Remember me
People diagnosed with cancer, among which breast and lung cancer are the most prevalent diagnoses globally [], often experience various problems and concerns that affect their quality of life and well-being across physical, psychological, social, and spiritual domains [-]. Understanding the fluctuations, interactions, and contextual variations of the multidimensional problems and concerns in patients’ daily lives is crucial to gain a comprehensive view of these patients’ quality of life and to optimize patient-centered care. Such insights could lead to, among others, improvements in drug schedules and personalized treatment decision-making [] and the identification of novel care intervention targets by identifying contexts or states that aggravate or buffer against certain problems and concerns [].
An effective way to gather insights into the daily and within-day variability of patients’ quality of life and well-being is the use of intensive longitudinal methods. Bolger and Laurenceau [] defined intensive longitudinal methods as “an umbrella term to encompass data collection methods that employ enough repeated measurements to model a change process for each subject.” The authors specify a minimum number of 5 sequential assessments, as it enables the estimation of linear models within each participant []. Examples of such methods are daily diaries and ecological momentary assessments (EMAs), also known as experience sampling methods (ESM). While predominantly developed in psychological research, these methods recently gained more attention in other fields and clinical practice, including oncology, due to advancements in handheld computer technologies that enable easier implementation than traditional pencil-and-paper approaches [-]. Despite easier implementation of these methods, researchers and clinicians in the field of oncology still lack a clear understanding of available options for intensive longitudinal monitoring, their opportunities, pitfalls, and feasibility in populations experiencing high symptom burden. This underscores the need for a structured overview of the use and capabilities of these methods.
Currently, no systematically conducted literature review exists on the use of intensive longitudinal methods in monitoring people with cancer. One systematic review [] provided the most recent overview of the use of EMA in people with cancer across 42 studies (23 and 8 studies included people with breast and lung cancer, respectively) and found considerable heterogeneity in the methodologies used. However, due to its inclusion criteria focusing solely on EMAs, a large group of studies monitoring patients on a once-daily basis was left out []. Furthermore, the review did not report on the barriers and facilitators that were encountered during the implementation of ESM, which is crucial information for optimal use in practice [].
ObjectiveWe aimed to describe to what extent intensive longitudinal methods with daily electronic assessments have been used among patients with breast or lung cancer, along with the methodologies used, associated outcomes, and influencing factors. We limited the scope of this review to these patient groups with the most prevalent cancer diagnoses for feasibility reasons to provide a more nuanced picture for these methods among these groups and to inform our own ongoing ESM project among these patient groups []. More specifically, we described (1) the characteristics of the populations with breast or lung cancer among whom intensive longitudinal methods with daily electronic assessments have been used; (2) the objectives, design, and methods used; (3) the results obtained (including study findings and response-related results); and (4) the identified barriers and facilitators for implementing these methods in clinical and research practice.
We conducted a scoping review using a systematic search strategy to gain insight into the extent, range, and nature of current evidence on the use of intensive longitudinal methods with daily electronic assessments in people with breast or lung cancer, rather than providing evidence for a specific research question as in systematic reviews [,]. This manuscript adheres to the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) [].
Eligibility CriteriaWe included articles that met the following criteria: articles that (1) performed in people diagnosed with breast or lung cancer through self-report or proxy responding; (2) included people aged ≥18 years; (3) used active intensive longitudinal methods, meaning the conscious reporting of experiences rather than passive data collection through wearables without conscious participant involvement []; (4) collected self-reports using electronic devices or allowed participants to choose between electronic and pen-and-paper self-reports, resulting in a partial sample that opted for electronic assessments; (5) applied a measurement period of >24 hours, with ≥5 planned assessments, including at least 1 assessment per day; and (6) included original full-text articles in English, Dutch, or French.
Articles were excluded if they met one or both of the following criteria: articles that (1) were conducted in people in complete cancer remission and (2) concerned reviews, meta-analyses, notes, letters to editors, conference abstracts, or study protocols.
Search StrategyThe initial literature search was conducted on April 7, 2022, and updated on January 19, 2024, both without restrictions for its time coverage. We searched 3 databases: PubMed, Embase, and PsycINFO. We consulted a librarian of the Vrije Universiteit Brussel for the development of the search strategy. Keywords included terms related to the population (eg, cancer) and methodology (eg, ecological momentary assessment and daily diary). The search strategy was validated in PubMed and translated to other databases. The full search strategy is provided in .
