Introduction
Preliminary research using Experience Sampling Methodology (ESM) began in the early 1980s to explore subjective reactions to experiences as they occur in the natural environment. ESM provides ecologically valid, real-life characterizations of thinking and behaviour (Hurlbert, Reference Hurlbert1997) by employing electronic signalling devices to prompt participants to complete self-reports at pre-selected randomized time-points. Thus, information on participants' environment, activities, thoughts, feelings or symptoms can be gathered ‘within the context and flow of everyday experience’ (de Vries, Reference de Vries and de Vries2006, p. 8). ESM data collection can be coupled with simultaneous physiological measures, such as changes in levels of stress hormones that accompany experiences. The reliability, validity and feasibility of ESM in clinical samples has been demonstrated by Csikszentmihalyi & Larson (Reference Csikszentmihalyi and Larson1987, Reference Csikszentmihalyi, Larson and de Vries2006), who showed that ESM data can discriminate between clinical and non-clinical groups, is more reliable than retrospective self-reports, is a stable measure over time, has good individual response consistency and results in satisfactory participant compliance. To date, ESM has been used to investigate conditions including schizophrenia (Myin-Germeys et al. Reference Myin-Germeys, Nicolson and Delespaul2001), bulimia (Larson & Asmussen, Reference Larson, Asmussen and de Vries1992), panic disorder (Dijkman-Caes & de Vries, Reference Dijkman-Caes and de Vries1991), personality disorder (Loewenstein et al. Reference Loewenstein, Hamilton, Alagna, Reid and de Vries1987) and major depressive disorder (MDD) (e.g. Myin-Germeys et al. Reference Myin-Germeys, Peeters, Havermans, Nicolson, de Vries, Delespaul and van Os2003b).
However, ESM studies examining the aetiology and impact of depression on everyday functioning are relatively new. ESM is ideally suited to test the predictions and inform the development of contemporary cognitive models of depression. This is not only because it can test theories and models in an ecologically valid way, but also because it offers a way to gain novel insights into the spontaneously occurring thoughts, feelings and physiological changes that accompany depression that would not be detected by retrospective self-report or cross-sectional laboratory-based studies. Although literature reviews of ESM in psychosis research have been conducted previously (Myin-Germeys et al. Reference Myin-Germeys, Delespaul and van Os2003a; Oorschot et al. Reference Oorschot, Kwapil, Delespaul and Myin-Germeys2009) and a recent review paper briefly examined ESM studies of depressed individuals (Myin-Germeys et al. Reference Myin-Germeys, Oorschot, Collip, Lataster, Delespaul and van Os2009), there has been no systematic critique of ESM in depression research. This paper aims to address this gap by reviewing the contribution of ESM in studies of clinical depression and considering how this technique might be used to advance our knowledge of depression and its treatment.
Method
Inclusion criteria were defined as:
(1) Published in English language, peer-reviewed journals.
(2) Published in the last 25 years.
(3) Use of ESM requiring participants to complete daily diaries (electronic or paper) in response to hearing pre-programmed auditory signals from a portable device.
(4) Include participants with a diagnosis of MDD.
(5) Statistical analyses performed using hierarchical or multi-level regression analyses. Hierarchical or multi-level regression analyses are ideally suited to ESM studies because of the nested structure of ESM data, with levels of nesting at ‘beep’, ‘day’ and ‘person’ level (Walls et al. Reference Walls, Jung, Schwartz, Walls and Schafer2006), and their ability to acknowledge that observations from the same participant are more similar than observations from different participants.
