Abstract
Mobile health (mHealth) technologies and applications are becoming more and more accessible. The increased prevalence of wearable and embeddable sensors has opened up new opportunities to collect health data continuously outside of the clinical environment. Meanwhile, wearable devices and smartphone health apps are useful to address the issues of health disparities and inequities. This study aims to identify different characteristics of individuals who use different mHealth technologies (wearable devices and smartphone apps) and explore the effectiveness and patterns of mHealth for impacting physical activities. We found that social determinants are significantly associated with the use of mHealth; mHealth is helping people to exercise more regularly and for a longer time. Smartphone app users are older while wearable device users are younger. Health disparities exist in mHealth use and physical activity level. Social determinants like education and income are associated with mHealth use and physical activity. The integration of passively-tracked patient-generated health data (PGHD) holds promise in increasing physical activities. Physical activity interventions that comprise wearable devices and smartphone apps may be more beneficial, since health goals, data visualization, real-time support and feedback, results interpretation, and group education could be embedded in the integrated “smart system”. These findings may be useful for stakeholders like wearable device and smartphone app companies, researchers, health care workers, and public health practitioners, who should work together to design and develop “precision mobile health” products with higher personalized and participatory levels, thus improving the population health.
Introduction
Wearable device
With the development of biomedical sensors and information technologies, there is an increasing number of wearable or portable devices that are available for users to monitor various parameters, such as physiological indicators, sedentary time, energy expenditure, etc.1,2 With these functions, wearable devices can not only collect and record health-related data but also assist individuals to maintain an active lifestyle. Studies have shown that wearable devices have the potential to increase the level of physical activity in diverse settings even as a stand-alone technique.3 Wearable devices have also been evolved and connected with smartphones on which all the data can be stored, displayed, and sent to the health care providers. Given the data collected from real-time tracking along with the advanced artificial intelligence (AI) algorithms,4 they can generate more specific, customized, and tailored responses back to users. These interactive means that increase physical activity levels and help maintain a structured lifestyle are increasingly promising in the information era. Also, most wearable devices are light and smart, without introducing technostress additional and burden to users.
Smartphone Apps
Smartphone apps have become an emerging technology to change people’s lifestyles and behaviors, such as motivating physical activities. But there have been contrasting findings in the past studies. For instance, Flores MG et al5 found that there was no significant association between smartphone app usage and physical activity and weight loss. Another study conducted by Coughlin SS et al6 found that there was a modest significant association between smartphone app usage and increased physical activity. A systematic review showed some evidence that the positive effects of smartphone apps were more significant over a short period of time7 (e.g. 3 months). That is, the intervention effect peaked at the beginning but dwindled as time went by. This phenomenon may be caused by the declined engagement with smartphone apps interventions,8 especially for those without supportive accountability9 or poor human-machine interaction design.
Physical activity
Physical inactivity is one of the top predictors that are related to acute or chronic disease and even mortality.10 Without sufficient physical activities, the prevalence of non-communicable diseases is becoming more prevalent.11 Declined physical activity level is associated with an increased likelihood of chronic diseases. In addition, sedentary behavior such as sitting for a long time is also associated with poor health status, e.g., poor functional independence. Regular activity can improve cardiorespiratory and muscular fitness, and mental and behavioral health, etc.12,13 The World Health Organization has proposed global recommendations on physical activity, which suggest adults participating in at least 75 minutes of vigorous physical activity or 150 minutes of moderate physical activity per week.14 However, over 25% of adults over the world have not achieved the bottom line of the recommendations.15 Emerging health technologies such as mobile health technologies can be effective strategies to improve the level of regular physical activity, thus achieving this essential global public health goal.
This study has two primary aims. The first is to describe the different characteristics of individuals who owned and used different mHealth technologies (wearable devices and smartphone apps). We consider social determinants and sociodemographic factors such as age, gender, educational level, income, race/ethnicity, marital status, BMI, etc. as important covariates. The second aim is to explore the effectiveness and patterns of mHealth for impacting physical activities controlling for significant factors of social determinants identified in Aim 1.
Methods
Mobile health user groups
The Health Information National Trends Survey (HINTS) is a nationally-representative survey that is administered every year by the National Cancer Institute, which provides a comprehensive assessment of the American public’s current access to and use of health information. The HINTS target population is civilian, non-institutionalized adults aged 18 or older living in the United States. The most recent version of HINTS administration is HINTS 5. The 1st round of data collection for HINTS 5 (Cycle 1) was conducted from January 25 through May 5, 2017; the 2nd round (Cycle 2) was conducted from January 26 through May 2, 2018; the 3rd round (Cycle 3) was conducted from January 22 to April 30, 2019.16 We combined data of 2017 (Cycle 1, N=3,285), 2018 (Cycle 2, N=3,504) and 2019 (Cycle 3, N=5,438) from the HINTS.17
Since participants who own wearable devices and smartphone apps may not use them, we just focused on usage rather than ownership. According to whether participants used health-related apps (“Tablet_AchieveGoal”) or electronic device to track/monitor health (“OtherDevTrackHealth”), we defined four user groups: (1) participants didn’t use apps or devices (No app and No device); (2) participants only used apps (App only); (3) participants only used health tracking devices (Device only); (4) participants used both apps and devices (App and Device).
