Digital Data Engagement and Health Behavior Diversity Among Smartwatch Users

Abstract

Smartwatches generate continuous personal health data, yet their value for self-care depends on users’ ability to engage with these data meaningfully. Using a cross-sectional survey of 838 adult smartwatch users in Taiwan, this study examined digital data engagement (DDE) as a data-specific construct linking smartwatch use to everyday self-care. DDE captured monitoring, interpretation, verification, and goal-oriented use of smartwatch-generated data, while health behavior diversity (HBD) represented the breadth of routine self-care practices across preventive, mind–body, physical activity/exercise, and restorative or expressive domains. The DDE-centered model showed that DDE was positively associated with health motivation, body awareness, and HBD. Health motivation showed the strongest association with HBD, whereas body awareness was not directly associated with HBD, suggesting that bodily cue awareness alone may not broaden self-care repertoires. HBD analysis further showed that smartwatch-enabled self-care was not merely exercise-oriented; preventive self-care had the highest item-adjusted coverage. Gender differences appeared mainly in domain composition, whereas health concern profiles were associated with both repertoire breadth and domain composition. These findings contribute a data-engagement perspective to smartwatch research and introduce HBD as a multidomain lens for evaluating everyday digital self-care beyond device access, tracking frequency, or single behavior change.

1. Introduction

Smartwatches are increasingly used for everyday health monitoring. They can collect physiological and behavioral data relevant to personal health and support monitoring, nudging, and predicting functions [1]. Access to wearable-generated data, however, does not necessarily support self-care. Users may receive device-generated feedback without knowing how to interpret it or how to relate it to daily routines. Research on patient-generated health data similarly emphasizes that data value depends not only on collection, but also on use, purpose, and contextual interpretation [2]. Applied to smartwatch-generated data, this suggests a gap between access to personal health data and the interpretive work needed to use such data in everyday self-care. The central issue is therefore not merely device access or the availability of tracked indicators, but how users engage with smartwatch-generated data as health-relevant information.

Self-tracking research suggests that personal data can support reflection and health management, although poorly supported reflection may also contribute to uncertainty or rumination [3,4]. This study examines the gap between data access and self-care use through digital data engagement (DDE), defined as users’ perceived engagement with smartwatch-generated data through monitoring, interpretation, verification, and goal-oriented use. Grounded in personal informatics and data sensemaking research, DDE captures how users relate personal data to lived experience, goals, and context [5].

DDE is positioned as a data-specific engagement construct. Rather than measuring smartwatch adoption, continued use, or tracking frequency, it examines how users review, interpret, verify, and apply smartwatch-generated health data in everyday self-care. This focus distinguishes DDE from related concepts in smartwatch and self-tracking research. Self-quantification emphasizes the measurement of the self and the production of numerical feedback, whereas DDE concerns how users make that feedback meaningful and usable. Wearable engagement is broader, often including device interaction, interface experience, cognitive absorption, and continued involvement with the technology [6]. Continued smartwatch use research explains why users keep using a device and has examined factors such as perceived usefulness, self-quantification, and goal pursuit motivation [7]. DDE assumes smartwatch use is already present and focuses on users’ engagement with the data itself. By foregrounding these data practices, DDE identifies a middle layer between wearable data availability and everyday self-care practice.

Smartwatch-generated data are often interpreted in relation to bodily cues and health goals. Body awareness refers here to users’ self-reported sensitivity to bodily cues. This definition is consistent with research on interoceptive awareness, which concerns the perception and interpretation of internal bodily signals [8]. Recent research also suggests that wearable trackers may be related to how users attend to their bodies, although such associations vary across users and contexts [9]. Health motivation refers to users’ motivation to maintain or improve health. It is relevant to smartwatch-enabled self-care, as users oriented toward health maintenance may attend more closely to health-related information. Prior work on activity trackers suggests that users’ existing motivation shapes how they perceive the informational and motivational affordances of tracking data [10]. On this basis, the present study examines DDE as a data-specific engagement construct associated with body awareness and health motivation.

Assessing smartwatch-enabled self-care also requires attention to the breadth of users’ self-care practices. Everyday self-care may involve multiple practices through which individuals maintain health, prevent illness, or manage health-related needs [11]. Research on multiple health behavior changes also suggests that health behaviors often co-occur and can be examined across behavioral domains rather than as isolated actions [12]. To capture this multidomain pattern, this study uses health behavior diversity (HBD), operationalized as self-care repertoire breadth. HBD refers to the number of distinct self-care practices a respondent reports performing regularly and is used as an indicator of self-care breadth rather than behavioral quality, intensity, or clinical effectiveness. Variation in HBD is further examined by gender and self-reported health concern profile, with attention to its overall level and domain-specific composition.

To address these issues, this study examines smartwatch-enabled self-care using a DDE-centered association model linking digital data engagement, health motivation, body awareness, and health behavior diversity. It contributes to smartwatch and digital health research by redirecting attention from device access and continued use to users’ engagement with wearable data as self-care-relevant information. First, it conceptualizes DDE as a data-specific construct that captures how users monitor, interpret, verify, and apply smartwatch-generated health data in everyday routines. Second, it introduces HBD as an indicator of self-care repertoire breadth, enabling smartwatch-enabled self-care to be examined beyond isolated health behaviors. By also examining health motivation and body awareness, the study distinguishes engagement with wearable data from users’ motivational orientation and attentiveness to bodily cues. This framing is relevant to digital health equity, preventive self-care, and digital health literacy, as the effective use of wearable-generated health data depends not only on access to indicators but also on users’ capacity to interpret, evaluate, and apply such data in everyday life [13,14].

