Real-Time Psychological Factors Using Ecological Momentary Assessment With Continuous Glucose Monitoring in Diabetes Mellitus: Scoping Review

Introduction

Diabetes self-management is a dynamic process that involves various biological, psychological, social, and behavioral factors, including sleep, exercise, stress, and diet [1-4]. The complex system of influences and moment-to-moment changes in self-monitored blood glucose pose challenges to researching these interactions. Over the past two decades, diabetes technologies have expanded substantially with innovative methods to monitor and treat diabetes, as well as improve understanding of how biopsychosocial interactions influence the momentary changes in blood glucose.

Ecological momentary assessment (EMA) is an increasingly used tool that allows for the capture of momentary psychosocial factors, including individual experiences and behaviors in real time within the natural environment. EMA has become more widely available in research settings through the emergence of web-, phone-, and wearable health technology-based devices [5]. EMA improves reliability by decreasing recall bias, maximizing ecological validity, and gathering information about the micro processes that influence daily behaviors [5]. Compared to other longitudinal study methods, EMA allows for frequent, brief, and accessible assessments that provide an in-depth analysis of behaviors, experiences, and emotions at the individual level [6]. This makes EMA an ideal technology to use for patient-based research [5,7] on chronic diseases with multiple self-management tasks such as diabetes [8].

Self-monitoring of blood glucose (SMBG) is among the diabetes self-management tasks required daily for many individuals with diabetes. SMBG has historically been done with invasive and uncomfortable capillary glucose checks, which often results in inadequate glycemic data [9,10]. Continuous glucose monitoring (CGM) has dramatically changed SMBG by capturing glucose readings every 1‐15 minutes without the need for fingerstick [11], improving the SMBG experience [12-14], and illuminating the dynamic nature of momentary glucose patterns [15,16]. This large amount of glucose data has also highlighted the inadequacies of historically limited measures of glycemic control, such as glycated hemoglobin. Despite the widespread use of hemoglobin A_1c (HbA_1c), analytical interferences and underlying medical conditions limit the accuracy of HbA_1c readings [17]. The granular glucose data obtained with CGM has allowed for more comprehensive CGM metrics that have now been standardized for clinical use and better describe the nuances found in momentary glucose fluctuations [18]. These CGM-derived “glucometrics” include the mean glucose, distributive measures such as the time above range (TAR; defined as glucose >180 mg/dL), time in range (TIR; defined as glucose 70‐180 mg/dL), time below range (TBR; defined as glucose <70 mg/dL), and measures of variability such as the coefficient of variation (CV) [18]. CGM allows researchers, clinicians, and individuals with diabetes to gain an understanding of how environmental, physiological, psychological, and behavioral influences impact glucose fluctuations [18].

Several reviews have examined the relationship between psychosocial factors and glucose. Mujis and colleagues [19] conducted a systematic review of 8 studies examining the association between glucose variability and mood in adults with diabetes. In this review, no clear association between glycemic variability and mood was found [19]. Another systematic review done by Nam et al [8] evaluated the relationships between psychosocial factors and diabetes self-management, finding that negative affect was associated with poor adherence to glucose monitoring and insulin injections, while stress was associated with binge eating behaviors. However, both reviews included studies that captured glucose using multiple modalities, including self-report and SMBG, which limits the availability and accuracy of glucose outcomes compared to the use of CGM.

Combining EMA and CGM methods provides a powerful opportunity to disentangle complex biopsychosocial influences and consequences of glycemic fluctuations in real or near-real time. Pairing these technologies can allow for higher validity in terms of patient responses [5] and can substantially lessen the amount of time between the measurements of glucose, psychological, and behavioral factors [20]. In a recent narrative review of EMA and CGM by Ehrmann et al [21], multiple research applications of these combined methodologies were noted, including a better understanding of the longitudinal variability in experiences and behaviors, delineation of objective versus subjective glycemia, and clarification of temporal associations between psychosocial factors and changes in glycemia. While these combined methods are increasingly used in research studies, Erhmann et al [21] also noted the potential for precision medicine clinical applications in the form of ecological momentary interventions such as just-in-time adaptive interventions.

