This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), 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.
Improvements in mobile telecommunication technologies have enabled clinicians to collect patient-reported outcome (PRO) data more frequently, but there is as yet limited evidence regarding the frequency with which PRO data can be collected via smartphone applications (apps) in breast cancer patients receiving chemotherapy.
The primary objective of this study was to determine the feasibility of an app for sleep disturbance-related data collection from breast cancer patients receiving chemotherapy. A secondary objective was to identify the variables associated with better compliance in order to identify the optimal subgroups to include in future studies of smartphone-based interventions.
Between March 2013 and July 2013, patients who planned to receive neoadjuvant chemotherapy for breast cancer at Asan Medical Center who had access to a smartphone app were enrolled just before the start of their chemotherapy and asked to self-report their sleep patterns, anxiety severity, and mood status via a smartphone app on a daily basis during the 90-day study period. Push notifications were sent to participants daily at 9 am and 7 pm. Data regarding the patients’ demographics, interval from enrollment to first self-report, baseline Beck’s Depression Inventory (BDI) score, and health-related quality of life score (as assessed using the EuroQol Five Dimensional [EQ5D-3L] questionnaire) were collected to ascertain the factors associated with compliance with the self-reporting process.
A total of 30 participants (mean age 45 years, SD 6; range 35-65 years) were analyzed in this study. In total, 2700 daily push notifications were sent to these 30 participants over the 90-day study period via their smartphones, resulting in the collection of 1215 self-reporting sleep-disturbance data items (overall compliance rate=45.0%, 1215/2700). The median value of individual patient-level reporting rates was 41.1% (range 6.7-95.6%). The longitudinal day-level compliance curve fell to 50.0% at day 34 and reached a nadir of 13.3% at day 90. The cumulative longitudinal compliance curve exhibited a steady decrease by about 50% at day 70 and continued to fall to 45% on day 90. Women without any form of employment exhibited the higher compliance rate. There was no association between any of the other patient characteristics (ie, demographics, and BDI and EQ5D-3L scores) and compliance. The mean individual patient-level reporting rate was higher for the subgroup with a 1-day lag time, defined as starting to self-report on the day immediately after enrollment, than for those with a lag of 2 or more days (51.6%, SD 24.0 and 29.6%, SD 25.3, respectively;
The 90-day longitudinal collection of daily self-reporting sleep-disturbance data via a smartphone app was found to be feasible. Further research should focus on how to sustain compliance with this self-reporting for a longer time and select subpopulations with higher rates of compliance for mobile health care.
Electronic health (eHealth) can be defined as the practice of medicine and public health using information and communication technology (ICT), such as computers, mobile phones, and satellite communications [
Prior to the eHealth era, the collection of PROs in the oncology field was based simply on patient recall at clinic visits or by asking patients to keep paper diaries outside the clinic. However, patient recall is inherently inaccurate and plagued by potential bias and the actual compliance with keeping paper diaries according to protocols has proved to be much lower than was expected [
The primary objective of this study was to determine the feasibility of using a smartphone app to collect sleep disturbance-related data from breast cancer patients receiving chemotherapy, with overall and individual reporting rates. We also sought to determine whether the patients stopped responding via the app after a short time by calculating a longitudinal reporting rate per day over a period of 90 days. A secondary objective was to elucidate the variables associated with a higher compliance rate in order to identify the optimal subgroups to include in future studies of smartphone-based interventions.
Patients who were planning to receive neoadjuvant chemotherapy for breast cancer were recruited for participation in this feasibility study when they were admitted to the Breast Cancer Center, Asan Medical Center, Seoul, South Korea, for 2 nights to evaluate the disease status for preoperative chemotherapy between March 2013 and July 2013. To be eligible for this study, patients had to indicate app avidity, be current iOS or Android smartphone users, and be able to read and understand Korean. In some cases, the app did not run properly due to unexpected incompatibility between the smartphones’ display specifications and the app.
