This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
Monitoring of patients may decrease treatment costs and improve quality of care. Pain is the most common health problem that people seek help for in hospitals. Therefore, monitoring patients with pain may have significant impact in improving treatment. Several studies have studied factors affecting pain; however, no previous study has reviewed the contextual information that a monitoring system may capture to characterize a patient’s situation.
The objective of this study was to conduct a systematic review to (1) determine what types of technologies have been used to monitor adults with pain, and (2) construct a model of the context information that may be used to implement apps and devices aimed at monitoring adults with pain.
A literature search (2005-2015) was conducted in electronic databases pertaining to medical and computer science literature (PubMed, Science Direct, ACM Digital Library, and IEEE Xplore) using a defined search string. Article selection was done through a process of removing duplicates, analyzing title and abstract, and then reviewing the full text of the article.
In the final analysis, 87 articles were included and 53 of them (61%) used technologies to collect contextual information. A total of 49 types of context information were found and a five-dimension (activity, identity, wellness, environment, physiological) model of context information to monitor adults with pain was proposed, expanding on a previous model. Most technological interfaces for pain monitoring were wearable, possibly because they can be used in more realistic contexts. Few studies focused on older adults, creating a relevant avenue of research on how to create devices for users that may have impaired cognitive skills or low digital literacy.
The design of monitoring devices and interfaces for adults with pain must deal with the challenge of selecting relevant contextual information to understand the user’s situation, and not overburdening or inconveniencing users with information requests. A model of contextual information may be used by researchers to choose possible contextual information that may be monitored during studies on adults with pain.
Monitoring involves repeated study of a question that requires collecting data [
Pain is an “unpleasant sensory and emotional experience associated with actual or potential tissue damage” [
The aim of this work is to determine what types of technologies have been used to monitor adults with pain and propose a model of context information relevant to patients with pain. For this, a systematic literature review (SLR) was conducted, which is a means to identify, evaluate, and interpret all relevant research available for a research question or topic [
A SLR was conducted following Kitchenham and Charter’s guidelines for performing SLRs [
A systematic search of published literature was conducted to analyze recent research about context information related to pain and technologies used to monitor adults with pain. The search was conducted electronically during October 2015 in the following digital libraries: ACM Digital Library, IEEE Xplore Digital Library, ScienceDirect, and PubMed. These libraries were chosen to cover medical and technological aspects. This review was limited to articles published between 2005 and 2015, and duplicate citations across databases were identified and excluded using the Papers software.
The keywords were identified by consulting with medical specialists on appropriate words, manually selecting publications related to the subject, and analyzing frequently used words. The set of keywords was (context-sensitive, context-aware, physiological, environment*) AND (monitor*, sens*, measure*) AND (pain). The asterisk operator (*) indicates that there may be more letters after the root word. With these keywords, the search string was built using Boolean AND and OR operators. The search string was input into each database and the keywords were restricted to be found in the abstract and/or document title and published on or after January 1, 2005. In total, 1758 articles were retrieved, with the following distribution according to the consulted database: ACM (n=113), IEEE (n=55), ScienceDirect (n=548), and PubMed (n=1042).
It is relevant to note that other keywords were tested in the search engines, most notably the word “context.” However, a large number of articles use “context” as the context of the study itself, so the words “context sensitive” and “context aware” were used instead.
A study was included in this review if it met the following inclusion criteria: (1) it presented a study of context information and pain; (2) the study was carried out on adults; (3) the article was peer reviewed and it was obtained from a journal, conference, or workshop; (4) it was published between January 1, 2005 and October 1, 2015; and (5) the study was published in English. Articles were excluded if they presented studies pertaining to animals, plants, robots, or children, or if the study was a literature review, mapping study, SLR, only presented as an abstract, or if it was not possible for any of the authors to download the full text of the article (no access through university subscriptions).
