Diabetes-Specific Quality of Life Changes Associated with a Digital Support Intervention: A Study of Adults with Type 1 Diabetes

Abstract

Although digital platforms have gained popularity in the delivery of diabetes interventions, few models have focused on type 1 diabetes (T1D), offer different support delivery mechanisms, and involve peer and health professional-led support. TRIFECTA is a six-month multi-modal digital support intervention that includes a 24/7 peer texting group, an “ask-the-expert” web-based portal, and professional-led virtual group-based interactive sessions. This study examined diabetes-specific quality of life (DSQoL) changes following TRIFECTA. DSQoL was measured using Type 1 Diabetes and Life, a self-report survey that allows for subscale analysis in different age groups. Among 60 adults with type 1 diabetes, improvements were observed for overall diabetes-specific quality of life, primarily driven by the 26–45 years cohort. Subscale analysis found DSQoL improved for emotional experiences and daily activities for adults 26–45 years old, and social isolation improved for adults 46–60 years old.

1. Introduction

As technology becomes more accessible, digital health platforms such as mobile apps provide opportunities for social support and care outside of the traditional healthcare setting. Not only have mobile apps for diabetes self-management support become increasingly popular [1,2,3,4], but a meta-analysis found that these interventions reduce A1c and improve quality of life markers (e.g., life satisfaction, worry of diabetes, mental health) [1]. Notably, the majority of diabetes mobile app interventions for type 1 diabetes (T1D) focus on self-management behaviors such as tracking blood sugar, meals, and physical activity; counting carbohydrates; calculating insulin etc. [5]. Moreover, when psychosocial endpoints such as quality of life are measured, the instruments can vary across studies and may not be specific to diabetes [1]. Among diabetes-specific quality of life (DSQoL) instruments utilized in mobile app interventions, none to date are customized to address unique issues and concerns for different age cohorts. As such, the objective of our study was to assess changes in DSQoL following participation in a digital support intervention using the Type 1 Diabetes and Life (T1DAL) survey, a quality-of-life instrument specific to T1D with different versions developed to be specific to age bands associated with different eras of adulthood.

2. Materials and Methods

This study was a single cohort pre-post research design approved by the Behavioral Research Ethics Board at the University of British Columbia. Participants were recruited from T1D-focused social media outlets, T1D advocacy groups, diabetes nonprofit organizations, diabetes education centers and clinics, previous T1D research studies, and word of mouth. Individuals interested in enrolling completed a registration survey and underwent screening for the following eligibility criteria (1) have T1D, (2) be 18 years of age, (3) have English proficiency, (4) have access to the internet, (5) living in British Columbia and (6) have a WhatsApp account (Version 2.19.343 and onwards depending on the latest release at the time)if registered for this support delivery modality. Those eligible provided informed consent online. Assessments were administered online at baseline and six months and measured sociodemographic characteristics, and health-related quality of life. To offset the costs of participation, participants were provided $20 and $40 e-gift cards at baseline and post-intervention, respectively.

2.1. Intervention

Described in greater detail in a previous publication [6], the six-month TRIFECTA intervention is based on the self-determination theory which asserts that individuals are motivated to meet three fundamental needs: autonomy, relatedness, and competence [7]. TRIFECTA invites participants (not researchers) to determine what type of support they want (three different support delivery platforms), how frequently they need support (e.g., daily, weekly, as needed), and when to access support (day, evening, weekday, weekend). In doing so, participants make self-directed choices (autonomy), engage with the broader TRIFECTA community (relatedness), and enhance their confidence in managing their diabetes and associated emotions (competence).

TRIFECTA was designed specifically for adults with T1D, as peer support is delivered by other adults with the same condition and self-management support is delivered by health professionals who specialize in T1D. This intervention offered three digital support platforms: (1) monthly group-based interactive Zoom sessions (Virtual Huddles) delivered by a health/mental health professional, (2) a 24/7 WhatsApp peer texting group and (3) an “Ask the Expert” web-based self-management support portal. Participants were provided with a welcome manual and instructed on how to use each platform. They were also encouraged to access any or all platforms as frequently as desired.

