Electronic Systems for Patients to Report and Manage Side Effects of Cancer Treatment: Systematic Review

1Increased efficacy of cancer treatments has led to a rising global population of people living with and beyond cancer. Effective multimodal cancer treatments can slow disease progression, ease the symptoms of the disease, and in some cases cure disease altogether. However, treatments can cause a vast array of side effects such as nausea, pain, fatigue, and diarrhea, which may negatively affect a patient’s quality of life (QoL) and may even become life-threatening, with severe cases such as neutropenic infections. Many cancer treatments are delivered in an ambulatory setting and methods of follow-up and support are highly variable dependent on disease, treatment type, and local practice and resources. Information is commonly provided by the health care team on expected and possible side effects, and patients are advised to seek help if symptoms become a cause for concern. However, patients may not always be able to fully absorb this information at the time it is provided or feel confident in making decisions on when additional hospital contact is necessary between routine clinical reviews. Furthermore, clinicians are mainly reliant on interpreting patient retrospective reports of treatment side effects to ensure safety of care and manage supportive medications. Side effects are not often documented in medical records in a consistent and comparable way.

2Over the past decade, there has been a dramatic increase in the number of electronic systems developed to support patients during and after cancer treatment by using patient-reported outcome measures (PROMs) to remotely assess symptoms. The routine use of PROMs in oncology care as a strategy to enhance symptom monitoring has demonstrated many benefits, such as improved communication between clinicians and patients, and better symptom awareness. Using electronic systems to collect and manage PROMs data has the potential to overcome some of the common challenges previously associated with collating data collected on paper. More recently developed systems can be accessed from any Web-enabled device, allowing patients to report symptoms from home using their own electronic devices such as computers, tablets, or mobile phones. This can be done in real time, rather than relying on retrospective reporting and potentially allows automated documentation of patient reports in the medical record.

3There is considerable variation in the features offered by symptom reporting systems. Some primarily focus on making symptom data routinely available to health professionals and provide alerts when severe symptoms have been reported. Others have been developed with a greater focus on patient self-management, delivering tailored and automated self-management advice when appropriate, and advising patients to contact their health care team when necessary. Some systems use a combination of both approaches and can also include additional features such as facilitating communication with medical teams or other patients.

4The availability or absence of certain features may affect how patients engage with systems. The terms “engagement” and “adherence” are often used interchangeably in this context. However, adherence suggests an optimal way to use a technology and this is not always easy to define. For the purposes of this review, we refer to engagement in a broad sense of levels of patient usage of the technology. Understanding the key components that can enhance patient engagement with electronic symptom reporting is potentially crucial for improving the development of future systems and encouraging their implementation into standard practice. There are many factors that are likely to have an impact, from individual differences, socioeconomic status and healthy literacy, to basic system usability. There is relatively little currently known about the underlying processes and particularly the role that the availability of systems features might play. However, there is evidence to suggest that individuals vary in the features that they value and use most. In addition, needs may change over time, as patients become more experienced with the system, but also with their disease and treatment.

5The presence or absence of system features is also likely to affect the level of patient benefit gained from using the system. For example, changes in behavior or disease outcome have been more often observed with interactive interventions in comparison with those that are purely educational. While the use of interactive online systems is associated with greater self-efficacy, better self-management, and more participation in health care, this may be associated only with specific features such as interactive communication and progress tracking features, and consultation and self-management support.

6Systematic reviews traditionally focus on high-quality evidence for a specific research question. However, increasingly, the value of taking a broader approach to inclusion is being recognized as important to answer complex research questions, particularly in the emerging field of online health interventions. With this in mind, the focus of this review was to take an inclusive approach to systematically review and describe the features and functions of existing systems. We also wanted to focus on understanding the level of evidence indicating whether key system features are associated with better patient system engagement and patient-centered outcomes.

7The aims of this systematic review are to (1) describe the features and functions of existing electronic symptom reporting systems developed for patients during cancer treatment, and (2) explore which features of these systems may be associated with patient engagement and outcomes. Specifically, we wanted to summarize (1) patient engagement and whether this is related to specific system features (eg, symptom monitoring, tailored self-management advice), and (2) patient-centered outcomes used to evaluate systems and whether better outcomes are associated with specific features.

