Risky self-management behaviors in adolescents with type 1 diabetes: Measurement validation for the Diabetes-Specific Risk-Taking Inventory (DSRI)

Abstract

Purpose:

Among persons with type 1 diabetes (T1D), adolescents often experience the greatest challenge achieving optimal treatment engagement and glycemic targets. Risk-taking behaviors often increase during adolescence and may interfere with engagement in T1D care. We developed the Diabetes-Specific Risk-Taking Inventory (DSRI) to assess risky T1D self-management behaviors in adolescents with T1D. In the current study, we aimed to examine the DSRI’s psychometric properties.

Methods:

We surveyed a national sample of 224 adolescents from the T1D Exchange registry (M age = 16.9 ± 1.1, 49% female, M A1c = 8.5% ± 1.3, 76.8% on insulin pumps) in a cross-sectional design. Participants completed the DSRI and measures of engagement, general risk-taking, and executive functioning and reported on incidence of severe hypoglycemia and diabetic ketoacidosis over the past year.

Results:

The DSRI demonstrated reliability (internal consistency: α = .89; test-retest reliability: r = 0.86, p < .01). Concurrent validity was demonstrated through significant associations between the DSRI and T1D engagement (r = −.75), general risk-taking (r = .57), executive dysfunction (r = .34), and report of severe hypoglycemia over the past year (r = .22). The DSRI accounted for unique variance in adolescents’ most recent glycated hemoglobin, above and beyond other variables, indicating its incremental validity.

Conclusions:

Overall, initial psychometrics suggest the DSRI is a reliable and valid measure of risks that adolescents may take with their T1D care. This innovative self-report measure has potential to be an actionable clinical tool to screen for high-risk behaviors not routinely assessed in T1D clinical care.

Introduction

Type 1 diabetes (T1D) is a prevalent and serious chronic illness among adolescents. In the United States, recent estimates suggest 187,000 youth under age 20 live with T1D, and more than 18,000 youth under age 20 are newly diagnosed with T1D every year.¹ T1D requires complex daily treatment, which includes insulin injections or use of subcutaneous insulin pump, blood glucose monitoring, monitoring of carbohydrate intake and physical activity, and correction of blood glucose fluctuations. Despite improvements in T1D treatment due to recent technological advances (e.g., insulin pump and continuous glucose monitor), adolescents and young adults are the only age group for which glycemic levels² and mortality have not improved.³ Among persons with T1D, adolescents have the lowest levels of treatment engagement and the highest glycated hemoglobin levels (A1c) of any other age group.² Low engagement in T1D treatment and high A1c are linked to increased risk for acute and long-term health complications and mortality.⁴

General risk-taking behaviors (e.g., drinking and driving, unprotected sex) peak in adolescence and are associated with morbidity and mortality in the general population.⁵ A perspective from developmental neuroscience depicts an increased propensity toward risk-taking in adolescence due to a maturational imbalance between the surge in activity between socio-emotional neural pathways (related to heightened reward sensitivity) and still developing executive functioning neural pathways (related to cognitive control, caution, and inhibition).^6–8 Adolescents with T1D are no exception, as they generally engage in similar levels of general risk-taking behaviors as their non-T1D peers.^9–11 Thus, risk-taking behavior increases around the same time that engagement in diabetes self-management generally decreases and A1c increases, potentially placing adolescents at elevated risk for deleterious outcomes related to their diabetes. However, little is known about how risk-taking behaviors influence engagement in diabetes self-care.

In a recent publication, we introduced the novel idea of illness-specific risk-taking, which we define as a type of risk-taking in which youth make decisions about disease management that put them at risk for suboptimal health outcomes.¹² For example, driving without first checking blood glucose might increase the odds for a teen having hypoglycemia and losing consciousness while driving, or going 24 hours without insulin could increase the odds of a teen experiencing an urgent medical event, such as diabetic ketoacidosis (DKA). Clinically, adolescents with T1D are known to take risks with their diabetes care, but these risky diabetes behaviors have not been empirically evaluated. Thus, what remains unknown is how frequently these types of risk-taking behaviors occur, why they occur, and how these behaviors contribute to A1c, glycemic excursions, and life-threatening acute complications (i.e., severe hypoglycemia and DKA) in adolescents with T1D.