Although a growing number of studies combining EMA and CGM data collected simultaneously are being conducted, there is a lack of consensus regarding the development of EMA protocols for question generation, delivery duration, delivery frequency, or delivery timing. Additionally, although CGM glucometrics have been standardized for clinical use, EMA-specific glucometrics have not been developed. Best practices regarding EMA-specific biopsychosocial outcomes and glucometrics are crucial to optimize the scientific rigor needed to ensure the validity and reliability of combined EMA and CGM studies, particularly if these data will be used for the development of future ecological momentary interventions. However, prior to the development of best practices, a more in-depth review of these studies is needed. Therefore, the main objective of this scoping review is to summarize the methods, outcomes, and results of studies that use both EMA and CGM data simultaneously as an initial step to standardizing future study protocols using these combined tools.

Methods

Protocol and Registration

The study protocol was developed in accordance with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) statement [22,23] and can be found in Multimedia Appendix 1 and Checklist 1. This protocol was not registered.

Eligibility Criteria

In line with the objective to summarize study methodologies of studies that analyzed concurrent EMA and CGM data in diabetes mellitus, studies were included if they were (1) peer-reviewed, original research studies; (2) recruited participants with diabetes; and (3) EMA and CGM data obtained and analyzed simultaneously. Studies were excluded if they (1) were non-original research articles (including exclusion of protocol papers, systematic reviews, meta-analyses, narrative reviews, or conference proceedings), (2) lacked CGM data or had asynchronous CGM data, (3) captured glycemic data using other modalities (including capillary, that is, fingerstick, glucoses or HbA_1c), and (4) were studies not published in English.

Information Sources

A literature review was performed using the PubMed, Elsevier Embase, and EBSCOhost CINAHL databases using the following key terms: “ecological momentary assessment AND blood glucose self-monitoring, diabetes mellitus, or glucose monitoring.” The search was conducted in May 2023 and again in July 2026, and articles published until December 2025 were considered for this review. Microsoft Excel version 2402 was used to organize and track the results during the review process.

Search Strategy

The search strategy was developed in conjunction with a medical informationist and includes the following search terms: ecological momentary assessment, blood glucose self-monitoring, continuous glucose monitoring, blood sugar, CGM, self-monitor, and diabetes. The full search strategies for PubMed, Elsevier Embase, and EBSCOhost are shown in Multimedia Appendix 2.

Selection of Sources of Evidence

The literature search resulted in a total of 782 studies from the 3 databases in 2 separate searches. After duplicates were removed, 610 studies remained. Two independent reviewers (KV and DA for the first search conducted in May 2023; KV and KM-S for the second search conducted in July 2026) reviewed study titles and abstracts of the 610 individual studies, and of these, 567 studies were excluded based on the prespecified exclusion criteria. Forty-three full-text studies were reviewed by two authors (KM-S and DA) independently; any disagreements were resolved by discussion with a third reviewer (KV), after which an additional 24 studies were excluded, for a total of 19 studies included in this review (as shown in the Results section).

Data Charting Process

Key characteristics of EMA and CGM methodology were identified based on consensus from the 3 authors (KV, KM-S, and DA) prior to beginning the data extraction process. Included studies were reviewed, and details on study characteristics, EMA protocols, EMA outcomes, CGM characteristics, CGM outcomes, and integrated EMA and CGM data analysis and outcomes were extracted by 2 authors (KM-S and KV for the first search, KM-S and DA for the second search), charted in Microsoft Excel, and then discussed as a group of 3 reviewers (KM-S, KV, and DA) to resolve discrepancies. Throughout the data extraction process, additional study details were added to capture other pertinent methodological details found throughout multiple studies based on consensus from 3 authors (KV, KM-S, and DA). These additional study details included participant compensation, EMA response rates, and statistical analysis plans.

Data Items

Study characteristics collected included study design (observational vs intervention), age range (target population age and sample age), diabetes type (type 1 and type 2), and sample size. EMA data collected included duration in days, prompts over the duration period (prompts per day, total responses over the study period), EMA response rate (percent completed), and psychological and/or behavioral outcomes as defined by each study. Similarly, CGM data included type of CGM device (blinded or personal), duration of wear, and glucose outcomes as defined by the study team (eg, mean glucose, TIR, TAR, TBR, SD, and CV). For integrated EMA and CGM data, study objectives, statistical methods, integrated EMA and CGM outcomes, and results were collected. A copy of this data extraction table is shown in Multimedia Appendix 3.

Critical Appraisal of Individual Sources of Evidence

In line with the descriptive objective of this study to explore the broad study methodologies that incorporate both simultaneous EMA and CGM data, a critical appraisal of individual sources of evidence was not conducted.

