Monitoring of Physiological Parameters to Predict Exacerbations of Chronic Obstructive Pulmonary Disease (COPD): A Systematic Review

Introduction: The value of monitoring physiological parameters to predict chronic obstructive pulmonary disease (COPD) exacerbations is controversial. A few studies have suggested benefit from domiciliary monitoring of vital signs, and/or lung function but there is no existing systematic review. Objectives: To conduct a systematic review of the effectiveness of monitoring physiological parameters to predict COPD exacerbation. Methods: An electronic systematic search compliant with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines was conducted. The search was updated to April 6, 2016. Five databases were examined: Medical Literature Analysis and Retrieval System Online, or MEDLARS Online (Medline), Excerpta Medica dataBASE (Embase), Allied and Complementary Medicine Database (AMED), Cumulative Index of Nursing and Allied Health Literature (CINAHL) and the Cochrane clinical trials database. Results: Sixteen articles met the pre-specified inclusion criteria. Fifteen of these articules reported positive results in predicting COPD exacerbation via monitoring of physiological parameters. Nine studies showed a reduction in peripheral oxygen saturation (SpO2%) prior to exacerbation onset. Three studies for peak flow, and two studies for respiratory rate reported a significant variation prior to or at exacerbation onset. A particular challenge is accounting for baseline heterogeneity in parameters between patients. Conclusion: There is currently insufficient information on how physiological parameters vary prior to exacerbation to support routine domiciliary monitoring for the prediction of exacerbations in COPD. However, the method remains promising.


Introduction
Chronic obstructive pulmonary disease (COPD) is a serious health matter, which significantly impacts the individual's quality of life. According to the World Health Organisation, in 2004, 65 million people were diagnosed with COPD globally [1]. In 2012, three million people died because of COPD [2], and thus COPD is anticipated to be the third leading cause of death by 2020 if no action is taken [3]. COPD, even when optimally managed, is associated with periodic deteriorations in respiratory health called exacerbations. Exacerbations are defined in the Global Initiative for Chronic Obstructive Lung Disease (GOLD) document "as an acute event characterised by a worsening of the patient's respiratory symptoms that is beyond normal day-to-day variations and leads to a change in medication" [4]. Exacerbations can lead to decline in the patient's overall function, causing hospitalisation, and/or death. Therefore, health care facilities, societies, and individuals have a common interest in better understanding how to prevent and manage exacerbations, reduce disease progression, and support patient self-management. To achieve this, early detection of exacerbations and prompt access to therapy and health services are needed. Detecting COPD exacerbation earlier will allow prompt initiation of treatment [4]; therefore facilitating faster recovery and outcomes. This may result in a reduced number of hospital admissions, and as well as a reduction in healthcare consumption.
It is recognised that whilst defined by changes in symptoms, exacerbations are also associated with alterations in physiological variables. In 2010, Hurst et al. [5] examined the ability of domiciliary pulse oximetry and peak flow to distinguish exacerbations from day to day fluctuations. They reported that changes in heart rate, peripheral oxygen saturation (SpO 2 %), and peak flow were significantly different just before and during an exacerbation. Rapid advancement in technology has offered numerous solutions targeting the management of chronic diseases (collectively known as tele-health). Tele-health is a form of distance communication between the patient and the healthcare provider for monitoring, communicating, managing or facilitating intervention [6]. Tele-health may monitor symptoms, and/or physiology parameters. Tele-health has shown some success in chronic disease management. The PROMETE study conducted in 2014 in Spain by Segrelles et al. reported a reduction in acute noninvasive ventilation (NIV) usage (p < 0.0001), emergency department (ER) visits (p = 0.001), admissions (p = 0.015) and bed days (p = 0.018) [7]. More recent studies in COPD have not been positive [8], perhaps reflecting the heterogeneity of COPD.
The objective of this systematic review was to summarise and report the value of domiciliary physiological monitoring in predicting exacerbations in patients with COPD.

Search Strategy
This systematic review (PROSPERO registration CRD42016046643) is compliant with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) standards [9]. The search was completed up to April 6, 2016. The search was performed in Medical Literature Analysis and Retrieval System Online, or MEDLARS Online (Medline), Excerpta Medica dataBASE (Embase), Allied and Complementary Medicine Database (AMED), Cumulative Index of Nursing and Allied Health Literature (CINAHL), and the Cochrane clinical trials database. The search terms used are detailed in the Appendix A, Tables A1 and A2. In addition to the electronic database search, the reference list of eligible articles was also screened.