Diabetes-specific risk-taking behaviors are not captured in any currently published diabetes-specific measures of adherence, general risk-taking, diabetes knowledge, or psychosocial functioning in youth with T1D. Moreover, having conversations with adolescents about risk-taking during a clinic appointment is likely difficult, because of time constraints or because they may involve disclosing sensitive information (e.g., drug and alcohol use).¹³ Thus, a questionnaire may provide a more feasible way to obtain reliable information about adolescent risk-taking.

To address this gap in available assessment tools and empirically examine diabetes-specific risk-taking behaviors, we developed and piloted the Diabetes-Specific Risk-Taking Inventory (DSRI).¹⁴ The DSRI is a self-report measure of risks that adolescents may take related to their T1D management. It was developed by a multidisciplinary team of experts (psychologists, endocrinologists, and nurse practitioners) and with input from adolescents with T1D. The DSRI could allow clinicians to create more individualized interventions to address unhealthy risk-taking behaviors. The DSRI also has the potential to be a powerful research tool that could help identify which diabetes risk-taking behaviors are most common and most detrimental to health and glycemic outcomes.

The objective of this study was to assess psychometric properties of the DSRI questionnaire. Specifically, we aimed to examine internal consistency, test-retest reliability, concurrent validity, external validity, and incremental validity of the DSRI. We hypothesized the following: 1) The DSRI will show good internal consistency (Cronbach alpha ≥ .80), reliability (inter-item correlations between 0.15 and 0.50), and test re-test reliability via a moderate correlation between DSRI scores at two time points; 2) The DSRI will demonstrate concurrent validity with established measures of related constructs including level of T1D engagement, general risk-taking behaviors, and executive functioning; 3) The DSRI will demonstrate external validity via correlations with A1c values, episodes of DKA, and episodes of severe hypoglycemia; 4) The DSRI will demonstrate incremental validity via a significant relation with A1c, even after controlling for adolescents’ T1D engagement, general risk-taking behaviors, and demographic factors that are known to be associated with differences in adolescent A1c.

Methods

Participants and Procedures

Participants were 224 adolescents with T1D (age range 15–18 years old) who were recruited in June and July of 2017 from a national clinic-based research registry for people with T1D: the “T1D Exchange” (T1Dx). The T1Dx registry includes a database of clinical information on people of all ages with T1D from participating diabetes clinics across the United States. Once a participant enrolls in the T1Dx registry, the T1Dx data coordinating center receives annual updates of information from the participant’s electronic medical record (e.g., A1c, number of hospitalizations, etc.). For the present study, T1Dx staff identified eligible participants based on the following inclusion criteria: age 15–18 years old; T1D duration > 1 year; had an A1c in the T1Dx database within the last year; most recent A1c ≥ 6.0% and ≤ 13.0%; and English speaking. The research team decided to exclude younger adolescents (age 14 and under) to ensure there would be opportunity for participants to endorse all items on the measure. In general, the opportunity for risk-taking increases in mid to late adolescence, as teenagers attend high school and spend more time away from parents. The study team also decided to exclude older adolescents (19 and older), because most teens in the United States graduate high school at the age of 18, and this transition marks a qualitatively different stage in life for many young people (e.g., transition to college or a job, transition from pediatric to adult care). For the current study, we decided to prioritize having a more age- homogenous group of participants to aid in evaluating the validity of the measure. On the basis of these inclusion criteria, T1Dx staff identified 1811 eligible youth from 54 clinical sites in the T1Dx registry.

The study team sent recruitment flyers via email to eligible participants. Interested families followed a link to the study’s REDCap^15,16 website to provide contact information. For participants under 18, staff sent REDCap consent forms to the parent and assent forms to the adolescent. For participants who were 18 years old, staff sent a link to the REDCap consent form, from which they could proceed to the survey. Participants completed a 15-minute online survey through REDCap. One month later, they received an automatic invitation from REDCap to complete a follow-up survey. Participants completed all questionnaires at baseline and only the DSRI at follow-up. Participants received $15 for the first survey and $35 for the follow-up survey. The Jaeb Center for Health Research, which was the data coordinating center for T1Dx, matched the participant to their information in the T1Dx database to retrieve their demographic information and A1c values over the previous year.