The included articles were selected through two steps. In the first step, title, publication venue, year of publication, and abstract for each article were collected in an Excel spreadsheet. Two reviewers assessed each publication (IR reviewed all articles; MM and PR each reviewed half) and applied the inclusion/exclusion criteria. Publications with two votes to include or exclude were automatically included or excluded. Publications with differing votes were sent to a third reviewer (VH), who analyzed it and determined whether the publication should be included or not.
In the second step, the primary and secondary reviewers (IR, CG) read the full text of a random sample of 10 publications. Each reviewer independently assessed whether the article should be accepted or rejected. Then, Cohen kappa was calculated, with a result of 1, which suggested that the inclusion/exclusion criteria were clear enough to be applied consistently [
Finally, an accelerated liberal approach was applied [
One reviewer (IR) extracted information from each publication using a predesigned Excel spreadsheet with 29 columns (eg, authors; study date; study purpose; country; contextual information; activity being monitored; main user; number of participants; study methodology, such as methods used, number and type of participants, activity; type of monitoring technology used). Quality of studies was not considered in this analysis.
In total, 1758 references were identified from the databases. After removal of duplicates, 1029 publications remained. These were analyzed for abstract and title, and 911 publications were excluded because they did not meet the inclusion criteria. A total of 113 publications were evaluated for full text and 87 publications satisfied the aforementioned eligibility criteria and were included in the final review. Out of these, 53 used technology to monitor pain. The selection flow diagram for this study is presented in
Selection flow diagram.
Most of the reviewed articles were published in journals (81/87, 93%) and only 6 of 87 (7%) in conference proceedings. The distribution of studies over the years is presented in
The selected studies collected information for several objectives. Several focused on pain relating to postures and movement in a work environment [
Regarding pain measurement, 53 of 87 investigations used pain scales. From these articles, the most frequently used scale was the Numeric Pain Rating Scale (27/53, 51%), followed by the Visual Analog Scale (24/53, 45%) and Verbal Numerical Scale (2/53, 4%). Validated questionnaires were also used: McGill Pain Questionnaire (7 articles), Brief Pain Inventory (5 articles), and Multidimensional Pain Inventory (1 article); 5 articles used their own questionnaires. Six articles used two or more methods to assess pain. Only 14 articles had a control group.
Number of publications per year.
The methodology used in each publication was analyzed. A quantitative methodology was applied in 79 of 87 studies (91%), none used only qualitative methodology, and 6 of 87 (7%) applied mixed methods. The following techniques for data collection were used: questionnaire, sensors (eg, heart rate monitor), diaries, interviews, and analysis of medical records.
The participants were asked to participate in experiments to collect data while they were monitored. The activities that participants underwent were classified into the following five categories:
Daily activity: monitoring the activities of a person in his/her daily life [
Specified activity: monitoring the activities of a person during an activity specified by the researchers, which were classified further into the following categories:
routine task: participants must perform a specified task (eg, reading, writing) [
physical activity: participants must engage in activities that require physical exertion (eg, lifting, cycling, walking) [
therapy: participants were evaluated while doing some type of therapy (eg, leg curls, music, behavioral therapy) [
Pain test: tests in which the participant feels pain (eg, hand dip tests in cold water and/or heat, or electrical stimuli) [
Display images: participants are shown images (eg, erotic, pleasant, gory images) [
Other: other activities [
The most frequent activities were pain test (14/87, 16%), therapy (13/87, 15%), physical activity (13/87, 15%), and daily activity (11/87, 13%). A summary of the methodology used in the included studies is presented in
Summary of study methodologies.
Classification of the existing knowledge in a domain can provide a better understanding of the relationships between the objects, identify gaps, and ease the sharing of knowledge [
Lienhard and Legner’s recent context model [
The 49 types of context information identified in this study are presented in
Activity: information collected from activities that require physical or mental effort by the user.
Identity: the user’s identifying characteristics (eg, roles, behaviors, and personality).
Wellness: information about a patient’s state of well-being (eg, quality of life, disability, comorbidity, and among others).
Environment: the patient’s surroundings (eg, noise, food, and music).