“OtherDevTrackHealth” in 2017 and 2018 was stated as “Other than a tablet or smartphone, have you used an electronic device to monitor or track your health within the last 12 months? Examples include Fitbit, blood glucose meters, and blood pressure monitors.” But in 2019, this question was assumed to be re-stated separately as two mutually exclusive questions: “WearableDevTrackHealth” (In the past 12 months, have you used an electronic wearable device to monitor or track your health or activity? For example, a Fitbit, Apple Watch, or Garmin Vivofit) and “OtherDevTrackHealth2” (In the last 12 months, have you used an electronic medical device to monitor or track your health? For example a glucometer or digital blood pressure device). Only when participants answered “No” in the former question, would they be presented with the latter. Based on the semantic analysis, we assumed “WearableDevTrackHealth” + “OtherDevTrackHealth2” was equivalent to “OtherDevTrackHealth” in Cycle 1 and Cycle 2.
Physical activities related outcomes
In this study, we studied how mhealth (health-related apps and/or health tracking devices) was associated with physical activities. We selected four physical activity outcomes for analysis as they were the common physical activity outcomes across 2017 to 2019: (1) number of days that a participant did exercises of moderate-intensity; (2) minutes for exercises of moderate-intensity per day; (3) minutes for exercises of moderate-intensity per week; (4) number of days that a participant did strength training. Outcome 3 was calculated by multiplying outcomes 1 and 2. Notably, for the Cycle 1 questionnaire in 2017, outcome 2 was listed separately as hours and minutes of exercises of moderate-intensity and thus we recalculated outcome 2 in Cycle 1 using “hour × 60 + minute” so that it could be consistent with outcome 2 in 2018 and 2019.
Sample characteristics
We included sociodemographic characteristics, such as age, self-reported gender, race and ethnicity, income level, education, marital status, and Body Mass Index (BMI) into the analyses. For race and ethnicity, we combined these two variables and defined “Non-Hispanic white”, “Non-Hispanic black”, “Hispanic”, “others”. Income level was recategorized as ≤ 20,000, 20,000~35,000, 35,000~50,000, 50,000~75,000, and > 75000 USD; education level was re-categorized as less than college (including the post high-school training), some college, college graduate and post-graduate degree; and marital status was re-categorized as single, in marriage or marriage-like relationship, divorced or widow or separated. For BMI, we considered the continuous scale and also the categorized BMI status as “underweight” (BMI > 18.5), “normal weight” (BMI 18.5~25), “overweight” (BMI 25~30), “obese” (BMI > 30).
Statistical analyses
As we combined data from 2017 to 2019, we first examined whether the sample characteristics were distributed differently among three years (Table 1), in which we used analysis of variance (ANOVA) for continuous variables (e.g., age and BMI) and Pearson’s Chi-Square test for categorical variables. For the afterward analysis, we used inverse probability weighting (IPW)18 to adjust for these potential issues of different distribution across years. In the IPW approach, a logistic regression model for the year with all sample characteristics involved as predictors was used to generate corresponding sample weights. The year 2019 was set as the reference level. By applying the resulting sample weights, data from 2017and 2018 could resemble data from 2019.
Table 1.
Characteristics of the population by year
| Demographic characteristics | 2017 (n=3285) | 2018 (n=3504) | 2019 (n=5438) | Pooled (n=12227) | P-value |
|---|---|---|---|---|---|
| Age (years), mean ± SD (n) | 56.34±16.14 | 57.02±16.73 | 56.93±16.89 | 56.80±16.65 | 0.19 |
| Missing (n) | 139 | 87 | 154 | 380 | |
| Gender (n) | |||||
| Male, n (%) | 1254(41.28) | 1310(40.65) | 2108(42.74) | 4672(41.74) | 0.14 |
| Female, n (%) | 1784(58.72) | 1913(59.35) | 2824(57.26) | 6521(58.26) | |
| Missing (n) | 247 | 281 | 506 | 1034 | |
| Race/Ethnicity, (n) | |||||
| Non-Hispanic White, n (%) | 1929(65.32) | 2046 (64.93) | 3129 (64.52) | 7104(64.85) | 0.97 |
| Non-Hispanic Black, n (%) | 427(14.46) | 457(14.50) | 701 (14.45) | 1585(14.47) | |
| Hispanic, n (%) | 418(14.16) | 458(14.54) | 730(15.05) | 1606(14.66) | |
| Other, n (%) | 179(6.06) | 190(6.03) | 290 (5.98) | 659(6.02) | |
| Missing (n) | 332 | 353 | 588 | 1273 | |
| BMI (kg/m2), mean ± SD (n) | 28.48±6.65 | 28.51±6.98 | 28.50 ±6.81 | 28.50± 6.82 | 0.98 |
| BMI Categories | |||||
| BMI ≤ 18.5, n (%) | 43 (1.35) | 70 (2.06) | 97 (1.83) | 210(1.77) | 0.35 |
| 18.5 < BMI ≤ 25, n (%) | 1007(31.57) | 1027(30.15) | 1614(30.53) | 3648(30.70) | |
| 25 < BMI ≤ 30, n (%) | 1081(33.89) | 1184(34.76) | 1827(34.56) | 4092(34.44) | |