Of the 1811 eligible youth, 385 families provided their contact information to the study team, and we successfully contacted 247 via phone. Of those contacted, three were ineligible and two declined participation. We sent the REDCap link for consent/assent forms to the 242 families that passed the phone screening, and 224 completed the forms and baseline survey. The REDCap data are housed on a secure website where only members of the research team are able access the data. Of those who completed the first survey, 165 adolescents also completed the 1-month follow-up survey. Participant demographics and clinical characteristics are summarized in Table 1.

Table 1.

Participant demographics and clinical characteristics.

	Percent (n)	Mean (SD)
Adolescent age, years		16.9 (1.1)
Adolescent sex, % female	49 (110)
Annual family income
< $25,000	6 (11)
$25,000-$34,999	7 (13)
$35,000-$49,999	11 (19)
$50,000-$74,999	12 (22)
$75,000-$99,999	19 (34)
≥ $100,000	44 (78)
Insurance status
Private insurance	84 (173)
Other insurance	16 (32)
No insurance	1 (2)
Race/Ethnicity
Non-Hispanic White/ Caucasian	88 (198)
Hispanic/Latino	5 (11)
Non-Hispanic Black/ African
American	3 (6)
Asian	2 (5)
“Other”	2 (4)
Diabetes duration		9.1 (3.6)
Insulin administration, % insulin pump	77 (172)
CGM use in the last year	55 (124)
Most recent % hemoglobin A1c		8.6 (1.3)
Episode of DKA in the last year (survey report):
“Yes”	7 (15)
“No”	69 (155)
“I don’t remember”	3 (7)
“I don’t know what DKA is”	21 (47)
Episode of severe hypoglycemia in last year (survey report):
“Yes”	9 (20)
“No”	90 (201)
“I don’t remember”	1 (3)

Note: Percentages do not always add up to 100 because of rounding. CGM = continuous glucose monitor, DKA = diabetic ketoacidosis

The Institutional Review Board at the Jaeb Center for Health Research approved this study and served as the central IRB with collaborating institutions operating under approved reliance agreements.

Measures

The Diabetes-Specific Risk-Taking Inventory (DSRI) was used to assess diabetes-specific risk-taking behaviors.¹⁴ We previously developed and piloted the DSRI with 13 diabetes health care providers and 30 adolescents (age 15–19 years) with T1D to establish face validity. ¹⁴ The DSRI is a self-report questionnaire that instructs adolescents to report the frequency with which they have engaged in each of 31 T1D-specific risk-taking behaviors “over the last year.” Answer options include “daily,” “weekly,” “monthly,” “every few months,” “once/twice a year,” “never,” “N/A,” or “prefer not to answer.” To score the DSRI, we calculated an average of the following values for each item: 0 for N/A to 6 for daily. To aid in interpretability, we then calculated the score as a percentage, by dividing the average by 6 (the maximum possible score) and then multiplying it by 100 to make it on a 0 to 100 scale, with a higher score indicating more frequent risk-taking behavior. The REDCap system did not permit participants to skip items but did offer a “prefer not to answer” response option for each question. We examined the number of participants who responded “prefer not to answer” for each item. No item had more than 3% of respondents choose “prefer not to answer.” Similarly, no adolescent responded “prefer not to answer” for more than two items on the DSRI. Items that a teen answered “prefer not to answer” were dropped when calculating the average (e.g., for those with two “prefer not to answer” responses, we calculated a mean from 29 items instead of 31). We assessed internal consistency (Cronbach alpha) as a part of the current study.

The Self-Care Inventory-Revised (SCI-R)¹⁷ was used to assess engagement in diabetes self-management behaviors over the last 2 months. Higher scores on this 15-item self-report questionnaire indicate more frequent diabetes self-management behaviors. The SCI-R has demonstrated good psychometric properties in previous studies¹⁷ and demonstrated good reliability in this sample (α = .80).

