Use of a personalised early warning decision support system for acute exacerbations of chronic obstructive pulmonary disease: results of the ‘Predict &amp; Prevent’ phase III trial

doi:10.21203/rs.3.rs-4616866/v1

Download PDF

Research Article

Use of a personalised early warning decision support system for acute exacerbations of chronic obstructive pulmonary disease: results of the ‘Predict & Prevent’ phase III trial

https://doi.org/10.21203/rs.3.rs-4616866/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Rationale Gold standard treatment for chronic obstructive pulmonary disease (COPD) includes a supported self-management plan to enable patients to recognise and treat acute exacerbations (AECOPD). The Predict & Prevent trial was designed to provide a definitive randomised clinical trial of a personalised early warning decision support system, COPDPredict^TM.

Methods Adults with ≥1 AECOPD or COPD admission were randomly assigned in a 1:1 ratio to use of a personalised early warning decision support system (COPDPredict^TM) or standard self-management plans with rescue medication (RM) (control). The primary outcome was number of hospital admissions for AECOPD at 12 months post-randomisation (intention to treat). Quality of life and healthcare utilisation were also assessed.

Results 789 patients were screened, of whom 90 were enrolled. Hospital admissions due to AECOPD at 12 months were lower with COPDPredict^TM: Incidence rate ratio (IRR) 0.64 (95% CI 0.19 to 2.17) albeit with insufficient evidence of a treatment effect (p=0.48). Exploratory Bayesian analysis and sensitivity analyses saw similar results. No significant differences were seen in inpatient days, visits to accident & emergency visits, and number of exacerbations. COPD Assessment Test (CAT) score benefits occurred at 3 and 6 months with COPDPredictTM (adjusted mean difference -3.8 points, 95% confidence interval (CI) -6.3 to -1.2, p=0.004 and -3.0 points, 95% CI -5.7 to -0.4, p=0.025 respectively) but was non-significant at longer periods (p>0.22), though this could be observed by chance as indicated by the time-point by treatment interaction (p=0.269). There was not enough evidence to indicate a statistically significant treatment effect on the other outcomes

Conclusions The quality of life data (CAT scores) suggests that 6 months usage of COPDPredict^TM period may be helpful to patients, with benefits exceeding the minimum clinically difference throughout that time.

Trial registration: NCT04136418

Chronic obstructive pulmonary disease

randomised controlled trial

self-management

digital health

clinical decision rules

Chronic obstructive pulmonary disease (COPD) affects 65m people worldwide, 2m of whom are in the UK, where it causes over 140,000 hospital admissions/year(1), breathlessness, cough and sputum production. Given its prevalence and impact on the health service, optimising care is therefore a priority for UK healthcare. This is highlighted by the fact that lung disease is one of Britain’s biggest killers, alongside heart disease and non-respiratory cancers, killing 115,000 people/year, the equivalent of one every five minutes(1). These mortality figures are roughly the same as a decade ago, whilst deaths from heart disease went down by 15% from 2008 to 2012(1). Furthermore, the UK has the 4th highest mortality rate from lung disease in Europe(1). Despite current therapies, patients continue to exacerbate, and these are a significant cause of death – 38% will die or be readmitted to hospital within 90 days of an exacerbation of COPD (AECOPD)(2) and both AECOPD and admissions causes significant morbidity and mortality. Prevention of AECOPD and the resultant admissions is therefore a priority for most health services.

Self-management is recognised to have a wide range of benefits in COPD, mainly on quality of life (QOL) (3), and respiratory related admissions(4). Whilst one meta-analysis hinted at greater effects in male patients, those with severe disease, and those with greater self-efficacy(5), results have not consistently shown specific groups to benefit more. However admissions are more common in severe disease(6, 7), such that ensuring all patients who are at risk of admission, either because they have been hospitalised previously, or because they have severe disease would prioritise those with most potential to benefit. However, implementation of standard self-management at discharge as part of a COPD care bundle has not impacted readmission rate in the UK(8), suggesting that the way in which we deliver self-management needs further optimisation. Previous studies have shown that coaching(9) and iterative reinforcement(4) may improve efficacy, but the feasibility of this within stretched health services is unclear. UK healthcare professionals (HCPs) have identified multiple training needs to deliver iterative support(10) and coaching takes time, typically more time than a normal appointment with a healthcare provider. Using digital technology to reinforce knowledge of day to day symptom variability with an alert at the time of deterioration to prioritise HCP input might be a reasonable substitute for this, helping patients to recognise the difference between day to day variability and AECOPD, and ensuring that fewer AECOPD events are untreated, whilst retaining the support of the health service. Unreported AECOPD events (symptom defined events which are not treated) are common and impact equally on symptoms and decline compared to treated events(11). Enhancing treatment rates may reduce severity of AECOPD, and perhaps admissions as a result.