Among the 67 patients initially recruited for participation, 29 were excluded for the following reasons: incompatibility of the app with their smartphone’s display specifications (n=14), not a current smartphone user or no app avidity (n=9), not interested in the research (n=5), and language barrier (n=1). A total of 38 patients were enrolled and provided informed consent to participate (IRB no. 2012-0709). Of these, 8 patients were not included in the analysis performed in this feasibility study because they did not start to use the app or did not report sleep-disturbance symptoms at all after providing consent to participate at admission (
The participants were presented with information regarding the study and the app, called “Pit-a-Pat”, which was developed for both the iOS and Android platforms, and has been described previously [
After a 2-night admission for workup, the participants were discharged and scheduled to visit the outpatient clinic 7 days later (on average) to receive neoadjuvant chemotherapy. The workup results for neoadjuvant chemotherapy, such as regional lymph node metastasis, hormone receptor expression, and planned chemotherapeutic agents regimens, were also collected.
Study design and participant flow.
Briefly, the app in this study was developed for cancer patients to self-report three kinds of health experience that may be caused by the effect of the diagnosis with breast cancer itself and the subsequent treatments: (1) sleep-disturbance symptoms related to mild depression, (2) acute symptoms related to cytotoxic chemotherapeutic agents, and (3) medication diary for antihormonal treatment such as tamoxifen and aromatase inhibitor. The sleep-disturbance symptoms related to mild depression were assessed using a six-item questionnaire: (1) time of falling asleep, (2) time of waking up, (3) number of awakenings during the total sleep period, (4) quality of total sleep on a 10-point rating scale, (5) present mood status on a 7-point rating scale, and (6) severity of present anxiety on a 10-point rating scale. The sleeping and waking times (in minutes) were stored as continuous variables. The 10-point rating scale for quality of total sleep and severity of present anxiety was displayed as a distress thermometer and the 7-point rating scale for mood status was displayed as faces with a range of expressions. The six items were displayed in three sequential pages as shown in
Participants were informed that they could start reporting their symptoms via the app at any time from the day following their provision of consent. Once the app was activated, push notifications were sent to the participants daily at 9 am and 6 pm, asking them to daily self-report sleep patterns, anxiety severity, and mood status via the smartphone app over a 90-day period. After entering appropriate answers to all six items without missing values, a participant could touch the send button and be considered as having completed a self-report of her sleep-disturbance symptoms on that specific day. The same push notification process was repeated for all participants during the study period, irrespective of their reporting rates. A database system collected the information about daily requests and inputs (ie, anonymized user ID, item ID, date and time of input, and input value), as well as baseline demographics.
The database system was connected to our electronic medical record system to enable clinicians to review the data. During the 90-day study period, clinicians could review the participants’ self-reporting data at the outpatient clinic at 3-week intervals, and managed them in the same way as the nonparticipating patients for the reported symptoms, since this was a study only related to feasibility. Participants were not provided with a smartphone for this study, instead being required to use their own device. Furthermore, they did not receive any reimbursement of expenses for extra data usage caused by utilization of the app, or any financial incentive for participating in the study.
At the end of the study period, the participants received a push notification requesting reasons for missing self-reporting, by choosing from the following possible responses: “The app didn’t work properly”, “I forgot”, “I didn’t think it was useful”, “I was too sick”, “I didn’t feel like it”, “It was inconvenient”, or “I was too busy”. Among the 30 participants who received 90 requests for self-reporting, 24 responded and their data were analyzed.
Screenshot of the app for self-reporting of sleep disturbance data.