Virtual huddles were hosted by health or mental health professionals and focused on various diabetes-related topics such as eating strategies during the pandemic, T1D and aging, mindfulness and meditation, T1D and financial freedom, etc. Participants were invited to submit questions prior to the session as well as post questions in the chat box during the Zoom session. Typically, sessions started with a brief presentation followed by a question-and-answer period and ended with an interactive group discussion.

The 24/7 WhatsApp peer texting platform offered a safe space to deliver and receive emotional, informational and community support. Participants posted messages on various topics including technology and device-related queries, benefits and limitations of different healthcare plans, low-carb recipes, challenges due to COVID-19 pandemic, etc. To ensure that clinical or medical advice was not exchanged, all communications were monitored by a diabetes nurse educator and registered psychologist.

The “Ask the Expert” web-based portal allowed participants to submit self-management questions to an interdisciplinary team of health and/or mental health professionals. Whichever team member opened the email first and was qualified to respond posted an answer within 48 h.

2.2. Measures

DSQoL was assessed using the Type 1 Diabetes and Life (T1DAL) self-report survey which includes different versions for the following age bands: 18–25 years; 26–45 years, 46–60 years, and over 60 years. Each version has 4 or 5 subscales and can cover areas such as emotional experiences and daily activities; support from others; financial considerations, etc. [8,9]. Each item was rated on a 5-point Likert scale based on experiences in the past four weeks with reverse scoring for negatively worded items. Higher T1DAL scores indicate higher DSQoL. In contrast to other DSQoL measures that pertain to all adults ≥18 years, the T1DAL survey addresses age-relevant T1D experiences across the lifespan. This age stratification allows for sub-group analysis.

2.3. Statistical Analyses

Descriptive statistics for sociodemographic characteristics and for primary outcomes were summarized with frequencies and percentages for categorical variables and means and standard deviations for continuous variables. Wilcoxon Signed-Rank nonparametric tests were used to determine examine the statistical significance of pre-post score change on T1DAL overall sample scores, on T1DAL age band subsample scores, and T1DAL subscale scores. Association among TRIFECTA engagement metrics (counts of Huddle sessions attended, Ask the Expert posts, Ask the Expert post views, and WhatsApp peer support messages) and T1DAL change scores for each age band were examined using Spearman nonparametric correlations corrected for false discovery rate using the Benjamini-Hochberg procedure. Independent Kruskal-Wallis tests examined the statistical differences between age cohorts as grouped in accord with the T1DAL measure on TRIFECTA engagement metrics. Significant Kruskal-Wallis tests were followed up with Dunn post-hoc tests to examine where significant pairwise differences existed between age cohorts. Nonparametric tests were used given small subsample sizes as these tests have less statistical assumptions about the distribution of scores and provide exact tests based on the sample. All analyses were performed in R version 3.6.2 statistical software [10].

3. Results

3.1. Participant Characteristics

Table 1. Sample sociodemographic and clinical characteristics.

		M	SD
Age		38.88	15.08
		N	%
Gender	Male	14	23.3%
	Female	45	75.0%
	Other	1	1.7%
Ethnicity	White	44	73.3%
	South Asian (Bangladeshi, Pakistani, Indian)	6	10.0%
	East Asian (Chinese, Japanese, Korean)	2	3.3%
	Other	5	8.4%
	No answer	3	5.0%
Marital Status	Married/Live with Partner	27	45.0%
	Never Married	26	43.3%
	Separated/Divorced/Widowed	4	6.7%
	No answer	3	5.0%
Birth Country	Canada	48	80.0%
	Other	12	20.0%
Education	Less than College	16	26.7%
	Bachelors Graduate	30	50.0%
	Masters/Doctoral Graduate	13	21.6%
	No answer	1	1.7%
Annual Income	$0–49,000	19	38.0%
	$50–69,000	11	22.0%
	≥$70,000	20	40.0%
	No answer	10	16.7%
Employment	Full- or Part-time	36	60.0%
	Retired	8	13.3%
	Unemployed	2	3.4%
	Student	8	13.3%
	No answer	6	10.0%
Extended Health Coverage		44	73.3%
Age of Diabetes Diagnosis	<10 years old	16	26.7%
	10–17 years old	14	23.3%
	18–34 years old	22	36.7%
	≥35 years old	8	13.3%
Glucose Monitoring	Continuous Glucose Monitoring (CGM)	39	65.0%
	Standard Meter	20	33.3%
	No answer	1	1.7%
Insulin Delivery	Multiple Daily Injections	23	38.3%
	Insulin Pump	33	55.0%
	Both	3	5.0%
	No answer	1	1.7%