| BMI > 30, n (%) | 1059(33.20) | 1125(33.03) | 1749(33.08) | 3933(33.10) | |
| Missing (n) | 134 | 133 | 231 | 498 | |
| Education | |||||
| High school or lower, n (%) | 1061(33.35) | 1135(32.87) | 1672(31.66) | 3868(32.46) | 0.35 |
| Some college, n (%) | 714(22.45) | 810(23.46) | 1199(22.70) | 2723(22.85) | |
| College graduate, n (%) | 828(26.03) | 910(26.35) | 1402(26.55) | 3140(26.35) | |
| Post-graduate degree, n (%) | 578(18.17) | 598(17.32) | 1008(19.09) | 2184(18.33) | |
| Missing (n) | 104 | 51 | 257 | 312 | |
| Income | |||||
| Income ≤ $ 20K, n (%) | 559(18.87) | 579(18.76) | 904(18.84) | 2042(18.83) | 0.66 |
| $ 20K< Income ≤ $ 35K, n (%) | 423(14.28) | 428(13.86) | 614(12.80) | 1465(13.51) | |
| $ 35K< Income ≤ $ 50K, n (%) | 386(13.03) | 404(13.09) | 630(13.13) | 1420(13.09) | |
| $ 50K< Income ≤ $ 75K, n (%) | 530(17.89) | 567(18.37) | 848(17.67) | 1945(17.93) | |
| $ 75K< Income, n (%) | 1064(35.92) | 1109(35.92) | 1802(37.56) | 3975(36.65) | |
| Missing (n) | 323 | 417 | 640 | 1380 | |
| Marital status, (n) | |||||
| Single, n (%) | 512(16.17) | 605(17.54) | 882(16.75) | 1999(16.82) | <0.01 |
| Married/Marriage-like, n (%) | 1751(55.31) | 1747(50.65) | 2847(54.05) | 6345(53.40) | |
| Separated/divorced/widowed, n (%) | 903(28.52) | 1097(31.81) | 1538(29.20) | 3538(29.78) | |
| Missing (n) | 119 | 55 | 171 | 345 | |
Abbreviations: BMI, Body mass index.
The following analyses aimed to explore participants’ usage of mHealth technologies and thus we excluded 816 participants whose group membership couldn’t be verified. We first examined whether each sociodemographic characteristic was distributed differently across four mHealth user groups (Table 2). After the sample size reduction (N=11411), for all the missing variables involved in this analysis, we applied multiple imputations (MI) for 20 iterations to guarantee sufficient imputation efficiency.19 Especially for categorical characteristics, we specified a discriminant model to impute the categorical variables. Multiple imputations and the post-imputation analyses were implemented using PROC MI and PROC MIANALYZE in SAS. As we applied the imputation approach, we also did sensitivity analyses on the original complete data and there were no significant conflicts in the findings between the complete data and imputed data analyses.
Table 2.
Characteristics of the four mHealth user groups
| Demographic characteristics | No App/Device (N=5154) | App only (N=2224) | Device only (N=1518) | App + Device (N=2515) | P-value | |
|---|---|---|---|---|---|---|
| Age (years), mean ± SD | 59.11±16.18 | 63.71±14.17 | 46.26±15.04 | 49.26±14.84 | <0.01 | |
| Missing (n) | 188 | 70 | 24 | 36 | ||
| Gender, n (%) | <0.01 | |||||
| Male | 2031(43.55) | 976(47.84) | 522(36.45) | 880(36.99) | ||
| Female | 2633(56.45) | 1064(52.16) | 910(63.55) | 1499(63.01) | ||
| Missing (n) | 490 | 184 | 86 | 136 | ||
| Race/Ethnicity, n (%) | <0.01 | |||||
| Non-Hispanic White | 3054(66.59) | 1348(68.60) | 863 (60.35) | 1524(64.14) | ||
| Non-Hispanic Black | 615 (13.41) | 262 (13.33) | 209 (14.62) | 357(15.03) | ||
| Hispanic | 660 (14.39) | 264(13.44) | 259 (18.11) | 305 (12.84) | ||
| Other | 257(5.60) | 91 (4.63) | 99(6.92) | 190(8.00) | ||
| Missing (n) | 568 | 259 | 88 | 139 | ||
| BMI (kg/m2), mean ± SD | 28.10±6.82 | 29.28±6.86 | 28.04±6.64 | 29.00±6.66 | <0.01 | |
| BMI Categories, n (%) | <0.01 | |||||
| BMI ≤ 18.5 | 102 (2.04) | 32(1.48) | 22 (1.48) | 27 (1.10) | ||
| 18.5 < BMI ≤ 25 | 1670 (33.39) | 549(25.45) | 513(34.45) | 684(27.77) | ||
| 25 < BMI ≤ 30 | 1701(34.01) | 745(34.54) | 517(34.72) | 865(35.12) | ||
| BMI > 30 | 1528(30.55) | 831(38.53) | 437(29.35) | 887 (36.01) | ||
| Missing (n) | 153 | 67 | 29 | 52 | ||
| Education, n (%) | <0.01 | |||||
| High school or lower, | 1919 (38.31) | 747(34.46) | 330(22.06) | 416(16.77) | ||
| Some college | 1138(22.72) | 550(25.37) | 354(23.66) | 530(21.37) | ||
| College graduate | 1178(23.52) | 510(23.52) | 477(31.89) | 876(35.32) | ||
| Post-graduate degree | 774(15.45) | 361(16.65) | 335(22.39) | 658(26.53) | ||
| Missing (n) | 145 | 56 | 22 | 35 | ||
| Income, n (%) | <0.01 | |||||
| Income ≤ $ 20K | 1022(22.72) | 379(19.36) | 191 (13.65) | 190(8.06) | ||
| $ 20K< Income ≤ $ 35K | 733(16.30) | 284(14.50) | 147(10.51) | 178(7.56) | ||
| $ 35K< Income ≤ $ 50K | 630(14.01) | 278(14.20) | 169(12.08) | 251(10.65) | ||
| $ 50K< Income ≤ $ 75K | 803(17.85) | 359(18.34) | 273(19.51) | 434(18.42) | ||
| $ 75K< Income | 1310(29.12) | 658(33.61) | 619(44.25) | 1303(55.31) | ||
| Missing (n) | 656 | 266 | 119 | 159 | ||
| Marital status, n (%) | <0.01 | |||||
| Single | 848(16.98) | 268(12.37) | 326(21.88) | 429(17.34) | ||
| Married/Marriage-like | 2436(48.79) | 1188(54.85) | 882(59.19) | 1575(63.66) | ||
| Separated/divorced/widowed | 1709(34.23) | 710(32.78) | 282(18.93) | 470(19.00) | ||
| Missing (n) | 161 | 58 | 28 | 41 |
Abbreviations: BMI, Body mass index.