2. Research Model and Hypotheses

Figure 1 presents a DDE-centered association model linking DDE, health motivation, body awareness, and HBD. Given the cross-sectional design, the hypothesized paths are interpreted as theoretically specified associations rather than causal effects. The proposed direction reflects the study’s focus on smartwatch-generated data engagement, while alternative ordering remains plausible.

The first two hypotheses examine whether DDE is associated with health motivation and body awareness. Users who report greater engagement with smartwatch-generated data may also report stronger motivation to maintain or improve health and greater sensitivity to bodily cues.

H1. 

Digital data engagement is positively associated with health motivation.

H2. 

Digital data engagement is positively associated with body awareness.

The next three hypotheses examine whether health motivation, body awareness, and DDE are associated with HBD. Health motivation may be associated with HBD because motivated users may report a broader range of self-care practices. Body awareness may be associated with HBD because users who attend to bodily cues may also report more varied self-care practices. DDE may be associated with HBD because users who review, interpret, verify, and use smartwatch-generated data may report broader self-care repertoires.

H3. 

Health motivation is positively associated with health behavior diversity.

H4. 

Body awareness is positively associated with health behavior diversity.

H5. 

Digital data engagement is positively associated with health behavior diversity.

Given the operationalization of HBD as self-care repertoire breadth, the study further examines whether HBD differs by gender and self-reported health concern profile. This exploratory analysis considers both total HBD and domain-specific patterns.

RQ1. 

How does health behavior diversity vary by gender and self-reported health concern profile?

3. Materials and Methods

3.1. Study Design and Participants

This study employed an online survey to examine associations among digital data engagement, body awareness, health motivation, and health behavior diversity among adult smartwatch users in Taiwan. Data were collected between 21 June and 21 July 2024 using a structured self-report questionnaire distributed through health-related boards on PTT, a major online forum in Taiwan. The study was approved by the Institutional Review Board of National Tsing Hua University (approval no. 11205ES076). The present article reports only data from the questionnaire component of the approved protocol. All participants provided informed consent before completing the online questionnaire. The literature review used to develop the conceptual framework and questionnaire items was conducted between December 2022 and April 2024, with an updated search conducted before manuscript revision in May 2026.

Eligible participants were adults aged 18 years or older who reported having used at least one health-related function on a smartwatch. Health-related smartwatch functions were defined broadly to include activity tracking, sleep tracking, heart rate monitoring, blood oxygen monitoring, and stress-related indicators. No restriction was imposed on smartwatch brand, model, or operating system.

3.2. Measures

The questionnaire assessed four main constructs: digital data engagement, body awareness, health motivation, and health behavior diversity. Items for the first three constructs were rated on a seven-point Likert-type scale (1 = strongly disagree, 7 = strongly agree), as shown in

Table 1. Measurement items for the main constructs.

Construct	Dimension	Questionnaire Items
Body Awareness	—	I am very aware of my bodily sensations.
		I immediately notice when I feel physical discomfort.
		I am highly sensitive to the internal signals of my body.
Health Motivation	—	I am highly motivated to maintain my physical health.
		I am strongly motivated to invest time and effort in maintaining my physical health.
		I have a strong desire to keep myself in good physical health.
		Maintaining an adequate level of physical health is very important to me.
		I strive to keep myself in the best possible physical condition.
Digital Data Engagement	Monitoring & routine review	It is important to me to understand the physiological or activity data tracked by my smartwatch.
		I use my smartwatch to monitor my physiological or activity patterns.
		I regularly review the data tracked by my smartwatch.
		I regularly analyze the data tracked by my smartwatch.
	Interpretation & verification	I can connect my smartwatch data with my own behavior.
		I look for relevant information to help me understand my smartwatch data.
		My smartwatch data are meaningful and help me better understand my daily life.
		My confidence in the device increases when my personal experience matches my smartwatch data.
		I actively look for the cause when my smartwatch data show unexpected changes.
	Goal-oriented data use	I set health-related behavior or activity goals based on my smartwatch data.
		I use multiple apps to monitor my physiological or activity data for more comprehensive health management.
		My smartwatch helps me track my progress toward my health management goals.

. Health behavior diversity was measured with a single multiple-response item that asked respondents to select the self-care practices they performed regularly in daily life.

The questionnaire was administered in Mandarin Chinese using Traditional Chinese characters. Items adapted from prior English-language instruments were translated and adapted by the research team to fit the smartwatch-enabled self-care context. The translated and adapted items were reviewed for semantic equivalence, conceptual consistency, and contextual appropriateness before survey administration.

3.2.1. Body Awareness

Body awareness was defined as respondents’ self-reported sensitivity to bodily cues. In this study, body awareness was conceptualized as bodily cue salience, reflecting users’ attentiveness to physical sensations, bodily discomfort, and internal bodily signals. Three items were used to assess generalized awareness of bodily sensations. The items were adapted from prior measures of body awareness and interoceptive attentiveness, with wording adjusted to fit the present smartwatch-enabled self-care context [15,16].

3.2.2. Health Motivation

Health motivation captured respondents’ motivation to maintain or improve their physical health. The items were developed to reflect perceived importance, desire, and effort investment in maintaining physical health, drawing on health-behavior and readiness-to-change perspectives [17].

3.2.3. Digital Data Engagement

Digital data engagement (DDE) was defined as users’ perceived engagement with smartwatch-generated data in everyday self-care. Drawing on personal informatics, digital health engagement, and data sensemaking research, DDE reflects the extent to which users monitor, interpret, verify, and apply wearable-generated data to support health-related understanding and self-care practices [18,19]. An earlier conference paper [19] drew on the same original survey response pool but used a differently screened analytic sample and a broader 20-item framework for exploratory user segmentation. The exact 12-item, three-dimensional higher-order DDE specification was not reported in that paper and is formally assessed in the present study.

