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.
Self-management plays an important role in maintaining good control of diabetes mellitus, and mobile phone interventions have been shown to improve such self-management. The Health Promotion Board of Singapore has created a caloric-monitoring mobile health app, the “interactive Diet and Activity Tracker” (iDAT).
The objective was to identify and describe short-term (8-week) trajectories of use of the iDAT app among patients with type 2 diabetes mellitus in a primary care setting in Singapore, and identify patient characteristics associated with each trajectory.
A total of 84 patients with type 2 diabetes mellitus from a public primary care clinic in Singapore who had not previously used the iDAT app were enrolled. The app was demonstrated and patients’ weekly use of the app was monitored over 8 weeks. Weekly use was defined as any record in terms of food entry or exercise workout entry in that week. Information on demographics, diet and exercise motivation, diabetes self-efficacy (Diabetes Empowerment Scale-Short Form), and clinical variables (body mass index, blood pressure, and glycosylated hemoglobin/HbA1c) were collected at baseline. iDAT app use trajectories were delineated using latent-class growth modeling (LCGM). Association of patient characteristics with the trajectories was ascertained using logistic regression analysis.
Three iDAT app use trajectories were observed: Minimal Users (66 out of 84 patients, 78.6%, with either no iDAT use at all or use only in the first 2 weeks), Intermittent-Waning Users (10 out of 84 patients, 11.9%, with occasional weekly use mainly in the first 4 weeks), and Consistent Users (8 out of 84 patients, 9.5%, with weekly use throughout all or most of the 8 weeks). The adjusted odds ratio of being a Consistent User, relative to a Minimal User, was significantly higher for females (OR 19.55, 95% CI 1.78-215.42) and for those with higher exercise motivation scores at baseline (OR 4.89, 95% CI 1.80-13.28). The adjusted odds ratio of being an Intermittent-Waning User relative to a Minimal User was also significantly higher for those with higher exercise motivation scores at baseline (OR 1.82, 95% CI 1.00-3.32).
This study provides insight into the nature and extent of usage of a caloric-monitoring app among patients with type 2 diabetes and managed in primary care. The application of LCGM provides a useful framework for evaluating future app use in other patient populations.
The prevalence of type 2 diabetes mellitus is expected to rise globally with an increasingly urbanized and aging population [
A meta-analysis of 22 trials attested to the possibility of significant reductions in glycosylated hemoglobin (HbA1c) levels (mean 0.5%; 95% CI 0.3-0.7) through the self-management of diabetes using mobile phone interaction [
Several studies have been conducted on the use of technology and mobile phones in diabetes management, including studies using interventional approaches—as opposed to control—whereby intervention groups received mobile phone reminders or feedback on self-monitoring of glucose levels [
“Latent-class growth modeling” (LCGM) is a statistical technique that exploits the existence of latent groups of individuals who share similar time trajectories of a particular trait, the characterization of which allows better understanding of the pattern of change in that variable [
The primary aim of our study was to assess iDAT app usage in patients with type 2 diabetes. More specifically, our goal was to identify and characterize short-term (8-week) trajectories of use of the iDAT app among patients with type 2 diabetes mellitus in a primary care setting in Singapore and to identify patient characteristics associated with different trajectories.
The study was conducted at one of the 18 public primary care clinics (polyclinics) located in the northeastern part of Singapore. It is a typical polyclinic, which managed almost 5000 patients with type 2 diabetes in 2013. Patients enrolled for the study had to meet all of the following inclusion criteria: (1) above 21 years of age, (2) type 2 diabetes mellitus diagnosed based on World Health Organization criteria [
Participants were enrolled over a 5-month period from November 2013 to March 2014. Patients attending the diabetes counselling and screening services for eye and foot complications at the polyclinic were approached. Patients who declined participation, did not feel comfortable using apps, or could not understand English were not recruited (
Patient use of the iDAT app was monitored weekly over a period of 2 months post-enrollment. There were no financial reimbursements to the patients for study participation. This study was approved by the SingHealth Centralized Institutional Review Board E (CIRB) (Ref: 2013/743/E), in accordance with all applicable regulations, and informed consent was obtained after the nature and possible consequences of the study were explained. Participants were informed when consent was taken and in the Participant Information Sheet that the email addresses used for iDAT registration would be collected and used to track app usage. This personal information, together with the other data collected, were to be kept confidential and only used on a need-to-know basis as approved by the CIRB.