Synthesis of Results

A full descriptive summary of the study characteristics, EMA data, CGM data, and integration of EMA and CGM data are shown in the table in Multimedia Appendix 4. A summary of the current evidence from the included studies, as well as key methodological or study gaps, is highlighted in the gap map shown in the Results section.

Results

Selection of Sources of Evidence

As shown in Figure 1, a total of 19 studies were included in this review. Nonduplicate studies from the literature search were excluded primarily due to being a review, abstract, or protocol paper; not using CGM to collect blood glucose data; and not including study participants with diabetes.

‎

Characteristics of Sources of Evidence

Study design characteristics across all studies are shown in Table 1. The majority of the studies were conducted in adult populations, while only 3 (19%) studies included adolescent populations. The weighted mean age of recruited participants was 44 years (range: 15.7‐61 years). Of the 19 studies, 14 (74%) were done in only type 1 diabetes, 2 studies were done in only type 2 diabetes, and 3 studies were done in both type 1 and type 2 diabetes populations. The sample sizes of the included studies varied widely from 18 to 602 participants (median 92, IQR 103). All studies were observational in nature.

Critical Appraisal Within Sources of Evidence

A critical appraisal of the included sources of evidence was not conducted as the objective of this study was a descriptive review of all possible methodologies incorporating concordant EMA measures of psychological and behavioral outcomes with blood glucose data measured via CGM.

EMA Psychological and Behavioral Outcomes

EMA delivery of psychological and behavioral assessments showed considerable variability across studies, as shown in Table 2. The median duration of EMA delivery was 14 (range 3‐18) days with a median of 5.5 (range 1-28) prompts delivered per day. The median EMA response rate was high at 90.2% (IQR 15.4%), with lower response rates in longer-duration EMA studies (range of 66% response for 15 days of EMA to 96% for EMA duration of 3 days). Compensation was inconsistently reported but varied substantially (range: total compensation US $0‐100; compensation per EMA prompt US $0.33‐2.80).

EMA outcomes were heterogeneous and included emotions, behaviors, cognition, symptoms, workload, and contextual factors. Emotional constructs such as affect, motivation, self-efficacy, engagement, distress, and mood were measured in 10 studies and included positive and negative affect, diabetes motivation and engagement, emotional and diabetes distress, mood states, and health-related quality of life [24-28,30,32,37,38,41]. Behavioral constructs were measured in 9 studies (activity, diabetes self-management, eating behaviors, and relationship and social interactions) [24,25,28,31,35,37,40-42]. Various cognitive tasks were measured in 2 studies [29,33]. Additional EMA outcomes included sleep, workload, impacts of hypoglycemia, and diabetes symptoms [30,34,36,39].

Glucose Monitoring and Outcomes

As shown in Table 3, of the 19 studies, 14 (74%) used blinded CGM; of these, 5 studies [28,33,34,37,38] allowed participants to also wear their personal CGM devices as well. The rationale for blinding was only reported in 1 study [24], which was to reduce reactivity to glucose values. CGM devices were worn for a median of 14 days with a range of 3‐18 days, coordinated with EMA protocols. CGM outcomes included both standardized and nonstandardized glucose-related outcomes. Just over half of the studies (58%) used any combination of standardized CGM glucometrics [18] including mean glucose, time in ranges (TAR, TIR, and TBR), and measures of variability such as SD or CV. Three studies used nonstandardized glucose outcomes as surrogate measures for disordered eating behaviors including insulin restriction [24,35] and binge-eating [25,35]. Other outcomes evaluated included hypoglycemic events [29], nocturnal hypoglycemia [30], 2-hour postprandial glucose [41], and glycemic variability prior to weight-focused eating behaviors [42].

Integrated EMA and CGM Data

A unique feature of the studies included in this review is the integration of EMA and glucose data measured via CGM (see Table 4). The purpose of integrating EMA and CGM data varied widely. Of the 19 studies, 8 (42%) evaluated the impacts of biopsychosocial factors, including psychological states [26,27,30,37,38,40,41], and workload [34] on CGM-derived glycemic outcomes or self-management behaviors. In contrast, 3 studies also evaluated the impact of CGM-derived glycemic outcomes on EMA-measured emotional states (such as diabetes distress [32,38]), mood [30], sleep [30], and cognition [33], recognizing that bidirectional associations were also possible. In the only 2 studies evaluating type 2 diabetes mellitus (T2DM) specific populations, both evaluated psychosocial domains in relation to glycemia. In the first study, integrated EMA and blinded CGM data were used to evaluate the effects of daily diabetes-specific social support from partners on daily glycemic outcomes, finding that higher levels of partner involvement improve short-term glycemic outcomes [31]. In the second study, the impact of psychosocial factors on eating behavior and CGM-derived glycemic outcomes was assessed, finding that stress led to dietary lapses with resulting higher daily glucose levels [41]. Similarly, in the one adolescent-only study, the effects of social interactions were evaluated on CGM-derived glycemic outcomes, finding that more social interaction improved TIR but also increased variability as measured by SD [40].

