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 JMIR mhealth and uhealth, is properly cited. The complete bibliographic information, a link to the original publication on http://mhealth.jmir.org/, as well as this copyright and license information must be included.
The 6-min walk test (6MWT) is a convenient method for assessing functional capacity in patients with cardiopulmonary conditions. It is usually performed in the context of a hospital clinic and thus requires the involvement of hospital staff and facilities, with their associated costs.
This study aimed to develop a mobile phone–based system that allows patients to perform the 6MWT in the community.
We developed 2 algorithms to compute the distance walked during a 6MWT using sensors embedded in a mobile phone. One algorithm makes use of the global positioning system to track the location of the phone when outdoors and hence computes the distance travelled. The other algorithm is meant to be used indoors and exploits the inertial sensors built into the phone to detect U-turns when patients walk back and forth along a corridor of fixed length. We included these algorithms in a mobile phone app, integrated with wireless pulse oximeters and a back-end server. We performed Bland-Altman analysis of the difference between the distances estimated by the phone and by a reference trundle wheel on 49 indoor tests and 30 outdoor tests, with 11 different mobile phones (both Apple iOS and Google Android operating systems). We also assessed usability aspects related to the app in a discussion group with patients and clinicians using a technology acceptance model to guide discussion.
The mean difference between the mobile phone-estimated distances and the reference values was −2.013 m (SD 7.84 m) for the indoor algorithm and −0.80 m (SD 18.56 m) for the outdoor algorithm. The absolute maximum difference was, in both cases, below the clinically significant threshold. A total of 2 pulmonary hypertension patients, 1 cardiologist, 2 physiologists, and 1 nurse took part in the discussion group, where issues arising from the use of the 6MWT in hospital were identified. The app was demonstrated to be usable, and the 2 patients were keen to use it in the long term.
The system described in this paper allows patients to perform the 6MWT at a place of their convenience. In addition, the use of pulse oximetry allows more information to be generated about the patient’s health status and, possibly, be more relevant to the real-life impact of their condition. Preliminary assessment has shown that the developed 6MWT app is highly accurate and well accepted by its users. Further tests are needed to assess its clinical value.
The 6-min walk test (6MWT) is a common clinical instrument for assessing patients’ functional capacity. It consists of instructing patients to walk as far as they can during 6 min, usually in a corridor [
The walked distance reflects exercise capacity determined by maximal cardiopulmonary exercise testing in patients with cardiopulmonary conditions and has a strong association with mortality in primary pulmonary hypertension [
Although the test is easy to perform, it involves costs and some practical limitations. To start with, it requires a dedicated corridor in the hospital, of length between 30 m and 50 m and no shorter than 15 m [
With the advent of affordable digital devices and mobile phones, it becomes possible to perform the test in (or near) the patient’s home, using sensors such as accelerometers or the global positioning system (GPS) to estimate the distance walked.
In this paper, we present a mobile phone app which enables patients to perform the 6MWT on their own, at their convenience or in the hospital setting, while augmenting the information collected during the test using off-the-shelf portable pulse oximeters.
The walked distance can be obtained using satellite positioning systems when outdoors and with inertial sensors when indoor.
Positioning systems like GPS are already widely used for estimating distance in the automotive sector. Modern GPS receivers provide a signal which is the result of heavy processing and is usually improved and smoothed with well-known techniques [
With regard to the indoor scenario, there is rich literature related to gait analysis with accelerometers [
In a study by Schimpl et al [
Gait analysis–based approaches have also been used for the 6MWT. For example, in a study by Schulte et al [
In addition to research papers, it is also worth mentioning the Apple Research Kit, an open-source software framework that allows developers to build mobile health (mHealth) apps with a set of already implemented use-cases. One of these use-cases is the
Our system was co-designed by a team of engineers, cardiologists, physiologists, and patients in a set of
It was also decided to make use of patients’ own phones, instead of providing them with dedicated ones. This was because we hypothesized that users would prefer using the devices with which they are familiar; however, this meant that both Android and iPhones had to be supported. This decision also allowed us to collect information about free-living physical activity, as gathered by the phone’s sensors or any wearable connected to it.