The COPDPredict™ device was developed to try and address this need. It acts like an advanced form of self-management by collecting symptom and spirometry data to flag to a patient, and their clinician, when an AECOPD event is imminent, and by means of this alert allow them to take action together to treat the event quickly. The system works by generating a personal baseline for each user, so that alerts occur when their parameters deteriorate, rather than relying on fixed values. Addition of biological data in the form of CRP further enhances its predictive ability. Prompt treatment may reduce the length and severity of an AECOPD, thus might prevent admission. Initial studies in a single UK site suggested a large impact on admission rate, in a previously frequently admitting cohort(12), such that a definitive trial to assess whether these results are generalisable was warranted. Here we report results of the attempt at that trial.

Study design and oversight

Predict & Prevent was an ethically approved (19/LO/1939) phase III, two arm, multi-centre, open label, parallel-group individually randomised clinical trial (NCT04136418) investigating the use of a personalised early warning decision support system (COPDPredict™) to predict and prevent acute exacerbation (AECOPD) in the United Kingdom (UK). All participants provided written informed consent. Trial steering and data monitoring committees provided independent oversight. The protocol has been published(13).

Participants

Eligible participants were adults who had COPD, defined as post-bronchodilator forced expiratory volume (FEV₁)/forced vital capacity < 0.7 and < lower limit of normal, in the stable state. Inclusion criteria were: adults at least 18 years old; ≥1 AECOPD in any 12-month period within the last 2 years or ≥ 1 hospital admission for AECOPD in the previous two years; exacerbation free for at least 6 weeks; ability to use intervention upon demonstration of the system to the patient at screening. Participants were excluded if they had life expectancy < 12 months or had co-morbidities that precluded the patient from using the intervention (for example significant cognitive impairment).

Randomisation and masking

Participants were randomly assigned (1:1) to receive either a standard self-management plan advising use of rescue medication containing 5 days of antibiotic and steroid treatment (control), or supported self-management using COPDPredict™, through a secure online central randomisation system provided by the Birmingham Clinical Trials Unit, with the use of minimisation to balance trial-group assignments according to baseline age (< 60, ≥ 60 years) and severity of disease as per Global Initiative for Obstructive Lung Disease at the time (GOLD)(Category B [{mMRC ≥ 2 or COPD Assessment Test (CAT) ≥ 10} and {AECOPD < 2}], Category D {mMRC ≥ 2 or CAT ≥ 10} and {AECOPD ≥ 2}])(14) and hospital centre.

Study interventions

The intervention COPDPredict™ is a CE marked Class I medical device, which comprises a patient-facing App pre-installed onto a Tablet device provided to participants randomised to this arm of the trial, clinician-facing web-based dashboard, CE-marked Bluetooth devices for assessment of physiology as well as CE-marked devices for collection of saliva. Participants using COPDPredict™ recorded daily well-being scores throughout the study; prompts to participants to complete their scores were issued and if compliance was < 75% over 2 weeks, then actions were taken by the research team and in App to encourage compliance. Participants also established a ‘personal baseline’ by entering stable state spirometry and blood CRP App at day 1 and day 7.

AECOPD events detected by deviation from baseline were alerted to patient and clinician by the App, requiring additional CRP measures and clinical review.

The comparator was the standard self-management plan (SSMP) alongside the use of rescue medication (RM) containing 5 days of antibiotic and steroid treatment.