DDE was operationalized as a multidimensional construct comprising three related dimensions: monitoring and routine review, interpretation and verification, and goal-oriented data use. Twelve DDE items were developed deductively from the theoretical definition of each dimension and reviewed by the research team for conceptual alignment before survey administration.

Monitoring and routine review was measured with four items assessing repeated checking, reviewing, and analysis of smartwatch-generated physiological or activity data. This dimension reflects routine self-tracking and self-quantification practices but is more specifically focused on users’ perceived engagement with data generated by the device [6,7].

Interpretation and verification were measured with five items assessing users’ efforts to connect smartwatch data with personal behavior and lived experience, seek additional information, evaluate unexpected data changes, and validate device feedback through comparison with subjective bodily experience. This dimension corresponds to the sensemaking work through which users move beyond passive data reception toward active interpretation of personal health data [5].

Goal-oriented data use was measured with three items assessing the use of smartwatch data for setting health-related goals, integrating data across applications, and tracking progress toward health management goals. This dimension reflects personal informatics research showing that effective self-management depends not only on numerical feedback, but also on goal setting, goal adaptation, and reflective use of personal data [7,20].

3.2.4. Behavior Diversity

Health behavior diversity (HBD) was operationalized as the number of distinct routine self-care practices endorsed by each respondent. HBD was treated as an observed count outcome representing the breadth of an individual’s self-care repertoire and is interpreted as a descriptive indicator of self-care breadth rather than as a direct measure of health status or behavioral effectiveness. HBD reflects the range of activities respondents reported using to maintain their physical and mental health. This operationalization follows multidomain and multiple-health-behavior perspectives, which view health practices as patterns of co-occurring behaviors embedded in daily life rather than as isolated actions [21,22].

A 16-item checklist of routine self-care practices, completed in a multiple-response (“select all that apply”) format, was developed for this study. The listed practices were grouped into four domains. Preventive self-care included maintaining a regular sleep schedule, healthy eating, and routine health check-ups (3 items), reflecting common descriptions of everyday preventive self-care [23]. Mind–body regulation included mindfulness meditation and yoga (2 items), consistent with evidence that mindfulness approaches support self-regulation and reduce stress-related physiological responses [24]. Physical activity and exercise included walking, running, strength training, swimming, ball sports, dancing, cycling, and hiking (8 items), consistent with the WHO 2020 physical activity guidelines [25]. Restorative and expressive well-being included singing, listening to music, and reading (3 items), with evidence linking arts and music engagement to well-being [26,27].

The total HBD score was calculated as the sum of endorsed practices, with higher scores indicating a broader self-care repertoire. Each practice was assigned equal weight because HBD was designed to capture repertoire breadth rather than the frequency, intensity, or health impact of individual practices. Domain-specific counts were also derived, and domain diversity was calculated as the number of domains in which a respondent endorsed at least one practice (range: 0–4). These indicators were used to examine whether self-care breadth differed by type of practice as well as by total count.

3.2.5. Heterogeneity Variables

Gender and self-reported health concern profile were measured as exploratory variables for describing variation in HBD. These variables were used to examine whether the level and domain composition of health behavior diversity differed across user subgroups.

Self-reported health concern profiles were derived from a multiple-response health concern checklist. Respondents were classified into four analytically interpretable profiles: no concerns, psychological/sleep-related concerns only, physiological/metabolic concerns only, and mixed concerns. Respondents who selected only “other” or whose responses could not be meaningfully assigned to one of the four profiles were coded as other-only or ambiguous. Because this category was small and substantively heterogeneous, health concern profile subgroup comparisons were limited to the four main profiles.

For classification purposes, poor sleep quality and poor emotional or mental state were coded as psychological/sleep-related concerns. High blood pressure, unstable blood glucose, unstable heart rhythm, high body fat, and menstrual irregularity were coded as physiological/metabolic concern. Given the binary gender categories available in the survey, menstrual irregularity was treated as not applicable for respondents classified as men when constructing the physiological/metabolic category. These profiles were used only as exploratory subgroup variables and should not be interpreted as clinical diagnostic categories.

3.3. Data Analysis

Data analysis proceeded in four steps. First, descriptive statistics were used to summarize participant characteristics and the distributions of the main study variables, including DDE, body awareness, health motivation, total HBD, domain diversity, and HBD domain-specific counts. HBD was treated as an observed count outcome representing self-care repertoire breadth. Potential common method bias was assessed using a Harman-type single-factor diagnostic. The 20 Likert-type indicators measuring DDE, health motivation, and body awareness were entered into an unrotated principal component analysis. This diagnostic was used to examine whether a single dominant method factor was indicated.

Second, partial least squares structural equation modeling (PLS-SEM) was performed using SmartPLS 4 (version 4.1.0.3) to evaluate the measurement model and test the DDE-centered research model. DDE was modeled as a reflective–reflective higher-order construct using a disjoint two-stage approach. Bootstrapping with 5000 subsamples was used to obtain standard errors, significance tests, and 95% confidence intervals. The structural model was assessed using standardized path coefficients, bootstrapped 95% confidence intervals, f² effect sizes, R² values, inner VIF values, and Q² predictive relevance. Predictive relevance was assessed using the blindfolding procedure with an omission distance of 7. Specific indirect effects from DDE to HBD through health motivation and body awareness were tested using bootstrapping.

Third, RQ1 was addressed through exploratory subgroup analyses of HBD by gender and self-reported health concern profile using IBM SPSS Statistics 27 (version 27.0.0.0). Gender differences were assessed using independent-samples t tests with Cohen’s d, and differences across the four main health concern profiles were assessed using one-way ANOVA with η². Profile comparisons were conducted among the classifiable subsample (n = 830), with respondents coded as other-only or ambiguous retained in the full analytic sample but not included in this subgroup analysis. When omnibus profile differences were significant, post hoc pairwise comparisons were used to interpret group differences. These subgroup analyses were exploratory and involved multiple comparisons; therefore, effect sizes and domain-level patterns were emphasized in the interpretation.