Physiological: data collected from the patient’s body (eg, heart rate, blood pressure, and skin conductance).
New context information model for pain monitoring.
Context category and |
Instruments | Articles, n | Correlation with |
No correlation |
|||||||||
Questions/ |
Medical |
Technical device | |||||||||||
Physical activity | Y | Y | 7 | 3 | 1 | ||||||||
Mental tasks | Y | Y | 1 | ||||||||||
Positions | Y | 3 | 2 | ||||||||||
Movements | Y | 4 | 1 | ||||||||||
Walk (gait) | Y | 2 | 1 | ||||||||||
Behavior | Y | 2 | 1 | ||||||||||
Personality | Y | 1 | 1 | ||||||||||
Role | Y | Y | 2 | 2 | |||||||||
Ethnicity | Y | 2 | 2 | ||||||||||
Coping | Y | 1 | 1 | ||||||||||
Quality of life | Y | 5 | 3 | 1 | |||||||||
Emotional state | Y | 15 | 7 | 1 | |||||||||
Comorbidity | Y | 3 | 1 | ||||||||||
Anxiety | Y | 10 | 3 | 2 | |||||||||
Depression | Y | 9 | 3 | ||||||||||
Stress | Y | 12 | 6 | ||||||||||
Fatigue | Y | 8 | 1 | 1 | |||||||||
Fear | Y | 8 | 1 | 1 | |||||||||
Muscle injury | Y | Y | 1 | ||||||||||
Sleep | Y | Y | 7 | 3 | |||||||||
Disability | Y | Y | Y | 14 | 1 | ||||||||
Catastrophism | Y | 6 | 2 | ||||||||||
Vibration | Y | 1 | 1 | ||||||||||
Music | Y | 1 | 1 | ||||||||||
Workload | Y | Y | 5 | 4 | |||||||||
Food | Y | Y | 1 | 1 | |||||||||
Setting | Y | 1 | |||||||||||
Voice | Y | 1 | |||||||||||
Social support | Y | 4 | 1 | ||||||||||
Noise | Y | 1 | |||||||||||
Location | Y | 3 | |||||||||||
Blood | Y | Y | 5 | 1 | 1 | ||||||||
Saliva | Y | Y | 3 | ||||||||||
Heart rate | Y | Y | 15 | 3 | 2 | ||||||||
Blood pressure | Y | Y | 8 | 1 | |||||||||
Skin conductance | Y | Y | 8 | 2 | |||||||||
Facial expressions | Y | 2 | 1 | ||||||||||
Flicker | Y | 2 | 1 | ||||||||||
Pupil | Y | Y | 1 | 1 | |||||||||
Muscular strength | Y | Y | 2 | ||||||||||
Muscle activity | Y | Y | 11 | 2 | |||||||||
Temperature | Y | Y | 2 | ||||||||||
Column compression | Y | Y | 2 | 1 | |||||||||
Cerebral activity | Y | Y | 4 | 2 | |||||||||
Asymmetry | Y | Y | 2 | 2 | |||||||||
Musculoskeletal symptoms | Y | Y | 1 | ||||||||||
Breathing | Y | Y | 5 | ||||||||||
Sensitivity | Y | 3 | 1 | ||||||||||
Pain tolerance | Y | 4 |
The context information that was collected more frequently was heart rate, disability, emotions, and depression. Most publications used physiological information (54/87, 62%) and wellness data (52/87, 60%) because these categories included heart rate, emotions, disability, and depression, which are frequently collected parameters. Next was environment information with 22% (19/87). Activity data were used in 18 of 87 publications (21%), corresponding to physical activities, mental tasks, and walking, among others. Finally was identity data with 9% (8/87).
The instruments used to collect each type of context information are also listed in
The proposed model may be used by researchers as a base taxonomy of possible information that may be monitored in adult patients with pain; however, it is not expected that any single device should monitor all this information. Rather, researchers may select information that is relevant to their specific study, choosing to focus on information that has been found to correlate with pain or otherwise choosing to fill gaps in the literature (eg, by studying whether some of the information that is frequently monitored has a relation to pain).