Given that these patients can desaturate significantly during exertion, it was decided to acquire pulse oximetry data during the 6MWT with a wireless sensor attached to the patient’s finger. To complement the observations made by the physiologist with these data, the mobile phone also had to be used during the test at the hospital. In addition, the clinician responsible for the patient’s care, in this case the cardiologist, had to be given an interface to review all the patient’s data collected by the system.
To summarize these requirements, the following use-cases were identified:
A patient performs the 6MWT in the hospital, while being monitored by a mobile phone app.
A physiologist supervising the 6MWT enters the observed outcomes on a tablet computer.
A patient performs the 6MWT outdoors, in a place of their choice.
A patient sends their physical activity data, as measured by passive monitors and activity trackers over the duration of a week.
A clinician reviews patient’s data on a website.
To support the abovementioned use-cases, we designed the client-server architecture shown in
The server includes a database and a website to collect patient data to be subsequently reviewed by clinicians. Physiologists can use a tablet computer with an app that allows them to review patients’ information and report the results of the 6MWT. The app also allows connection to a wireless pulse oximeter to retrieve peripheral arterial oxygen saturation (SpO2) and heart rate values, while the patient is performing the test.
Patients are provided with a mobile phone app, downloaded onto their phones, which allows both indoor and outdoor 6MWT. The indoor test is performed on a walkway of a known length, for example, in a hospital clinic. The outdoor test can be performed in any place where there is a GPS signal of sufficient strength. At the end of each test, the data are sent to the server to be reviewed by clinicians. Patients can also send data about passive activity monitoring using HealthKit for iOS and Google Fit for Android. These can compute steps and activity through either mobile phone sensors or other connected apps.
Patient data are protected by means of well-established techniques, that is, users are authenticated with a username and password, and data are transmitted over an http encrypted channel.
Architecture of the 6-min walk test (6MWT) system. It includes 4 scenarios: (a) 6MWT at home, where patients perform a 6MWT in their home setting using their mobile phone with the app and a wireless pulse oximeter; (b) 6MWT in the hospital, where patients perform the test while being observed by a physician and with pulse oximetry data being collected through a tablet app; (c) activity tracking data retrieved by Google Fit or HealthKit transmitted for subsequent analysis; (d) data review performed by a physician through a Web interface.
To compute the distance walked, we developed and tested 2 algorithms: one for the indoor scenario and one for the outdoor scenario.
The accuracy of the algorithms was estimated by performing a set of indoor and outdoor 6MWTs, with the app running on a mobile phone held in one hand and a trundle wheel held in the other hand. Different types of walking styles (from slow to fast), path curviness (from straight to U-turns), and hand shakiness (from still to slightly shaky) were simulated.
Most tests were executed by researchers in lab settings; however, to fine tune the indoor algorithm in real-life conditions, we also asked some patients to hold the mobile phone while performing the 6MWT during a regular clinic visit. Only the distance and an approximate age range were collected.
Accuracy was calculated using the mean, median, and standard deviation of the difference between the reference values and the outputs from our algorithm; the mean, standard deviation, minimum and maximum of the absolute difference; and the Pearson correlation between estimated values and the reference values. In addition, Bland-Altman plots were also generated.
Details about the algorithms are provided as follows.
We tried to implement the algorithm described by Capela et al [
The underlying concept is to detect changes of 180° in the mobile phone azimuth signal (an example of such a signal is shown is
After calibration, samples of the compass signal are acquired every 500 ms. If the current sample is more than
The parameters to be optimized in this algorithm are
Once U-turns are detected, the walked distance is computed by multiplying the number of U-turns by the length of the lap. Any residual time between the last U-turn and the end of the test is accounted for by multiplying it by the median of the detected times between U-turns.