The feasibility of this self-reporting system was analyzed by calculating individual patient-level reporting rates, defined as the total number of days in which self-reporting was completed divided by 90 for each participant, and a longitudinal day-level reporting rate, defined as the total number of participants who completed self-reporting divided by 30 on a single specific day from the start of self-reporting. A cumulative longitudinal day-level reporting rate was also defined as the mean of all longitudinal day-level reporting rates from the starting day until the specified day. For example, the longitudinal day-level reporting rate at day 34 indicates the number of participants self-reporting on the 34th day since the starting day divided by 30 and the cumulative longitudinal day-level reporting rate at day 34 indicates the mean value of all longitudinal day-level reporting rates from day 1 to day 34. A longitudinal compliance curve was plotted by connecting the 90 longitudinal day-level reporting rates for each day over the 90-day study period and a cumulative longitudinal compliance curve with 90 cumulative longitudinal day-level reporting rates over time. For the cumulative dropout probability curve, a dropout event was defined as when a cumulative longitudinal day-level reporting rate of a participant decreases first to below 28.6%, which means that she reported less than twice per week, which is almost the same as the weekly biodata collection.
For a priori threshold criteria for good compliance, prior feasibility studies in related contexts were reviewed [
The chi-square test and
Between March 2013 and July 2013, 67 patients were initially eligible for participation. However, 14 patients (20%, 14/67) could not participate in this study because of inherent technological limitations in the development of the app causing incompatibility with their smartphone’s display specifications. Among the 53 women who did not experience technical problems in using the app, 9 (17%, 9/53) showed no interest in either their smartphone or the app and 5 (9%, 5/53) refused to participate in this study because they were not interested in the research. Of the 38 eligible patients who were actually enrolled into this feasibility study, 8 (21%, 8/38) did not start to use the app. The final analysis thus included 30 patients (mean age 45 years, SD 6; range 36-65 years), all 30 of whom were Korean women who had been diagnosed with breast cancer within 4 weeks prior to enrollment; 73% (22/30) were 50 years old, 47% (14/30) had an educational attainment of college level or higher, and 43% (13/30) were currently employed. The baseline BDI and EQ5D-3L scores were 11.5 (SD 8.8) (range 0-35) and 0.92 (SD 0.09) (range 0.56-1.00), respectively. For tumor characteristics, lymph-node metastases were confirmed in 77% (23/30), 60% (18/30) were hormone-responsive breast cancer, and those with any kind of distant metastases were excluded. During the 90-day study period, all participants received adriamycin-based (87%, 26/30) or epirubicin-based (13%, 4/30) combinational chemotherapeutic agents, of whom 77% (23/30) received additional docetaxel because of positive lymph-node metastasis after this feasibility study (
Baseline demographics of patients (n=30).
Characteristic | Total |
Lower compliance rate |
Higher compliance rate |
|
||
n (%) | ||||||
|
||||||
|
Mean (SD) years |
|
45 (6) | 46 (5) | 45 (8) | NSa |
|
Range, years |
|
36-65 | 38-55 | 36-65 |
|
|
≤49 years |
|
22 (73) | 11 (73) | 11 (73) | NS |
|
≥50 years |
|
8 (27) | 4 (27) | 4 (27) |
|
|
NS | |||||
|
Up to high school |
|
16 (53) | 10 (67) | 6 (40) |
|
|
College or greater |
