AbstractDisturbances in sleep and circadian rhythm are commonly reported in the pe-riod preceding deterioration in mental state in people with psychosis. Sleep disturbance may therefore serve as an early sign of impending relapse, for prompting timely, preventative interventions. Also, establishing a temporal association between sleep and symptomatic deterioration would lend support to the hypothesis that sleep plays a causal role in generating psychotic symp-toms, and may in itself form an important therapeutic target. I aim to ex-plore both of these questions through a digital phenotyping approach, which harnesses continuous, real-time data from mobile and wearable devices, in real-world settings.
In the ﬁrst chapter of this thesis incorporating publication, I summarise the ev-idence for sleep-circadian disruption during the relapse prodrome in schizophre-nia, and incorporate literature from the bipolar disorder ﬁeld. I explore the neurobiological mechanisms by which sleep reduction and circadian misalign-ment may drive psychopathology, and provide an overview of how digital tech-nologies are being harnessed as clinical tools in psychiatry.
The second chapter, a systematic review and meta-analysis of case-control actigraphy studies, seeks to understand sleep-circadian dysfunction across the psychosis spectrum by comparing actigraphy parameters in schizophrenia and bipolar disorder, in the non-acute phase of illness. I show that in both disor-ders, patients show signiﬁcantly longer total sleep time and time in bed, with lower daytime activity. However, sleep is also more disturbed, with greater sleep latency and fragmentation, pointing to a range of diﬃculties pervading all aspects of the 24 hour cycle. Eﬀect sizes were greater in schizophrenia; higher antipsychotic dose was associated with longer sleep time, and greater use of sedative medication was associated with reduced sleep latency, suggest-ing that treatment eﬀects play a role in these ﬁndings. This sets the scene for the subsequent examination of dynamic changes, across phases of illness.
To date, sleep-circadian and psychopathology variables have not been studied for long enough to allow episodes of deterioration and relapse to be captured.
Increasingly ubiquitous mobile and wearable technologies may allow continu-ous sampling of subjective and objective sleep-circadian variables in the home environment, over extended periods. In chapter three, I describe the ﬁndings of a pilot study, undertaken to develop and test the feasibility and accept-ability of Sleepsight, a digital platform designed speciﬁcally for this objective. Although there were high rates of engagement with the technology, greater negative symptom burden was associated with poorer completion of smart-phone diaries. Importantly, when the study was conducted in a transparent, collaborative manner, concerns over privacy and security were not signiﬁcant barriers to engagement and adherence. This tool was then used in a sub-stantive study of 36 people with psychosis, each participating for 12 months, in order to capture natural ﬂuctuations in mental state, including relapse episodes.
In chapter four, I examine the association between subjective sleep-circadian variables and psychopathology data using the Diﬀerential Time Varying Ef-fects Model (DTVEM), a novel analytic approach which elucidates the lag structure between sleep variables and mental state. The results clearly sug-gest that poorer subjective sleep quality and reduced subjective sleep duration precede deterioration in symptoms by 8-12 days. Although an inverse direc-tion of eﬀect of psychopathology on sleep was also found, this was shorter, and of smaller magnitude. These ﬁndings support the hypothesis that sleep disturbance precedes deterioration in psychopathology.
In chapter ﬁve, I describe how raw digital phenotype data are classiﬁed into sleep-circadian variables using a preprocessing pipeline, and employ a range of analytic approaches to explore these data. Visualisation of rest-activity patterns demonstrated a wide range of sleep-circadian phenotypes, some with striking patterns of circadian dysfunction including relative co-ordination and non-24 hour rhythms. A second analysis using the DTVEM method, with ob-jective data, sustained the ﬁnding that sleep reduction preceded symptomatic deterioration. Group-level comparisons between relapsing and non-relapsing participants did not demonstrate signiﬁcant diﬀerences in clinical or sleep-circadian variables. However, visualising these variables in the participants who relapsed, centred on relapse event, suggested that a reduction in sleep duration and decrease in activity accompany the period before, during and after relapse. Mathematical modelling was able to simulate major circadian disturbances in a subset of participants, and crucially showed that increas-ing daylight in the model restored normal circadian rhythmicity. Some of the technological limitations currently inherent in the approach are also discussed.
Taken together, these ﬁndings suggest that sleep reduction, fragmentation and poorer sleep quality are features of schizophrenia in remission, and are also associated with deterioration and relapse. The evidence supports these changes playing a causal role in the development of psychopathology. Despite technical challenges relating to data quality, these ﬁndings also suggest that meaningful objective data can be collected from smartphone and wearable devices, which show promise for use as decision-support tools.
|Date of Award||1 Jun 2021|
|Supervisor||James Maccabe (Supervisor), Derk-Jan Dijk (Supervisor) & Maarten de Vos (Supervisor)|