If a step counter is available, it is used to improve the residual distance estimation. Specifically, instead of using the median
The source code of the algorithm is provided in the
Example of mobile phone azimuth signal. The first seconds show the calibration phase, after which U-turns are detected when the difference between near angles becomes greater than the set threshold within a short time window.
The outdoor algorithm works by using the localization information provided by the GPS system embedded in the phone (an example is shown in
As the mobile phone GPS system needs some time to be fully connected, we let the user wait until a
After this simple signal quality step, positioning samples start to be collected. Every
The
If step counting is available, we use it to exclude samples for which the number of steps does not increase. For example, a perfectly still mobile phone will produce positioning samples with jitter around them because of noise, and the algorithm will sum up the distances between them. By using the step counter, it is possible to identify when the user is still, thus not accumulating those erroneous distances.
The source code of the algorithm is provided in the
Example of a positioning trace (in red) retrieved from the mobile phone. The walking man figure indicates the starting point of the test; the flag indicates its end. Comparing the trace with the underlying picture shows that the position is sometimes affected by an error, for example, near tall buildings which reflect the signal or because of trees obscuring the global positioning system satellite’s signal.
Maximum and mean error of the distance estimation versus the sampling period of the localization signal selection_period computed on all available tests. The 5 seconds value minimizes both mean and maximum error.
To understand user aspects such as the usability and technology acceptance of our system, we organized a discussion group to collaboratively analyze one of the first prototypes that had been developed. Different types of stakeholders were invited to the group including patients, physiologists, physicians, and engineers. Participants were informed that the outcomes of the discussion could be used for scientific publications, and patients were required to sign an informed consent form.
To structure the content of the discussion group, we used the mHealth technology acceptance model form [
The discussion group was led by one researcher, who explained the system and its capabilities and asked the attendees questions. The content was split into 5 parts. In the first part, 2 questions were asked to understand what the current limitations and needs were in relation to a conventional 6MWT:
Q1. What are the most annoying things about the 6MWT as it is done now?
Q2. How does the test compare with your normal level of fitness; do the test’s results adequately reflect the way you feel?
After a discussion of the answers to these 2 questions, a general presentation of the system was given (part 2), after which, other questions were asked about the overall concept (part 3):
Q3. What do you think are the advantages of the system you have just seen?
Q4. What are the disadvantages?
The fourth part of the discussion consisted in letting patients download the app on their phone and use it in a test run, while the research team recorded difficulties, technical issues, and general comments.
Finally, in part 5, a further set of questions were asked about usability and acceptance:
Q5. Do you find the system easy to use?
Q6. Would you suggest any changes to it?
Q7. Do you see yourself performing the test at home?
Q8. Would you need someone to help you?
Q9. Would you use the indoor or the outdoor test?
Q10. How often do you think you will be doing it?
Q11. Do you see yourself using the app in the long term (2 years or more)?
These questions were mapped to the constructs under analysis as follows: perceived ease of use: Q5, self-efficacy: Q7 and Q8, response cost: Q4, response efficacy: Q3, perceived vulnerability: Q1 and Q2, intention to adopt: Q9, Q10, and Q11.
The discussion was audio-recorded for later analysis.
We developed 2 apps, 1 for the patient and 1 for the physiologist, as well as a server for the back-end system.
Having decided to use the patients’ own mobile phones, we needed to support both Android and iOS operating systems. We therefore implemented the patients’ app using the Apache Cordova framework. In addition to Cordova, the app makes use of the Ionic framework (first version), which uses Angular as the front-end JavaScript framework.
To retrieve the data about passive monitoring, we connected the app to Google Fit on Android and HealthKit on iOS. Both systems provide an
For ambulatory pulse oximetry, we chose the Nonin WristOx (only compatible with Android) and Creative Medical PC68B (compatible with both Android and iOS) because they are wrist-worn, with a finger probe, and because of their Bluetooth wireless connectivity.
The outdoor 6MWT use-case is shown in
The server system consists of a nonrelational database (ArangoDB), a REST API developed with Nodejs, and a front-end website developed with Angular. The Web interface (
Screenshots of the patients’ app. (a) Home page, (b) instructions about how to perform the test, (c) connection to the pulse oximeter and baseline measurements at rest, (d) estimation of the distance during walk, (e) total distance estimation and recovery at rest, (f) Borg scale questionnaire.