Following imputation, the weighted versions of Poisson models (Table 3) for physical activities and logistic model (Table 4) of predicting wearable device usage were implemented separately by using imputed datasets with sample weights generated from the IPW approach above. We used Rubin’s rule20 to combine these post-imputation analysis results for valid inference because naively averaging the estimates can bring in unreasonably small standard errors that come from the non-missing replicates across imputed datasets. In addition, since outcome 2 was input by self-entry, unrealistic extreme values in outcome 2 and 3 were detected in data screening. We removed those outliers if the input values were greater than the upper 95% quantiles (120 minutes per day for outcome 2 or 630 minutes per week for outcome 3).
Table 3.
Poisson models for physical activities by mHealth user groups, controlling for social determinants
| Demographic characteristics | Days for Moderate Exercises in a Week | Minutes for Moderate Exercises per week | Minutes for Moderate Exercises per week | Days for Strength Training in a week | ||||
|---|---|---|---|---|---|---|---|---|
| Est. | 95% CI | Est. | 95% CI | Est. | 95% CI | Est. | 95% CI | |
| Age | 1.00 | (1.00, 1.00) | 1.00 | (1.00, 1.00) | 1.00 | (1.00, 1.00) | 1.00 | (1.00, 1.00) |
| Gender | ||||||||
| Female vs. Male | 0.85 | (0.84, 0.88) | 0.84 | (0.84, 0.86) | 0.80 | (0.79, 0.82) | 0.75 | (0.73, 0.78) |
| Race and Ethnicity | ||||||||
| Non-Hispanic White | Ref | Ref | Ref | Ref | ||||
| Non-Hispanic Black | 0.99 | (0.95, 1.02) | 1.01 | (0.99, 1.03) | 1.01 | (0.98, 1.03) | 1.10 | (1.04, 1.16) |
| Hispanic | 0.96 | (0.93, 1.00) | 1.05 | (1.03, 1.07) | 1.02 | (1.00, 1.04) | 1.07 | (1.02, 1.13) |
| Other | 0.87 | (0.83, 0.91) | 0.99 | (0.87, 0.92) | 0.85 | (0.84, 0.86) | 0.90 | (0.84, 0.97) |
| Education | ||||||||
| High school or lower | Ref | Ref | Ref | Ref | ||||
| Some college | 1.08 | (1.04, 1.12) | 1.12 | (1.08, 1.14) | 1.12 | (1.09, 1.14) | 1.08 | (1.02, 1.14) |
| College graduate | 1.12 | (1.08, 1.15) | 1.15 | (1.13, 1.16) | 1.16 | (1.14, 1.19) | 1.16 | (1.11, 1.22) |
| Post-graduate degree | 1.15 | (1.11, 1.20) | 1.16 | (1.14, 1.19) | 1.20 | (1.17, 1.21) | 1.22 | (1.15, 1.28) |
| Income | ||||||||
| Income ≤ $ 20K | Ref | Ref | Ref | Ref | ||||
| $ 20K< Income ≤ $ 35K | 0.99 | (0.94, 1.04) | 1.04 | (1.00, 1.08) | 1.03 | (0.99, 1.07) | 0.90 | (0.83, 0.97) |
| $ 35K< Income ≤ $ 50K | 1.04 | (0.99, 1.09) | 1.16 | (1.13, 1.21) | 1.11 | (1.07, 1.14) | 1.02 | (0.94, 1.09) |
| $ 50K< Income ≤ $ 75K | 1.12 | (1.07, 1.17) | 1.20 | (1.15, 1.23) | 1.21 | (1.16, 1.25) | 1.08 | (1.01, 1.16) |
| $ 75K<Income | 1.13 | (1.08, 1.19) | 1.26 | (1.22, 1.30) | 1.30 | (1.26, 1.34) | 1.15 | (1.06, 1.23) |
| Marital status | ||||||||
| Single | Ref | Ref | Ref | Ref | ||||
| Married/Marriage-like | 0.94 | (0.90, 0.98) | 0.93 | (0.91, 0.95) | 0.89 | (0.87, 0.90) | 0.89 | (0.84, 0.93) |
| Separated/divorced/widowed | 0.94 | (0.91, 0.98) | 0.97 | (0.95, 0.99) | 0.92 | (0.90, 0.94) | 0.99 | (0.93, 1.05) |
| BMI Status | ||||||||
| BMI ≤ 18.5 | 0.93 | (0.85, 1.02) | 0.85 | (0.80, 0.91) | 0.84 | (0.77, 0.91) | 1.09 | (0.96, 1.26) |
| 18.5 < BMI ≤ 25 (ref) | Ref | Ref | Ref | Ref | ||||
| 25 < BMI ≤ 30 | 0.89 | (0.87, 0.91) | 0.92 | (0.91, 0.93) | 0.87 | (0.86, 0.88) | 0.85 | (0.82, 0.89) |
| BMI > 30 | 0.67 | (0.65, 0.69) | 0.74 | (0.73, 0.75) | 0.64 | (0.64, 0.66) | 0.71 | (0.68, 0.74) |
| Groups | ||||||||
| No App/Wearable Device | Ref | Ref | Ref | Ref | ||||
| App only | 1.04 | (1.01, 1.07) | 1.03 | (1.02, 1.05) | 1.04 | (1.02, 1.05) | 1.05 | (1.00, 1.11) |
| Wearable Device only | 1.15 | (1.12, 1.20) | 1.15 | (1.14, 1.17) | 1.22 | (1.20, 1.23) | 1.33 | (1.26, 1.39) |
| App and Wearable Device | 1.24 | (1.20, 1.27) | 1.22 | (1.21, 1.23) | 1.27 | (1.26, 1.28) | 1.30 | (1.23, 1.36) |
Abbreviations: Est., Estimate; CI, confidence interval; BMI, Body mass index.