Search strategy
The studies included in this review were identified by systematically searching, up to July 2009, the following five databases: PsychARTICLES; PsycINFO; AMED (Allied and Contemporary Medicine); Ovid Medline; CINAHL (Cumulative Index to Nursing & Allied Health Literature). To reflect the fact that there exists a range of other techniques with a family resemblance to ESM, such as ecological momentary analysis (EMA) and electronic diary studies, the keywords ‘ecological momentary analysis’, ‘EMA’ and ‘electronic diary’ were also employed in the search. Although there are important differences, as well as similarities, between ESM and EMA (Hektner et al. Reference Hektner, Schmidt and Csikszentmihalyi2007), for the purposes of the present review we treated EMA as a form of ESM. The search string employed in the present study was (‘experience sampling’ OR ‘ESM’ OR ‘ecological momentary analysis’ OR ‘EMA’ OR ‘electronic diary’ OR ‘momentary assessment’) AND (‘depression’ OR ‘depressive disorder’).
Results
Studies that sampled the participants' experience only once per day (e.g. Stetler & Miller, Reference Stetler and Miller2005; Cohen et al. Reference Cohen, Gunthert, Butler, Parrish, Wenze and Beck2008) were deemed not to sample the ambulatory nature of participant's states in sufficient detail to be comparable with the majority of other studies identified by our search and were hence excluded. Other studies (e.g. Putnam & McSweeney Reference Putnam and McSweeney2008; Silk et al. Reference Silk, Dahl, Ryan, Forbes, Axelson, Birmaher and Siegle2007; Forbes et al. Reference Forbes, Hariri, Martin, Silk, Moyles, Fisher, Brown, Ryan, Birmaher, Axelson and Dahl2009) that only sampled experience in order to create an ecologically valid measure of a given construct (e.g. affect) were also excluded. ESM studies that failed to use hierarchical or multi-level regression analyses (Barge-Schaapveld et al. Reference Barge-Schaapveld, Nicholson, van der Hoop and DeVries1995; Conrad et al. Reference Conrad, Wilhelm, Roth, Spiegel and Taylor2008) in addressing topics that later studies subsequently addressed using such analyses were also excluded from this review. However, the study of Whalen et al. (Reference Whalen, Silk, Semel, Forbes, Ryan, Axelson, Birmaher and Dahl2008), which failed to employ hierarchical or multi-level regression analyses (instead using a mixed method analysis of variance), was included as it made a contribution to the field by assessing a novel variable (caffeine use) not reported on by other studies. After this process, 19 published articles were selected for inclusion in this review. Table 1 provides details of the characteristics and main findings of these studies.
Table 1. Summary of Experience Sampling Methodology (ESM) studies of depression
MDD, Major depressive disorder; PA, positive affect; NA, negative affect; mQoL, momentary quality of life; BD, bipolar disorder; NAP, non-affective psychosis.
All participants were aged >18 years unless stated otherwise.
Aspects of MDD as explored using ESM
Although all 19 studies reviewed collected data within the ‘flow of everyday experience’ (de Vries, Reference de Vries and de Vries2006, p. 8), six broad themes were identified.
Theme 1: Methodological issues
Husky et al. (Reference Husky, Gindre, Mazure, Brebant, Nolen-Hoeksema, Sanacora and Swendsen2010) examined acceptance rates, compliance rates, fatigue effects and reactivity to ESM techniques in individuals with MDD in order to establish the feasibility of this methodology in this population. All patients with MDD who were invited to enter the ESM study agreed to participate. Altogether, 95% completed at least 50% of the assessments and 70% completed 95% of the assessments. A practice effect was found, with MDD patients completing assessments more quickly as the study proceeded. The authors concluded that the results supported the feasibility and validity of this methodology in patients with MDD. However, it should be noted that Husky et al. used a lower frequency of sampling (five beeps per day) than most of the other studies (10 beeps per day) reviewed here. This may mean their compliance rates are not generalizable to the other studies reviewed in this paper.