The technologies used to monitor adults with pain were studied to learn about current trends and challenges regarding pain monitoring.
A technological device includes a user interface (ie, the representation of a system with which a user can interact) [
Graphical user interfaces represent information through an image-based representation in a display [
A tangible user interface is a user interface in which a person uses a physical object to interact with digital information [
A wearable user interface is a device that is worn on the body (eg, embedded in clothing or accessories) [
For each of the reviewed articles that presented monitoring technologies, the type of user interface (graphical user, tangible user, or wearable user as previously defined), the target user (either the patient him/herself or the health care worker), what type of information was monitored (according to the categories defined by our model), and the type (whether available commercially or as a research prototype) are listed in
The most common type of interface used to monitor adults with pain was a wearable user interface (37/53, 70%), followed by tangible user interface (22/53, 42%) and graphical user interface (9/53, 17%). Regarding wearable user interfaces, the body part where most devices were placed was the trunk (17/37, 46%) and arms (including hands; 15/37, 41%). Naturally, this was related to the type of condition that was being studied (eg, back pain was more frequently monitored through devices placed on the trunk).
The target users of these technologies were most often the patients themselves (47/53, 89%) and/or health care professionals (22/53, 42%). The devices were worn on the patients’ bodies (eg, electrodes, sensors). Most studies used these technologies not to monitor users, but rather to conduct measurements in controlled or supervised environments. Tangible user interfaces in these articles were mostly oriented toward health care professionals and not patients because they used medical equipment such as scanners or blood tests, which require special training to operate.
Using the previous classification of activities used for evaluation, the three activities that were most frequently done to evaluate technological devices to monitor adults with pain were daily activity and physical activity with 19% each (10/53), followed by pain test with 17% (9/53). Using technology allows researchers to monitor patients during their daily activities, which provides more realism and a richer context for evaluation.
Five challenges in terms of monitoring adults with pain were found:
When monitoring is in real contexts, the user of the device must be the patient. This may generate usability challenges when users have low digital skills, as well as other technical challenges such as battery life.
Many contextual factors may influence pain and, as previously stated, current sensors allow measuring a large amount of information, but it is not yet clear which types of information to monitor for a particular evaluation.
Monitoring technology usually sends reports to health care professionals, whereas almost no feedback is given to the patients to help them understand their pain patterns, triggers, and how to adjust their activities accordingly, for example. A possible explanation for this is that medical-grade health monitors that can provide feedback to patients are rigorously tested and highly regulated [
Most studies do not collect environmental information from the patient, although there are already sensors on the market to capture this type of information (eg, noise, humidity, temperature).
Increasingly, researchers have been taking advantage of available sensor technology and implementing tangible and wearable devices to monitor adults with pain in a mobile way. However, most studies did not collect data in real contexts, rather focusing on laboratory or controlled experiments.
The results were analyzed to see whether trends could be found (ie, whether changes could be detected over the time period of the review), especially regarding study methods, evaluation activities, technologies, and collected context information. No significant differences were found in the contextual data that were collected over the years nor in the types of technology used or evaluation methods.
Technology used to collect context information.
Body part and usera | Interfaceb | Context datac | Type | |||||||||||||||||||||||
WUI | TUI | GUI | Wellness | Environment | Physiological |
Activity |
||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Y | Commercial/research | ||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Y | Commercial/research | ||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Y | — | ||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
Y | ||||||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | — | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Commercial | |||||||||||||||||||||||
[ |
Y | Y | Y | Commercial | ||||||||||||||||||||||
[ |
Y | Y | Y | Y | Commercial/research |
a HCW: health care worker.
b GUI: graphical user interface; TUI: tangible user interface; WUI: wearable user interface.
c No article monitoring identity data.