Note. Income currency is in Canadian dollars. M = Mean; SD = Standard Deviation; N = Count; % = relative frequency.

presents a summary of baseline participant demographic and diabetes-related information. Sixty participants were enrolled in this study and approached to complete baseline and follow-up measures. The sample was predominantly White (73%) and female (75%). Nearly half were married or living with a partner and most of the remaining half had never been married (see Table 1). Nearly three-quarters of the sample had a bachelor’s degree or higher and most were employed either full- or part-time. Most participants were using insulin pump and continuous glucose monitoring (CGM) technologies to manage diabetes.

3.2. Pre-Post Changes in T1DAL Scale

Table 2. DSQoL changes pre- and post- digital support intervention.

	Baseline			Follow-Up				95% CI
	N	Mean	SD	n	Mean	SD	d	Lower	Upper	p-Value
Overall T1DAL	57	48.86	13.96	58	51.97	14.29	0.219	−0.147	0.584	0.039 *
T1DAL Age Band 18–25	11	50.15	12.94	11	50.72	12.06	0.044	−0.792	0.880	0.999
Emotional Experiences & Daily Activities	11	30.44	14.52	11	29.98	18.28	−0.026	−0.862	0.810	0.678
Handling Diabetes Well	11	64.77	26.26	11	72.73	13.77	0.310	−0.526	1.146	0.322
Peer Relationships	9	72.22	10.64	9	71.67	25.37	−0.018	−0.942	0.906	0.778
Healthcare Experiences	11	65.91	25.52	11	64.2	26.38	−0.063	−0.899	0.773	0.766
T1DAL Age Band 26–45	30	50.53	12.9	31	55.39	14.62	0.344	−0.158	0.846	0.022 *
Emotional Experiences & Daily Activities	30	38.07	14.05	32	42.26	15.31	0.281	−0.217	0.779	0.028 *
Family Relationships	32	72.14	18.04	32	76.04	20.27	0.200	−0.290	0.691	0.532
Peer Relationships	32	69.73	22.06	27	75.69	17.62	0.288	−0.225	0.800	0.253
Healthcare Experiences	30	60	25.75	32	66.67	27.76	0.245	−0.253	0.744	0.059
Financial Considerations	29	40.8	29.32	30	42.5	28.23	0.058	−0.452	0.569	0.717
T1DAL Age Band 46–60	8	45.1	10.37	8	47.66	14.38	0.175	−0.805	1.155	0.547
Emotional Experiences & Daily Activities	7	27.55	18.87	7	29.08	18.14	0.077	−0.971	1.125	0.933
Handling Diabetes Well	8	70.31	17.28	8	66.41	11.05	−0.200	−1.180	0.780	0.341
Support from Others	6	81.25	11.18	7	71.43	24.7	−0.410	−1.501	0.680	0.581
Social Isolation	8	34.38	14.56	8	44.53	22.52	0.456	−0.524	1.436	0.049 *
Financial Considerations	8	34.38	16.37	7	44.64	18.55	0.550	−0.464	1.565	0.106
T1DAL Age Band Over 60	8	44.63	21.7	8	44.77	14.07	0.006	−0.975	0.986	0.641
Emotional Experiences & Social Isolation	8	36.9	25.83	7	38.31	16.88	0.056	−0.959	1.071	0.675
Handling Diabetes Well	8	63.12	15.1	8	59.38	13.74	−0.243	−1.223	0.737	0.438
Support from Others	8	51.04	27.25	8	56.25	24.3	0.189	−0.791	1.169	0.590
Financial Considerations	6	34.72	39.94	6	41.67	31.18	0.153	−0.979	1.285	0.999