Two studies examined how data collected by ESM compared with data obtained from retrospective reports. A first study examined data collected by these two methodologies on levels of positive affect (PA) and negative affect (NA). Ben-Zeev et al. (Reference Ben-Zeev, Young and Madsen2009) found that levels of PA and NA, reported by both MDD patients and healthy controls during an ESM study, differed from their retrospective reports. Both groups' retrospective reports overestimated the level of PA and NA that they reported in the ESM stage of the study. MDD patients were more inaccurate at recalling levels of NA (but not PA) than healthy controls. It was also found that, whereas controls recalled more PA than NA, MDD patients did not show this bias, but neither did they recall more NA than PA. In a later study, Ben-Zeev & Young (Reference Ben-Zeev and Young2010) examined this effect for specific symptoms of depression. Patients with MDD were found to have a greater negative retrospective bias (i.e. they claimed in retrospect to have experienced higher levels of such symptoms than they actually reported in the ESM phase) than healthy controls for anhedonia, suicidality, sadness and confusion. Healthy controls were found to have positive retrospective biases (i.e. they claimed in retrospect to have experienced lower levels of such symptoms than they actually reported in the ESM phase) for detachment, helplessness, self-control and anhedonia, which patients with MDD did not display.
The three methodological studies of ESM with patients with MDD indicate that the method is tolerable, provides data that appear to better represent their internal states than retrospective methodologies and provides additional information in terms of the nature of their retrospective biases.
Theme 2: PA, NA and stress
All studies that reported data on absolute levels of PA and NA found that individuals with MDD reported higher levels of NA and lower levels of PA than healthy controls (Barge-Schaapveld et al. Reference Barge-Schaapveld, Nicolson, Berkhof and DeVries1999; Myin-Germeys et al. Reference Myin-Germeys, Peeters, Havermans, Nicolson, de Vries, Delespaul and van Os2003b; Peeters et al. Reference Peeters, Nicolson, Delespaul, deVries and Berkhof2003c, Reference Peeters, Berkhof, Delespaul, Rottenberg and Nicolson2006; Bower et al. Reference Bower, Bylsma, Morris and Rottenberg2010). By comparing the diurnal mood variation of PA and NA in adults with MDD and healthy controls, Peeters et al. (Reference Peeters, Berkhof, Delespaul, Rottenberg and Nicolson2006) found that moment-to-moment variability of mood state was more pronounced in those with more severe depression. Depressed participants' PA levels increased more during the course of the day, peaked later and showed reduced inter-individual variation. Their NA levels showed increased diurnal variation with decreasing NA levels during the course of the day, greater inter-individual and more moment-to-moment variation than healthy controls.
Peeters et al. (Reference Peeters, Nicolson, Delespaul, deVries and Berkhof2003c) found that individuals with MDD reported similar frequencies of negative daily events compared with healthy controls, but that they appraised these events as being more stressful, unpleasant and important. Yet despite making such appraisals, MDD patients showed significantly smaller (‘blunted’) NA increases in response to negative daily events than healthy controls. However, negative daily events had a more persistent effect on NA levels for MDD participants. A later study by the same group (Peeters et al. Reference Peeters, Berkhof, Rottenberg and Nicolson2010) found that diminished emotional reactivity to everyday life events predicted relatively higher depression severity levels after 1 month of a pharmacotherapy and psychotherapy intervention. Furthermore, individuals who exhibited reduced NA reactivity to negative daily events were also less likely to have recovered 18 months later.
Myin-Germeys et al. (Reference Myin-Germeys, Peeters, Havermans, Nicolson, de Vries, Delespaul and van Os2003b) also examined perceived stress and affect in patients with MDD. They found that MDD patients reported more stress due to ongoing activities and their social environment than did controls. They also found, in seeming contradiction to Peeters et al. (Reference Peeters, Nicolson, Delespaul, deVries and Berkhof2003c), that the MDD group had a greater increase of NA in response to stress resulting from ongoing activities (but not social context) than healthy controls. Myin-Germeys and colleagues proposed that their results differed to Peeters and colleagues due to the different operationalizations of stress (activity-related stress versus negative daily events respectively). However, Wichers et al. (Reference Wichers, Schrijvers, Geschwind, Jacobs, Myin-Germeys, Thiery, Derom, Sabbe, Peeters, Delespaul and van Os2009a), in a study reported in more detail in Theme 6, found (using a negative daily events definition of stress) that MDD was associated with greater levels of NA in response to negative daily events. Hence, the difference between event- and activity-related stress cannot fully explain the incongruence of the results of Myin-Germeys et al. (Reference Myin-Germeys, Peeters, Havermans, Nicolson, de Vries, Delespaul and van Os2003b) with those of Peeters et al. (Reference Peeters, Nicolson, Delespaul, deVries and Berkhof2003c). This highlights the need for clear, comparable operationalizations of affect and stressor types in ESM studies.