Table 4.
Characteristics significantly associate with being a wearable device user
| Demographic characteristics | Est. | Odds Ratio | 95% CI |
|---|---|---|---|
| Age (years)* | -0.05 | 0.95 | (0.95,0.96) |
| Gender* | |||
| Female vs. Male | 0.39 | 1.48 | (1.35,1.62) |
| Race and Ethnicity* | |||
| Non-Hispanic White | Ref | ||
| Non-Hispanic Black | 0.29 | 1.34 | (1.18, 1.52) |
| Hispanic | 0.05 | 1.05 | (0.92, 1.20) |
| Other | 0.26 | 1.30 | (1.09, 1.55) |
| Education* | |||
| High school or lower | Ref | ||
| Some college | 0.40 | 1.50 | (1.32,1.70) |
| College graduate | 0.54 | 1.71 | (1.51,1.93) |
| Post-graduate degree | 0.72 | 2.05 | (1.78, 2.35) |
| Income* | |||
| Income ≤ $ 20K | |||
| $ 20K< Income ≤ $ 35K | 0.20 | 1.22 | (1.02,1.46) |
| $ 35K< Income ≤ $ 50K | 0.40 | 1.49 | (1.25,1.78) |
| $ 50K< Income ≤ $ 75K | 0.61 | 1.84 | (1.56,2.18) |
| $ 75K<Income | 0.91 | 2.48 | (2.11,2.92) |
| Marital status* | |||
| Single | |||
| Married/Marriage-like | 0.33 | 1.39 | (1.23,1.58) |
| Separated/divorced/widowed | 0.17 | 1.19 | (1.03,1.38) |
| BMI Status* | |||
| BMI ≤ 18.5 (Underweight) | -0.39 | 0.68 | (0.47,0.99) |
| 18.5 < BMI ≤ 25 (Normal) | Ref | ||
| 25 < BMI ≤ 30 (Overweight) | 0.29 | 1.33 | (1.19,1.48) |
| BMI > 30, (Obese) | 0.36 | 1.43 | (1.28,1.60) |
Abbreviations: Est, estimate; CI, confidence interval; BMI, Body mass index. *: P < 0.01
Data were managed and analyzed using SAS 9.4 (SAS Institute., Cary, NC). Categorical variables were compared using the chi-square test, and multivariable analysis used logistic regression. The statistical significance threshold was set at less than .05 for 2-sided tests.
Results
There are 12227 respondents in the combined HINTS sample. Table 1 presents the respondents’ sociodemographic and health-related characteristics from Cycle 1 to 3. In this pooled population, most respondents were female (58.26%), with a mean age at 56.8 years old, non-Hispanic whites (64.85%), overweight (34.44%), having high school or lower degree (32.46%), married or marriage-like (53.40%), having a higher annual income (36.65%). The mean BMI is 28.50, which indicates that most respondents need a healthier lifestyle. There are no significant differences among the three cycles of HINTS across the three years except for marital status (p<0.01).
In order to explore participants’ usage of mHealth technologies, Table 2 presents the characteristics of the four mHealth user groups. There were 2515 respondents (22.0%) who reported having used both wearable devices and health apps, 2224 (19.5%) only used health apps, and 1518 (13.3%) only used wearable devices. More than one-third of respondents (45.2%) neither used a wearable device or a health app. There are statistically significant differences across the four user groups (p<0.01).
In the “No app and No device” group, their mean age was 59.11 years old; most of them were female (56.45%), non-Hispanic whites (66.59%), and married or marriage-like (48.79%); BMI was overweight (34.01%); having a high school or lower degree (38.31%) and higher annual income (29.12%); the mean BMI was 28.10. The “App only” group was the oldest group with a mean age of 63.71 years old; most of them were female (52.16%), non-Hispanic whites (68.60%), and married or marriage-like (54.85%); BMI was obese (38.53%); having high school or lower degree (34.46%) and a higher annual income (33.61%); the mean BMI was 29.28. In the “Device only” group, their mean age was 46.26 years old; most participants were female (63.55%), non-Hispanic whites (60.35%), and married or marriage-like (59.19%); BMI was overweight (34.72%); having a college or graduate degree (31.89%) and a higher annual income (44.25%); the mean BMI was 28.04. Within the “App and Device” group, their mean age was 49.26 years old; most of them were female (63.01%), non-Hispanic whites (64.14%), and married or marriage-like (63.66%); BMI was obese (36.01%); having a college or graduate degree (35.32%) and a higher annual income (55.31%); the mean BMI was 29.00.