Recruitment and study flowchart.
Questions evaluating "interactive Diet and Activity Tracker" (iDAT) app usefulness, current diet and exercise, motivation to improve diet, and motivation to exercise.
Diabetes Empowerment Scale-Short Form (DES-SF) questions developed and validated by the Michigan Diabetes Research and Training Center.
The iDAT app (
The app is free to download through Apple’s iTunes/App Store and Android’s Google Play. It functions as a calorie counter, helping users to balance calories consumed with calories burned on a daily basis. The “Meal” section allows the user to input food consumed via a food database with their estimated calories, including local ethnic foods. The “Workout” section enables the user to tap on their smartphones’ Global Positioning System (GPS) to monitor fitness workouts and calculate estimated calories burned. Workouts can be added manually or by using the app’s “Step Counter”. Other functions include social features such as Facebook-sharing and a “Weight and Goal” feature that allows users to set a weight loss goal and track weight loss over time.
"interactive Diet and Activity Tracker" (iDAT) screen captures.
Demographic variables and clinical characteristics at baseline were summarized as mean with standard deviation for continuous variables and counts and percentages for categorical variables. HPB provided the iDAT app backend information in the form of weekly use. To summarize this data, any record in terms of food entry or exercise workout entry in a week was considered as usage for that week.
A statistical analysis software (SAS) macro, PROC TRAJ, was used to apply LCGM to analyze weekly iDAT app usage data and to identify the latent groups characterizing the iDAT app use trajectories for the cohort. LCGM uses maximum likelihood to estimate model parameters [
Of 153 patients approached, 84 who consented and satisfied the inclusion/exclusion criteria were enrolled (
As we prioritized the enrollment of newly diagnosed patients, only 21% (18/84) of the enrolled participants had been diagnosed with diabetes more than a year prior to enrollment. Therefore, most of the participants had “diet only” treatment without medications (31%, 26/84) or were using one diabetes medication but not insulin (42%, 35/84).
When asked to rate how healthy their diet was on a scale of 0-9 (0-very unhealthy and 9-very healthy), the participants reported a mean score of 4.8 (SD 1.9). Their reported exercise frequency ranged from 25% (21/84) who stated they have “not exercised for the past year” to 7% (6/84) who indicated that they exercise “between 1 to 3 times per month”. They were generally quite motivated to improve their diet and exercise, giving similar mean scores of 7.3 (SD 1.5) and 6.7 (SD 1.5) respectively when asked to rate their motivation on a 0-9 scale.
Most owned Android-based smartphones (70%, 59/84). Most used their smartphones and apps frequently, with 87% (73/84) indicating that they used their smartphones more than 5 times a day and 76% (64/84) used apps more than 5 times a day. After being shown how to use iDAT, they gave a positive baseline rating for its usefulness with mean score of 6.7 (SD 1.5) on a 0-9 scale.
Characteristics of enrolled patients at baseline.