Another ESM study examined factors that predicted levels of NA and PA in MDD patients. Using ESM, Bower et al. (Reference Bower, Bylsma, Morris and Rottenberg2010) found that sleep quality predicted levels of PA in both healthy controls and patients with MDD, but did not predict levels of NA in either of these groups once diagnostic status had been controlled for. Going beyond simple measures of affect, Barge-Schaapveld et al.'s (Reference Barge-Schaapveld, Nicolson, Berkhof and DeVries1999) ESM study attempted to look at a more general measure of well-being, quality of life and, in particular, ‘moment-to-moment quality of life’ (mQoL) as assessed by the question: ‘In general, how is it going with you right now?’. The authors found that, compared with healthy controls, depressed participants had more frequent and severe physical complaints, lower levels of mQoL and less activity enjoyment. Although there were no group differences in time spent in social, leisure or maintenance activities, depressed participants were more likely than controls to be doing nothing and less likely to be engaged in work (despite similar employment rates in both groups). Analyses showed that PA and activity enjoyment were associated with higher mQoL, whereas NA and physical complaints were associated with lower mQoL in both groups. Among depressed participants there was more moment-to-moment, day-to-day and inter-individual variability in mQoL.
Finally, Whalen et al. (Reference Whalen, Silk, Semel, Forbes, Ryan, Axelson, Birmaher and Dahl2008) used EMA to investigate interrelations between affect and caffeine use in youths with MDD. Their study found that at baseline youths with MDD used more caffeine than healthy controls. Over 8 weeks of cognitive behavioural therapy (CBT) and/or psychopharmacology, MDD patients experienced a four-fold decrease in caffeine consumption across treatment, even though they were not explicitly told to abstain from caffeine. The authors suggest that youths with MDD may use caffeine to help treat symptoms of depression. The authors also examined the relationship between caffeine consumption and NA. The daily level of caffeine consumption predicted levels of all four types of NA on the same day (sadness, anger, nervousness and upset). Additionally, an interaction was found between caffeine consumption and diagnostic group in predicting nervousness, with daily caffeine use being positively associated with daily nervousness for youths with MDD, but not for healthy individuals.
In summary, it appears that altered diurnal mood variation and emotional reactivity, lower PA and higher NA, alongside more moment-to-moment variability in levels of affect are characteristic of depression. In addition to producing findings specifically pertaining to NA and PA, these studies also show how ESM can be used to assess other variables (e.g. quality of life, nutritional intake and levels of specific forms of stress) that are likely to prove of relevance to our understanding of MDD.
Theme 3: Cortisol secretion
Hormones secreted from the hypothalamic-pituitary-adrenal axis play a role in the aetiology of mood disorders (e.g. Holsboer, Reference Holsboer2001) and it is understood that cortisol helps restore homeostasis following stress. ESM has been used to assess such moment-to-moment physiological changes and concurrent affect.
Participants in Peeters et al.'s (Reference Peeters, Nicolson and Berkhof2003a) study collected salivary cortisol at the same time as making ESM self-reports. Participants with MDD, unlike healthy controls, showed no significant increase in cortisol following negative events. Subanalyses showed that cortisol responses to negative events were significantly lower in men with MDD (compared with women with MDD) and also lower in MDD participants with a family history of mood disorder. Although participants in this study with MDD did not differ from healthy controls in their diurnal pattern of cortisol levels, in a more detailed examination of cortisol levels and variability over the course of the day by Peeters et al. (Reference Peeters, Nicolson and Berkhof2003b) analysis of within-person variability among MDD patients revealed a more erratic pattern of cortisol secretion throughout the day, particularly in patients with more severe or recurrent episodes. Yet, although small elevations in cortisol secretion were associated with more severe depressive symptoms, there was no evidence of hypercortisolism in MDD patients, compared with healthy controls.