Among the four mHealth user groups, respondents in the “Device only” group were the youngest and healthiest, their age and BMI were the lowest. Respondents who used wearable devices were more likely to have higher education levels since most respondents in both the “Device only” group and “App and Device” group had at least a college degree. Respondents who used apps were more likely to have higher BMI. Most respondents in both the “App only” group and “App and Device” group were obese. As shown in Table 2, there were significant differences across the four user groups in terms of age, gender, race/ethnicity, BMI, education, income, marital status when controlling for all the other variables.
Poisson models were employed to characterize the physical activity outcomes and compare the four user groups’ performance. Except for age that was not statistically significant in the model for days per week in moderate exercises, all of the other covariates in all four models were found to be significant by F-tests. Poisson model estimates shown in Table 3 are in exponential scale, which can be interpreted as the multiple of the estimate of the reference level. For the moderate exercise outcomes, the “App and Device” group performed the best in terms of frequency and length of time spent. The days per week doing moderate exercise for “App and Device” users are 1.24 times the days for “No app and No device” participants. The length of time spent in moderate exercises per day of “App and Device” users is 1.22 times the length of “No app and No device” participants. The length of time for moderate exercises per week of “App and Device” users is 1.27 times the length of “No app and No device” participants. For the frequency of strength training, the “Device only” group performed the best. Days for strength training for “Device only” users are 1.33 times the days for “No app and No device” group users. We speculated that mHealth technologies especially wearable devices helped individuals to do physical activities more regularly and maintain sustainability.
In general, induvial using wearable devices are performing better regarding physical activity. Smartphone apps seem to be less effective in promoting physical activity. Regarding the effect of social determinants on physical activity, there are significant differences across the four race/ethnicity groups in terms of exercise magnitude and exercise pattern. For example, white respondents had a higher frequency of doing moderate exercise but black respondents spent a longer time for it each day. We found that days of doing moderate exercises in the white population were 0.99 times the days of the black population; the minutes spent for moderate exercises in the black population were 1.01 times the length of white respondents per day and 1.01 times in a week span. Hispanic population exercised in moderate intensity with the longest time for both per day and per week. The black population had the highest frequency of strength training, and it was 1.10 times the days of the white population. The minority population performed better than the white population on physical activity. Other than race/ethnicity, we found that health disparities exist in other social determinants of health.21 Users with higher education and higher income level were doing more exercise and strength training. Participants maintaining normal weight (18.5 < BMI ≤ 25) had the highest frequency and longest time for moderate exercises. However, for underweight participants (BMI ≤ 18.5), they did more strength training to gain weight. Therefore, participants with different social determinants had different physical activity preferences (moderate exercise vs. strength training) and patterns (multiple days for short time vs. one day for a long time). Functions of smartphone apps and wearable devices could be further improved to fit in a person’s exercise style.
Since we found that people who used wearable devices had a more significantly higher level of physical activity, we further examined the patterns and characteristics of wearable device usage. We divided the sample into two groups based on whether the respondents had used wearable devices. Table 4 demonstrates the characteristics that are significantly associated with being a wearable device user. We found that females (OR 1.48, 95% CI 1.35-1.62) and black respondents (OR 1.34, 95% CI 1.18-1.65) were more likely to use wearable devices, which aligns with the finding from Table 2. Respondents with higher education and higher income were more likely to use wearable devices. Respondents who were married or near marriage (OR 1.39, 95% CI 1.23-1.58) were more likely to use wearable devices compared to those of another marital status, which was probably because the household financial burden for them was lower and the cost of using device could also be lower due to the family sharing. Interestingly, respondents who had higher BMIs were more likely to use wearable devices, however, since we don’t have follow-up data, we could not specify whether wearable devices helped them to reduce the BMI level.
Discussion
Participants who did not use mHealth technologies and applications made up nearly half of the population. There were substantial differences among mHealth users in terms of social determinants. These distinctive characteristics may be useful to tailor health interventions accordingly. Since participants who own wearable devices and smartphone apps may not use them, we just focused on usage rather than ownership. The perceived utility of wearable devices and smartphone apps may facilitate the design and development of personalized features and products. Our findings demonstrate that smartphone app users are older while wearable device users are younger. In terms of social determinants, health disparities exist in mHealth usage and physical activity level. Social determinants like education and income are associated with mHealth usage and physical activity since poor social determinants result in less likely owning such digital tools. Comprising wearable devices and smartphone apps may be more beneficial if health goals, data visualization, real-time support and feedback, results interpretation, and group education could be embedded into an integrated “smart system”.