The risk-taking subscale of the Risk-Taking and Self-Harm Inventory for Adolescents (RTI-A)¹⁸ was used to measure adolescent self-report of the frequency (“never (0),” “once (1),” “more than once (2),” “many times (3)”) with which they engage in general risk-taking behaviors such as risky driving and substance use. A total score is calculated, with a higher score indicating more frequent risk-taking behavior. The RTI-A risk-taking subscale has demonstrated good internal consistency in a large sample of youth aged 11 to 18 years (α = .85)¹⁸ as well as in the present study (α = .80).

The Behavior Rating Inventory of Executive Functions, 2^nd Edition, Short-Form Self-Report (BRIEF-2 SF-SR)^19,20 is a 12-item measure of general executive functioning difficulties in youth aged 5 to 18 years. The BRIEF-2 SF-SR has excellent psychometric properties and uses t-scores based on normative data to report values.²⁰ Higher t-scores indicate greater executive dysfunction. The BRIEF-2 SF-SR demonstrated good reliability in the current sample (α = .85).

Demographic and clinical characteristics including adolescent age, sex, race/ethnicity, diabetes duration, mode of insulin administration (pump vs. injections), and insurance status (private, public, or none) were provided by the adolescent.

Diabetes health outcomes.

This study used A1c values collected from each adolescent’s electronic medical record, as reported to the T1Dx registry. All adolescents had an A1c result within the last 10 months prior to the date of baseline survey completion. We used the most recent A1c value for the current study. Higher A1c values indicate higher average blood glucose over the previous 2 to 3 months. Adolescents also self-reported on whether they had an episode of DKA over the last year, and whether they had an episode of severe hypoglycemia (a low blood glucose event that required the help of someone else to treat it) over the last year. Response options for both variables were dichotomous (was there an episode in the last year: yes or no), and both variables were analyzed separately from one another.

Statistical Analysis

All analyses were conducted using Stata v15. We evaluated descriptive statistics to characterize the sample and the DSRI summary data. Reliability and validity were assessed in accordance with published psychometric measure development and validation guidelines.²¹

First, we sought to identify characteristics of the DSRI, by evaluating associations with key demographic and clinical variables using Pearson correlations (age and diabetes duration), independent samples t-tests (sex, insurance status, current use of insulin pump, and current use of a continuous glucose monitor), and ANOVAs (race/ethnicity, and annual family income). Second, reliability of the DSRI was assessed via inter-item correlations and Cronbach’s alpha. We also assessed test-retest reliability by examining correlations between DSRI scores from the baseline survey and the 1-month follow-up. Third, to assess concurrent validity, we assessed Pearson correlations with each measure of similar constructs (general risk-taking behaviors and executive functioning difficulties) and dissimilar constructs (engagement in T1D self-management). We also examined associations when the data was stratified by sex and age to determine whether concurrent validity of the measure was similar across age and sex differences, considering both factors have been associated with differences in general risk-taking behavior. To compare correlations between these groups (2 groups based on sex: male and female, 4 groups based on age: 15, 16, 17, and 18 years old), we used meta-analyses, including chi-squared test for heterogeneity. To assess external validity, we evaluated the DSRI in relation to A1c, reported episodes of DKA, and episodes of severe hypoglycemia. Finally, we examined incremental validity using a step-wise linear regression model with A1c as the dependent variable and adolescent age and sex, the SCI-R, RTI-A, and DSRI as independent variables. All statistics were evaluated using p < .05 to denote significance.

Results

Descriptive Statistics

Participant demographic and clinical characteristics are summarized in Table 1. Average time between hemoglobin A1c value and survey completion was 123.3 days, or approximately 4.1 months. Hemoglobin A1c values ranged from 6.0% to 14.1%. Of the possible score range on the DSRI (0–100), the mean score in this sample was 36.8 ± 12.0 (range = 15.7–85). Respondents endorsed the full range of responses (i.e., 0–6) on most of the 31 items. For 3 of the items (“Waited until you were out of insulin before telling your parents or getting more from the pharmacy”, “Gone without taking insulin for at least 24 hours”, and “Gotten drunk to the point where you could not take care of your diabetes”) none of the participants chose the most frequent response option (daily). For one item (Told your doctor you had taken insulin when you really had not”), none of the participants endorsed the two most frequent response options (daily or weekly). Finally, there was also an item on which none of the participants endorsed the “not applicable” option (“Told your parents you had taken insulin when you really had not”). Item means ranged from 0.8 to 4.3 (SE = 0.0–0.2), and kurtosis ranged from −1.5 to 14.5 (SE = 0.32); total score skewness = 0.9 and kurtosis = 1.3, indicating normal skewness and positive kurtosis for the DSRI total score. See Table 2 for the frequencies of responses to each item.