Inclusion Criteria
The studies included in this review met the following criteria: (1) Stable COPD; (2) Domiciliary monitoring; (3) Monitoring any physiological variables; (4) Reporting statistical analysis of the measured physiological variables; (5) Prediction of exacerbations via physiological variables.
The main outcome of interest was variation in physiological parameters before and during COPD exacerbations, and the ability of measuring changes in physiological variables to provide early detection of COPD exacerbations.

Data Collection
Screening of the titles and abstracts was performed by the first author to eliminate all non-relevant studies. Titles and abstracts potentially relevant were read in full-text to evaluate if they were eligible or not. In addition to screening and evaluating for eligibility, the reference list of the eligible articles was screened. The second author confirmed the eligibility. Disagreement on five studies between authors was resolved after discussion.

Quality Assessment
The quality assessment was performed by each author individually based on two different modified scales, the Cochrane tool [10] and Newcastle-Ottawa scale [11]. The Cochrane quality assessment tool consists of seven questions to evaluate randomised studies included in this review. The Newcastle-Ottawa scale consists of seven questions used to assess cohort and non-randomised studies included in this review. The assessment was performed by each author individually and any disagreement was solved by discussion.

Synthesis of Results
The primary purpose of this systematic review was to assess the feasibility of predicting COPD exacerbations by domiciliary monitoring of physiological parameters. Because of significant methodological heterogeneity between included studies, meta-analysis was not conducted. However, a narrative synthesis of the results of the studies was performed and full details of the included studies are reported in Tables 1 and 2.   If abnormal values of respiratory rate and % triggered breaths were reported for two days or more, or abnormal values of NIV daily usage were reported for three days or more out of five. Abnormal values were defined as "value of a parameter was >75th or <25th percentile, the day was recorded as abnormal value' ('high value' > 75th, 'low value' < 25th).
Hamad et al., 2016 [25] Admission to hospital, or started antibiotics or/and steroids.

Results
The systematic review search generated 3377 articles, 345 were excluded due to duplication. After screening the titles and abstracts, 28 articles out of 3032 were potentially relevant to the inclusion criteria. After that, full-text screening of the 28 articles was conducted to assess eligibility, which resulted in 13 relevant articles. The reference list of the relevant articles was also examined which resulted in identification of three further articles giving 16 in total ( Figure 1).
Of the 16 articles that met the pre-specified inclusion criteria, all the studies were conducted prospectively, and in seven different countries: one each in Australia, Denmark, France, Italy, Netherlands, four in Spain, and eight in the United Kingdom. Most of the articles were published in 2015 (5/16), with three in 2012, two each in 2009 and 2013, and one each was published in 2000, 2010, 2014 and 2016. The sample size and duration of the studies varied from three months to fifteen months except for one study, which was run for 30 months. The sample size varied from 16 to 183 participants (eight studies <50 patients, five studies ≥50 patients, and three studies >100 patients). Fifteen studies were on COPD patients only (at different disease stages), and one was on heart failure and chronic lung disease patients [16]. Full details of the included studies are reported in Tables 1 and 2. The sample size and duration of the studies varied from three months to fifteen months except for one study, which was run for 30 months. The sample size varied from 16 to 183 participants (eight studies <50 patients, five studies ≥50 patients, and three studies >100 patients). Fifteen studies were on COPD patients only (at different disease stages), and one was on heart failure and chronic lung disease patients [16]. Full details of the included studies are reported in Tables 1 and 2.

Quality Assessment
Among the 16 identified articles, four studies were randomised clinical trials and 12 were cohort studies. The four studies evaluated using the modified Cochrane risk of bias tool [10] were ranked as being at high risk of bias. The 12 studies evaluated by the modified Newcastle-Ottawa scale [11] were all ranked as moderate quality except for one, which was ranked as low quality.

Heart Rate and Oxygen Saturation
Most of the included studies 14/16 monitored the participant's vital signs and assessed the capability of vital signs to predict COPD exacerbation. Although heart rate (HR) and oxygen

Quality Assessment
Among the 16 identified articles, four studies were randomised clinical trials and 12 were cohort studies. The four studies evaluated using the modified Cochrane risk of bias tool [10] were ranked as being at high risk of bias. The 12 studies evaluated by the modified Newcastle-Ottawa scale [11] were all ranked as moderate quality except for one, which was ranked as low quality.