The increased prevalence of wearable devices and smartphone apps has opened up new opportunities to collect health data continuously outside of the clinical environment.22 Although people have long been journaling and logging their activities as a means of managing various aspects of health, the influx of low-cost wearables make it possible to track multiple measures (e.g. heart rate, step count) passively in real-time and at high-frequency intervals so that self-report bias and recall bias can be alleviated. The health-related data that is created, recorded, or collected by people with the intention of introducing the information to the clinic to help address their health concerns, has been termed patient-generated health data (PGHD).23 This data can be captured with little burden on the users, enabling nearly continuous data streams over extended periods of observation. The use of PGHD holds promise in increasing physical activities. Studies have shown that the effects of most behavioral changes are short-term.24 As trackers, wearable devices can be an effective tool to generate PGHD in real time and in a long time course.25 The integration of passively-tracked PGHD also affords clinicians more evidence to diagnose and treat illnesses and supports communication with patients for improved treatment adherence and health outcomes. Integrating these data into EHR will assist healthcare professionals in monitoring individuals’ health status and providing timely support without introducing large resource expenditures. Accompanied with smartphone apps, wearable devices can better motivate and manage individual health.26 A systematic review found that when combined with other intervention techniques, such as education or counseling, the improvement of physical activity participation is greater than wearable devices alone.27 In addition, incentives may be useful to motivate physical activity adherence.28 These features could be added through connecting the smartphone apps. Furthermore, wearable devices have a greater effect on decreasing the time of sedentary behaviors. Human-machine design and interactions should be utilized to develop strategies which can boost user engagement and aid sustainable effectiveness.29
Future research may be directed to understanding the association between the intervention effect size and time course, and studying which features of mhealth can lead to sustainable engagement. Additional efforts should be made to expand the time course of the positive effects. Specifically, software engineers should work with researchers to design and implement features that can maintain user engagement using reimbursement, rewards, competition among friends, etc. Personalized and customized design, tailored information, and regular assessment are also useful strategies to engage users. That being said, smartphone apps should not incorporate too many functions that target multiple health behavior changes, because this kind of design will introduce technostress and pressure to users. Focusing on specific health behavior is more likely to lead to longer-term effects.
We recommend that future iterations of HINTS may consider categorizing the types and functions of wearable devices, medical devices, and smartphone apps in the survey at a more granular level. Because motivations for using these technologies and tools may vary by the specific functions and services, this information would be useful to further investigate the patterns among mobile health, social determinants, and physical activity.
Limitations
For the analyses, we have identified and controlled for the social determinants for physical activities but we didn’t explore the interaction between user groups and these covariates and time, which together with stratification analyses could be possible steps in future analyses. We used Poisson models to handle the skewness of the, which was less intuitive for interpretation. In addition, this sample was confronted with missing data regarding physical activity outcomes and covariates. The missing data may not be missing completely at random. Those who didn’t respond to questions regarding physical activities may be less active and thus our estimates may be subject to bias. Imputing categorical variables was risky. Therefore, more advanced methods for imputation or weighting are needed to handle the missing data.
This study is the largest and most recent nationally representative study of mobile health usage across 3 cycles (2017-2019) of HINTS. However, because the survey was cross-sectional, we couldn’t examine causal inferences among variables. The survey did not specify the types and main functions of wearable devices and apps; some devices were just used for physiological indicators monitoring,1 and some apps may just focus on mental health, which may not be associated with physical activity. Meanwhile, given the limitations of the dataset, we did not have the information about mHealth usage frequency and duration. Despite these limitations, this study contributes to a better understanding of the effects and patterns among mHealth, social determinants, and physical activities. These findings may be useful for stakeholders like wearable device and smartphone apps companies, researchers, health care workers, and public health practitioners to work together to design and develop “precision mobile health” products with higher personalized and participatory levels, thus improving the population health.
Conclusion
This nationally representative cross-sectional study incorporates a wide range of participants including different age strata, race and ethnicity, and other important social determinants. We find that social determinants are significantly associated with the use of mHealth and individuals using mHealth have more regular physical activity habits. Physical activity interventions comprising wearable devices and smartphone apps can probably be promising in promoting regular physical activity. This study is a good start for further causal inference analyses between mHealth technologies and physical activity level. All the findings may have clinical and public health relevance as improved physical activity levels can contribute to the population health. Combining the advantages of wearable devices and smartphone apps would be useful to integrate PGHD to EHR and make mobile health generate more benefits to a broader population.