Table 2.

Items and response options for the DSRI. Frequencies of endorsed responses for each item are indicated in columns.

Instructions: Over the last year, how often have you?
		6 Daily (%)	5 Weekly (%)	4 Monthly (%)	3 Every few months (%)	2 Once/twice a year (%)	1 Never (%)	0 N/A (%)	Mean ± SD
1	Guessed the number of carbs in a snack/meal you were eating when the nutrition information was right there?	24.6	33.5	13.4	12.5	6.3	9.4	0.4	4.4 ± 1.6
2	Decided not to wear a diabetes ID?	49.6	7.1	4.9	6.7	8	14.3	8	4.1 ± 2.2
3	Engaged in physical activity without wearing a diabetes ID?	49.1	15.6	3.6	4.5	6.3	13.8	5.8	4.3 ± 2.2
4	Engaged in physical activity without first checking your blood sugar?	17.9	32.1	15.6	13.4	7.1	13.4	0.4	4.1 ± 1.6
5	Participated in an organized sport or activity without telling the coach you have diabetes?	2.2	0.9	4.5	5.8	7.6	69.2	9.8	1.4 ± 1.2
6	Ate without first checking your blood sugar?	18.3	37.1	17.9	8.9	10.3	6.7	0.9	4.3 ± 1.5
7	Felt your blood sugar might be low and did not treat it?	0.4	5.8	7.1	10.7	14.3	60.3	0.9	1.8 ± 1.3
8	Had high blood sugars and did not check ketones?	17	26.8	20.1	17	10.3	7.1	1.3	4.0 ± 1.5
9	Felt nauseous and/or vomited and did not check ketones?	3.1	3.1	3.1	6.3	20.5	57.1	6.7	1.7 ± 1.3
10	Gone without checking your blood sugar for at least 24 hours?	1.3	5.8	6.7	5.8	17	62.1	1.3	1.8 ± 1.3
11	Told someone you checked your blood sugar when you really had not?	4	15.6	16.1	17.4	16.5	29.9	0.4	2.8 ± 1.6
12	Entered made-up blood sugar numbers in your log book?	2.2	4.5	2.7	5.4	11.2	54.5	19.6	1.3 ± 1.3
13	Reported a made-up blood sugar number to someone?	5.4	9.4	12.1	13.4	19.6	38.4	1.3	2.6 ± 1.6
14	Taken insulin without checking your blood sugar first?	14.3	30.4	14.3	11.6	14.3	14.7	0.4	3.8 ± 1.7
15	Ate without taking short-acting insulin to cover carbs (except when BG is low)?	6.7	21	19.2	17	13.4	20.1	2.2	3.3 ± 1.7
16	Taken less insulin than you knew you needed?	0.9	12.5	13.4	17	11.6	42.9	1.8	2.4 ± 1.6
17	Taken more insulin than you knew you needed?	2.7	14.3	11.2	15.6	17	37.9	0.9	2.5 ± 1.6
18	Waited until you were out of insulin before telling your parents or getting more from the pharmacy?	0	2.7	4.5	4.9	10.7	75.9	0.9	1.5 ± 1.0
19	Waited until your glucagon had expired before telling parents or getting more from the pharmacy?	0.9	0.9	1.3	3.6	6.3	77.2	9.4	1.1 ± 0.9
20	Gone without taking insulin for at least 24 hours?	0	1.8	1.8	2.7	8.9	83.5	0.9	1.2 ± 0.7
21	Told your parents you had taken insulin when you really had not?	1.8	5.8	8.5	9.8	17.4	56.3	0	1.9 ± 1.3
22	Told your doctor you had taken insulin when you really had not?	0	0	3.1	2.2	8.9	84.4	1.3	1.2 ± 0.6
23	Driven a car without first checking your blood sugar?	19.2	12.9	7.1	8	8.5	23.7	20.5	2.8 ± 2.2
24	Driven a car without fast-acting carbs within reach?	7.1	6.3	4.5	6.7	7.6	48.7	19.2	1.8 ± 1.8
25	Drank alcohol without eating extra carbs?	0.4	0.9	3.6	4	7.1	53.6	30.4	1.0 ± 1.0
26	Drank alcohol when no one around knew you had diabetes?	0.4	1.3	1.3	1.8	1.8	62.9	30.4	0.9 ± 0.9
27	Drank alcohol without wearing a diabetes ID?	1.8	2.2	3.1	5.8	6.3	46	34.8	1.1 ± 1.3
28	Gone to sleep after drinking alcohol with no plan for checking blood sugars during the night?	0.4	2.2	2.7	3.6	6.7	50.9	33.5	1.0 ± 1.1
29	Gotten drunk to the point where you could not take care of your diabetes?	0	0.9	1.8	0.4	4	60.7	31.7	0.8 ± 0.8
30	Used drugs (for example: pills that were not prescribed to you, marijuana, etc) when no one around knew you had diabetes?	0.9	0	1.8	1.3	1.3	68.8	25.9	0.9 ± 0.8
31	Had sex without first checking your blood sugar?	2.7	5.4	7.1	3.1	3.1	50	27.2	1.4 ± 1.5