Heart Rate and Oxygen Saturation
Most of the included studies 14/16 monitored the participant's vital signs and assessed the capability of vital signs to predict COPD exacerbation. Although heart rate (HR) and oxygen saturation (SpO 2 %) were monitored in 10/16 studies [5,7,13,15,[18][19][20][21]23,25], 7/10 studies did not report any statistical analysis for the HR and SpO 2 % variation. However, they concluded with the possibility that heart rate and/or SpO 2 % may be useful in detecting deterioration. Four studies (three at moderate quality, and one at high risk of bias) reported a significant variation (p ≤ 0.05) in HR and/or SpO 2 % prior to the onset of COPD exacerbation [5,18,23,25]. In Hurst et al. [5], the magnitude of the fall in SpO 2 % two days into the exacerbation was −1.24 standard deviation (SD) and the rise in HR was +3.09 SD above the patient's baseline. Martin-Lesende et al. [18] reported the difference between the mean values monitored over the whole study period, which were for SpO 2 % 93.1% (2.2 SD), and for HR 77.8 min −1 (14.6 SD); Moreover, the mean values monitored over the five days prior to cause-specific admission were for SpO 2 % 91.0% (4.6 SD) and for HR 84.2 min −1 (17.1 SD), p = 0.003 for both. There was therefore a typical rise in HR of 7 min −1 and fall in SpO 2 % of 2%. Burton et al. [23] reported that the magnitude of SpO 2 % fall and HR rise was approximately 1 SD (SpO 2 % fall from 93.6% to 92.4%, and HR increased from 87.4 min −1 to 93.7 min −1 ).

Respiratory Rate
The works of Yanez and Borel, which were ranked as moderate quality [17,24], evaluated variations in respiratory rate prior to an exacerbation. In both, the change was statistically significant (p ≤ 0.05). Importantly Yanez et al. reported an increase in the mean respiratory rate one to five days prior to hospitalisation due to an acute exacerbation. At 48 h, the mean respiratory rate increased by 2.3 min −1 (15% from baseline) with 72% sensitivity and 77% specificity (area under the curve (AUC) = 0.76, p < 0.05) for detecting exacerbation, whilst the rise noted 24 h prior to hospitalisation at 4.4 min −1 (30% from baseline) had a 66% sensitivity and 93% specificity (AUC = 0.79, p < 0.05) for exacerbation detection. At five days before hospitalisation, the mean respiratory rate rose from 15.2 ± 4.3 min −1 to 19.1 ± 5.9 min −1 (p < 0.05) suggesting a longer window for preventing hospitalisation. However, in contrast, Martin Lesende [18] did not see significant change in the respiratory rate five days before hospitalisation. Mohktar [21] included respiratory rate with daily monitored variables, but no analysis was reported.

Blood Pressure and Temperature
Four studies of 16 (two at high risk of bias and two at moderate quality) included blood pressure monitoring [7,15,18,21], but there was no evidence indicating changes in blood pressure was as a variable with high predictive capacity for exacerbation (p-value not significant). Likewise, body-temperature was monitored in 4 out of of 16 studies. Martin-Lesende [18] compared the mean temperature in the overall follow-up period, 35.9 • C (0.4SD), to the mean of five days, 35.5 • C (1 SD), prior to cause-specific admission. Changes in body temperature resulted in 27.8% of alerts (only 5.6% of alerts were due to an increased temperature over 37 • C). Hamad [25] reported increased body-temperature in 9 out of 98 exacerbations.
Five studies (two at high risk of bias and three at moderate quality) out of 16 [13,15,19,20,22] did not provide sufficient statistical analysis of changes in vital signs despite reporting these variables. For example, Pedone [19] evaluated the capability of a tele-monitoring system for lower hospitalisation rates, and to identify COPD exacerbation onset. The researchers did not report whether the result was statistically significant but noted a 33% reduction in the risk of hospitalisation. Pedone also noted a fall in SpO 2 % in three days preceding the onset of an exacerbation, which therefore led to prediction of COPD exacerbation. Furthermore, Jensen [15] tried to develop an algorithm to enhance the prediction of COPD exacerbation. The four variables heart rate, systolic blood pressure, diastolic blood pressure, and oxygen saturation were monitored and classified into 273 features. Jensen reported that their system was able to distinguish ten COPD exacerbations with 70% sensitivity, 95% specificity, and 0.73 AUC.
Considered together, SpO 2 % was the most studied variable before an exacerbation episode, and the variable which has been reported to have the highest predictive capacity although the magnitude of change is typically small (1%-2%).