Of the 19 studies, 4 used integrated EMA and CGM data to evaluate disordered eating behaviors. In one study, CGM-derived glycemic variability was used to predict disordered eating behaviors measured via EMA [42]. In the other 3 studies, EMA-derived measures of negative affect were evaluated for correlations with either insulin restriction [24] or binge-eating behaviors [25] based on study-specific metrics from blinded CGM, and disordered eating profiles were developed [35].

Finally, investigators also used integrated EMA and CGM data for study protocol [29] or mobile app [36] optimization. In the study by Mascarenhas Fonseca et al [29], EMA-derived cognitive measures were delivered at low frequency/longer duration versus high frequency/shorter duration to determine which prompt period was superior in finding the most blinded CGM-derived hypoglycemic events, finding that low frequency/longer duration prompts were associated with a higher capture rate for hypoglycemic events [29]. In terms of the statistical analysis approach, due to the inherent repeated measures within EMA studies, most studies used mixed models for analysis.

Methodological Gaps in Studies With Combined EMA and CGM Data

Despite the increased use of studies combining EMA and CGM quantitative data sources, there is little published guidance on protocol development. A review of the existing literature demonstrates notable methodological gaps, which are highlighted in Table 5. While one study evaluated the duration of EMA protocols specifically for future study optimization [29], consensus regarding the duration of EMA data collection and EMA question prompt frequency to minimize response burden and maximize response validity has not been reached. Additionally, there are very few EMA-specific outcome measures, and the psychometric properties of EMA questions derived from previously validated patient-response outcome measures used in the included studies were either not fully evaluated or not reported on by study teams.

Similarly, methodological gaps exist in CGM data collection as well. The majority of studies used blinded CGM for data collection though the rationale for CGM type was only reported in one study [24]. Additionally, over half of the studies (58%) used standardized CGM outcome measures though the duration of CGM data collection varied widely, potentially limiting the accuracy of these outcomes as standard duration for CGM data collection is 14 days [43]. The studies that used nonstandardized CGM outcomes may have been more specifically tailored to the momentary nature of the EMA data though the validity of these outcome measures is unclear.

The largest methodological design gaps involve the studied populations. Combined EMA and CGM studies have been primarily done in adults with type 1 diabetes mellitus (T1DM). Additionally, only 3 of the studies included adolescent populations and none of the studies specifically focused on older adults with diabetes. Whether or not associations found generalize to other populations such as individuals with T2DM or across the lifespan are not clear based on the available evidence.

Discussion

Summary of the Evidence

This review summarizes 19 studies that correlate EMA-measured psychosocial, cognitive, or behavioral factors with CGM-derived glucose data. Although the included studies primarily focused on adults with T1DM, other study design characteristics varied widely including sample size, EMA delivery duration, EMA frequency, and EMA biopsychosocial outcomes. Most of the studies captured glucose using participant-blinded data though CGM outcomes were nearly equally divided into either standard CGM glucometric outcomes or nonstandardized outcomes.

Study objectives varied widely but included the impacts of psychological factors on momentary glucose; the effects of momentary glucose on emotional states, mood, personal behaviors, sleep, and cognition; and optimization of study protocols or mobile apps. Advanced statistical models were used in nearly all studies to integrate EMA and glucose data.

Despite the wide variability in EMA protocols including the duration and frequency of prompts, the included studies overall demonstrated high levels of engagement with EMA. Mirroring the findings of one of the included reviews by Mascarenhas Fonseca et al [29], higher completion rates were seen in studies with shorter EMA duration versus longer EMA duration, perhaps reflective of higher participant burden. Based upon this review, EMA prompts may be optimized with durations between 3 and 7 days, which is an important consideration to reduce missing data and the resulting self-selection bias that has been found in EMA data [44]. However, Masacarenhas Fonesca et al [29] also found that although shorter EMA duration with higher frequency prompts led to higher overall completion rates, longer EMA duration with lower frequency prompts captured more hypoglycemia events, which was the primary glycemic outcome in this study. This suggests that different combinations of EMA duration and prompt frequency may be appropriate depending on the study purpose but should be evaluated and reported by study teams.