Study Selectionprovides an overview of the selection procedure. Most duplicates were automatically detected and removed using EndNote (version 20; Clarivate) []. Screening followed a 2-step process. First, 2 researchers (JG and KdN) independently screened titles and abstracts and labeled them as relevant, irrelevant, or potentially relevant for inclusion. Additional duplicates not detected by EndNote were removed during this step. Second, both reviewers screened the full texts of relevant and potentially relevant studies for final inclusion. JG and KdN resolved discrepancies in both steps through discussion and consensus and consulted a third and fourth reviewer (LP and LVdB), if necessary. JG screened articles found during the updated search. We used Rayyan (Qatar Computing Research Institute) [] for reference management and manual removal of duplicates.
Figure 1. PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) flow diagram. ILM: intensive longitudinal method. Data Extraction and SynthesisJG extracted data into a precreated MS Excel (version 16; Microsoft Corp) spreadsheet. To ensure consistency and accuracy of the initial search, KdN independently completed the data extraction form for a random 10% sample of included articles. JG and KdN discussed and resolved discrepancies. JG extracted updated search results. The data extraction form was revised throughout the review process. It included study characteristics (ie, authors, year, country, and overarching study); sample characteristics (ie, sample size, site of primary tumor, stage of disease, mean age, proportion of female participants, and comparison group characteristics); study aims and design; system characteristics (ie, device, application, and operation system); daily questionnaire characteristics (ie, number of items, constructs measured daily, existing measurement instruments, or sources used); sampling schedule characteristics (ie, number of monitoring periods, duration of the monitoring periods, type of sampling scheme [ie, fixed or random signal-contingent, event-contingent, or interval-contingent] [], daily prompt frequency, and approximate time interval between prompts); supportive features for participants; response-related results (ie, participation rate, attrition rate, proportion of completed prompts, and monetary incentives); and main study findings. We listed the barriers and facilitators for the implementation of the used method in research and clinical practice per study.
We have presented the study and sample characteristics, system and sampling schedule characteristics, and response-related results in the Results section, grouping articles reporting on the same study. We conducted content analysis on the extracted barriers and facilitators, inductively categorizing the content in themes and subthemes.
Of the 1311 identified articles, we screened 253 (19.3%) full-text articles for eligibility. We included 52 articles, describing 41 unique studies ().
Population CharacteristicsAll the 41 studies were conducted in high-income countries, except for 1 (2%) study in Türkiye [] (all study and sample characteristics are listed in ) [-]. We included 21 (51%) studies [-] reporting on samples of people with mixed primary tumor sites (including breast and lung cancer), 16 (39%) studies [,-] on samples of people with breast cancer only, and 4 (10%) studies [-] on a sample of lung cancer only (). A total of 7 (17%) studies included patients’ partners [,,,,-,]. While 26 (63%) studies were conducted in people at differing stages of disease, of which 11 (42%) included up to stage III [,-,-,,,] and 15 (58%) included up to stage IV [,,,,,,-,-,,,,], 6 (15%) studies [,,,,-,] specifically focused on people with stage IV cancer. Sample sizes ranged from 4 [] to 344 [] participants, with a mean of 54.3 (SD 56.4). The mean ages were 51 (SD 4.7) years for patients with breast cancer, 65 (SD 2.8) years for patients with lung cancer, and 58 (SD 5.7) years for patients with mixed primary tumor sites. None of the studies used proxy responding.
Table 1. Study and sample characteristics of the included studies (n=41).CharacteristicsFrequency of studies, n (%)Reference, yearPrimary tumor sitesOf the 41 studies, 30 (73%) [-,,,,-,-,,-,-] used intensive methods in observational study designs, whereas 11 (27%) [,,,-,,,,,] used them in interventional studies. While 38% (20/52) of the articles [,,-,,,-,-,,,-] focused on the intensive method’s feasibility, usability, or validity, other articles investigated the prevalence or trajectory of measured variables or relationships between those variables [,,,-,-,,,,-,,,], the effectiveness of the intensive methods as an intervention [,,,,], or the effectiveness of other interventions [,,] or introduced a novel statistical approach [].