Outcomes

The primary outcome was number of hospital admissions at 12 months post-randomisation where the primary reason for admission was AECOPD. Secondary clinical outcomes at 6 and 12 months included CAT score, total number of inpatient days, visit to Accident and Emergency (ED) because of symptoms related to COPD, number of AECOPD, association between symptom control markers and clinical judgement (hereafter referred to as ‘appropriate self-management’), and secondary clinical outcomes at 12 months post randomisation included FEV₁ predicted % and blood C-reactive protein (CRP) levels (intervention arm only).

Sample size

Based on previous published data for estimated admissions in the control group(15), and a desire to detect a difference of 1 admission between groups, assuming a standard deviation of 2.6³ and using t-test for equal variances, 144 participants per group were required to provide 90% power with a type I error rate of 5% (two-sided). We planned to include 384 participants to account for ≤ 25% attrition. Due to difficulties in recruitment partly due to the impact of the pandemic, the eventual sample size was lower and 12-month follow-up was not achievable for all participants.

Statistical analysis

A comprehensive statistical analysis plan (SAP) (Supplementary material) was drawn up before the final analysis. All analyses were performed in SAS (version 9.4) or Stata (version 17.0). Categorical data were summarised with frequencies and percentages. Continuous variables were summarised with means and standard deviations unless skewed, where medians and IQRs were presented. Participants were analysed in the treatment group to which they were randomly allocated (intention to treat). All estimates of differences between groups were presented with 95%, two-sided confidence intervals (CIs), adjusted for the minimisation variables. The estimate mean difference of the CAT score was also adjusted for the CAT baseline score.in these adjustments age was treated as a continuous variable and all covariates were treated as fixed effects. The primary outcome was analysed using a negative binomial regression model (due to over dispersion of the data [Dispersion (95%CI) = 1.5 (0.2 to 10.7)] to generate an adjusted incidence rate ratio (aIRR). Centre was removed due to convergence issues. Time in years from randomisation was added as an offset variable because of variable follow up time. To maximise the utility of the collected data and to aid interpretation, exploratory analysis using Bayesian methods estimated the adjusted IRR of the number of hospital admissions for AECOPD at 12 months, with its 95% highest posterior density (HDP) credible intervals (Crl).

For the CAT score outcome measure, adjusted mean differences and associated 95% CIs were estimated at the time-points (3, 6, 9 and 12 months) using mixed effect repeated measures model(16). The time-point as a categorical variable and the time-point by treatment interaction were also included. The total number of inpatient days outcome was categorised in three groups (Inpatient days = 0, 0 < Inpatient days < = 6, Inpatient days > 6) and an adjusted odds ratio (aOR) was estimated using an ordinal logistic regression model because data were highly skewed. Binary secondary outcomes measured at a single timepoint were analysed using a log-binomial model or a Poisson regression model where there were convergence issues to generate an adjusted risk ratio (aRR). Visits to ED and number of AECOPD were analysed as per the primary outcome. Methods to determine appropriate self-management are described in the supplementary methods. CRP level was summarised among those who had at least one, only one and two or more exacerbations. FEV₁ predicted % was only summarised by group due to the high volume of missing data.

Sensitivity and supportive analyses of the primary outcome are described in the supplementary methods. No sub-group analyses were performed because the sample size was too small.

One interim analysis of effectiveness and safety endpoints was performed on behalf of the data monitoring committee during the period of recruitment. This analysis was done with the use of the Haybittle–Peto(17) principle and hence no adjustment was made in the final p values to determine significance.

Screening and randomisation occurred from Oct 7 2020 to Sep 30, 2022 (Fig. 1); 45 participants (50%) were allocated to SSMP and 45 (50%) to COPDPredict™. At the end of the follow up on Mar 31, 2023 90 participants reached 6 months post randomisation and 64 reached 12 months post randomisation. 80 participants were included in the primary analysis but extra supportive/sensitivity analysis was conducted by imputing the missing data (supplementary material) to check the robustness of primary results. Baseline characteristics at enrolment are reported in Table 1; patients generally had severe COPD, by spirometry or GOLD, and 4 patients in each group regularly used non-invasive ventilation (NIV). 99% of participants were of white ethnic origin. Nine (20%) of 45 participants in the SSMP group and 13 (29%) of 45 in the COPDpredict group (p = 0.327; supplementary table S1) experienced a serious adverse event but none of them were deemed related to the intervention.