|
14 (47) | 5 (33) | 9 (60) |
|
|
NS | |||||
|
Married |
|
26 (87) | 13 (87) | 13 (86) |
|
|
Single |
|
1 (3) | 0 (0) | 1 (7) |
|
|
Divorced |
|
3 (10) | 2 (13) | 1 (7) |
|
|
NS | |||||
|
Yes |
|
28 (93) | 15 (100) | 13 (87) |
|
|
No |
|
2 (7) | 0 (0) | 2 (13) |
|
|
.03 | |||||
|
Yes (of any kind) |
|
13 (43) | 10 (67) | 3 (20) |
|
|
No (eg, stay-at-home mother) |
|
17 (57) | 5 (33) | 12 (80) |
|
|
||||||
|
Mean (SD) |
|
11.5 (8.8) | 11.9 (8.5) | 11.1 (9.4) | NS |
|
Range |
|
0-35 | 1-33 | 0-35 |
|
|
≤15 |
|
21 (70) | 10 (67) | 11 (73) | NS |
|
≥16 |
|
9 (30) | 5 (33) | 4 (27) |
|
|
||||||
|
|
|
|
|
|
NS |
|
|
Mean (SD) | 69.6 (16.1) | 68.3 (16.0) | 70.9 (16.6) |
|
|
|
Range | 38-99 | 45-95 | 38-99 |
|
|
|
|
|
|
|
NS |
|
|
Mean (SD) | 0.92 (0.09) | 0.92 (0.07) | 0.91 (0.11) |
|
|
|
Range | 0.56-1.00 | 0.74-1.00 | 0.56-1.00 |
|
|
|
|
|
|
|
NS |
|
|
No problems | 29 (97) | 15 (100) | 14 (93) |
|
|
|
Problems | 1 (3) | 0 (0) | 1 (7) |
|
|
|
|
|
|
|
NS |
|
|
No problems | 30 (100) | 15 (100) | 15 (100) |
|
|
|
Problems | 0 (0) | 0 (0) | 0 (0) |
|
|
|
|
|
|
|
NS |
|
|
No problems | 29 (97) | 15 (100) | 14 (93) |
|
|
|
Problems | 1 (3) | 0 (0) | 1 (7) |
|
|
|
|
|
|
|
NS |
|
|
No problems | 19 (63) | 11 (73) | 8 (53) |
|
|
|
Problems | 11 (37) | 4 (27) | 7 (47) |
|
|
|
|
|
|
|
NS |
|
|
No problems | 14 (47) | 6 (40) | 8 (53) |
|
|
|
Problems | 16 (53) | 9 (60) | 7 (47) |
|
|
NS | |||||
|
Negative |
|
7 (23) | 4 (27) | 3 (20) |
|
|
Positive |
|
23 (77) | 11 (73) | 12 (80) |
|
|
NS | |||||
|
Positive |
|
18 (60) | 6 (40) | 12 (80) |
|
|
Negative |
|
12 (40) | 9 (60) | 3 (20) |
|
|
NS | |||||
|
ACf#4 |
|
7 (23) | 4 (27) | 3 (20) |
|
|
AC#4 followed by docetaxel#4 |
|
19 (64) | 7 (46) | 12 (80) |
|
|
FECg#3 followed by docetaxel#3 |
|
4 (13) | 4 (27) | 0 (0) |
|
aNot significant.
bBDI: Beck’s Depression Inventory.
cHRQOL: health-related quality of life.
dEQ5D-3L: EuroQol Five Dimensional Questionnaire.
eVisual analog scale.
fCombination of doxorubicin+cyclophosphamide chemotherapy.
gCombination of 5-fluorouracil+epirubicin+cyclophosphamide chemotherapy.
In total, 2700 daily push notifications were sent to the 30 participants via their smartphones during the 30-day study period; 1215 responses were received containing completed self-reporting sleep-disturbance data (overall compliance rate=45.0%; 1215/2700). The median individual patient-level reporting rates from the 30 participants was 41.1% (range 6.7-95.6%), with the following distribution: 6 (20%, 6/30), 9 (30%, 9/30), 6 (20%, 6/30), 7 (23%, 7/30), and 2 (7%, 2/30) ranging from 0% to 20%, 21% to 40%, 41% to 60%, 61% to 80%, and 81% to 100%, respectively (
The longitudinal day-level reporting rate on day 1 is theoretically 100%, because that day is the first day when all participants started to use the app. The longitudinal day-level reporting rates from day 1 to day 90 were calculated at daily intervals and plotted as a longitudinal compliance curve. The longitudinal compliance curve for this app decreased rapidly to about 50.0% at day 34 after the start of self-reporting, and continued to decrease steadily thereafter, reaching a nadir of 13.3% at day 90. The cumulative longitudinal compliance curve (representing overall compliance) revealed a steady decrease by about 50% at day 70, and continued to fall to 45% on day 90. Cumulative dropout probabilities showed that 10% (3/30) of all participants self-reported less than twice per week on average (cumulative longitudinal compliance rate <28.6%) during the first 28 days, 20% (6/30) during the first 56 days, and 33.3% (10/30) during the 90-day study period (
Distribution of individual patient-level reporting rates.