Characteristic | Total recruited |
||
|
|||
|
Age (years), mean (SD) |
|
48.2 (8.5) |
|
|
||
|
|
Male | 43 (51) |
|
|
Female | 41 (49) |
|
|
||
|
|
Chinese | 45 (54) |
|
|
Malay | 23 (27) |
|
|
Indian | 10 (12) |
|
|
Others | 6 (7) |
|
|
||
|
|
Single | 10 (12) |
|
|
Married | 70 (83) |
|
|
Divorced / Separated | 4 (5) |
|
|
||
|
|
Retired | 6 (7) |
|
|
Homemaker | 7 (8) |
|
|
Unemployed | 1 (1) |
|
|
Employed | 70 (83) |
|
|
||
|
|
Secondary and below | 39 (46) |
|
|
Post-secondary (‘A’ levels, technical) | 10 (12) |
|
|
Diploma | 18 (21) |
|
|
Degree and above | 17 (20) |
|
|||
|
BMI (kg/m2) |
|
29.1 (6.1) |
|
Height (cm) |
|
163.7 (8.7) |
|
Weight (kg) |
|
78.3 (18.9) |
|
|
||
|
|
Systolic | 130.5 (18.5) |
|
|
Diastolic | 77.6 (10.9) |
|
|||
|
|
||
|
|
New (less than 1 year) | 66 (79) |
|
|
Long-term (more than 1 year) | 18 (21) |
|
|
||
|
|
Diet only | 26 (31) |
|
|
On 1 diabetes medicine (without insulin) | 35 (42) |
|
|
On 2 diabetes medicines (without insulin) | 16 (19) |
|
|
On insulin | 7 (8) |
|
|||
|
Healthy diet score (0-9), mean (SD) |
|
4.8 (1.9) |
|
|
||
|
|
No | 63 (75) |
|
|
Ex-smoker | 8 (10) |
|
|
Yes | 13 (15) |
|
|
||
|
|
Non-drinkers | 53 (63) |
|
|
Used to drink | 5 (6) |
|
|
Regular/social drinkers | 26 (31) |
|
|
||
|
|
None in the past year | 21 (25) |
|
|
Few times per year | 13 (15) |
|
|
1-3 times per month | 6 (7) |
|
|
Once per week | 18 (21) |
|
|
2-3 times per week | 18 (21) |
|
|
Daily | 8 (10) |
|
|||
|
|
||
|
|
Apple | 25 (30) |
|
|
Android | 59 (70) |
|
|
||
|
|
More than 5 times /day | 73 (87) |
|
|
Less than 5 times /day | 11 (13) |
|
|
||
|
|
More than 5 times /day | 64 (76) |
|
|
Less than 5 times /day | 20 (24) |
|
|
iDATausefulness score (0-9), mean (SD) | 6.7 (1.5) |
|
|||
|
Diet motivation score (0-9) |
|
7.3 (1.5) |
|
Exercise motivation score (0-9) |
|
6.7 (1.5) |
|
DES-SFb(1-5) |
|
4.1 (0.5) |
aiDAT: interactive Diet and Activity Tracker
bDES-SF: Diabetes Empowerment Scale-Short Form
Using the LCGM approach and applying goodness-of-fit criteria, weekly iDAT app use was best characterized as three latent trajectory groups (
A total of 78.6% (66/84) of study participants were Minimal Users with a typical usage pattern of no iDAT input or iDAT input only during the first 2 weeks post-recruitment; 11.9% (10/84) were Intermittent-Waning Users with a typical input pattern of an occasional weekly input, mainly in the first 4 weeks post-recruitment. The remaining 9.5% (8/84) were Consistent Users with a typical input pattern of weekly input throughout all or most of the 8-week post-recruitment period.
Weekly "interactive Diet and Activity Tracker" (iDAT) app use trajectory groups identified using latent class growth modeling.
Univariate (
Univariate polytomous logistic regression for baseline predictors of iDAT app use trajectory group membership (odds ratios are calculated with Minimal Users as the reference group).