Monitoring Lung Function to Predict Exacerbations
Lung function, particularly spirometry, is a valuable test for diagnosing COPD and evaluating disease progression. A few studies assessed the usefulness of lung function variables in predicting acute exacerbation. Eight studies (two at high risk of bias, one as low quality, and two at moderate quality) of 16 [5,7,[12][13][14]16,21,23] monitored either the peak expiratory flow rate (PEFR), or the forced expiratory volume in one second (FEV 1 ), or both. Three studies [12,13,23] measured FEV 1 and PEFR at different frequencies (per day/per week). Seemungal et al. [12] reported data from 101 COPD patients on PEFR, FEV 1 and vital capacity (FVC) on the day of exacerbation onset, and showed significant changes (p < 0.001). The analysis of 504 COPD exacerbations revealed a fall in the median PEFR of 8.6 (interquartile range (IQR) 0 to 22.9) L/min, FVC of 76.0 (IQR −40.4 to 216.4) mL, and FEV 1 of 24.0 (IQR −16.1 to 84.3) mL. Burton et al. [23] reported a strong correlation between FEV 1 and PEFR and a 0.1 L reduction in FEV 1 was associated with a change in the symptom score.
Sund et al. at low quality and Mohktar et al. at moderate quality [14,21] focused only on FEV 1 . Sund [14] detected 55/75 exacerbations using monitoring, and 6/55 exacerbations were detected only via FEV 1 (defined as a 10% fall in FEV 1 for ≥2 consecutive days). Three studies [5,7,16] examined predicting COPD exacerbations with daily monitoring of PEFR. Segrelles [7] did not report detailed PEFR data, but reported that PEFR and SpO 2 % were the most predictive variables. Hurst [5] reported a statistically significant variation in PEFR before and during an acute exacerbation with a maximal −2.97 SD fall in PEFR four days into the exacerbation. However, Berge [16] reported a significant decrease in the mean of PEFR during an exacerbation episode, which was back to baseline in two weeks.

Monitoring Respiratory Sounds to Predict Exacerbations
In 2015 Fernandez-Granero at moderate quality [22] reported a study demonstrating that 25 out of 33 COPD exacerbations could be detected via monitoring patient's respiratory sounds at home. Each participant was asked to record his/her respiratory sounds daily by placing a microphone on the suprasternal notch. Exacerbation episodes were detected 5 ± 1.9 days prior to the exacerbation onset with a sensitivity of 73.76% and 97.67% specificity.

Alarm limits
A challenge in COPD is the variation between patients and how to set alarm limits for an individual patient. Of the 16 articles included in this review, only eight studies (three at high risk of bias, one at low quality and two at moderate quality) [5,13,14,[18][19][20][21]25] mentioned that they had customised the alarm limits for each individual. Methods used were reported in six out of the eight studies. Cooper [13] monitored the participants for two weeks to identify the normal range for each and personalise the alert limits. Sund [14] set a baseline for each participant by taking the median and the mean after monitoring symptoms and FEV 1 for 14 days (exacerbation-free). In the Hurst study [5], heart rate, oxygen saturation, and peak expiratory flow rate assessed for 30 days (symptom-free). These established a baseline of the selected variables with ±SD. Pedone [19] customised the limits based on the participant's "clinical situation". Harding [20] personalised each participant's limits by applying a probability density function after monitoring the participant for six weeks, or having 40 sets of recorded daily data. Mokhtar [21] personalised the limits range in a different way; they took the median (50th percentile), lower (25th percentile), and upper (75th percentile). They then adjusted the lower limits to be 25th percentile minus 1.5 times the interquartile, and the upper limits to be 75th percentile plus the 1.5 times the interquartile. There are no studies comparing different methods of personalising alarm limits.

Monitoring Characteristics
The approach pursued by the 16 studies in monitoring physiological signs were heterogeneous with regard to the type of equipment or instrument used to monitor and assess the participant's data. In some studies, a mobile/tablet app was used to communicate with the participant [19,20], and transfer data. Some studies set up a monitoring station for each individual with different devices [7,[13][14][15][17][18][19][21][22][23][24][25], where the data were transmitted automatically through an Internet modem. If a red flag was raised or threshold breached, a notification alert was sent to the system operator in real time. In two other studies, another form of monitoring was used. A diary card for symptoms and vitals were provided to participants, and a visit was arranged to collect the data [5,12,16].