Fourth, because total HBD was a count variable, Poisson regression was estimated as a robustness check for the HBD paths in the structural model.

4. Results

4.1. Sample Characteristics

A total of 926 responses were received. During revision, the original response pool was re-evaluated using a refined screening procedure incorporating the eligibility and response-quality criteria described below, yielding a final analytic sample of 838. All reported analyses were conducted using this final sample. Excluded cases comprised respondents who did not use a smartwatch (n = 30) or were younger than 18 years (n = 7). Additional exclusions included cases with identical response patterns and response sequence indicators (n = 21), cases showing an extreme dominance of a single response option or extreme long-string patterns of identical answers (n = 23), and cases completed in under two minutes (n = 7). Exclusion categories were applied hierarchically and counted as mutually exclusive. As shown in

Table 2. Sample characteristics of the analytical sample.

Characteristic	Category	Total, n (%)	No Concern	Psychological/Sleep-Related	Physiological/Metabolic	Mixed Concerns	Other-Only/Ambiguous
Gender	Men	398 (47.5%)	85	115	78	116	4
	Women	440 (52.5%)	63	107	101	165	4
Age	19–25	166 (19.8%)	24	42	29	71	0
	26–35	316 (37.7%)	62	76	65	110	3
	36–45	299 (35.7%)	57	84	70	83	5
	46–55	44 (5.3%)	4	15	13	12	0
	56 and above	13 (1.6%)	1	5	2	5	0

Note. Percentages may not sum to 100.0% because of rounding.

, the sample was predominantly young to middle-aged. The largest age group was 26–35 years (37.7%, n = 316), followed by 36–45 years (35.7%, n = 299) and 19–25 years (19.8%, n = 166). Women accounted for 52.5% of the sample (n = 440), and men accounted for 47.5% (n = 398). Regarding self-reported health concern profiles, the mixed-concern profile was the largest group (33.5%, n = 281), followed by psychological/sleep-related concerns only (26.5%, n = 222), physiological/metabolic concerns only (21.4%, n = 179), no concerns (17.7%, n = 148), and other-only or ambiguous responses (1.0%, n = 8). The eight respondents classified as other-only or ambiguous were retained in the full analytic sample but excluded from health concern profile subgroup analyses because the category was small and substantively heterogeneous.

4.2. Measurement Model Assessment

DDE was specified as a reflective–reflective higher-order construct composed of three lower-order dimensions: monitoring and routine review, interpretation and verification, and goal-oriented data use. In the first stage, the three lower-order dimensions showed acceptable measurement quality. Indicator loadings ranged from 0.732 to 0.877, Cronbach’s alpha ranged from 0.802 to 0.883, composite reliability ranged from 0.881 to 0.919, and AVE ranged from 0.664 to 0.738 (

Table 3. Measurement model results.

Construct Level	Construct/Dimension	Indicators	Loadings	Cronbach’s α	CR	AVE
Lower order	Monitoring and routine review	4 items	0.832–0.877	0.883	0.919	0.738
Lower order	Interpretation and verification	5 items	0.732–0.865	0.872	0.908	0.664
Lower order	Goal-oriented data use	3 items	0.787–0.876	0.802	0.881	0.713
Higher order	DDE	3 lower-order dimensions	0.918–0.946	0.919	0.949	0.860
Reflective construct	Body awareness	3 items	0.875–0.901	0.866	0.918	0.788
Reflective construct	Health motivation	5 items	0.845–0.885	0.918	0.939	0.754

). These scores were then used as indicators of the higher-order DDE construct in the second stage.

In the second stage, the higher-order DDE construct also demonstrated strong measurement quality. The three lower-order dimensions loaded strongly on DDE, with loadings ranging from 0.918 to 0.946. DDE showed high internal consistency and convergent validity. Body awareness and health motivation also showed acceptable measurement quality. HBD was modeled as a single-item observed count construct; reliability statistics are therefore not applicable to HBD.

Discriminant validity was evaluated using HTMT. In the first-stage model, the HTMT values were 0.925 between monitoring and routine review and interpretation and verification, 0.863 between monitoring and routine review and goal-oriented data use, and 0.967 between interpretation and verification and goal-oriented data use. Two values exceeded 0.90, while the remaining value exceeded the more conservative 0.85 criterion, indicating limited discriminant validity when the lower-order dimensions are treated as standalone constructs. Accordingly, the dimensions are interpreted as closely interrelated facets of the higher-order DDE construct rather than as fully independent constructs; this interpretation is considered together with their conceptual relatedness and strong second-stage loadings. In the final higher-order model, HTMT values among the main constructs ranged from 0.195 to 0.637, below commonly used thresholds. The Harman-type single-factor diagnostic showed that the first unrotated component explained 49.39% of the total variance, below the commonly used 50% threshold. This result suggests that a single dominant method factor was not indicated.

4.3. Structural Model Results

PLS-SEM was used to test the five hypothesized associations in the DDE-centered research model.

Table 4. PLS-SEM structural model, predictive relevance, and indirect effects. Panel (A). Structural path results. Panel (B). Explained variance and predictive relevance. Panel (C). Specific indirect effects.