|
|
Minimal Users (n=66) | Intermittent-Waning Users |
Consistent Users |
Overall |
||||
|
|
n (%) or |
n (%) or |
OR (95% CI) |
|
n (%) or |
OR (95% CI) |
|
|
Gender (female) | 30 (45%) | 4 (40%) | 0.8 (0.2-3.1) | .75 | 7 (88%) | 8.4 (1.0-72.1) | .052 | .14 | |
Age, years | 47.8 (8.7) | 47.2 (7.7) | 1.0 (0.9-1.1) | .82 | 52.0 (7.7) | 1.1 (1.0-1.2) | .20 | .41 | |
Body Mass Index (BMI) | 29.1 (6.3) | 31.0 (5.7) | 1.0 (0.9-1.2) | .38 | 26.3 (4.2) | 0.9 (0.8-1.1) | .20 | .27 | |
Glycosylated hemoglobin (HbA1c) | 8.9 (2.5)a | 7.2 (1.6) | 0.7 (0.4-1.0) | .07 | 8.5 (3.0)b | 0.9 (0.7-1.3) | .71 | .19 | |
iDATcusefulness score (0-9) | 6.6 (1.6) | 7.2 (1.1) | 1.4 (0.8-2.3) | .21 | 7.4 (1.2) | 1.5 (0.9-2.7) | .15 | .20 | |
Healthy diet score (0-9) | 4.6 (1.9) | 5.2 (1.5) | 1.2 (0.8-1.7) | .39 | 6.1 (2.1) | 1.6 (1.0-2.5) | .045 | .11 | |
Diet motivation score (0-9) | 7.1 (1.6) | 7.7 (0.9) | 1.4 (0.8-2.3) | .25 | 8.3 (1.0) | 2.1 (1.0-4.5) | .055 | .10 | |
|
.32 | ||||||||
|
Few times or none per year | 29 (44%) | 4 (40%) | 0.6 (0.1-2.7) | .49 | 1 (13%) | 0.1 (0.1-1.1) | .06 |
|
|
1-4 times per month | 20 (30%) | 2 (20%) | 0.4 (0.1-2.6) | .36 | 2 (25%) | 0.3 (0.1-2.0) | .23 |
|
|
More than once a week | 17 (26%) | 4 (40%) | Ref |
|
5 (62%) | Ref |
|
|
Exercise motivation score (0-9) | 6.4 (1.5) | 7.4 (1.3) | 1.8 (1.0-3.1) | .049 | 8.3 (0.9) | 3.9 (1.6-9.6) | .003 | .004 | |
DES-SFd(1-5) | 4.0 (0.5) | 4.4 (0.3) | 6.6 (1.4-29.8) | .02 | 4.3 (0.5) | 4.0 (0.8-19.4) | .09 | .02 |
an=56, not all patients had HbA1clevels at baseline. bn=7, not all patients had HbA1clevels at baseline.
ciDAT: interactive Diet and Activity Tracker
dDES-SF: Diabetes Empowerment Scale-Short Form
Multivariate polytomous logistic stepwise regressionafor baseline predictors of iDATbapp use trajectory group membership with Minimal Users group as reference category.
|
Intermittent-Waning Users |
Consistent Users |
Overall |
||
|
OR (95% CI) |
|
OR (95% CI) |
|
|
Gender (female) | 1.21 (0.28-5.20) | .80 | 19.55 (1.78-215.42) | .02 | .052 |
Exercise Motivation Score (0-9) | 1.82 (1.00-3.32) | .05 | 4.89 (1.80-13.28) | .002 | .003 |
aSLE (Significance Level to Enter)=SLR (Significance Level to Remove)=0.20.
biDAT: interactive Diet and Activity Tracker
To our knowledge, this is the first study to apply LCGM to delineate trajectories of app usage. We were able to distinguish usage patterns of a caloric-monitoring mobile health app into three latent trajectory groups: Minimal (76.8%), Intermittent-Waning (11.9%), and Consistent Users (9.5%). While a majority of patients did not use or rarely used the app, about 20% used the app, with close to 10% using the app on a regular basis during the 8-week post-enrollment period. The adjusted odds of being a Consistent User, as opposed to a Minimal User, were significantly higher for females and for subjects with higher exercise motivation scores at baseline. The adjusted odds of being an Intermittent-Waning User were also significantly higher for those with higher exercise motivation scores at baseline. The application of LCGM allowed us to delineate distinct trajectories of iDAT app usage and then identify predictors of specific patterns of app use.