Table 1

Baseline characteristics by group and overall
		Standard Self-management plan (N = 45)	COPDPredict™ Self-management system (N = 45)	Overall (N = 90)
Minimisation variables
FEV₁% predicted¹, n(%)	< 50% predicted	17 (65.4)	17 (68.0)	34 (66.7)
	≥ 50% predicted	9 (34.6)	8 (32.0)	17 (33.3)
	Mean (SD, N)	43.2 (14.3, )	40.9 (16.6)	42.1 (15.4)
Severity of COPD¹, n(%)	B: (mMRC ≥ 2 or CAT ≥ 10) and AECOPD < 2	6 (13.3)	5 (11.1)	11 (12.2)
Severity of COPD¹, n(%)	D: (mMRC ≥ 2 or CAT ≥ 10) and AECOPD ≥ 2	39 (86.7)	40 (88.9)	79 (87.8)
Age (years)	< 60	14 (31.1)	13 (28.9)	27 (30.0)
	≥ 60	31 (68.9)	32 (71.1)	63 (70.0)
	Mean (SD)	63.3 (9.3)	66.0 (8.9)	64.7 (9.1)
Gender, n(%)	Male	24 (53.3)	22 (48.9)	46 (51.1)
Gender, n(%)	Female	21 (46.7)	23 (51.1)	44 (48.9)
MRC score, n(%)	1	2 (4.4)	1 (2.2)	3 (3.3)
	2	5 (11.1)	9 (20.0)	14 (15.6)
	3	12 (26.7)	14 (31.1)	26 (28.9)
	4	21 (46.7)	17 (37.8)	38 (42.2)
	5	5 (11.1)	4 (8.9)	9 (10.0)
Respiratory features, n(%)	Bronchiectasis	8 (17.8)	8 (17.8)	16 (17.8)
	Chronic bronchitis	6 (13.3)	4 (8.9)	10 (11.1)
	Asthma	13 (28.9)	12 (26.7)	25 (27.8)
	Emphysema	22 (48.9)	29 (64.4)	51 (56.7)
Years of COPD	Mean (SD)	10.1 (7.0)	10.6 (7.1)	10.3 (7.0)
Exacerbations in the last year	Mean (SD)	3.1 (1.9)	3.5 (2.3)	3.3 (2.1)
Hospitalisations in the last year	Mean (SD)	0.5 (1.0)	1.0 (2.4)	0.8 (1.8)
Pulmonary rehabilitation, n(%)	Yes	28 (62.2)	26 (57.8)	54 (60.0)
Pulmonary rehabilitation, n(%)	No	17 (37.8)	19 (42.2)	36 (40.0)
Medical History², n(%)	Anxiety	20 (44.4)	22 (48.9)	42 (46.7)
	Depression	17 (37.8)	15 (33.3)	32 (35.6)
	Osteoporosis	12 (26.7)	12 (26.7)	24 (26.7)
	GORD	24 (53.3)	19 (42.2)	43 (47.8)
	Diabetes	5 (11.1)	7 (15.6)	12 (13.3)
Smoking status, n(%)	Never	1 (2.3)	0 (0)	1 (1.1)
	Current	14 (31.8)	16 (35.6)	30 (33.7)
	Stopped < 6 weeks	0 (0)	2 (4.4)	2 (2.3)
	Stopped ≥ 6 weeks to < 1 year	3 (6.8)	2 (4.4)	5 (5.6)
	Stopped ≥ 1 year	26 (59.1)	25 (55.6)	51 (57.3)
Among smokers and ex-smokers - Smoking pack years	Mean (SD)	39.9 (18.2)	43.1 (18.6)	41.5 (18.4)
Data shown as mean (standard deviation (SD) or n (%)
¹ The minimisation variable, ‘severity of disease’ measured by the FEV₁% predicted values, was amended after 51 participants were randomised in the trial because the tactic of assessing disease severity changed and FEV₁% predicted stopped being measured. After that ‘severity of disease’ measured to fit the gold model for COPD, as of which we have four categories:
• Category A (Modified Medical Research Council (mMRC) Dyspnoea scale 0–1 and CAT < 10) and (AECOPD < 2)
• Category B (mMRC ≥ 2 or CAT ≥ 10) and (AECOPD < 2)
• Category C (mMRC 0–1 and CAT < 10) and (AECOPD ≥ 2)
• Category D (mMRC ≥ 2 or CAT ≥ 10) and (AECOPD ≥ 2)
Categories A and C were not applicable in this trial according to the trial’s inclusion criteria. The 51 participants who had been randomised up to the time point of this change were retrospectively allocated into one of the of the two categories (B or D) of the gold model for COPD and then using the minimisation technique randomise the subsequent participants using the new minimisation variable.
² BMI was calculated using the formula \(\frac{weight \left(Kg\right)}{{height \left(m\right)}^{2}}\).
³ Responses are not mutually exclusive; percentages may total more than 100%.