(A)
Hypothesis	Path	β	SE	t	f2	95% CI	Inner VIF
H1	DDE → Health motivation	0.589	0.027	21.83 ***	0.531	[0.535, 0.641]	1.00
H2	DDE → Body awareness	0.477	0.031	15.61 ***	0.295	[0.415, 0.535]	1.00
H3	Health motivation → HBD	0.276	0.041	6.73 ***	0.049	[0.196, 0.357]	1.77
H4	Body awareness → HBD	−0.031	0.041	0.76	0.001	[−0.113, 0.048]	1.50
H5	DDE → HBD	0.134	0.041	3.29 **	0.013	[0.051, 0.212]	1.62
(B)
Endogenous Construct		R2		Adjusted R2		Q2
Health motivation		0.347		0.346		0.260
Body awareness		0.228		0.227		0.176
Health behavior diversity		0.125		0.122		0.122
(C)
Indirect Path		β	SE	t		95% CI	Result
DDE → Health motivation → HBD		0.163	0.025	6.429 ***		[0.115, 0.214]	Supported
DDE → Body awareness → HBD		−0.015	0.020	0.756		[−0.055, 0.022]	Not supported

(A): Note. ** p < 0.01. *** p < 0.001. (C): Note. β = standardized path coefficient. CI = percentile-based 95% bootstrap confidence interval. Q² was obtained using blindfolding with an omission distance of 7. Inner VIF values ranged from 1.00 to 1.77, indicating that collinearity was not a concern in the structural model. *** p < 0.001.

presents standardized path coefficients, standard errors, t values with significance markers, 95% confidence intervals, f² effect sizes, and inner VIF values. Inner VIF values ranged from 1.00 to 1.77, indicating that collinearity was not a concern in the structural model. Explained variance and predictive relevance are reported in Panel B.

DDE was positively associated with health motivation (β = 0.589, SE = 0.027, t = 21.83, p < 0.001, 95% CI [0.535, 0.641]) and body awareness (β = 0.477, SE = 0.031, t = 15.61, p < 0.001, 95% CI [0.415, 0.535]), supporting H1 and H2. Health motivation was positively associated with HBD (β = 0.276, SE = 0.041, t = 6.73, p < 0.001, 95% CI [0.196, 0.357]), supporting H3. Body awareness was not directly associated with HBD (β = −0.031, SE = 0.041, t = 0.76, p = 0.445, 95% CI [−0.113, 0.048]), so H4 was not supported. DDE was positively associated with HBD (β = 0.134, SE = 0.041, t = 3.29, p = 0.001, 95% CI [0.051, 0.212]), supporting H5.

The model explained 34.7% of the variance in health motivation, 22.8% of the variance in body awareness, and 12.5% of the variance in HBD. The Q² values were positive for all endogenous constructs, indicating predictive relevance.

Overall, the structural model provided partial support for the proposed DDE-centered framework. DDE was associated with health motivation, body awareness, and HBD, while health motivation was also associated with HBD. Body awareness, however, was not directly associated with HBD in the model.

4.4. Domain Composition and Subgroup Variation in HBD

This section addresses RQ1 by examining HBD as both a total count and a domain-structured self-care repertoire. We first analyzed the domain composition of HBD and then compared HBD patterns by gender and self-reported health concern profile.

4.4.1. Domain Composition of Health Behavior Diversity

The 16 self-care practices were grouped into four domains: preventive self-care, mind–body regulation, physical activity/exercise, and restorative/expressive well-being.

Table 5. Health behavior diversity across self-care domains.

HBD Domain	No. of Items	Mean	SD	Raw Share of HBD (%)	Item-Adjusted Coverage (%)
Preventive self-care	3	1.23	0.94	31.2	41.0
Mind–body regulation	2	0.19	0.43	4.8	9.5
Physical activity/exercise	8	1.83	1.29	46.5	22.9
Restorative/expressive well-being	3	0.69	0.85	17.4	22.9
Total HBD	16	3.94	2.12	100.0	24.6

Note. HBD = health behavior diversity. Raw share was calculated as the domain mean divided by the total HBD mean and multiplied by 100. Item-adjusted coverage was calculated as the domain mean divided by the number of items in that domain and multiplied by 100. This value indicates the average proportion of practices endorsed within each domain and accounts for unequal numbers of checklist items across domains. Total HBD item-adjusted coverage was calculated as the total HBD mean divided by the total number of checklist items.

summarizes each domain’s contribution to total HBD.

As shown in Table 5, physical activity/exercise contributed the largest raw share of HBD, followed by preventive self-care, restorative/expressive well-being, and mind–body regulation. However, because the four domains contained different numbers of items, raw shares were interpreted together with item-adjusted coverage. From this perspective, preventive self-care showed the highest item-adjusted coverage, whereas mind–body regulation showed the lowest coverage.

These findings suggest that HBD should not be interpreted simply as an exercise-oriented measure. Rather, it captures a multidomain self-care repertoire in which routine preventive practices were most consistently endorsed after adjusting for the number of items in each domain.

4.4.2. Gender and Health Concern Profile Differences in HBD Level and Domain Composition

We next examined whether HBD varied by gender and self-reported health concern profile. The analysis examined both HBD level, represented by total HBD and domain diversity, and domain composition, represented by the four domain-specific counts. Gender comparisons used the full analytic sample of 838 respondents. Health concern profile comparisons used the classifiable subsample of 830 respondents, excluding respondents coded as other-only or ambiguous.

Table 6. Group differences in HBD level and domain composition. Panel (A). HBD outcomes by gender. Panel (B). HBD outcomes by health concern profile.