Stetler et al.'s (Reference Stetler, Dickersin and Miller2004) study examined the relationship between cortisol secretion and ‘social zeitgebers’ among clinically depressed individuals and healthy controls. A ‘zeitgeber’ is any external cue that entrains (i.e. modifies in order to align with external environmental cues) the internal time-keeping systems of organisms. Contrary to expectation, there were no significant group differences in diurnal cortisol secretion or regular activities (social or non-social). However, analyses revealed an uncoupling of the relationship between regular activity and diurnal cortisol patterns in the depressed group. Whereas regular activities (especially those of a social nature) were important in entraining cortisol secretion among healthy controls, this was not found for depressed participants.
Summarizing the findings from these studies examining cortisol secretion in MDD using ESM is difficult. However, it is clear that ESM has established an altered pattern of cortisol reactivity in MDD, with subtle differences in the relationship between cortical secretion and stressful or social events in depression compared with that seen in non-depressed adults.
Theme 4: Antidepressant treatment
Antidepressant medication remains the most common treatment for depression in the UK (Hollinghurst et al. Reference Hollinghurst, Kessler, Peters and Gunnell2005), despite the ready accessibility and efficacy of alternative treatments and concerns over the dangers of antidepressants. Barge-Schaapveld & Nicolson (Reference Barge-Schaapveld and Nicolson2002) utilized ESM in a clinical drug trial. Participants diagnosed with MDD were given an antidepressant or placebo treatment, and a healthy control group was employed. Treatment-related increases in frequency and severity of side-effects were associated with lowered mQoL, as assessed by ESM, especially among treatment drop-outs. Despite greater clinical improvement in depression at week six, participants receiving antidepressant treatment did not report greater increases in mean mQoL ratings vis-à-vis placebo participants. However, drug treatment did stabilize mQoL fluctuation and led to increased activity. Participants' decisions to prolong treatment depended on clinical improvement, mQoL changes and some specific early side-effects. By week 18, although remitted participants' global QoL normalized (as measured by retrospective questionnaires), mQoL differed from that of healthy controls. These results suggest that everyday quality of life takes longer to return to normal than retrospective measures suggest.
More recently, Wichers et al. (Reference Wichers, Barge-Schaapveld, Nicolson, Peeters, de Vries, Mengelers and van Os2009b) performed a randomized controlled trial in which MDD patients received either imipramine or placebo for 6 weeks. At baseline, patients with MDD were found, using ESM, to have higher levels of activity-related stress and lower levels of activity-related reward than controls. Emotional reactivity to such events also differed, with patients displaying an increased NA response to activity-related stress as compared with controls (replicating Myin-Germeys et al. Reference Myin-Germeys, Peeters, Havermans, Nicolson, de Vries, Delespaul and van Os2003b). After treatment it was found that stress-sensitivity (responding with NA to negative stimuli) decreased more and reward experience (responding with PA to pleasant stimuli) increased more in the treatment group than the placebo group. The authors also reported the notable finding that ‘increases in Reward Experience discriminated between treatment responders and non-responders, independently from changes in appraisal and thus appeared to be a necessary condition for response to treatment, in contrast to changes in Stress-Sensitivity’ (p. 928).
ESM studies evaluating the effects of antidepressant treatment have hence demonstrated the utility of using ESM to better understand the diverse and complex effects of pharmacological agents on internal states and behaviour.