Screenshots of the server Web interface. (a) The form physiologists fill in when observing a 6-min walk test (6MWT), (b) an example of an outdoor 6MWT results (heart rate and oxygen saturation charts are omitted).
A total of 79 indoor and outdoor tests were performed. Lab tests were undertaken by researchers, all males, aged 30, 33, and 37. The distance estimated in regular 6MWT clinics was collected 18 times from both male and female volunteers and with an age span of 15 to 85 years.
The accuracy of the algorithm is reported separately for the indoor and outdoor scenarios.
The accuracy of the indoor algorithm was estimated using results from 49 tests. The characteristics of the tests are shown in
The difference between the algorithm’s estimates and the measurements taken from the trundle wheel are summarized in
Summary characteristics of the indoor tests.
Characteristics | Value |
Number of tests | 49 |
Number of different phones tested | 11 |
Walked distance measured by trundle wheel (m), mean (SD) | 381.79 (103.90) |
Steps, as estimated by the phone’s pedometer, mean (SD) | 574.10 (146.30) |
Accuracy metrics for the indoor algorithm. By difference, we mean the difference between the estimated distance as computed by the app and the reference distance, as measured by the trundle wheel.
Accuracy metric | Value |
Mean difference (m) | −2.01 |
Median difference (m) | −1.51 |
Standard deviation of the difference (m) | 7.84 |
Correlation | 0.99 |
Mean absolute difference (m) | 5.55 |
Standard deviation of the absolute difference (m) | 5.84 |
Minimum absolute difference (m) | 0 |
Maximum absolute difference (m) | 23.68 |
Bland-Altman plot of the difference between the estimated distance walked and the absolute distance. The Shapiro-Wilk test confirms the normality of the data (0.91).
The characteristics of the outdoor tests are shown in
The accuracy metrics are listed in
Characteristics of the outdoor tests.
Characteristic | Value |
Number of tests | 30 |
Number of different phones tested | 8 |
Walked distance measured by trundle wheel (m), mean (SD) | 437.99 (147.82) |
Steps, as estimated by the phone’s pedometer, mean (SD) | 696.5 (78.31) |
Accuracy metrics of the outdoor algorithm. By difference, we mean the difference between the estimated distance as computed by the app and the reference distance, as measured by the trundle wheel.
Accuracy metric | Value |
Mean difference (m) | −0.80 |
Median difference (m) | −0.63 |
Standard deviation of the difference (m) | 18.56 |
Correlation | 0.99 |
Mean absolute difference (m) | 13.39 |
Standard deviation of the absolute difference (m) | 12.65 |
Minimum absolute difference (m) | 0 |
Maximum absolute difference (m) | 47.27 |
Bland-Altman plot of the difference between the estimated distance walked and the ground truth. The Shapiro-Wilk test confirms the normality of the data (0.97).
The discussion group mentioned in the Methods section was held shortly after the first version of the app was ready. The attendees were 2 engineers, 1 cardiologist, 1 nurse, 2 physiologists, and 2 patients with pulmonary hypertension. One engineer led the discussion, while the other took notes.
From the initial discussion, the main issues with the way that the 6MWT is currently performed in hospital were identified as follows:
The corridor being used in the hospital was not ideal, as it was usually busy with other people and patients being moved on trolleys, both of which may affect the walking pace.
Patients’ performance might depend on their health status on that particular day and may not reflect their average status.
The test is only performed rarely (once or twice a year), and episodes of health deterioration may be missed.
Younger patients might underperform as opposed to the older ones who may try harder in the hospital test, which might not reflect real-life conditions.
White coat syndrome may cause anxiety in some patients and affect their performance.
Overall, patients are stressed and rushed in hospital.
After the app and monitoring system were presented, the following advantages were identified:
The system allows the patient to perform the test in a more comfortable environment and more often than the hospital tests.