(A)
Outcome		Male (n = 398), M (SD)		Female (n = 440), M (SD)		p	Cohen’s d
HBD total		3.90 (1.98)		3.97 (2.25)		0.639	−0.032
Domain diversity		2.15 (0.79)		2.33 (0.92)		0.003	−0.208
Preventive self-care		1.18 (0.96)		1.28 (0.91)		0.110	−0.111
Mind–body regulation		0.12 (0.37)		0.25 (0.48)		<0.001	−0.301
Physical activity/exercise		2.08 (1.24)		1.61 (1.29)		<0.001	0.372
Restorative/expressive well-being		0.53 (0.74)		0.83 (0.92)		<0.001	−0.367
(B)
Outcome	No Concern (n = 148), M (SD)	Psych/Sleep (n = 222), M (SD)	Physio/Metabolic (n = 179), M (SD)	Mixed Concern (n = 281), M (SD)	F	p	η2
HBD total	4.05 (2.28)	3.63 (2.04)	3.74 (2.02)	4.23 (2.16)	3.98	0.008	0.014
Domain diversity	2.20 (0.89)	2.17 (0.88)	2.15 (0.78)	2.40 (0.87)	4.37	0.005	0.016
Preventive self-care	1.34 (1.03)	1.17 (0.92)	1.13 (0.86)	1.27 (0.94)	1.84	0.139	0.007
Mind–body regulation	0.15 (0.38)	0.19 (0.45)	0.17 (0.42)	0.22 (0.46)	1.12	0.340	0.004
Physical activity/exercise	1.91 (1.39)	1.59 (1.31)	1.94 (1.21)	1.91 (1.25)	3.56	0.014	0.013
Restorative/expressive	0.65 (0.80)	0.68 (0.83)	0.50 (0.79)	0.83 (0.91)	5.66	0.001	0.020

(B): Note. HBD = health behavior diversity. Health concern profile comparisons were conducted among respondents classified into one of the four main profiles (n = 830). Psych/sleep = psychological/sleep-related concerns; Physio/metabolic = physiological/metabolic concerns. Other-only or ambiguous cases were retained in the full analytic sample but not included in this subgroup analysis. η² was calculated as the ratio of between-group sum of squares to total sum of squares. Exact p values are reported; p values smaller than 0.001 are reported as p < 0.001.

reports group differences in total HBD, domain diversity, and domain-specific HBD counts.

Gender differences were examined using independent-samples t tests with Cohen’s d reported as the effect size. Total HBD did not differ by gender. Men reported a mean HBD of 3.90, whereas women reported a mean of 3.97, and the effect size was negligible. However, gender differences were observed in domain diversity and domain composition. Women reported slightly higher domain diversity and greater engagement in mind–body regulation and restorative/expressive well-being. Men reported a higher count in physical activity/exercise. Preventive self-care did not differ significantly by gender. These results indicate that gender differences were expressed less through the overall number of self-care practices and more through the domains in which self-care was enacted. The effect sizes were small and should be interpreted cautiously.

Health concern profile differences were examined among the 830 respondents who could be classified into one of the four main profiles. Total HBD differed across profiles, although the effect size was small. Respondents in the mixed-concern profile reported the highest total HBD, whereas those in the psychological/sleep-related profile reported the lowest total HBD. Post hoc comparisons indicated that the mixed-concern group reported higher total HBD than the psychological/sleep-related group.

Domain diversity also differed across profiles. The mixed-concern group showed the broadest domain coverage and reported higher domain diversity than the psychological/sleep-related and physiological/metabolic groups. At the domain level, significant differences were observed for physical activity/exercise and restorative or expressive well-being, but not for preventive self-care or mind–body regulation. The psychological/sleep-related group reported lower physical activity/exercise breadth than the physiological/metabolic and mixed-concern groups. The mixed-concern group also reported higher restorative or expressive well-being practices than the physiological/metabolic group.

Overall, health concern profiles were associated with both HBD level and domain composition, but the effect sizes were small. These findings should therefore be interpreted as exploratory evidence of repertoire variation rather than as strong subgroup effects.

4.5. Poisson Regression Robustness Check

As a robustness check for the HBD paths in the structural model, a Poisson regression was estimated because HBD was operationalized as an observed count outcome. The likelihood-ratio test for the Poisson model was significant, χ²(3) = 125.722, p < 0.001, indicating that DDE, health motivation, and body awareness jointly improved the prediction of HBD counts compared with an intercept-only model. Goodness-of-fit diagnostics indicated no meaningful overdispersion, with deviance/df = 0.977 and Pearson χ²/df = 0.998. DDE remained positively associated with HBD, B = 0.073, SE = 0.023, Wald χ² = 10.372, p = 0.001, IRR = 1.076, and 95% CI [1.029, 1.125]. Health motivation was also positively associated with HBD, B = 0.163, SE = 0.024, Wald χ² = 44.448, p < 0.001, IRR = 1.177, and 95% CI [1.122, 1.235]. Body awareness was not significantly associated with HBD, B = −0.019, SE = 0.021, Wald χ² = 0.769, p = 0.380, IRR = 0.982, and 95% CI [0.941, 1.023]. These results were consistent with the PLS-SEM findings.

5. Discussion

5.1. From Device Use to Data Engagement

The results shift attention from smartwatch use as a matter of access, adoption, or tracking frequency to the work users do with the information the device provides. The coherence among the three DDE dimensions is important in this respect. Monitoring and routine review, interpretation and verification, and goal-oriented use did not appear as merely adjacent activities; empirically, they formed a coherent orientation toward wearable data. This high coherence indicates that the lower-order DDE dimensions should not be interpreted as fully independent constructs. Rather, they appear to operate as interrelated facets of an integrated orientation toward smartwatch-generated health data. This pattern suggests that smartwatch data may become relevant to self-care when users connect device feedback with bodily experience, evaluate its plausibility, and relate it to health goals.

This view is consistent with accounts of personal informatics as lived and situated. Rooksby et al. [28] describe personal tracking as a practice in which data are woven into routines and contexts rather than simply received as feedback. The present study extends this perspective by proposing a quantitative construct for the otherwise difficult-to-measure space between data availability and self-care practice. DDE does not capture the full richness of data sensemaking, but it identifies a tractable layer of engagement in which users monitor, evaluate, verify, and use smartwatch-generated information as health-relevant data. This interpretation supports modeling DDE as a higher-order construct: for many users, routine review, interpretation, verification, and goal-oriented data use may be mutually reinforcing rather than strictly sequential tasks.