References
- 1.Ye J, Li N, Lu Y, Cheng J, Xu Y. A portable urine analyzer based on colorimetric detection. Analytical Methods. 2017;9(16):2464–2471. [Google Scholar]
- 2.Zhang J, Fu R, Xie L, et al. A smart device for label-free and real-time detection of gene point mutations based on the high dark phase contrast of vapor condensation. Lab on a Chip. 2015;15(19):3891–3896. doi: 10.1039/c5lc00488h. [DOI] [PubMed] [Google Scholar]
- 3.de Vries HJ, Kooiman TJ, van Ittersum MW, van Brussel M, de Groot M. Do activity monitors increase physical activity in adults with overweight or obesity? A systematic review and meta‐analysis. Obesity. 2016;24(10):2078–2091. doi: 10.1002/oby.21619. [DOI] [PubMed] [Google Scholar]
- 4.Ye J, Yao L, Shen J, Janarthanam R, Luo Y. Predicting mortality in critically ill patients with diabetes using machine learning and clinical notes. BMC Medical Informatics and Decision Making. 2020;20(11):1–7. doi: 10.1186/s12911-020-01318-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Mateo GF, Granado-Font E, Ferré-Grau C, Montaña-Carreras X. Mobile phone apps to promote weight loss and increase physical activity: a systematic review and meta-analysis. Journal of medical Internet research. 2015;17(11):e253. doi: 10.2196/jmir.4836. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Coughlin SS, Whitehead M, Sheats JQ, Mastromonico J, Smith S. A review of smartphone applications for promoting physical activity. Jacobs journal of community medicine. 2016;2(1) [PMC free article] [PubMed] [Google Scholar]
- 7.Wang Q, Egelandsdal B, Amdam GV, Almli VL, Oostindjer M. Diet and physical activity apps: perceived effectiveness by app users. JMIR mHealth and uHealth. 2016;4(2):e33. doi: 10.2196/mhealth.5114. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Schoeppe S, Alley S, Van Lippevelde W, et al. Efficacy of interventions that use apps to improve diet, physical activity and sedentary behaviour: a systematic review. International Journal of Behavioral Nutrition and Physical Activity. 2016;13(1):127. doi: 10.1186/s12966-016-0454-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Yardley L, Spring BJ, Riper H, et al. Understanding and promoting effective engagement with digital behavior change interventions. American journal of preventive medicine. 2016;51(5):833–842. doi: 10.1016/j.amepre.2016.06.015. [DOI] [PubMed] [Google Scholar]
- 10.Brown WJ, Bauman AE, Bull F, Burton NW. Development of Evidence-based Physical Activity Recommendations for Adults (18-64 years) Report prepared for the Australian Government Department of Health, August 2012. 2013. [Google Scholar]
- 11.Reiner M, Niermann C, Jekauc D, Woll A. Long-term health benefits of physical activity–a systematic review of longitudinal studies. BMC public health. 2013;13(1):1–9. doi: 10.1186/1471-2458-13-813. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Warburton DE, Nicol CW, Bredin SS. Health benefits of physical activity: the evidence. Cmaj. 2006;174(6):801–809. doi: 10.1503/cmaj.051351. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Ye J. Pediatric Mental and Behavioral Health in the Period of Quarantine and Social Distancing With COVID-19. JMIR pediatrics and parenting. 2020;3(2):e19867. doi: 10.2196/19867. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.World Health Organization. Global recommendations on physical activity for health. World Health Organization; 2010. [PubMed] [Google Scholar]
- 15.World Health Organization. Global action plan on physical activity 2018-2030: more active people for a healthier world. World Health Organization; 2019. [Google Scholar]
- 16.Finney Rutten LJ, Blake KD, Skolnick VG, Davis T, Moser RP, Hesse BW. Data resource profile: The national cancer institute’s health information national trends survey (HINTS) International journal of epidemiology. 2020;49(1):17–17j. doi: 10.1093/ije/dyz083. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Nelson D, Kreps G, Hesse B, et al. The health information national trends survey (HINTS) development, design, and dissemination. Journal of health communication. 2004;9(5):443–460. doi: 10.1080/10810730490504233. [DOI] [PubMed] [Google Scholar]
- 18.Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Statistics in medicine. 2015;34(28):3661–3679. doi: 10.1002/sim.6607. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Rubin DB. Multiple imputation after 18+ years. Journal of the American statistical Association. 1996;91(434):473–489. [Google Scholar]
- 20.Royston P. Multiple imputation of missing values. The Stata Journal. 2004;4(3):227–241. [Google Scholar]
- 21.McHugh M, Ye J, Maechling CR, Holl JL. 2020. Anchor Businesses in the United States. [Google Scholar]
- 22.Ye J. The role of health technology and informatics in a global public health emergency: practices and implications from the COVID-19 pandemic. JMIR Medical Informatics. 2020;8(7):e19866. doi: 10.2196/19866. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Shapiro M, Johnston D, Wald J, Mon D. RTI International, April. 2012. Patient-generated health data. [Google Scholar]
- 24.Cradock KA, ÓLaighin G, Finucane FM, Gainforth HL, Quinlan LR, Ginis KAM. Behaviour change techniques targeting both diet and physical activity in type 2 diabetes: A systematic review and meta-analysis. International Journal of Behavioral Nutrition and Physical Activity. 2017;14(1):1–17. doi: 10.1186/s12966-016-0436-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Bradley SM. Use of Mobile Health and Patient-Generated Data—Making Health Care Better by Making Health Care Different. JAMA Network Open. 2020;3(4):e202971–e202971. doi: 10.1001/jamanetworkopen.2020.2971. [DOI] [PubMed] [Google Scholar]
- 26.Lyons EJ, Lewis ZH, Mayrsohn BG, Rowland JL. Behavior change techniques implemented in electronic lifestyle activity monitors: a systematic content analysis. Journal of medical Internet research. 2014;16(8):e192. doi: 10.2196/jmir.3469. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Gal R, May AM, van Overmeeren EJ, Simons M, Monninkhof EM. The effect of physical activity interventions comprising wearables and smartphone applications on physical activity: a systematic review and meta-analysis. Sports medicine-open. 2018;4(1):42. doi: 10.1186/s40798-018-0157-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Finkelstein EA, Haaland BA, Bilger M, et al. Effectiveness of activity trackers with and without incentives to increase physical activity (TRIPPA): a randomised controlled trial. The lancet Diabetes & endocrinology. 2016;4(12):983–995. doi: 10.1016/S2213-8587(16)30284-4. [DOI] [PubMed] [Google Scholar]
- 29.Ye J, Zhang R, Bannon JE, et al. Identifying Practice Facilitation Delays and Barriers in Primary Care Quality Improvement. The Journal of the American Board of Family Medicine. 2020;33(5):655–664. doi: 10.3122/jabfm.2020.05.200058. [DOI] [PubMed] [Google Scholar]