84% and 98% of participants randomised to COPDPredict™ were adherent (≥ 75%) with symptom completion in the App at 2 and 4 weeks post-randomisation respectively. Patients ability to self-manage, as determined by whether they treated themselves when their symptoms indicated AECOPD was similar between arms, and showed little difference over time (supplementary results table S2). Unreported AECOPD events where symptoms indicated AECOPD but no action was taken occurred in 24.4% of patients in the control arm, but only 12.8% with COPDPredict™ at 3 months. By 12 months both groups had improved, with unreported untreated events occurring in 7.2% of patients in the control arm and 4.7% with COPDPredict™ (supplementary results)

Primary outcome

Most patients (85% in both groups) did not experience a hospital admission. Adjusted IRR (95% CI) of participants experiencing hospitalisation for AECOPD was lower in the intervention group compared to usual care group: 0.64 (0.19 to 2.17), however there was insufficient evidence to indicate a treatment effect (p = 0.478) (supplementary table S3). Sensitivity and supportive analyses had minimal impact on effect estimates (supplementary tables S4-6) and Bayesian analysis, whether using optimistic or pessimistic priors, did not alter the significance (supplementary results).

Secondary outcomes

CAT score benefits occurred in the intervention arm at 3 and 6 months (adjusted mean difference − 3.8 points (-6.3 to -1.2); p = 0.004 and − 3.0 points (-5.7 to -0.4); p = 0.025 respectively) but ceased to be statistically different at longer periods (p > 0.22), however, this observed tailing off of effect over time may be due to the increase in missing responses or to random factors as indicated by the time-point by treatment interaction (p = 0.269) (Table 2 and Fig. 2). There was insufficient evidence to indicate a significant treatment effect on the other outcomes (supplementary table S3).

Table 2

Quality of life differences between groups
CAT Components		Standard Self-management plan (N = 45)	COPDPredict^™ Self-management system (N = 45)		Adjusted Mean Difference^1,2 (95% CI, p-value)
Baseline	Mean (SD)	25.1 (6.4)	23.8 (7.2)	p = 0.269	NA
Month 3	Mean (SD)	26.1 (6.3)	21.5 (8.1)		-3.78 (-6.32 to -1.24), p = 0.004
Month 6	Mean (SD)	25.6 (7.5)	21.3 (6.5)		-3.04 (-5.70 to -0.38), p = 0.025
Month 9	Mean (SD)	26.1 (6.1)	22.1 (7.2)		-1.76 (-4.56 to 1.03), p = 0.215
Month 12	Mean (SD)	25.8 (5.0)	22.6 (7.6)		-0.82 (-3.89 to 2.24), p = 0.596
¹ Adjusted mean difference < 0 favours COPDPredict™ Self-management system group.
² Adjusted comparisons taking into account treatment and all minimisation variables including the baseline score. Time-point by treatment interaction parameter also included in the model.

The trial was significantly impacted by the pandemic, and as such was under-powered to definitively exclude effects of the intervention. Whilst no effect on the primary outcome was seen, there were some encouraging results regarding QOL, which improved more in the intervention group. It was also notable that very few patients were having unreported (untreated) AECOPD events after 12 months of the trial, though the difference between arms was small.