Theme 5: Work performance
Depression is associated with absenteeism and reduced productivity at work as well as redundancy (Adler et al. Reference Adler, McLaughlin and Rogers2006). Wang et al. (Reference Wang, Beck, Berglund, McKenas, Pronk, Simon and Kessler2004) used ESM to examine the impact of MDD on work performance in service workers. MDD was found to be related to decrements in task focus and productivity at work. Wang and colleagues argued that the impact of depression on job performance was greater than the impact of depression on absenteeism and subsequently estimated that the costs of lost productivity exceed those of effective treatment. A major impact of this study is the economic evidence it provides for improving treatment provision for depressed employees.
Theme 6: Genetic risk factors
Levinson's (Reference Levinson2006) review of the role of genetics in depression describes MDD as moderately heritable and notes an overlapping genetic susceptibility with the personality trait neuroticism. It is recognized that the tendency to develop NA in response to stress is closely related to neuroticism (Kendler et al. Reference Kendler, Kuhn and Prescott2004). This is sometimes referred to as a ‘mood bias towards NA’ (Erickson et al. Reference Erickson, Drevets and Clark2005). Wichers et al.'s (Reference Wichers, Myin-Germeys, Jacobs, Peeters, Kenis, Derom, Vlietinck, Delespaul and van Os2007a) ESM examined the contribution of genetics in mood bias towards NA in response to stress. Non-depressed twins with a co-twin diagnosed with MDD exhibited greater NA toward daily life stress than individuals with a healthy co-twin, suggesting that genetic liability to depression may, in part, be expressed by a heightened negative affective response to everyday stress. Wichers et al. (Reference Wichers, Myin-Germeys, Jacobs, Peeters, Kenis, Derom, Vlietinck, Delespaul and van Os2007b) later re-analysed some of the data collected during this study and found some evidence that PA acts as a protective factor in NA reactivity. Furthermore, their findings indicated that shared genetic vulnerability to depression matters less among individuals who experience more day-to-day PA.
A later study by Wichers et al. (Reference Wichers, Schrijvers, Geschwind, Jacobs, Myin-Germeys, Thiery, Derom, Sabbe, Peeters, Delespaul and van Os2009a) attempted to determine the mechanism through which genes had an effect on the increased affective reaction to stress by examining if there was a differential effect of earlier stressful events on the stress-reactivity of those at high and low genetic risk of depression. Non-depressed twins at high genetic risk of depression (i.e. with a co-twin with a history of depression) developed higher levels of daily life stress-sensitivity after exposure to prenatal (birth weight) and post-natal (childhood adversity and adult recent negative life events) stress than those at low genetic risk. The authors argued that genes associated with depression may act by accelerating the process of stress-sensitization following stress exposure over the life course.
The findings of Wichers and colleagues' studies on ESM and genetic risk have indicated that a genetic liability to depression may, in part, operate through a heightened negative affective response to everyday stress and an accelerated process of stress-sensitization. Furthermore, using ESM has highlighted that PA may have a protective effect in those with such a genetic liability. However, we note that Wichers and colleagues' all-female samples meant the findings are not directly generalizable to male populations.
Discussion
In total, 19 ESM studies examining MDD were found and six research foci identified: methodology; positive affect, negative affect, and stress; cortisol secretion abnormalities; the side-effects and benefits of antidepressant treatment; work performance; and genetic risk factors in depression. The results of this review show that ESM studies of depression have already made important contributions to our understanding of MDD.
Support for existing theories
ESM has contributed to MDD research by supporting and adding strength to existing theories. The results of Peeters et al. (Reference Peeters, Nicolson and Berkhof2003a, Reference Peeters, Nicolson and Berkhofb), for example, corroborate the dysregulation hypothesis of cortisol abnormalities in MDD (Siever & Davis, Reference Siever and Davis1985). In addition, the work of Wichers et al. (Reference Wichers, Myin-Germeys, Jacobs, Peeters, Kenis, Derom, Vlietinck, Delespaul and van Os2007b) supports the ‘broaden-and-build theory’ of the protective function of PA in depression (Fredrickson, Reference Fredrickson2001). In view of these findings, one of ESM's strengths within MDD research is its ability to strengthen existing theory with ecologically valid evidence.