Note. Overall scores for the T1DAL is presented along with the scores for each age band on the T1DAL. d corresponds to Cohen’s d effect size for repeated samples with Hedges’ correction for smaller sample sizes applied. The 95% Confidence Interval (CI) is also presented for this d statistic. p-value corresponds to the Wilcoxon Signed Rank test for T1DAL overall score, T1DAL age band scores, and subscale scores. These tests and p-values determined if the change score from baseline to follow-up assessment is statistically significant for each outcome. * designates statistical significance with α = 0.05. 60 participants were enrolled and completed baseline measures while 59 completed follow-up measures. Within the sample, 11 participants were 18–25-year-olds, 33 participants were 26–45-year-olds, 8 participants were 46–60-year-olds, and 8 participants were >60-year-olds. N’s listed in table may vary based on missing item data that could have precluded total or subscale score calculation. SD = Standard Deviation; T1DAL = Type 1 Diabetes and Life measure. Underlining is used to demarcate separate age cohorts of the T1DAL. Bold typeface is used to demarcate overall scores while regular typeface demarcates subscales of each respective measure.

presents the means and standard deviations for the T1DAL at baseline and 6-months. Table 2 also shows the p-values of the Wilcoxon Sign Rank test for the overall T1DAL score, T1DAL scores by age band, and for the subscale scores. As shown in the table, overall T1DAL scores demonstrated a significant increase over time with a small effect size (d = 0.219). T1DAL scores changed more for 26–45-year-old participants, particularly regarding improvements in the Emotional Experiences & Daily Activities subscale of the T1DAL.

3.3. TRIFECTA Engagement Metrics & Associations with T1DAL Scales

The descriptive statistics associated with engagement in the TRIFECTA peer support platforms were reported in a prior publication [6]. Descriptive statistics for each platform engagement metric were as follows: Huddle Sessions M = 2, SD = 2; Ask-the-expert Posts M = 1, SD = 2; Ask-the-expert Views M = 19, SD = 18; and WhatsApp Messages M = 22, SD = 36. Change in overall T1DAL score for 26–45-year-old’s was significantly associated with Ask-the-expert Views (ρ = 0.574, B-H adj. p-value = 0.008) and with Huddle Sessions attended (ρ = 0.515, B-H adj. p-value = 0.016). Counts of WhatsApp Messages and Ask-the-expert Posts were not associated with T1DAL change scores for this age band. No other associations for other age bands were found to be statistically significant.

When examining association of age cohort on TRIFECTA engagement metrics, age cohort was associated with Huddle Session count (p = 0.038) and Ask the Expert post views (p = 0.039). For Huddle sessions, 26–45-year-olds engaged more often than 18–25-year-old’s (M = 2.06 vs. 0.91; p = 0.038) and Over 60-year-olds engaged more often than 18–25-year-old’s (M = 3.12 vs. 0.91; p = 0.005). For Ask the Expert post views, 46–60-year-olds engaged more often than 18–25- year-old’s (M = 26.12 vs. 10.27; p = 0.014) and Over 60-year-olds engaged more often than 18–25- year-old’s (M = 27.00 vs. 10.27; p = 0.016). Age cohort was not significantly associated with Ask the Expert post counts or with WhatsApp peer messages counts.

4. Discussion

To date, DSQoL surveys for adults have been developed and validated for all adults ≥18 rather than tailored to age-bound subgroups such as young adults, middle-aged adults, and older adults. This is one of the few studies that has utilized a DSQoL instrument (T1DAL) customized across the adult lifespan [9]. As such, we can examine, more meaningfully, life domains distinctly relevant to each age group (e.g., 18–24; 25–45; 26–59; ≥60) as well as how these cohorts respond, differentially, to the same intervention. Following the TRIFECTA intervention, overall improvements in DSQoL were detected for the total sample, primarily driven by DSQoL elevations in the 26–45-year cohort. Subscale analysis found adults 26–45 years old reported higher levels for emotional experiences and daily activities, and additionally, adults 46–60 experienced reduced social isolation.