Note: Please contact the first author with requests to use this measure in any research or clinical setting. Percentages may not add up to 100% because of some participants responding, “prefer not to answer.”

The DSRI score was significantly and positively correlated with age (r = 0.21, p < .05; 15-year-olds M = 33.9 ± 10.4; 16-year-olds M = 35.8 ± 11.8; 17-year-olds M = 39 ± 14.1; 18-year-olds M = 39.5 ± 10.7). There were no significant differences in DSRI score based on sex, race/ethnicity, T1D duration, insurance status, pump use, continuous glucose monitor use, or family income.

Reliability

Data indicated excellent internal consistency for the DSRI, α = .89, while inter-item correlations ranged from .20 to .64 with an average inter-item covariance of .47. The DSRI also showed good test-retest reliability (r = 0.86, p < .01).

Validity

The DSRI demonstrated concurrent validity, via significant correlations with the SCI-R (r = −0.74, p < .05), the RTI-A (r = 0.57, p < .05), and the BRIEF-2 SF-SR (r = 0.28, p < .05; Table 3). When stratified by sex, these correlations remained similar, except for the RTI-A (χ²(1)= 13.45, p<.01), which demonstrated a significantly stronger association with the DSRI among females (r = 0.72, 95% CI [0.61, 0.80]) versus males (r = 0.38, 95% CI [0.20, 0.53]). When stratified by age, the correlations for each age were not significantly different.

Table 3.

Correlation matrix of associations among DSRI score, demographic and clinical characteristics, and measures of concurrent and external validity.

		1	2	3	4	5	6
1	DSRI
2	Age	.21*
3	Duration	.10	.17*
4	A1c	.40**	.11	.16
5	SCI-R	−.74**	−.24**	−.09	−.42**
6	RTI-A	.57**	.16	.01	.24*	−.38**
7	BRIEF-2 SF-SR	.28**	−.03	.04	.19*	−.28**	.25**

DSRI = Diabetes-Specific Risk-Taking Inventory; Duration = duration of T1D; A1c = glycosylated hemoglobin A1c; SCI-R = Self-Care Inventory-Revised; RTI-A = Risk-Taking Inventory for Adolescents; BRIEF-2 SF-SR = Behavior Rating Inventory of Executive Functions, 2^nd Edition, Short Form Self-Report

Evidence supporting the DSRI’s external validity included significant correlations with most recent A1c (r = 0.40, p < .05). Additionally, adolescents who reported experiencing a severe hypoglycemic event in the last year also reported significantly higher DSRI scores (M = 45.4 ± 17.3) compared with those who reported no hypoglycemic event (M = 36.1 ± 11.1; t(220) = 2.18, p < .05). However, the DSRI total score was not significantly different between those who did (M = 38.9 ± 19.6) and did not (M = 36.3 ± 11.5) report having DKA in the last year. The DSRI demonstrated adequate incremental validity, in that it accounted for unique variance in adolescents’ most recent A1c, above and beyond the SCI-R, the RTI-A, sex, and age (see Table 4).