Intermittent vs. Continuous Monitoring
In the reviewed articles, 16 studies monitored the participants' physiological parameters and symptoms intermittently. The frequency of monitoring/recording was varied, some once daily or multiple times daily. However, in four studies [7,13,14,23], participant's data were monitored less than daily (different frequencies per week). In addition to that, sometimes measurements taken were restricted to morning, however, in Harding et al. [20], the stipulated time for measurements recording was based on the patient's preference.

Discussion
We have conducted the first systematic review examining the utility of monitoring physiological variables to predict exacerbations of COPD. In general, and as discussed below, the studies are small and heterogeneous using different variables and different protocols. The need for better healthcare solutions in people diagnosed with chronic diseases is real. COPD imposes burdens on individuals and health care organisations. Whilst the methods hold promise, further adequately powered studies are required to properly define the utility of physiological monitoring to predict exacerbations.
In this systematic review, sixteen articles met the inclusion criteria, which were compliant with PRISMA. Five studies out of 16 [13,15,19,20,22] did not provide sufficient statistical data to draw conclusions consistent with the results of other studies, despite reporting changes in physiological variables (no p-value). The methodological quality of the studies was variable but generally low with 12 cohort studies ranked as moderate or low quality, and four trials ranked as having a high risk of bias.
We have described those studies that showed positive results in predicting/detecting an exacerbation episode via monitoring of physiological parameters. Although this approach appears to be promising, further well-designed clinical trials are required to investigate the true magnitude and time-course pre, during, and post an exacerbation episode of changes in physiological parameters. Understanding the extent of the magnitude of change for each variable is critical in using this knowledge for early exacerbation detection. In three studies [5,18,23] the magnitude of the change in heart rate and SpO 2 % reported was an increase of around 5 min −1 for heart rate and a fall by 1%-2% for SpO 2 %. Two studies [17,24] reported an increase in the respiratory rate before the onset of COPD exacerbation/hospitalisation. These findings all support the hypothesis that monitoring of vital signs can detect respiratory deterioration. However, the question arises as to whether these variables can be reliable enough. Moreover, to answer that question we need to better understand the relationship between physiological signs and symptoms. This has been confirmed in some of the above mentioned studies [5,12,14]. Hurst combined peak expiratory flow (PEF) with a symptom score to provide optimal exacerbation detection [5].
Having demonstrated that monitoring physiological variables has the theoretical potential to detect COPD exacerbations, the second step is implementation of this in a clinical environment-Tele-monitoring. To enable healthcare providers and patients to feel secure managing COPD and detecting acute exacerbations with no anticipated harm, an intelligent interface to provide live communication is essential. In the above mentioned studies, various designs were employed. However, the optimal technique/algorithm still requires more investigation. Despite the fact that tele-health offers the possibility for the clinician and the patient to be connected and monitored in a 'virtual clinic', the accuracy and specificity of this discipline are still uncertain. Developing an algorithm to detect an exacerbation is important because that would facilitate the services provided via tele-health. A particular challenge is around alarm thresholds. To increase the value of tele-health in self-management, a customised threshold for each patient is essential as this will help to decrease false alarms, and differentiate between true deterioration and day-to-day variation. Six studies had addressed this issue by specifying the alarm settings for each individual [5,13,14,[19][20][21], but using different methods and the optimal way to set individual patient alarms remains an open question.
Even though most of the reviewed studies exhibited some significant positive results in the efficacy of physiological parameters in predicting/detecting COPD exacerbation, there are insufficient data to draw a secure conclusion in this review. This is due to the diversity of the designs, methods, and sample size of studies. The demand for technology to meet the needs of the COPD patient and society are increasing. Further clinical trials are needed to achieve that.

Strength and Limitations
In this systematic review, a number of limitations can be considered. First, non-English studies (abstract and full text) were excluded. Second, only one author performed the screening of titles and abstracts, which may have increased the risk that studies were excluded inappropriately. Thirdly, the definitions of exacerbation vary across the studies, which can make comparison between studies challenging. The major strength of this study is that, to our knowledge, there is no pre-existing review conducted regarding the usefulness of monitoring physiological signs to predict COPD exacerbation.

Conclusions
Monitoring of physiological parameters may be useful in assisting earlier detection of COPD exacerbations but further, robust studies are required to confirm this. A particular challenge is how to set alarm limits for individual patients given the heterogeneity inherent in COPD and COPD exacerbations.