Recruitment to the study was challenging. The first sites were initiated in early 2020, and when the pandemic began the study was paused by most sites. Patients were also reluctant to be seen, particularly in a hospital setting, because of fear of Covid-19(18). Initially a large proportion of patients were screened out due to co-morbidities which recruiting sites thought could impact self-management; we felt it was important that the trial population reflected the typical COPD patient, in whom heart disease, anxiety and depression are common(19) – 76% of severe COPD patients have at least 1 comorbidity and 40% have 3 or more. Retraining of sites to emphasise the pragmatic nature of this trial resulted in fewer exclusions of this nature. We also experienced issues with AECOPD rate precluding enrolment; this was twofold with the most frequently exacerbating patients never experiencing enough weeks of stability for the COPDPredict™ system to establish a good personal baseline, and patients’ AECOPD falling with social isolation from the pandemic(20), which made them ineligible. We were concerned that lack of confidence in using digital technology might be a factor causing poor uptake, but this was rare (3.5% of patients), and during the study adherence to daily use of the device was equally high as in the initial single site study (98%) by 4 weeks. This is consistent with the wider literature suggesting that patients valued digital support during the pandemic(18).

Our observation of improved quality of life in participants using COPDPredict™ supports the theory that digital reinforcement of self-management may be effective. The minimum clinically important difference (MCID) in CAT score is 1 point(21), and the 95% CI showed that at 3 months intervention arm patients were improved at this level. By 6 months on average patients had improved by more than the MCID, but the bottom end of the 95% CI was less than − 1, and by 12 months there was no significant difference. Statistically, however, there was no time interaction proven. Interestingly maintaining regular e-diary use over longer time periods has been shown to be challenging(22), such that short to medium term use might be more practical.

Other forms of digital education, partially designed to support self-management have been tested in COPD. Real world testing of MyCOPD in rural Scotland showed no effect on admission or healthcare utilisation, and actually suggested an increase in requirement for home visits overall, although there were benefits seen in the 17% of users who engaged highly with the system, reading multiple modules and entering symptom data frequently(23). Our results are consistent with this finding, since patients were required to enter data frequently to remain in the trial and were seen more often at home; healthcare utilisation overall within the trial and consequent cost-effectiveness will be the subject of a separate economic analysis. Home visits were not usual care in all of our sites for AECOPD, with telephone contact more usual to support patient decision making, however the study protocol required face to face contact, in particular for CRP measurement to be obtained, so this observation does not necessarily imply a shift from hospital to community care for AECOPD. Similar results were obtained in a multimorbid population in Canada, where admissions were numerically but not statistically lower, and emotional wellbeing improved significantly.

Quality of life drops markedly in COPD after exacerbation and after hospitalisation(24), such that differences may be easier to detect initially, but it is also conceivable that in the early stages after hospitalisation a more intensive tool such as COPDPredict™ may be more useful; this is supported by the observation that proportionally twice the number of patients were self-managing appropriately by 3 months in the intervention compared to control group. Since more than 90% of AECOPD were appropriately managed by patients at 12 months this implies a time based phenomenon to learning to self-manage, which could be achieved over a long duration with low intensity support, but more quickly with the higher intensity support that COPDPredict™ could provide. This suggests that using it after initial hospital admission may be the most beneficial option, since this is the period when patients feel most unwell and are at highest risk of readmission. Ongoing real-world studies of COPDPredictTM adoption may help answer this question. These study designs may also be more appropriate for tests of digital interventions.

Conceptually COPDPredict™ was designed as a partnership between patients and the health service, such that the device alerts both patient and staff to a deterioration, so that the health service may respond. In a paternalistic medical model this implies someone would attend the patient upon deterioration, and there is an assumption that this engagement will help; if the interaction improves their well-being then an admission might be avoided. However, this requires that the health service be available 24 hours a day, 7 days a week, and for users to respond promptly to alerts. We worked hard with sites to design solutions that would allow them to run the trial, even in the absence of a routine care service that ran in this manner, however we had to be pragmatic and allow within our protocol a response time of 48 hours in the alerts, to account for 5 day services. It is possible that this limited effectiveness of the intervention in prevention of admission, since timelines of AECOPD symptoms suggest that deterioration occurs quickly in this time(22, 25). Other limitations to the work including failure to reach the desired sample size.