8Details of the protocol were registered on the International Prospective Register of Systematic Reviews (PROSPERO) database. There were no major deviations from the protocol. However, study selection, data extraction, and data synthesis comprised two stages: (1) identifying and characterizing available systems, and (2) summarizing data on patient engagement and patient-centered outcomes. This staged approach was not initially planned but was necessary in order to meet the aims of the review.

9The review question was refined using Population, Intervention, Comparator, Outcomes, Study design (PICOS) criteria (Table 1), and eligibility criteria were developed based on this. For Stage 1, we wanted an overview of all systems available, so all relevant publications including published abstracts, protocols, and qualitative studies were included. However, discussion papers or systematic reviews were excluded. For Stage 2, in order to review evidence available on patient engagement and any patient-centered outcomes, we wanted to include feasibility studies with any evaluation data of patient use, rather than restricting criteria to randomized controlled trials (RCTs) only. Only full papers were included in this stage. Criteria were piloted by 2 researchers (LW and KA) on a subset of 10 randomly selected papers and subsequently refined and clarified before the next stage.

10Studies were identified from systematic searches of Medline, Embase, PsycInfo, Web of Science, Cochrane Central Register of Controlled Trials, and the Health Technology Assessment databases in March 2016. Due to the nature of the review, results were limited to those published after 2000. No restrictions were imposed on language of publication. Searches were updated on September 12, 2017. Reference lists of relevant publications were screened to identify papers not picked up by the electronic searches. In addition, citations of selected key papers were searched.

11A detailed example of the search strategy used for Medline is outlined in Textbox 1. This search strategy was adapted for each of the databases.

12For initial screening, a decision for inclusion was made based on title and where available, abstract. This was carried out by one researcher (LW) only, and for this reason, a cautious approach erring on the side of over-inclusion was used. Following this, 2 researchers (LW and KA) independently assessed all remaining papers for relevance. Disagreements were resolved by consensus after referring to the protocol. All discussions and decision making were documented. Where there was insufficient information to make a decision, authors were contacted for further information. If no response was received within 2 weeks, a final decision was made based on available information.

13Neoplasms/

14oncolog*.mp.

15cancer patient*.mp.

161 or 2 or 3

17Medical Informatics/

18Telemedicine/

19Mobile Applications/

20Smartphone/

21Self Report/

22Self Care/

23Self-Assessment/

24(electronic adj2 (Patient report* or Patient-report* or Self report* or Self-report* or Self manage* or Self-manage* or Self monitor* or Self-monitor* or Symptom report* or Symptom-report* or Symptom manage* or Symptom-manage*)).mp.

25(online adj2 (Patient report* or Patient-report* or Self report* or Self-report* or Self manage* or Self-manage* or Self monitor* or Self-monitor* or Symptom report* or Symptom-report* or Symptom manage* or Symptom-manage*)).mp.

26(web* adj2 (Patient report* or Patient-report* or Self report* or Self-report* or Self manage* or Self-manage* or Self monitor* or Self-monitor* or Symptom report* or Symptom-report* or Symptom manage* or Symptom-manage*)).mp.

27(remote* adj2 (Patient report* or Patient-report* or Self report* or Self-report* or Self manage* or Self-manage* or Self monitor* or Self-monitor* or Symptom report* or Symptom-report* or Symptom manage* or Symptom-manage*)).mp.

285 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15

294 and 16

30Limit 17 to (humans and yr=“2000 -Current”)

31For Stage 1, basic data were extracted on authors, title, year of publication, and country of origin, in addition to the name (if any given) and type of system being described (eg, Web-based or mobile app). If the system did not already have a descriptive name, an arbitrary name was assigned (eg, System A). A preliminary list of common features was created based on existing knowledge and further developed throughout data extraction until a comprehensive list of common or important features was identified. Data were extracted from each publication on the presence of each feature. This was coded as “Yes” only if it was explicitly described in the publication, otherwise it was coded as a “#” For abstracts, if it was unclear whether or not a feature was present by information available in an abstract, this was classed as “Unable to determine.” Where information was lacking, authors were not contacted for information. However, searches were undertaken for other publications related to the same system.

32For Stage 2, data were extracted from studies with some form of system evaluation (eg, patient use of system or evaluation of efficacy). This included data on the number of patient participants, baseline demographics, disease and treatment type, duration of the evaluation, methods used to assess engagement, and actual usage or adherence. Where available, data were also extracted on any patient-centered outcomes used and results of evaluation.