New findings from ESM
A number of new insights into MDD have been generated by ESM studies. The work of Peeters and colleagues and other groups using ESM has captured characteristic details of the profile and persistence of NA and PA that retrospective designs would be incapable of revealing. For example, Peeters et al.'s (Reference Peeters, Nicolson, Delespaul, deVries and Berkhof2003c) finding that individuals with MDD do not report more frequent negative events, but do report fewer positive events, contradicted much of the previous research in this area, gathered using retrospective designs (e.g. Ravindran et al. Reference Ravindran, Griffiths, Waddell and Anisman1995). Furthermore, Peeters and colleagues also found that although individuals with MDD may react with less NA to daily negative events they recover more slowly, prolonging the impact of negative events in such individuals. The importance of this result was supported by the results of their later ESM work (Peeters et al. Reference Peeters, Berkhof, Rottenberg and Nicolson2010) showing that individuals with MDD with diminished reactivity to negative everyday life events were less likely to have recovered 18 months later.
The mechanisms through which a genetic component to depression may operate has also benefited from the use of ESM methodologies. For example, Wichers et al. (Reference Wichers, Schrijvers, Geschwind, Jacobs, Myin-Germeys, Thiery, Derom, Sabbe, Peeters, Delespaul and van Os2009a) have found that a genetic predisposition to depression may operate through the development of higher levels of daily life stress-sensitivity. This offers the potential for the advancement of a more integrated biopsychosocial model of depression and has clinical implications for psychological interventions aimed at preventing the onset of depression among individuals where there is a significant familial or genetic risk of onset (see below).
Limitations of ESM in depressed populations and of existing ESM studies
A number of potential limitations to using ESM to collect data in specifically depressed populations still remain to be considered. For example, depression is known to be associated with ruminative thinking (e.g. Kim et al. Reference Kim, Yu, Lee and Kim2011) as well as high trait levels of self-focused attention (Dunn et al. Reference Dunn, Dalgleish, Lawrence and Ogilvie2007). It is plausible that participation in ESM studies may further exacerbate such thinking/attention patterns, which, in addition to the ethical issues this raises, may result in artefacts in the data collected. However, as we note below, this potential impact of ESM may actually be used to aid the therapeutic process. Furthermore, as CBT for depression includes helping patients correctly identify their emotions (e.g. anger, anxiety, sadness, shame, etc.), the stage of therapy a patient is at may impact upon the type of affective states or cognitions they report, introducing further artefacts into data collected through ESM.
A number of limitations of existing ESM studies of MDD are also apparent. First, they could be improved by employing a greater range of outcome measures, going beyond the currently predominant affective (e.g. NA and PA) and physiological (e.g. cortisol) measures, to more fully understand MDD. Studies such as Barge-Schaapveld et al. (Reference Barge-Schaapveld, Nicolson, Berkhof and DeVries1999), which assessed mQoL, and Whalen et al. (Reference Whalen, Silk, Semel, Forbes, Ryan, Axelson, Birmaher and Dahl2008), which assessed the relationship between affect and caffeine use, have already indicated the utility of a diverse range of measures.
A second shortfall of existing ESM studies is their failure to assess the specific cognitions that are central to cognitive theories of depression. For example, at the heart of Beck's (Reference Beck1976) model of depression is the cognitive triad of negative automatic thoughts, systematic logical errors and depressogenic schemas. All items in this triad could potentially be assessed using ESM to tap into ongoing examples of these and their relationship to affect and depressive symptomatology. It is hence imperative that future ESM studies attempt to test the hypothesis that negative automatic thoughts precede negative mood. In order to address other causal questions, it would also be interesting to use ESM to examine specific cognitions and behaviours pretreatment, as well as change following psychological interventions such as CBT. Such studies could usefully investigate whether these therapies are actually working as expected in terms of the changing relationship between thought and emotion.