The structural results nevertheless qualify the behavioral implications of DDE. Although DDE was positively associated with HBD, its direct effect size was very small (f² = 0.013). Engagement with smartwatch data may be relevant to broader self-care repertoires, but it is unlikely to be sufficient on its own. Health motivation showed a stronger association with HBD and carried the significant indirect association between DDE and HBD, suggesting that motivational conditions may be more consequential for self-care repertoire breadth than data engagement alone. The model also explained a modest proportion of variance in HBD (R² = 0.125), indicating that the constructs included in the model account for only part of the variation in self-care repertoire breadth. Everyday self-care is likely shaped by broader personal, social, and contextual conditions. This interpretation is consistent with personal informatics work that treats engagement with personal data as cyclical and changing rather than as a linear movement from tracking to behavior change [29]. For design and digital health research, the implication is not simply to provide more indicators, but to support users in contextualizing data, evaluating device feedback, and connecting data to routines and goals. In this respect, DDE offers a way to examine how wearable data may become part of everyday self-care, including forms of care that are not reducible to activity or performance metrics [30].

5.2. Motivation, Bodily Awareness, and the Limits of Awareness-to-Action Assumptions

The findings complicate a familiar premise in personal informatics and smartwatch-enabled self-care: that greater bodily awareness or access to feedback will necessarily correspond to a broader range of health practices. Prior work has similarly cautioned that self-insight does not necessarily lead to behavior change, especially when users lack the conditions needed to interpret or act on personal data [31]. In the present study, DDE was associated with both health motivation and body awareness, suggesting that engagement with smartwatch data is tied not only to technical use but also to users’ orientation toward health and their attentiveness to bodily cues. Rather than making smartwatch data self-care-relevant on their own, these associations suggest that wearable feedback may become meaningful when users relate it to bodily states, evaluate its plausibility, and connect it to health-related concerns or goals.

The associations with HBD were more selective. Among the direct paths to HBD, health motivation showed the strongest positive association with self-care repertoire breadth. DDE showed a smaller direct association, and body awareness was not directly associated with HBD. This distinction is theoretically important because it separates awareness from the broader conditions through which self-care repertoires are organized and sustained. Health motivation also carried the significant indirect association between DDE and HBD, whereas the indirect association through body awareness was not supported. This pattern suggests that smartwatch data engagement may be most relevant to self-care breadth when it is accompanied by users’ motivation to maintain or improve health. These results are interpreted as indirect associations rather than evidence of causal mediation. Health behavior theories have long treated motivation, planning, capability, and opportunity as related but distinct components of action [32,33]. The present findings extend this distinction to smartwatch-enabled self-care. Bodily awareness may make physical states more salient, but broader self-care repertoires also require reasons to act, feasible options, and routines into which action can be incorporated.

The nonsignificant path from body awareness to HBD should therefore not be read as evidence that embodiment is irrelevant. It points instead to a more specific role for bodily awareness in smartwatch use. Body awareness may help users judge whether device-generated signals fit lived experience or deserve further attention, but it does not appear to broaden self-care practices on its own. This interpretation is consistent with personal informatics research that treats self-tracking as a route to self-knowledge rather than a straightforward mechanism of behavior change. For design and practice, the finding cautions against assuming that stronger alerts, more indicators, or more physiological feedback will necessarily support broader self-care. Smartwatch systems may be more useful when they help users interpret what a signal means, connect it to feasible forms of action, and adjust goals in relation to everyday constraints.

5.3. Self-Care Repertoires and Contextualized Smartwatch-Enabled Self-Care

In this study, HBD is used to capture the breadth of users’ self-care repertoires, allowing self-care to be examined as a multidomain configuration rather than as a single target behavior or an exercise-only outcome. Although health practices are often evaluated through activity-centered metrics, everyday self-care may extend across preventive, mind–body, activity-based, and restorative or expressive domains. The results support this repertoire-oriented interpretation [34]. Physical activity and exercise contributed the largest raw share of HBD, but preventive self-care showed the highest item-adjusted coverage, indicating that routine maintenance practices were widely embedded in participants’ repertoires. This pattern suggests that participants’ self-care repertoires were not organized solely around exercise-oriented practices; rather, they incorporated diverse practices that may support everyday health maintenance across physical, psychological, and restorative domains. Conceptualizing HBD as repertoire breadth highlights how self-care is organized across domains and makes visible differences in repertoire structure that would be obscured by single-behavior or exercise-only outcomes. In this sense, HBD is used throughout as an indicator of repertoire breadth rather than as a measure of health status, behavioral quality, or clinical effectiveness.

The subgroup findings further show that self-care breadth and self-care composition are analytically distinct. Gender did not significantly differentiate total HBD, but it did differentiate the structure of self-care repertoires. Women reported greater domain diversity and greater breadth in mind–body regulation and restorative/expressive practices, whereas men reported greater breadth in physical activity/exercise. This pattern does not imply that one gender engages in more self-care than the other. Rather, it indicates that similar overall self-care breadth may be organized through different domain profiles. Because these gender differences were small in effect size, they should be interpreted as compositional differences in self-care repertoires rather than as strong gender effects. This interpretation is consistent with gender and health research showing that health behaviors are embedded in gendered routines, identities, and social expectations [35,36].