33Data were extracted using the online Systematic Review Data Repository. The form was piloted on 10 randomly selected papers and further refined. For Stage 1, three additional researchers (KA, BC, MA) each double-coded a number of allocated publications, totaling 36% (27/77) of the overall included publications. A high level of agreement (86%) was found. Discrepancies were resolved by referring back to the protocol and additional publications where available. For Stage 2, the same 3 researchers again each double-coded a proportion of the included publications totaling 46% (13/29) and 100% agreement was found.

34Quality was assessed using the Downs and Black checklist for nonrandomized studies and was undertaken alongside data extraction. It was deemed appropriate to assess only studies that included some feasibility/evaluation data, that is, publications included in Stage 2. Studies were given a score along a possible range of 0-26.

35A narrative synthesis was undertaken using the guidelines outlined by the Economic and Social Research Council. Microsoft Excel was used to manage data. For Stage 1, information from multiple publications relating to the same systems was pooled to form a description of features. Where information was conflicting due to earlier and later iterations, the most recent description was used. For Stage 2, information was collected on how patient engagement was assessed for any feasibility study or trial that included these data. For trial studies, information was collected on primary and secondary study outcomes and any results recorded. We then summarized these data to explore any relationships with system features identified in Stage 1.

36An overview of search and selection procedures is outlined in Figure 1. A total of 6727 publications were identified after removal of duplicate publications, including two publications identified from secondary searches (ie, citation and reference lists). All publications were in English. We assessed 279 publications for eligibility, and a total of 202 papers were excluded at this point based on predefined eligibility criteria (intervention, eg, not home-based or Web-based, n=132; population, eg, patients not on active treatment, n=41; discussion paper or systematic review, n=19; or abstract unavailable, n=10). We included 77 publications in Stage 1 of the review (ie, systems descriptions). A large proportion (23/77, 30%) of these publications were abstracts. The reasons for exclusions are outlined in Figure 1. Those 8 publications categorized under “Other” included 2 summary papers giving an overview of development and evidence for a system, a description of standard usability testing, a cost-effectiveness analysis, a content analysis of email communication within a system, a discussion of design approaches and methodology, an evaluation focusing on blood monitoring, and one publication where we were not able to access the full paper and did not receive a response from the authors when this was requested. We identified 29 publications for inclusion in Stage 2 of the review (ie, patient engagement and evaluation of systems). These were 21 feasibility studies and 8 controlled trials (7 randomized and 1 nonrandomized).

37The 77 publications referred to 41 individual systems. Most originated from the United States (19/41, 46%) or the United Kingdom (6/41, 15%), and all publications were available in English. Systems were commonly Web-based (24/41, 56%), 27% (11/41) were mobile apps, 2 were both mobile and Web-based (2/41, 5%), and 22% (9/41) were Web-enabled mobile devices purposely designed for symptom reporting and were provided to patients for the duration of the study.

38Seven common system features were identified. Figure 2 outlines each of the features and its prevalence in the 41 identified systems. Features could be categorized broadly as supporting patients to monitor and manage their own symptoms, and to communicate with health professionals and one another, or supporting clinicians to monitor and manage patient symptoms.

39Table 2 provides an overview of each identified system and its associated publications, in addition to the presence or absence of each of the features identified in Figure 2.

40Along a possible range of 0-26, the overall median quality assessment score of studies using the Downs and Black checklist was 17.0 (mean 16.2, SD 5.3, range 2-24). For trials described in the section on patient-centered outcomes, the median score was higher at 20.0 (mean 20.4, SD 2.6, range 17-24).

41Table 3 summarizes data on patient engagement from the 29 included studies (ie, 21 feasibility studies, 7 RCTs, and one non-RCT). All 21 feasibility studies (100%) reported some data on patient engagement, although there was variation in how engagement was defined and measured. Three of the eight trials (38%) did not report any data on patient engagement.

42Of the 29 studies, the most common method of assessing engagement was the number of symptom report completions or number of times the system was accessed (12/29, 41%). This was given as an overall figure for the whole sample, as an average per patient, or with a breakdown of the variance. Nine studies (9/29, 31%) assessed adherence by number of actual completions/accesses in comparison to the number of expected completions/accesses. This was reported as median or mean adherence of the overall sample for the duration of the study period, or with a breakdown of adherence at different time points. Only 2 studies studies (2/29, 7%) categorized patients as users or nonusers dependent on predefined criteria. Four studies (4/29, 14%) combined results of patients reporting from home and in clinic. Not all studies reported on actual usage, and some used evaluation questionnaires with or without semistructured interviews to assess acceptability to patients.