Table 4.

Results of linear regression examining the incremental validity of the DSRI in relation to A1c, above and beyond other, established measures of engagement in T1D self-management, executive functioning, and general risk-taking.

	Model 1	Model 2
	Coefficient (95% CI) p-value
SCI	−0.04 (−0.06, −0.03) < 0.001	−0.02 (−0.04, 0.001) 0.066
RTI-A	0.03 (−0.01, 0.08) 0.116	0.00 (−0.05, 0.05) 0.991
Sex	0.01 (−0.32, 0.35) 0.944	0.04 (−0.29, 0.37) 0.817
Age	−0.01 (−0.17, 0.14) 0.860	−0.02 (−0.17, 0.13) 0.779
DSRI	--	0.04 (0.01, 0.06) 0.002
R2	0.162	0.198

F test comparing R² for Model 1 vs. Model 2 = p = 0.002 DSRI = Diabetes-Specific Risk-Taking Inventory; SCI = Self-Care Inventory; RTI-A = Risk-Taking Inventory for Adolescents

Discussion

The purpose of the current study was to conduct an initial psychometric evaluation of the first questionnaire to assess T1D-specific risk-taking behaviors among adolescents, the Diabetes-Specific Risk-Taking Inventory (DSRI). Our results provide evidence for excellent internal consistency across items and reliability over time. Additionally, the DSRI demonstrated good validity via associations with established measures of similar constructs (i.e., engagement in T1D care and general risk-taking behaviors) and glycemic outcomes (i.e., A1c and severe hypoglycemic events). The only glycemic outcome that was not associated with the DSRI, contrary to our hypothesis, was incidence of DKA in the last year. The base rate of DKA in our sample was low, limiting our power to detect an effect. Fifteen adolescents reported having at least 1 DKA episode within the last year, which amounts to an incidence rate of 88.2 cases per 1000 person years, whereas national data suggests incident rates for adolescents are more likely about 102.4 to 132.8 incidence rate per 100 person years for males and females, respectively²². Additionally, self-report in adolescents may not be the most reliable method of determining DKA incidence-rates, considering 47 adolescents (21.0% of our sample) selected the response option “I don’t know what DKA is” (see Table 1). Future studies may strive to identify alternative methods for collecting information on incidence of DKA. Still, it is possible that in future studies the DSRI may continue to not be associated with DKA, considering the many, complex factors that can lead to DKA. It is possible other glycemic outcomes, such as time outside of range or mean amplitude of glycemic excursion, collected via blood glucose monitor or continuous glucose monitor, might provide a more sensitive measure of glycemic outcomes related to T1D-specific risk-taking. Consistent with our last hypothesis, the DSRI demonstrated incremental validity in its association with A1c, suggesting that it may be assessing a construct that is unique and not currently assessed by established measures. Taken together, the study findings suggest that the DSRI is an acceptable, valid, and reliable measure of T1D-specific risk-taking behaviors. These findings generally provide preliminary support for our previously published model on adolescent T1D-specific risk-taking.¹²

In addition to the main findings of the study, it is notable that the association between DSRI scores and general risk-taking behaviors differed based on participant age and sex. The finding that older age was associated with more frequent T1D-related risky behaviors is consistent with the general risk-taking literature, suggesting that, like general risk-taking behaviors, diabetes-specific risk-taking behaviors may increase with age.²³ This conclusion is further supported by the finding that correlations between the DSRI and general risk-taking behaviors were not different when stratified by age. Similarly, it is interesting that the association between general risk taking and T1D-specific risk-taking behaviors was significantly stronger for females than for males. When the scatterplots were examined, we identified seven male participants who reported more-frequent general risk-taking behaviors but less-frequent T1D-specific behaviors, indicating this subgroup of participants take risks (e.g., they might frequently use alcohol), but not may not take risks as frequently with their T1D management (e.g., despite drinking alcohol, they do not drink without someone around knowing they have T1D). More research is needed to confirm this finding, and to further understand what factors (e.g., beliefs and attitudes) might contribute to these decisions and why this pattern appears to be a more prevalent among males than females.