COPDPredict™ did not reduce severe AECOPD events, but 6 months usage may be helpful to patients, with benefits exceeding the minimum clinically difference in quality of life throughout that time.

Ethics approval and consent to participate

Predict and Prevent was ethically approved (19/LO/1939) and all patients gave informed consent.

Consent for publication

No individual consent for publication was required since only group summary data is presented.

Availability of data and materials

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request

Competing interests

AMT has had grants and/or honoraria from Chiesi, AstraZeneca, GSK, Sanofi and Boehringer Ingelheim. No other authors have competing interests to declare.

Funding

The National Institute of Health Research (NIHR), who funded the study via the i4i scheme, had no role in study design, data collection, data analysis, data interpretation, writing of the report, or the decision to submit the results for publication.

AMT is also funded by the NIHR Midlands Patient Safety Research Collaboration (PSRC). The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. She has also received grant funding from NIHR HTA and EME.

Authors' contributions

EG performed data analysis

RLM designed the analysis strategy

ST and LD collected the data, via the recruiting sites, and managed the study

SJ, NG and AMT acquired funding and designed the study

AMT drafted the manuscript, collected data and oversaw the study

All authors reviewed the manuscript prior to submission

Acknowledgements

We would like to acknowledge staff at all sites recruiting to the study, specifically University Hospitals Birmingham, University Hospitals Coventry and Warwickshire, University Hospitals Derby, Royal Alexandra Hospital and Barts Health NHS Foundation Trust.