43Due to the variation in the methods of reporting, it was not possible to determine if there was any overall association between engagement and specific system features.

44All the trials used some measure of patient-centered outcome to evaluate system efficacy, most commonly validated QoL and symptom and psychosocial outcome measures. Table 4 outlines each trial, the intervention and comparator groups, outcomes reported, and a summary of the results.

45CASSY and STAR interventions both demonstrated improvements in overall QoL. However, in addition to the online component, CASSY included access to a collaborative care coordinator with experience in cognitive behavioral therapy and psycho-oncology, which is likely to have contributed to the efficacy. In the STAR study, patients were allocated to computer-experienced and inexperienced groups prior to randomization and only the computer-experienced group had access to the system from home. Results are pooled, making it difficult to assess efficacy for our purposes. No significant impact on QoL was found for WebChoice.

46An overall reduction of symptom distress was found in the studies assessing Electronic Self-Report Assessment-Cancer (ESRA-C) and WebChoice. However, in addition to the online intervention, ESRA-C also included a communication coaching component to improve symptom disclosure to physicians. System B was found to have significant positive impact on the general physical complaints subscale compared to the control group.

47Advanced Symptom Management System (ASyMs) and Comprehensive Electronic Cancer Support System for the Treatment of Cancer Related Symptoms (CaSSY) both demonstrated positive impact on levels of fatigue while System K demonstrated a lesser decline in functional activity in contrast to the control group, but this was not significant. Both ASyMs and System K were evaluated using the same measure as used to assess symptoms in the intervention, which may have affected results.

48WebChoice and System B both demonstrated a positive impact on self-efficacy. However, for System B, this was assessed only as a subscale of a main measure. System K reported an improvement in patient empowerment; however, this was assessed using a single item regarding using the Internet for information seeking, which is unlikely to be a reliable measure.

49CASSY and WebChoice demonstrated significant reductions in depression in intervention compared to control groups. System B demonstrated no difference on the depression subscale of a QoL measure but a significant impact on state anxiety and fear related to specific head and neck problems. WebChoice demonstrated no impact on social support. QoC Health Inc was primarily assessed on number of hospital contacts but also included patient scores of convenience and satisfaction using a simple 5-point Likert scale and found an impact for convenience, but not for patient satisfaction.

50Due to the considerable variation in outcomes used and study design, it was not possible to assess any relationships between outcomes and system features.

51The main aim of this review was to systematically describe and assess the features and functions of current systems available for patients to report and manage side effects of cancer treatment. We also wanted to focus on understanding the level of evidence indicating whether key system features are associated with better patient system engagement and patient outcomes.

52In Stage 1 of the review, we identified a total of 41 individual systems. There was significant variation between systems, though published descriptions of systems were often limited. We developed a taxonomy of features that were then classified into those supporting clinicians to deliver patient care in an innovative way and those aimed to support patients to better self-manage their condition and identify when medical input may be needed. This was successfully applied to describe the presence or absence of common system features.

53The review of features highlighted some interesting findings. It was surprising to note that while over half (58%) of systems had the facility for health care providers to monitor patient data over time, fewer than half (46%) included the facility for patients to monitor and review their own data. Given the available evidence suggesting that self-monitoring is generally beneficial to support patients’ self-management, this feature could be very important to improve efficacy of systems and in most cases, may be relatively easy to implement. Similarly, less than half of the systems (41%) included a feature for delivering advice to support patients to self-manage symptoms and less than a third provided patients with access to general educational information. The two least common features were facilities to support communication between patients and health care providers (15%) and communication between patients themselves, respectively (10%). Previous research has indicated that these features are highly valued and utilized by patients. It is likely that these features are less common due to complexities in their implementation and maintenance. For example, it may be difficult to engage busy clinicians to respond to patient communication in this way, and there are ethical considerations around the need to moderate patient forums that are endorsed by a health care facility.

54In Stage 2 of the review, we found little agreement on how patient engagement with systems was defined, measured, or reported, which meant it was not possible to compare levels of engagement across studies or make any conclusions on relationships with system features. Our review also indicated heterogeneity in terms of outcomes used to evaluate systems. Even of those that focused on symptoms or global QoL, the variation in methods and measures used made meaningful comparison impossible.