Patients can see for themselves how they are progressing.
The system can alert in the case of very low oxygen saturations.
In terms of disadvantages:
When the test is performed outdoors, the weather can affect the patient’s performance.
Changes in altitude, as a result of walking up an incline, can affect the results of the test.
During the dry-run test of the system with 2 patients, the following observations were made:
The pulse oximeter generates a sound when a low oxygen saturation value is measured, but this can be disabled.
A patient had problems understanding how to wear the pulse oximeter.
Patients were able to install the app correctly and could understand its structure easily.
One patient tried to send their activity tracking data but did not have Google Fit installed.
Patients struggled to log in, because complex passwords with capital and lower case letters were originally assigned to them.
During a test, the pulse oximeter produced artefactual values.
A physiologist asked to be able to discard the results of a test if the data recorded did not appear to be accurate enough to them.
With respect to the usability of the app, the patients’ comments were as follows:
The app is usable and easy to understand.
It would be better to show how many seconds are left until the end of the test, rather than how many seconds have elapsed since the start.
One patient asked for a sound to be generated at the end of the test to allow them not to have to look at the mobile phone screen all the time.
Regarding the willingness to use the app, the following points were noted:
Both patients said that they would use the app regularly and would not need any help to do so.
One patient would prefer performing the test indoors during winter because the cold weather affects their breathing, whereas the other patient only wanted to use the outdoor version.
As indoor tests require a long passageway, it was suggested that shopping malls could be used for these tests.
One patient would like to place the mobile phone on an armband or in a pocket during the test.
Patients identified scenarios for which they could perform the test while doing some other activity, for example, taking children to and from school.
Both patients agreed to use the app in the long term.
The results from our tests show that, in both the indoor and outdoor scenarios, the difference between the distance estimated by the app and the ground truth is always below 54 m, when 50 m is considered to be the clinically significant threshold for detecting changes in disease state [
If we compare our accuracy results with those reported in the literature, for the outdoor scenario, an average error of a few meters, up to a maximum of 20 m, was reported by Gray et al [
As limitations to our approach, we should mention that most of the accuracy tests were performed by researchers in a lab environment. Although we tried to simulate worst-case scenarios, it is possible that the distance estimation algorithm may perform worse
In terms of user aspects, the results from the discussion group suggest that the way the 6MWT is performed in hospitals has significant limitations (a
In terms of
Given the positive answers provided for its associated constructs,
Although the indications of the discussion group were generally supportive of the system, there are risks that can affect its actual use; particularly, the fact that it has to be either used outdoors, which is limited by the weather, or indoors but in a long corridor, which is limited by the availability of space. In addition, the integration with the sensor and external apps makes the overall user experience more cumbersome, a fact that may affect some less tech-savvy users.
The system described in this paper allows patients to perform the 6MWT at a place of their convenience, thus allowing more information to be generated about their general health status, more frequently, and possibly reflecting their general health status better. The algorithms implemented to estimate the distance walked from either the compass or GPS showed good agreement with the reference trundle wheel measurements, with errors below the clinically significant threshold. Our preliminary user validation also indicates that the app is usable and has the potential to be well accepted among patients.
The system will need to be tested with more patients to assess its feasibility in a real-world scenario. A clinical trial is currently being run with 30 pulmonary hypertension patients to understand the relationship between tests undertaken using the app and conventional in-hospital tests. Further studies will then be needed to assess the clinical significance of the tests in the community and the relationship between passive activity monitoring, for example, through wearables and in-hospital 6MWT results.
Javascript implementation of the indoor distance estimation algorithm.
Javascript implementation of the outdoor distance estimation algorithm.
6-min walk test
chronic obstructive pulmonary disease
global positioning system
mobile health
peripheral artery disease
The research described in this paper was supported by the NIHR Biomedical Research Centre, Oxford, and by an EPSRC grant (EP/EP/N024966/1—Intelligent Wearable Sensors for Predictive Patient Monitoring).
None declared.