Another area requiring future development results from the fact that ESM studies are currently limited by existing statistical analysis methods. The rich time-series data generated by ESM has so far been analysed in fairly rudimentary ways. Analysis of unevenly spaced observations within a multi-level modelling framework will require new statistical techniques that are still being developed.
Treatment
It has been argued that current treatment guidelines for depression are inadequately informed by recent research findings (Luyten et al. Reference Luyten, Blatt, van Houdenhove and Corveleyn2006) and the findings of ESM studies of MDD can be seen to be one area that has yet to be fully developed in terms of their treatment potential. Indeed, the ESM studies reviewed here have a number of implications for the treatment of MDD.
ESM studies of NA and PA patterns in MDD (e.g. Peeters et al. Reference Peeters, Nicolson and Berkhof2003a) have clear implications for treatment. For example, psychological interventions for MDD should focus on stress management training (particularly in those at high genetic risk of developing MDD), as well as behavioural activation and increasing exposure to pleasant events. This conclusion is reinforced by genetic studies of MDD employing ESM. For example, studies (e.g. Wichers et al. Reference Wichers, Myin-Germeys, Jacobs, Peeters, Kenis, Derom, Vlietinck, Delespaul and van Os2007b) showing that PA may act as a protective factor against a genetic predisposition to MDD also suggest that behavioural activation and increasing exposure to pleasant events may play an important role in preventing the development of MDD. ESM studies of response to antidepressant treatment, which show response to be associated with an increase in reward experience (e.g. Wichers et al. Reference Wichers, Barge-Schaapveld, Nicolson, Peeters, de Vries, Mengelers and van Os2009b), also support this proposal. It can also be seen that ESM studies showing that MDD patients' retrospective reports of their NA and PA are more inaccurate than those of healthy controls (Ben-Zeev et al. Reference Ben-Zeev, Young and Madsen2009; Ben-Zeev & Young, Reference Ben-Zeev and Young2010) highlight the importance of tools utilized in CBT, such as mood and thought records (e.g. Greenberger & Padesky, Reference Greenberger and Padesky1995), in allowing patients and clinicians to form an accurate historical representation of the patient's cognitive and affective states.
Importantly, future research should examine the feasibility of using ESM itself as a clinical intervention in MDD. Indeed, several studies have already described therapeutic approaches using ESM. For example, Newman et al. (Reference Newman, Kenardy, Herman and Barr1997) gave patients with panic disorder palm top computers and at random intervals through the day they prompted participants to complete a breathing retraining exercise or a programme to help them alter their thinking and remain in the present situation, resulting in clinically significant improvements.
None of the ESM studies reported here considered the impact of prolonged self-reporting on participants' awareness, knowledge and experience of internal phenomenon. As ESM prompts individuals to think about their thoughts and feelings as they go about their day-to-day life, it is plausible that it could be used to develop individuals' awareness of cognitive and emotional processes. In this respect, ESM could potentially be used as an intervention tool by helping patients with MDD become more mindful of physical symptoms, unhelpful thinking patterns and external triggers. In particular, ESM appears ideally suited to use with mindfulness-based cognitive therapy for MDD (Segal et al. Reference Segal, Williams and Teasdale2002), which employs mindfulness techniques to create awareness of negative thinking patterns and disengagement from ruminative depressive cycles. Furthermore, the potential of ESM to encourage an awareness of moments that are temporally limited and contextual, which may reduce a maladaptive tendency to have overgeneralized autobiographical memory in depressed individuals (Williams & Scott, Reference Williams and Scott1988), may also be a fruitful area for research.
Conclusions
ESM studies have made a significant contribution to our understanding of depression, consolidating existing theories, uncovering new, clinically relevant information and identifying questions requiring further exploration. A number of key ways in which ESM could be used to further increase our understanding of MDD have been suggested here. Importantly, this review also suggests that future studies should investigate the feasibility of ESM as a clinical tool in psychological therapies.
Declaration of Interest
None.