55Due to the heterogeneous nature of reporting engagement and outcomes, we were unable to explore any relationships with system features. Our findings are similar to other reviews undertaken in this area, which have also found that poor assessment and reporting of patient engagement with systems makes comparison between studies difficult. Brower et al made quantifiable and comparable reports of engagement as part of their inclusion criteria for their review, and results indicated that facility for communication with other patients may be a very influential factor in patient engagement and needs careful consideration during system design. However, other oncology specific reviews have found that methods of assessing and reporting patient engagement were too heterogeneous to make meaningful conclusions. We identified only 8 trials (7 randomized and 1 nonrandomized) that evaluated systems, none of which reported any analysis on relationships between engagement and outcomes, and 3 of which did not report any data on patient engagement at all. This does not seem to be unique to oncology. Donkin et al set out to review the impact of patient engagement with e-therapies across a range of disease groups and similarly found that this is not a link that is routinely explored.

56Robust evidence supporting the value of systems for patient-centered outcomes was limited, with a large proportion of feasibility studies identified and even fewer RCTs. While all trials used some measure of patient-centered outcome to evaluate systems, a wide range of assessment tools were used, again making comparison difficult. In addition, 2 studies used the same measure for symptom assessment as part of the intervention, as for the outcome measure. Only 3 trials reported any measure of self-efficacy or patient empowerment, one of which used a study-specific nonvalidated measure, and another that was assessed using a subscale of a global QoL measure. There is an array of evidence to suggest that online interventions can have a positive impact on self-efficacy and patient activation levels. Growing evidence suggests that self-efficacy and patient activation play a significant role in symptom management and quality of life throughout cancer treatment and are associated with an array of improved health behaviors and health outcomes and lower use of hospital resources. The reviewed systems generally demonstrate positive outcomes for patients as has been found in other reviews.

57To our knowledge, this is the first systematic review in this field to identify and characterize all available systems for patients to report and manage side effects of cancer treatment, in addition to evidence on patient engagement and patient-centered outcomes.

58In order to meet the aims of the review, we included many publications that provided limited information about the system evaluated and some of which were of poor quality. However, we felt that this was necessary in order to meet the aims of the review and evaluate all evidence. Due to limitations on available resources, the initial stage of study selection (ie, assessment of titles and abstracts) was undertaken by a single reviewer. This is a limitation of our methodology and may have resulted in some bias of inclusion. To address this, a cautious approach erring on the side of over-inclusion was adopted, in order for records to be fully assessed by 2 researchers in the next stage of the review.

59Due to the heterogeneous nature of study designs and methods of reporting engagement and outcomes, we were unable to explore any relationships with system features. This is a field of research that is still in its infancy, and the large number of feasibility studies and abstracts identified are likely to be indicative of this. The search was last updated September 2017. Due to the fast-moving nature of this field of research, it is likely that additional publications will be available by the time of publication. This is a common limitation of systematic reviews that is particularly pertinent with reviews of technology. We did identify a number of protocols for planned quality trials that may contribute to a more in-depth understanding of associations between system features, adherence, and outcomes in the future. In addition, we have not explored how issues with implementing systems into clinical practice may have affected the efficacy of systems. A discussion of these issues is outside the scope of this review but has been well-documented elsewhere.

60There is a real need for evidence-based guidance on developing, evaluating, and reporting systems. Based on this systematic review, we propose a taxonomy for characterizing system features to guide future development, improvement, and implementation of such systems. More work is needed to develop guidance for standardized reporting of patient engagement both in feasibility studies, and in evaluation trials. This is a complex and multifaceted issue, and it is important that barriers and facilitators to engagement are shared to help the evolution of more sustainable and valuable systems. Similarly, the development of guidance for the evaluation of systems is necessary. Variation in approaches to design and implementation will rightly affect outcomes chosen to evaluate efficacy. However, there is enough commonality between systems to call for a set of recommended core outcomes to be developed. More work is needed to develop this, and this is something we will work towards in the future. However, based on this review we recommend that all system evaluations include (1) a description of the system using our taxonomy of system features, (2) measures of feasibility and engagement, (3) patient-centered outcomes focusing on QoL and symptom improvement, in addition to those focusing on self-efficacy and patient activation, and (4) a measure of health economics. This will facilitate synthesis of evidence in order to improve the design of systems and make them practically useful for both patients and clinicians.