Data Collection MethodsDaily Measured ConstructsOf the 41 studies, 30 (73%) [-,-,,,-,-,-,,,-] used items adapted from previous studies or scales (study details are listed in ) [-]. Some of the most frequently recurring questionnaires were the Common Toxicity Criteria Adverse Events grading system [,-,,,], EORTC Core Quality of Life Questionnaire (EORTC-QLQ-C30) [,,,,], and the Positive and Negative Affect Schedule-Expanded scale [-,,,,]. Measured constructs covered physical, psychological, and social domains; behaviors and intentions; daily events; sleep quality; and general quality of life. The physical domain was the most assessed domain, with the most frequently measured constructs being pain [,,,,,,,-,-,,,,,,-,] and fatigue [,-,,,,,-,,,,]. Anxiety [,,,,,,,,-,,,] and depression [,,,,,,,,] were the most frequently measured constructs in the psychological domain, and social support [,,,,] and communication [,,,,,,] were the most frequently measured constructs in the social domain. Frequently measured behavioral constructs included medication use [,,-,,,] and physical activity [,,,,,].
Sampling Schedule CharacteristicsOf the 41 studies, 23 (56%) [-,-,,-,,,-,,,,,,] required patients to fill in the questionnaire once per day, while 6 (15%) studies [,,-,,] required 2 completions daily, and 7 (17%) studies [,-,-,,,] required 3-6 completions daily (). Moreover, 5 (12%) studies [,,,,] did not report the specific amount.
Table 2. Data collection methods used in the included studies (n=41).Data collection methodsFrequency of studies, n (%)Reference, yearSampling scheduleaMultiple options possible per study.
Out of the 41 studies, 22 (54%) studies [,-,-,-,,,-,] applied signal-contingent sampling (ie, prompting respondents to complete the questionnaire) and 6 (15%) studies [,,,,,] applied interval-contingent sampling (ie, instructing respondents to complete the questionnaire at certain intervals), while 15 (37%) studies [-,-,,,,,,,,] did not specify the sampling method. Furthermore, 6 (15%) studies used event-contingent sampling on top of the other sampling methods; of these, 4 (67%) studies [-,,] instructed patients to complete the assessment when experiencing adverse events, 1 (17%) study [] required the patients to assess when rescue medication was taken, and 1 (17%) study [] prompted patients when a physiologically measured stress threshold was reached. Out of 22 signal-contingent sampling studies, 13 (59%) [,-,,-,,,,,] prompted patients at fixed times, with times between prompts ranging from 3 to 24 hours. Moreover, 36% (8/22) of the studies [,,-,-,-] prompted patients at random times, of which 5 (62%) [,-,,,] randomly prompted within a fixed time block (eg, between 9 AM and midnight). Minimum time intervals between randomly timed prompts ranged from 30 minutes to 3 hours [-,-].
Of the 41 studies, 7 (17%) [-,,-,,,] had multiple data collection periods for each patient. While the most common data collection period lengths were 7 days [,,-,,] and 14 days [,,,,,-,,,], ranging from 1 to 336 days [,], 10 (24%) studies [,,,,-,,,,] mentioned differing study lengths for each patient (eg, based on patients’ next chemotherapy visit) [].
System CharacteristicsData collection devices and software varied substantively in the included studies (n=41), with 11 (27%) studies [,,,,,,,,-,,,,] using smartphones, 1 (2%) study using smartwatches [,], and 8 (20%) studies [,,,,,-,,,] using handheld computers for self-report assessments. Other studies used basic telephone and SMS text messaging functionality [,,,,,], internet functionality [-,,-,,,,,], and used a specifically developed device [,,]. A total of 19 (46%) studies [,,,,,,,,,-,-,,] used patients’ devices, whereas 12 (29%) studies [-,,,,,,,,-,,,] provided devices to patients.
Different types of software were used, including smartphone apps [,,,,,,,,,,,], browser-based surveys [-,,-,,,,,,], SMS text messaging [,,], interactive voice responding systems [,,,], and other specifically developed software applications [,,,,-,,,,,].
Questionnaire LengthSome studies (7/41, 17%) [,,-,,-,-] used conditional items that were presented when a certain response was given to previous items and different questionnaires depending on the timing of the prompt (eg, the use of morning prompts to assess sleep quality [,]). Most studies (20/41, 49%) [-,-,-,-,-,,,-,,-] had questionnaire lengths ranging between 1 and 20 items, with the longest being 84 items (including conditional items) [-]. Several studies (13/41, 32%) [,,-,,,,,,] did not provide complete information on the number of items.