There is strong evidence that good self-management in the chronic care of diabetes leads to better outcomes of the condition [
Our study protocol initially aimed at enrolling newly diagnosed diabetes patients, defined as patients in their first year following a diagnosis of diabetes. We felt these patients would benefit most from using the iDAT app, since they would likely be learning new diets and making lifestyle changes. In addition, there have been few studies focused on patient self-motivation in newly diagnosed diabetes [
The younger relative age of our study cohort could also be attributed in part to the larger proportion of the younger generation owning mobile phones or being familiar with app usage. The larger representation of Indians and Malays among our study participants was consistent with the demographic profile of patients with diabetes in Singapore’s multi-ethnic population [
In appraising patient clinical characteristics, it was not surprising that average BMI in our study was in the “high risk category”. Obesity is a well-known risk factor for diabetes and urbanized Singapore has a rising obesity trend [
While the medical literature has not been clear in reflecting the differences in app usage between genders, this topic has been thoroughly analyzed in marketing research studies so that app development could be directed toward a targeted audience. Their results have shown that, while well-known and popular apps like Facebook and Twitter have equal gender usage, there are differences in the type of apps that males and females download or use [
Our findings also showed that patients with higher exercise motivation scores had greater app usage. There are many barriers to initiating or increasing an exercise routine, so patients who indicate higher exercise motivation may be more determined to take active steps toward improving their diabetes, including more diligent use of the app.
Implementation of this study in a primary care environment void of external pressure or add-on facilitation such as regular reminders, reimbursements, or financial incentives for participants to use the iDAT app, underpins its strength. The study provides insight into the potential of a typical mobile phone app to reach out to a target group of users in a patient population. We believe our results provide a good indication of the extent and pattern of use of this caloric-measuring app based largely on self-motivation, in a naturalistic “real-world” setting.
The analysis was limited by the fact that the app database could only provide information on app usage on a weekly basis. This limitation was considered during the study design process and was accepted on the basis of what we felt were realistic expectations for participant compliance and diligence in entering data. A database with daily app input would likely have enabled a more detailed picture of usage patterns, assuming adequate participant compliance for daily data entry.
The study has practical implications and applications. Health care providers who recommend health-related apps alongside diet and exercise instructions should be aware that only 2 in 10 are likely to use the apps and only 1 in 10 is likely to be a consistent user. Males and those with lower motivation for exercise are less likely to be frequent users of such apps. Further research is needed to understand the user’s psychological construct in the three trajectory groups, which will influence their app adoption. The design, features, and functionalities of the respective app are other potential factors that can facilitate or hinder the user’s engagement with the app and this requires further investigation.
This inaugural study using LCGM as the modality of analysis is limited by the relatively small study sample and short length of observation. However, the information gathered, especially the variations in uptake of the app across the three trajectory groups will inform the design and sample size estimation of a future study to determine the effectiveness of a caloric-measuring mobile phone app on clinical outcomes among users with diabetes.
Our successful, novel application of the statistical method, LCGM, provides insightful analysis of longitudinal data to determine app utility among a target population. For selected patients with diabetes, the iDAT app can serve as an adjunct tool to facilitate lifestyle changes in conjunction with the usual modality of counselling.
Bayesian information criterion
body mass index
Centralized Institutional Review Board
Diabetes Empowerment Scale-Short Form
Global Positioning System
glycosylated hemoglobin
Health Promotion Board
interactive Diet and Activity Tracker
Apple operating system
latent-class growth modeling
significance level to enter
significance level to remove
The Singapore Health Promotion Board provided the iDAT app backend usage data. The Michigan Diabetes Research and Training Center provided use of the DEF-SF, a project supported by Grant Number P30DK092926 (MCDTR) from the National Institute of Diabetes and Digestive and Kidney Diseases.
None declared.