Self-reported health concern profile emerged as a relevant but modest contextual dimension of smartwatch-enabled self-care. The mixed-concern group showed the highest total HBD and the greatest domain diversity, whereas the psychological/sleep-related and physiological/metabolic groups generally reported lower HBD levels. The clearest pattern was that respondents with mixed concerns reported broader and more domain-diverse self-care repertoires than some single-concern groups, although the effect sizes were small. This pattern suggests that health-related practices are better understood as patterned configurations rather than isolated behavioral choices. The findings align with person-centered studies showing that self-care and health-promoting behaviors can form distinct profiles with different implications for self-rated health, sleep quality, and mental health risk [37]. Health concern profile is therefore better understood as a subgroup lens for observing variation in self-care repertoire breadth and composition, rather than as evidence of a simple linear relationship between greater health concern and more self-care.

This study shifts attention from whether users perform a single target behavior to how they organize self-care across multiple domains. Although HBD is operationalized as a count of endorsed practices, it is interpreted here as a repertoire-breadth indicator that describes how users compose everyday self-care across domains. This contributes to smartwatch and digital health research by extending the focus beyond device use, activity tracking, or isolated behavior change toward the broader organization of self-care in daily life. This interpretation also suggests that smartwatch-based health promotion should not define success only as increasing steps, exercise frequency, or adherence to a single behavioral target. Instead, interventions may need to help users recognize the structure of their existing self-care repertoires, identify domains that are underdeveloped or overly narrow, and connect wearable feedback to multiple forms of everyday self-care. This implication is particularly important because wearable feedback can shape users’ perceptions, affect, and health-related mindsets, sometimes independently of actual activity levels [38].

5.4. Limitations and Future Research

This study has several limitations. First, its cross-sectional, online self-report design limits causal interpretation. The observed relationships among DDE, health motivation, body awareness, and HBD should therefore be understood as associations rather than temporal or causal effects. Common method variance, recall bias, and social desirability bias also cannot be fully ruled out. Although the Harman-type single-factor diagnostic did not indicate a dominant single-factor structure, this test does not eliminate the possibility of common method bias. Future longitudinal or experience-sampling studies could examine how DDE, health motivation, body awareness, and self-care repertoires develop over time.

Second, the study did not include objective smartwatch data. Respondents reported their perceived engagement with smartwatch-generated data and self-care practices, but these reports were not validated against device logs, sensor-derived indicators, or objectively recorded behaviors. Future research could combine survey data with device logs, passive sensing, ecological momentary assessment, or qualitative interviews to examine how users interpret and act on wearable data in everyday contexts.

Third, the sample was recruited from health-related PTT boards in Taiwan and should not be treated as representative of all smartwatch users. The sample may overrepresent digitally active users with relatively high interest in health information. Respondents aged 56 years or older were also few, limiting generalizability to older adults. Future studies should include more diverse age groups, health conditions, socioeconomic backgrounds, digital literacy levels, and cultural or healthcare contexts. The rapid evolution of wearable technologies, including software updates, new sensors, and AI-supported feedback, may also affect the generalizability of the findings to future generations of smartwatch systems.

Fourth, the present 12-item, three-dimensional DDE specification has not yet been validated in an independent sample. Further validation is therefore needed across populations, devices, and health contexts. The questionnaire was also translated and adapted through an in-house review without independent back-translation; residual cross-language differences therefore cannot be ruled out.

Finally, HBD was operationalized as an equal-weight count of routine self-care practices. This approach captures repertoire breadth, but not the frequency, duration, intensity, quality, consistency, or clinical effectiveness of those practices. A broader HBD score should therefore not be interpreted as evidence of better health or more effective self-care. The subgroup analyses by gender and health concern profile were also exploratory and should be interpreted as descriptive patterns rather than confirmatory subgroup effects. Future research should examine how self-care repertoire breadth relates to behavioral quality, sustained change, well-being, and health outcomes.

6. Conclusions

This study used digital data engagement to revisit what is at stake when we talk about smartwatch-enabled self-care. Rather than focusing only on device adoption or tracking frequency, the study examined how adult smartwatch users engage with smartwatch-generated health data through monitoring, interpretation, verification, and goal-oriented use. From this perspective, the central issue is not simply access to wearable data, but whether users can interpret, evaluate, and connect such data to everyday self-care. The findings show that DDE was positively associated with health motivation, body awareness, and HBD. However, its direct association with HBD was small, while health motivation showed a stronger association with self-care repertoire breadth and carried the significant indirect association between DDE and HBD. Body awareness, although associated with DDE, was not directly associated with HBD. These findings suggest that smartwatch-enabled self-care depends not only on wearable data engagement, but also on motivational and interpretive conditions that help users relate personal health information to feasible everyday practices.

Health behavior diversity, in turn, shifts attention from single behaviors to self-care repertoires. By showing how preventive, mind–body, activity-based, and restorative practices are combined, HBD makes visible differences in how people organize self-care across domains, genders, and health concern profiles. Gender differences were expressed mainly in domain composition rather than total HBD, whereas health concern profiles were associated with both repertoire breadth and domain-specific patterns. These subgroup findings were exploratory and modest in effect size, but they demonstrate the value of examining smartwatch-enabled self-care beyond exercise-centered or single-behavior metrics. In this sense, HBD is useful not as a measure of health status or clinical effectiveness, but as a way to describe the breadth and organization of everyday self-care practices.

The study points toward a capacity-oriented understanding of digital self-care. Expanding device access or adding further physiological indicators is unlikely to be sufficient if users cannot interpret what the data mean in their own circumstances or lack feasible ways to act on that information. For designers and health professionals, the challenge is to support the interpretive, motivational, and organizational work through which people weave wearable data into sustainable repertoires of self-care, rather than to optimize a single activity metric. For researchers, the task is to follow these repertoires over time and across contexts, so that evaluations of digital health interventions can move beyond sole reliance on step counts or isolated behaviors and speak more directly to the complexity of everyday self-care.