The biopsychosocial outcome measures varied broadly, as was shown in prior reviews of combined EMA and CGM data [21]. Although the majority of these studies developed EMA questions based on validated patient-reported outcome measures, these question items have not been validated for use in EMA protocols. As Ehrmann et al [21] noted in their review, standardization of EMA-item development, validation, and response cutoffs that are clinically meaningful is needed to ensure the scientific rigor for determining accurate associations between biopsychosocial factors and glucose prior to use in precision medicine applications such as ecological momentary interventions [21].

There was also significant heterogeneity in CGM duration and CGM outcomes measured in the studies included in this review. Slightly over half of the included studies used clinically standardized CGM outcome measures. According to the International Consensus on Time in Range [18], 14 days of CGM use correlates the strongest with 3-month data for key glucometrics [45,46] and is now the standard duration on the ambulatory glucose profile from CGM downloads [47]. Analysis of timeframes less than 14 days for standardized CGM glucometric outcomes, which was done in 5 studies [26,31,33,36,40], may result in glycemic data that is less representative for that individual. On the other hand, analysis of 14-day glucose data may not be in line with the momentary nature of EMA data collection. The remaining studies used nonstandardized glucose outcomes, which may better reflect the momentary nature of the study question but have limited validity. The broad range of glucose metrics, particularly nonstandardized metrics, also creates challenges with translating results to direct clinical care such as associations with longer-term diabetes outcomes such as retinopathy [48]. Validation of EMA-specific CGM metrics is important to ensure these outcomes are clinically meaningful, particularly as individuals without diabetes can experience glucose elevations detected on CGM that do not have any clinically important consequences [49].

Limitations

Although this review provides a comprehensive overview of the methods, outcomes, and results of original research studies that used both EMA and CGM data simultaneously in participants with diabetes, there are a number of limitations. First, although the EMA and CGM outcomes varied widely, the populations that these studies applied to focused primarily on adults. Therefore, gaps in using combined EMA and CGM methods for research remain in children, adolescents, and older populations. EMA may be challenging with children due to lack of access to digital health technologies [50] and ethical, cognitive, and developmental considerations for directly assessing child behavior and emotional states [51]. Although there is some evidence that EMA may be successfully implemented in children as young as 7 years [52], these studies have not been used in conjunction with CGM. The lack of EMA studies in older adult populations may be reflective of lower technology acceptance or confidence in this population. Previous literature has shown that feelings of inadequacy, anxiety, and overall lack of exposure surpass older participants’ willingness to adopt new technologies [53-55]. In addition, all but one of the included studies enrolled individuals with T1DM (92%) and only 4 studies (31%) enrolled participants with T2DM. This may be due to the need for frequent SMBG, which is recommended for all individuals with T1DM due to the need for insulin therapy. In contrast, only about one-quarter of individuals with T2DM require insulin, which is the main indication for SMBG in T2DM [56]. Future studies may benefit from expanding the populations enrolled to include a broader participant pool to ensure associations between psychosocial factors and glucose are generalizable.

Another limitation is the small sample size of original research studies using both EMA and CGM data together in one study design. Due to this, a meta-analysis was not feasible. Additionally, the significant heterogeneity in study design limits conclusions that can be definitively drawn for ideal EMA protocol design recommendations such as biopsychosocial question development and the duration and timing of question delivery, as well as CGM glucose outcomes. However, nearly all of the included studies were published just in the last 5 years, highlighting an increased use of these combined methodologies and highlighting the urgent need for standardizing EMA-specific biopsychosocial outcomes and EMA-specific CGM outcomes to ensure validity and reliability of this data. Nevertheless, this review provides a very detailed overview of the methodology and outcomes of original research studies that combine biopsychosocial EMA and CGM data.

Conclusions

Together, EMA and CGM can be used to disentangle the complex interactions between biopsychosocial factors and glucose data. However, given the broad application of these methods and lack of associated clinical relevance, best practices surrounding study design regarding biopsychosocial EMA protocols and CGM metrics are urgently needed to ensure scientific rigor and accuracy. This review provides a comprehensive overview of the studies using combined EMA and CGM methods in diabetes, which is an important stepping stone in the development of these best practices.