Changes in compliance over time.
The distribution of individual patient-level reporting rates depicted in
The interval from enrollment to first self-reporting (ie, lag time) was also investigated in this study. As shown in
Average individual patient-level reporting rates in each subgroup classified according to the clinicopathologic variables (n=30).
Subgroup classified according to each variable | n (%) | Individual patient-level reporting rate (%) | |||
Mean (SD) | Range |
|
|||
|
NSa | ||||
|
Young age (≤49 years) | 22 (73) | 46.0 (23.8) | 7.8-94.4 |
|
|
Old age (≥50 years) | 8 (27) | 42.4 (33.1) | 6.7-95.6 |
|
|
NS | ||||
|
Up to high school | 16 (53) | 40.3 (24.8) | 6.7-95.6 |
|
|
College or greater | 14 (47) | 50.3 (27.3) | 7.8-94.4 |
|
|
NS | ||||
|
Married | 26 (87) | 45.6 (26.4) | 6.7-95.6 |
|
|
Single | 1 (3) | 50.0 |
|
|
|
Divorced | 3 (10) | 37.8 (31.8) | 7.8-71.1 |
|
|
NS | ||||
|
Yes | 28 (93) | 43.6 (26.1) | 6.7-95.6 |
|
|
No | 2 (7) | 65.0 (21.2) | 50.0-80.0 |
|
|
.006 | ||||
|
Yes (of any kind) | 13 (43) | 30.7 (19.2) | 6.7-71.1 |
|
|
No (eg, stay-at-home mother) | 17 (57) | 55.9 (25.7) | 10.0-95.6 |
|
|
NS | ||||
|
No anxiety (BDI≤15) | 21 (70) | 46.2 (27.8) | 7.8-95.6 |
|
|
Anxiety (BDI≥16) | 9 (30) | 42.2 (22.7) | 6.7-80.0 |
|
|
NS | ||||
|
No problems | 19 (63) | 41.2 (27.5) | 6.7-95.6 |
|
|
Problems | 11 (37) | 51.6 (23.0) | 24.4-80.0 |
|
|
NS | ||||
|
No problems | 14 (47) | 49.6 (29.9) | 7.8-95.6 |
|
|
Problems | 16 (53) | 41.0 (22.3) | 6.7-78.9 |
|
|
NS | ||||
|
Localized | 7 (23) | 41.6 (31.4) | 13.3-95.6 |
|
|
Advanced | 23 (77) | 46.0 (24.9) | 6.7-94.4 |
|
|
NS | ||||
|
Positive | 18 (60) | 52.8 (24.8) | 6.7-95.6 |
|
|
Negative | 12 (40) | 33.2 (24.2) | 7.8-94.4 |
|
|
NS | ||||
|
ACd#4 | 7 (23) | 41.6 (31.4) | 13.3-95.6 |
|
|
AC#4 followed by docetaxel#4 | 19 (64) | 50.7 (24.6) | 6.7-94.4 |
|
|
FECe#3 followed by docetaxel#3 | 4 (13) | 23.9 (11.1) | 7.8-33.3 |
|
aNot significant.
bBDI: Beck’s Depression Inventory.
cEQ5D-3L: EuroQol Five Dimensional Questionnaire.
dCombination of doxorubicin+cyclophosphamide chemotherapy.
eCombination of 5-fluorouracil+epirubicin+cyclophosphamide chemotherapy.
Intervals from enrollment to first self-reporting. X and O indicate day of enrollment and day of start of self-reporting, respectively.
A total of 24 participants responded to the question regarding their primary reason for missing self-reporting. The most common response, which accounted for 38% (9/24) of all responses, was that “The app didn’t work properly.” Other reasons included “I forgot” (29%, 7/24) and “I didn’t think it was useful” (21%, 5/24). Minor reasons were “I was too sick” (8%, 2/24) and “I didn’t feel like it” (4%, 1/24). No missing data were due to either “It was inconvenient” or “I was too busy” (
Reasons for missing a self-reporting event and average individual patient-level reporting rates (n=24).
