; which is a non-contact monitor of nocturnal respiratory patterns; in diagnosing SDB by deriving measures of the Apnoea Hypopnea Index (AHI) and percentage overnight CSR from the SM signals. I found that SM was good in terms of diagnostic accuracy with an area under receiver operator characteristic curve (ROC) of 0.82 (p=0.02) for an AHI threshold >15, but only moderately so for % overnight CSR>0, with an area under ROC curve of 0.72 (p=0.06). In the second study, I examined the changes that occur in SM derived respiratory parameters over a long period of monitoring and found that the AHI, quantity of CSR, Total Sleep Time (TST) and Respiratory Rate (RR) were highly variable with Intra-Class Correlation (ICC) measures of 0.32, 0.39, 0.25, 0.36 respectively over a period of 12 months. Relying on data from a year rather than a single night resulted in misclassification of patients into a different severity group of SDB during 35% of the follow up period and placed patients into a different treatment group during 21% of this period. I also observed that a high proportion (59%) of patients studied had a mean AHI that was consistently above the accepted threshold for treatment (AHI>15). This was consistent even over a shorter follow up period of 2 weeks suggesting that a single night measure of the AHI may not be a sufficient risk assessment of SDB in heart failure patients. In the final study, I have investigated the predictive value of the SleepMinderTM for acute decompensation of heart failure (ADHF) using algorithms derived from its signals. I found that the SM was not accurate for this purpose, performing with a sensitivity and specificity of 0.38 and 0.71, respectively. In summary this study has demonstrated that the SleepMinderTM device provides a novel screening method, which is convenient for the detection of sleep disordered breathing in patients with CHF. It performs with a good diagnostic accuracy and is acceptable to these patients due to its non-contact operation. Algorithms derived from its signals however cannot be used to predict acute decompensation of chronic heart failure. Further, longitudinal analyses of nocturnal respiratory patterns in these patients have demonstrated that the Apnoea Hypopnea Index (AHI) is highly variable over a prolonged period of monitoring and a mean value rather that a single night measurement may be a more appropriate risk assessment tool for SDB. This requires confirmation.