While the current study had many strengths, such as assessment of multiple types of reliability and validity, there are some limitations. The current study included a self-selected sample, considering that families indicated they were interested in the study and that we reached 247 out of 1811 eligible families (about 14%). Additionally, families must agree to have their information included in T1Dx database in the first place; thus, we had no way to recruit families who had not agreed to share their information with the T1Dx registry. The current study did not examine all types of validity and reliability. Because of the significant resources required to assess all psychometric properties (e.g., multiple time points, large sample sizes), we prioritized aspects of validity and reliability that we felt were most salient at this stage of measure development. Because of our cross-sectional design, we could not evaluate predictive validity. Additionally, our study’s sample size and the longer length of the DSRI precluded us from conducting a factor analysis to examine the underlying structure of the DSRI. Therefore, in the future, it may be helpful to conduct a larger examination of the DSRI to better understand its construct validity. A future factor analysis would also allow for potential reduction of the number of items on the DSRI, as a shorter measure might be more useful in a clinical setting, considering time constraints of most diabetes clinic visits.

Compared to the T1D exchange registry, our sample was similar in many ways to the characteristics reported for T1D exchange-wide studies. For example, participants in our sample had a wide range of A1c levels (6.0% to 14.1%) and an average A1c (8.6%) that was only slightly lower than that obtained for adolescents (age 16) in the exchange (8.9%)²⁴. However, the racial and ethnic diversity of participants in the current study was limited and does not reflect estimated proportions of youth with T1D of various racial and ethnic backgrounds in the United States,²⁵ or in the T1D exchange ²⁴. Also, families in our sample reported relatively high socioeconomic status and high pump use (77%)²⁴. Our findings might therefore have limited generalizability to the general population of youth with T1D, as it is possible the DSRI might perform differently in more-diverse samples. In future studies of the DSRI, it may be helpful to use purposeful recruiting methods to ensure a more-diverse sample. Finally, we are not able to assess whether the DSRI will be sensitive to intervention or treatment.

Despite these limitations, the current study adds to an emerging area of research that aims to explore the role of T1D-specific risk-taking in the health and well-being of youth with T1D. Certainly, further research is necessary to examine longitudinal trajectories and causal associations between risky T1D behaviors and other outcomes. Additionally, future research may aim to assess if/how risky T1D behaviors change over time. The DSRI was developed for and tested within a sample of older adolescents (ages 15–18), though it is possible it could be adapted for younger and older persons with T1D. Future research could determine whether the DSRI could be used in an expanded age range (e.g., under 15 or over 18 years old), or if it needs to be adapted for other age groups (e.g., young adults who may have different types of risky situations and behaviors).

Future research is also needed to develop and test interventions that aim to reduce or prevent risky T1D behaviors for youth with T1D. As far as we know, there are no published studies on measures or interventions that specifically aim to assess and/or address risk-taking behaviors for adolescents with chronic illnesses. Fortunately, though, research on general adolescent risk-taking behavior has advanced, and behavioral specialists have developed effective programs to reduce or prevent unhealthy adolescent risk-taking behaviors in the general population (e.g., screening, brief intervention and referral to treatment, motivational interviewing, an intervention targeting emotion regulation²⁶). It is likely that these interventions will need to be modified to include information and education that address T1D-specific risk-taking behaviors. However, the DSRI could become an important tool for evaluating any new interventions developed to reduce and/or prevent T1D-specific risk-taking behaviors.

This study extends the current literature examining factors related to glycemic outcomes among adolescents with T1D.² The finding that the DSRI was more strongly associated with self-management and glycemic outcomes than the general risk-taking measure support the importance of condition-specific measures when assessing diabetes outcomes. As well, adolescents are not “little adults”²⁷ and require a nuanced approach to understanding the many factors that may uniquely contribute to their engagement in T1D management and their health outcomes. Behavioral researchers have already created and validated questionnaires for adolescents that assess diabetes distress,²⁸ health-related quality of life,²⁹ T1D skill level,³⁰ T1D-related eating problems,³¹ readiness for independent T1D self-care,³² and T1D-related strengths.³³ Therefore, the DSRI adds to this diverse library of assessment tools for adolescents with T1D and may provide a novel target for interventions that aim to improve glycemic health in adolescents with T1D.

Data Sharing Statement

Data available on request from the authors.