As such, we can examine, more meaningfully, life domains distinctly relevant to each age group (e.g., 18–24; 25–45; 26–59; ≥60) as well as how these cohorts respond, differentially, to the same intervention. Following the TRIFECTA intervention, overall improvements in DSQoL were detected for the total sample, primarily driven by DSQoL elevations in the 26–45-year cohort. Subscale analysis found adults 26–45 years old reported higher levels for emotional experiences and daily activities, and additionally, adults 46–60 experienced reduced social isolation.

Consistent with our findings, among 193 children 10–18 years with T1D, Iafusco and colleagues [11], also reported elevations in overall DSQoL following a 2-year digital health intervention involving monthly synchronous chat line with peers and health professionals [11]. Although, when stratifying by age cohort, we found only the cohort with the largest number of participants (26–45 years) reported increases in DSQoL. This finding could be related to having more statistical power due to larger subsample size or due, in part, to differential engagement by age. In fact, the 26–45 year old cohort attended more Huddle sessions compared to their 18–25 year old counterparts. Interestingly, previous research also found higher utilization rates for digital health management interventions among “middle-aged adults” compared to “young adults” [12]. Subsequent research should examine a possible dose response relationship with DSQoL change across age bands with larger subsample sizes to verify this preliminary finding.

Not surprisingly, research has found that, during COVID-19, adults with T1D reported elevations in social isolation [13,14]. However, with access to 3 digital support modalities through TRIFECTA, our 46–60 year old group experienced reductions in social isolation. Similarly, Comtois and colleagues also found usage of commercial mental health support mobile apps during COVID to be associated with improved psychosocial outcomes [15]. Interestingly, social isolation scores remained unchanged for adults older than 60 years. However, previous research suggests that older adults derive greater meaning from interactions with close friends and family and, thus, may be less reliant on social networks [16,17]. Moreover, because the consequences of isolation and loneliness can be worse for younger adults compared to older adults [18], it is possible that the TRIFECTA intervention had a more positive impact on the former. Nonetheless, efforts to engage older adults than 60 years should be considered as research suggests that this age group is less likely to utilize digital health or social services [19]. Offering virtual orientations or instructional videos that demonstrate how to navigate the respective digital interventions may boost participation.

Digital support interventions such as TRIFECTA may disproportionately attract younger and more technology savvy end users. In fact, 68% of our sample were between the ages of 18 to 45 years while only 27% were 46 years and older. While research shows that older adults are aware of online resources for mental health support [20], their difficulty navigating mobile apps and websites [20,21,22], preference for in-person versus virtual interactions [20,21], and concerns around privacy and security [21,23] can potentially produce selection bias in recruitment. However, emphasizing the benefits of interpersonal connection and social support for this age group can overcome these obstacles to engagement [21,22].

This study is not without its limitations. First, in the absence of a control group, we are not able to conclude that any improvements observed are directly related to the intervention itself. Second, potential bias including recruiting a self-selected sample, natural changes in DSQoL over time, testing effect, and confounding variables (rural vs. urban residence) may affect the outcome. Third, while using an age band-specific instrument increases specificity, statistical power is reduced substantially, particularly for the 46–60 age cohort (n = 8). Achieved power based on chosen statistical test, sample size, and estimated effect size was only 0.37 indicating high chance of type II error. Thus, results should be interpreted with caution. Finally, because our sample was largely White and female, results cannot be generalized to the larger T1D population nor can results reflect non-lockdown conditions. Future studies should recruit a more sociodemographically diverse sample size across age bands.

5. Conclusions

Digital health interventions have the potential to improve psychosocial outcomes for the T1D community particularly those who are geographically marginalized or have limited access to traditional face-to-face self-management support. By examining the unique impact of these interventions on different age groups, we can better customize interventions for the challenges experienced at different developmental stages of life.