Functional and exercise capacity, fatigue, and accelerometry in the early post-ICU recovery period

Abstract

Research on post-critical illness outcomes has grown significantly, revealing that many survivors experience long-term physical, mental, and cognitive challenges, accompanied by persistent and burdensome symptoms, such as fatigue. Regaining functional capacity is a top priority for patients, yet the experience of fatigue and low energy, despite the availability of several measuring tools, is often overlooked in research. Several objective methods are used to assess functional and exercise capacity, with the cardiopulmonary exercise test (CPET) considered the gold standard, while accelerometry/actigraphy is emerging as a valuable tool for assessing physical activity and sleep in clinical research. This thesis explores the utility of cardiopulmonary exercise testing (CPET), accelerometry, and the assessment of fatigue and energy in post-critical illness patients. It aims to address the following primary research questions: (1) To what extent are CPET and accelerometry currently utilised in this population? and (2) What is the feasibility of conducting CPET, implementing long-term wrist-worn accelerometry, and assessing fatigue and energy levels using the Visual Analogue Scale for Fatigue in patients shortly after ICU discharge? The background and context were reviewed in Chapter 1.

Chapter 2 describes a systematic literature review of existing studies on CPET testing in post-critical illness survivors, focusing on exercise-related oxygen consumption rates and causes of exercise limitations. Eight studies involving 207 patients, were analysed. Study quality ranged from fair to good, with variations in reporting. CPET timing ranged from 2 to 160 weeks post-hospitalisation. The pooled mean submaximal VO2 at 1 month was 14 (95% CI: 13.2, 14.7) ml/kg/min, and VO2 at the anaerobic threshold was 10.3 (95% CI: 9.7, 10.8) ml/kg/min, representing 56% (95% CI: 51%, 62%) of the predicted value. General deconditioning was the primary cause of functional impairment. No significant adverse events occurred during CPET. These findings suggest the underutilisation of this method in critical illness research and highlight a lack of knowledge regarding functional capacity assessment through CPET during the ICU-to-hospital discharge period.

Chapter 3 describes a scoping literature review, to examine the utility of accelerometer sensors published during the last 5 years in survivors of critical illness in the context of the post-ICU period. Eleven non-interventional studies were included; all conducted at varying times post-ICU discharge. There was significant variation in participant numbers, demographics, clinical characteristics, device brands, and body position. Most studies focused on physical activity behaviour, with less than 7 days of continuous monitoring, and notable inconsistency in reported behavioural measures and acceleration data analysis. These findings highlight the scarcity of data on long-term accelerometry use and the absence of a standardised approach to analysing and reporting acceleration data, limiting comparability between studies.

Chapter 4 describes a prospective clinical service evaluation to inform future studies by assessing functional status with the Barthel Index (BI) and perceived fatigue levels using the Visual Analogue Scale-Fatigue (VAS-F) in patients receiving routine care and rehabilitation after ICU discharge. Over two months, 39 out of 45 screened post-ICU patients were included, mostly male (64%), with a median age of 54 years (IQR [37, 61]) and a median ICU stay of 2.6 days (IQR [1.6, 3.9]). At one week post-ICU, 56% were highly dependent in basic daily tasks, while 28% were independent. Clinically significant fatigue was observed in 60% of patients, with a mean score of 6.2 (95% CI: [5.1, 7.3]). These findings suggest that the VAS-F is suitable for use in the early post-ICU period, and that some patients may be capable of performing CPET during their index hospital stay.

Chapter 5 describes a prospective non-interventional clinical study to: 1) Explore the utility of long-term continuous accelerometry/actigraphy in post-critical care patients; 2) Assess the feasibility of cardiopulmonary exercise testing in the early post-critical illness period; 3) Measure basic daily activity levels at ICU discharge and pre-hospital discharge; 4) Assess perceived fatigue and energy levels at ICU discharge and pre-hospital discharge; 5) Explore associations between physical activity, sleep (from accelerometer data), and fatigue, energy, and functional capacity. Twenty participants were recruited from 55 eligible patients, with a median age of 49 years (IQR [33, 61]), a median APACHE II score of 19 (IQR [12, 20]), and a median post-ICU stay of 10 days (IQR [8, 18]). Eighty percent of participants had at least 7 days of valid accelerometer data, with a mean recording duration of 32 (±16) days. Only 4 participants completed CPET. At enrolment, 65% of participants exhibited varying degrees of functional dependency, which improved by hospital discharge, with a mean BI score increase of 3.75 points (95% CI: 2.4, 4.9, p = 0.0004). Clinically relevant fatigue was present in 60% of participants at enrolment, decreasing to 10% at discharge, while low energy was reported by 55%, decreasing to 40% by discharge. Functional status was positively correlated with daily step count (rho = 0.5, p = 0.03) and negatively correlated with sedentary time (r = -0.5, p = 0.01). Sleep efficiency also showed a positive correlation (rho = 0.62, p = 0.0009). Clinically significant fatigue was associated with time spent in the supine position (tau = 0.6, p = 0.04). These findings suggest long-term accelerometry is feasible and informative for assessing recovery, while CPET is challenging to complete within the proposed timeframe.

Chapter 6 explored the agreement between behavioural measures produced by two accelerometer data analysis algorithms, the Activinsights GENEActiv R Markdown analysis tool (GENEA) and the Generalized Estimation of Activity and Posture (GGIR), on a single dataset. A total of 420 days of monitoring, each with over 20 hours of usable data per day, were analysed. For all physical activity and sleep behaviour measures assessed, Bland-Altman analysis revealed modest systematic bias with wide 95% limits of agreement, exceeding the proposed clinically meaningful difference in all cases except daily step count during hospitalisation. These findings suggested poor agreement in end-user output between the two raw acceleration data analysis approaches, questioning their validity in this population with low levels of physical activity.

In an individual case study series, chapter 7 explored whether clinical episodes of physical therapy undertaken during post-ICU hospital ward stay were adequately represented by relevant accelerometry-derived activity obtained using a freely available classification model. From the data collected in the previous study, 6 participants had a median of 3 days (range: 1-12) of valid accelerometer data on days with documented rehabilitation sessions in the medical records. A total of 25 days were analysed. In all cases, the activity suggested by the activity classification algorithm showed poor correspondence with the activity reported by the physiotherapist in the medical notes.

In Chapter 8, I discuss the thesis findings and how they contribute to the understanding of the challenges associated with performing CPET during hospitalisation and the feasibility of long-term accelerometry for continuous monitoring of physical activity and sleep behaviour. I highlight the need for further research to improve the accuracy and application of these tools in post-critical illness recovery.

COVID-19 Impact statement

Similar to the experience of many researchers in the clinical field, the COVID-19 pandemic and the resulting unprecedented measures affected the conduct and timeline of this PhD project. Key impacts included:

•Delays in participant recruitment and data collection: Due to restrictions imposed by both the ethics committee and research and development offices, reviews of all non-COVID-19-related research were postponed, creating a significant backlog of studies awaiting approval. Consequently, participant recruitment only commenced after pandemic-related constraints were lifted, which occurred markedly later than initially anticipated.

•Changes to study protocol: In the immediate post-pandemic period, recruitment proved challenging, likely due to the psychological aftermath of social distancing and isolation. As a result, several adaptations to the study protocol were required, which themselves took a prolonged time to be reviewed and approved by the responsible regulatory bodies.

Despite these challenges, the research proceeded with appropriate adjustments and continued to adhere to all ethical and safety standards. However, due to these unforeseen circumstances, only half of the targeted study sample was ultimately reached.