This SLR found 49 different types of context information used to monitor adults with pain and categorized them into a new five-dimensional model of context information that includes activity, identity, wellness, environmental, and physiological data. Several types of context information have been studied to see whether they correlate to pain; although publication bias tends to skew data toward positive results, we found that some contextual information has not been found to correlate to pain (eg, sleep), whereas other (eg, emotional state) has an increasing amount of evidence of its correlation to pain. This review did not find trends in the contextual information that has been presented in previous research (ie, it has not changed substantially in the 10 years of the review). Therefore, although there is potential for new sensors to allow monitoring new contextual information, there is a degree of independence between the contextual information that is of interest in the monitoring of patients with pain and sensor availability.
A total of 53 studies presented technological devices to collect context information, using wearable, tangible, and graphical user interfaces. Even though several approaches aim to capture context automatically (eg, through sensors), the proposed model makes evident that contextual information also requires manual input because patient-supplied information is relevant (eg, in the identity category of the model). Although sophisticated technologies exist for inferring emotions through facial expressions captured by video [
Recent advances in the miniaturization of biosensors, wearable technology, and microelectronics have enabled continuous ambulatory monitoring of physiological signals [
The selected articles were found to have diverse types of evaluations, spanning hours, days, or weeks, and with diverse sample sizes. The most frequent activities that participants underwent were a pain test, therapy, physical activity, and daily activity. Generally, studies with a longer duration use specific activities, whereas daily activities are used when the sample size is small, possibly because daily activities are more complex to evaluate when the period of time or sample size is larger. Usually, evaluation periods are short (less than one day), and most studies involve young people or adults, but not seniors. Only seven studies (8%) had patients older than 50 years; however, it has been found that prevalence of chronic pain does vary with age, increasing as patients age [
Researchers have mostly studied patient-supplied context information, and infrequently contextual information from environmental factors or patient activities. For example, studies have suggested that environmental problems may greatly affect health [
Several systematic reviews related to pain have been undertaken, but they have focused on pain management (eg, therapy effectiveness [
This study aims to provide information about contextual data that may be monitored through technological devices. Nevertheless, this area is fraught with interesting challenges. One is preserving the privacy of patients [
This study only used four specific databases and only in English; therefore, more regional contributions may have been missed, which may explain our low rate of studies in Africa, Asia, and Latin America. This search was restricted to 10 years, partly to uncover recent technological proposals; however, important contextual information may have been discussed in older research articles. Also, only the word “pain” was included in the search string, omitting related words (eg, “misery” or “spasm”), which may have uncovered additional literature on this topic.
A SLR was conducted with the goal of studying technologies to monitor adults with pain and relevant contextual information. Eighty-seven articles were reviewed in depth and 49 types of context information were found and organized into a five-dimension model of contextual information. Most contextual information was related to patient-supplied data and few were collected from the environment or patient’s activities. Regarding technology, wearable user interfaces are used most often to collect data and monitor patients. Nevertheless, not all information may be monitored through sensors automatically—some data must be user-supplied because some information from the patient is subjective (eg, pain intensity, fear, and emotional state).
systematic literature review
IR was supported by CONICYT-PCHA/Doctorado Nacional/2014-63140077, CONICIT and MICIT Costa Rica PhD scholarship grant and Universidad de Costa Rica. VH was funded by CONICYT/FONDECYT 1150365 (Chile). CF and MM were supported by PhD Scholarship Program of CONICYT Chile (CONICYT-PCHA / Doctorado Nacional / 2013-21130661) and (CONICYT-PCHA / Doctorado Nacional / 2012-21120544), respectively. No other funding sources were provided.
IR and VH together developed the concept of this systematic literature review. POR provided guidance about the systematic literature review protocol, and MC provided guidance about medical aspects of this work. MM conducted the literature search, and IR, VH, POR, and Maíra Marques reviewed abstracts. IR, CG, and CF worked in the accelerated liberal approach. IR and VH conducted the analyses and contributed to the interpretation of the findings. IR wrote the first draft of the manuscript and Valeria Herskovic contributed to changes in the manuscript. All authors reviewed the final document.
None declared.