The battle for. breath – the impact of lung disease in the UK. British Lung Foundation; 2016.
Echevarria C, Steer J, Heslop-Marshall K, Stenton SC, Hickey PM, Hughes R, et al. The PEARL score predicts 90-day readmission or death after hospitalisation for acute exacerbation of COPD. Thorax. 2017;72(8):686–93.
Jolly K, Majothi S, Sitch AJ, Heneghan NR, Riley RD, Moore DJ, et al. Self-management of health care behaviors for COPD: a systematic review and meta-analysis. Int J Chron Obstruct Pulmon Dis. 2016;11:305–26.
Schrijver J, Lenferink A, Brusse-Keizer M, Zwerink M, van der Valk PD, van der Palen J, et al. Self-management interventions for people with chronic obstructive pulmonary disease. Cochrane Database Syst Rev. 2022;1(1):CD002990.
Jonkman NH, Westland H, Trappenburg JC, Groenwold RH, Bischoff EW, Bourbeau J, et al. Do self-management interventions in COPD patients work and which patients benefit most? An individual patient data meta-analysis. Int J Chron Obstruct Pulmon Dis. 2016;11:2063–74.
Hurst JR, Vestbo J, Anzueto A, Locantore N, Mullerova H, Tal-Singer R, et al. Susceptibility to exacerbation in chronic obstructive pulmonary disease. N Engl J Med. 2010;363(12):1128–38.
NRAP. Clinical outcomes October 2018 – March 2020 Summary report. Royal College of Physicians; 2023.
Stone PW, Adamson A, Hurst JR, Roberts CM, Quint JK. Does pay-for-performance improve patient outcomes in acute exacerbation of COPD admissions? Thorax. 2022;77(3):239–46.
Long H, Howells K, Peters S, Blakemore A. Does health coaching improve health-related quality of life and reduce hospital admissions in people with chronic obstructive pulmonary disease? A systematic review and meta-analysis. Br J Health Psychol. 2019;24(3):515–46.
Blackmore C, Johnson-Warrington VL, Williams JE, Apps LD, Young HM, Bourne C, et al. Development of a training program to support health care professionals to deliver the SPACE for COPD self-management program. Int J Chron Obstruct Pulmon Dis. 2017;12:1669–81.
Jones PW, Lamarca R, Chuecos F, Singh D, Agusti A, Bateman ED, et al. Characterisation and impact of reported and unreported exacerbations: results from ATTAIN. Eur Respir J. 2014;44(5):1156–65.
Patel N, Kinmond K, Jones P, Birks P, Spiteri MA. Validation of COPDPredict: Unique Combination of Remote Monitoring and Exacerbation Prediction to Support Preventative Management of COPD Exacerbations. Int J Chron Obstruct Pulmon Dis. 2021;16:1887–99.
Kaur D, Mehta RL, Jarrett H, Jowett S, Gale NK, Turner AM, et al. Phase III, two arm, multi-centre, open label, parallel-group randomised designed clinical investigation of the use of a personalised early warning decision support system to predict and prevent acute exacerbations of chronic obstructive pulmonary disease: 'Predict & Prevent AECOPD' - study protocol. BMJ open. 2023;13(3):e061050.
Global Initiative for. Obstructive Lung Disease [Available from: www.goldcopd.com.
Pinnock H, Hanley J, McCloughan L, Todd A, Krishan A, Lewis S, et al. Effectiveness of telemonitoring integrated into existing clinical services on hospital admission for exacerbation of chronic obstructive pulmonary disease: researcher blind, multicentre, randomised controlled trial. BMJ. 2013;347:f6070.
Brown H, Prescott R. Applied mixed models in medicine. 2nd ed. John Wiley & Sons Ltd; 2015.
Peto R, Pike MC, Armitage P, Breslow NE, Cox DR, Howard SV, et al. Design and analysis of randomized clinical trials requiring prolonged observation of each patient. I. Introduction and design. Br J Cancer. 1976;34(6):585–612.
Madawala S, Quach A, Lim JY, Varatharaj S, Perera B, Osadnik C et al. Healthcare experience of adults with COPD during the COVID-19 pandemic: a rapid review of international literature. BMJ Open Respir Res. 2023;10(1).
Crapo JD, Gupta A, Lynch DA, Turner AM, Mroz RM, Janssens W, et al. Baseline characteristics from a 3-year longitudinal study to phenotype subjects with COPD: the FOOTPRINTS study. Respir Res. 2023;24(1):290.
Reschen ME, Bowen J, Novak A, Giles M, Singh S, Lasserson D, et al. Impact of the COVID-19 pandemic on emergency department attendances and acute medical admissions. BMC Emerg Med. 2021;21(1):143.
Kon SS, Canavan JL, Jones SE, Nolan CM, Clark AL, Dickson MJ, et al. Minimum clinically important difference for the COPD Assessment Test: a prospective analysis. Lancet Respiratory Med. 2014;2(3):195–203.
Ellis P, Parekh G, Duvoix A, Watson L, Sharp A, Mobeen F, et al. Characteristics of alpha-1 antitrypsin deficiency related lung disease exacerbations using a daily symptom diary and urinary biomarkers. PLoS ONE. 2024;19(2):e0297125.
Cooper R, Giangreco A, Duffy M, Finlayson E, Hamilton S, Swanson M, et al. Evaluation of myCOPD Digital Self-management Technology in a Remote and Rural Population: Real-world Feasibility Study. JMIR Mhealth Uhealth. 2022;10(2):e30782.
Machado A, Barusso M, De Brandt J, Quadflieg K, Haesevoets S, Daenen M, et al. Impact of acute exacerbations of COPD on patients' health status beyond pulmonary function: A scoping review. Pulmonology. 2023;29(6):518–34.
Aaron SD, Donaldson GC, Whitmore GA, Hurst JR, Ramsay T, Wedzicha JA. Time course and pattern of COPD exacerbation onset. Thorax. 2012;67(3):238–43.

No competing interests reported.

PredictPreventsupplementv1.0.docx

Download PDF

Reviews received at journal
14 Aug, 2024
Reviewers agreed at journal
12 Aug, 2024
Reviews received at journal
17 Jul, 2024
Reviewers agreed at journal
17 Jul, 2024
Reviewers invited by journal
16 Jul, 2024
Editor assigned by journal
24 Jun, 2024
Submission checks completed at journal
24 Jun, 2024
First submitted to journal
21 Jun, 2024

You are reading this latest preprint version

Use of a personalised early warning decision support system for acute exacerbations of chronic obstructive pulmonary disease: results of the ‘Predict & Prevent’ phase III trial

Status:

Version 1

Abstract

Figures

Introduction

Methods

Study design and oversight

Participants

Randomisation and masking

Study interventions

Outcomes

Sample size

Statistical analysis

Results

Primary outcome

Secondary outcomes

Discussion

Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1