Supportive FeaturesOf 41 studies, 17 (41%) [,,,,-,,,,,,,] provided supportive features; 9 (22%) studies [,-,,,] offered automated self-care advice to patients based on their responses directly after response submission, for instance, offering advice for managing reported symptoms, with severe symptoms triggering advice to contact a health care professional []. Also, 9 (22%) studies [,,,-,,,] automatically contacted health care professionals based on symptom severity (ie, clinician alerts). Some studies (2/41, 5%) [,,] differentiated between different severities to indicate varying levels of need for immediate intervention (eg, amber and red alerts). A total of 6 (15%) studies [,,-,,] combined automated self-care advice and clinician alerts. One study [] alerted clinicians based on responses given on domains other than physical symptoms, namely psychological variables (ie, depressive mood and anxiety) and distress caused by symptoms. Other supportive features included providing the opportunity to clinicians to view a summary or visualization of responses given by the patient [,,,,] and providing patients with informational modules [,], modules allowing communication with clinicians [,], response summaries [,], and relaxation reminders [].
Study-Reported FindingsFindings Concerning Methodological EvaluationsIntensive longitudinal methods that sampled once daily [,-,,,,] or multiple times per day [-,,,] were deemed feasible and acceptable for patients. These findings applied to various system characteristics, such as interactive voice response and SMS text messaging systems [,,] and smartphone apps [,,]. Compliance decreased over time in a 90-day study [], with higher compliance among unemployed women. Patients believed in the method’s ability to improve symptoms [], symptom management [,], and communication with clinicians []. Moreover, patients had positive views on the usability of the methods [,,,,,] and felt reassured by using them [,].
Health care professionals had a positive view of the methods [] and found them reassuring for patients, especially during out of hours [], and clinically useful [,,]. In addition, health care professionals thought that the methods could be helpful aids in timely interventions [] and for assessing [] and managing symptoms [,]. However, one study [] reported that quality of life data was not used for making treatment decisions, and other studies [,,] reported that clinicians rarely contacted the patients after receiving clinical alerts or monitored their responses. In one study [], health care professionals mentioned that reduced complexity of the system was needed to promote its utility.
Some studies (5/41, 12%) [,,,,] compared intensive longitudinal methods with other scales and found agreement between the methods, such as depression ratings and Patient Health Questionnaire-9 [,]. One study [] found a lack of agreement between the intensive methods and the Short Form Health Survey [], but this concordance improved with higher compliance rates.
Findings Concerning Prevalence and Covariability of ConstructsSeveral studies (16/41, 39%) examined the prevalence and covariability of constructs ranging across multiple topics. For instance, 7 studies [,,,-,] reported findings related to the social dynamics between patients and their partners. One study [] found greater reports of relationship interference when patients experienced more pain and lower arousal mood. Moreover, partners were more likely to provide support when patients experienced more tiredness and less active mood resulting from pain []. Another study on this topic [] found that partners’ reports of support provision were positively associated with feelings of relationship intimacy reported by patients.
Overall, studies investigated various topics such as physical activity, affect, and physical symptoms. For instance, studies [,] showed associations between sedentary behavior, affective valence, and fatigue at different time points, analog to other studies [,] that found within-person associations between physical activity and same-day affect, fatigue, pain, and others.
Findings Concerning the Intensive Methods as an InterventionOf the 41 studies, 7 (17%) [,,,,,,] investigated the impact of intensive longitudinal methods as an intervention tool to improve symptoms, for instance, by providing automated self-care advice to patients or alerting clinicians when a certain symptom threshold was reached []. Patients in the intervention groups reported lower distress [], lower fatigue, and higher levels of hand-foot syndrome [] than those in the control groups. Patient-reported benefits included improved communication with health care professionals and symptom management and reassurance that symptoms were being monitored at home []. After the intervention, patients reported increased quality of life [,], lower anxiety and drowsiness, lower pain [], and higher self-care efficacy [] than at the baseline. One study using clinician alerts [] found no improvements in symptom severity, explained by clinicians rarely contacting patients after alerts.
Response-Related ResultsOf the 41 studies, 21 (51%) [-,,,,,-,-,-,,-,-] reported participation rates ranging from 23.6% to 90.3% (mean 52.9, SD 3.4; ; ) [-]. Overall, 17 (41%) studies [,,,,,,-,,,,,,,,] reported
Comments (0)