Reason | n (%) | Individual patient-level reporting rates (%) |
The app didn’t work properlyb | 9 (38) | 49.6 (24.6) |
I forgot | 7 (29) | 32.4 (23.0) |
I didn’t think it was useful | 5 (21) | 57.8 (26.6) |
I was too sick | 2 (8) | 42.2 (40.9) |
I didn’t feel like it | 1 (4) | 33.3 |
It was inconvenient | 0 (0) | NAc |
I was too busy | 0 (0) | NAc |
No response | 6 | 45.0 (31.2) |
aMeans among the subgroups were not significantly different by ANOVA (analysis of variance).
bTemporary dysfunctions such as delay or failure of log-on or abnormal shutdown of the app during the self-reporting.
cNot available.
This study has shown that the collection of patient self-reporting symptoms via a smartphone at daily intervals for a relatively long-term period may be feasible in women with a recent diagnosis of breast cancer and receiving ongoing chemotherapy. The overall rate of compliance was 45.0% and the median value of individual compliance (ie, individual patient-level reporting rates) was 41.1%. About half of the participants did not self-report at around the 5th week after starting to use the app and only 13.3% of the participants could be expected to self-report at the 90th day. We found that in this population, depression, HRQOL status, and demographic characteristics such as age and educational level did not affect compliance, but the results suggest that women who were not currently in employment and those who started to use the app on the day immediately after enrollment exhibited greater compliance with daily self-reporting.
During the past decade, the greatest upsurge of mHealth research occurred between 2007 and 2008, when the new generation of smartphones, such as the iPhone and similar devices, were introduced [
Comparison of longitudinal compliance rates according to different time units.
The lower compliance observed among women with a current job may be due to failure to immediately respond to push notifications and simple forgetfulness. Self-reporting during the hour immediately following the push notification (9-10 am and 7-8 pm) accounted for almost half of all responses, which is consistent with the proportion of women in this cohort without a current job. However, in women with current employment, the pattern of self-reporting had two wider peaks (
Push notifications and distribution of self-reporting time.
The compliance rate was higher for the subgroup without a lag time in this study (mean 51.6%, SD 24.0 and 29.6%, SD 25.3, respectively;
In this feasibility study, we also sought to determine the degree of persistence of compliance and the optimal time interval for self-reporting of daily changing symptoms, such as the sleep-disturbance data found in this study. Although it is not possible to form categorical conclusions regarding these parameters from the findings of the present single study, the results will help toward the development of future strategies for collecting repetitive biodata. Researchers working on smartphone-based health care can expect interactive and responsive communications from 50% of women receiving chemotherapy at the 5th week, from 20-50% between the 6th and 10th week and from just 10-30% thereafter (
The limitations of this study include the enrollment only of Korean women with breast cancer at a single urban tertiary cancer center. Other populations may exhibit different self-reporting characteristics in terms of compliance and should be independently evaluated. In addition, these findings should be generalized only after taking into account that this study was conducted under special circumstances in South Korea in which there is easy and ubiquitous access to wireless networks and the penetration of smartphones is the second highest in the world, after the United Arab Emirates (
In conclusion, the findings of this study suggest that 90-day, longitudinal collection of daily self-reporting sleep-disturbance data via a smartphone app is feasible. Further research is needed to determine how to sustain the compliance with a self-reporting program over a longer period of time and to select subpopulations with higher compliance rates for mobile health care.
Top 15 countries with highest smartphone penetration in 2013 (from our mobile planet by Google).
application
Beck’s Depression Inventory
electronic health
electronic PROs
EuroQol Five Dimensional Questionnaire
health-related quality of life
information and communication technology
mobile health
operating system
patient-reported outcome
This study was supported by a grant from the National R&D Program for Cancer Control, Ministry for Health and Welfare, Republic of Korea (no. 1220170).
None declared.