Featured Article: Comparison of Diabetes Management Trajectories in Hispanic versus White Non-Hispanic Youth with Type 1 Diabetes across Early Adolescence

Abstract

Objective

Ethnic minority youth with type 1 diabetes (T1D) often have poorer glycemic control and lower rates of adherence compared to White Non-Hispanic (WNH) youth. Variables such as family conflict, autonomy support, and youth regimen responsibility have been shown to change over adolescence and impact diabetes management. However, these factors have been investigated in predominantly White samples. Few studies have examined potential differences in these variables and their trajectories for Hispanic youth over early adolescence.

Methods

Youth with T1D (178 WNH and 33 Hispanic youth participants), as well as their maternal caregivers (174 WNH and 32 Hispanic maternal caregivers), completed measures of diabetes-specific autonomy support, diabetes-related family conflict, regimen responsibility, and blood glucose monitoring frequency at 4 timepoints over a 3-year period.

Results

At baseline, Hispanic youth had significantly poorer glycemic control, more family conflict, and fewer blood glucose checks on average compared to WNH youth. Similar to WNH youth, Hispanic youth have increasing independence for regimen tasks and decreasing parent autonomy support during this developmental period. However, while Hispanic youth had worsening diabetes management during early adolescence (as did WNH youth), Hispanic parents reported a more gradual change in youth’s diabetes management over early adolescence.

Conclusions

This study presents an important contribution to the existing literature on youth with T1D. Findings suggest potential strengths and targets for Hispanic youth navigating diabetes management during the adolescent period. It is important to continue to investigate the trajectories of ethnic minority youth with diabetes.

With the rising incidence of type 1 diabetes (T1D), Hispanic youth are being diagnosed with diabetes more than ever. In fact, the SEARCH for Diabetes in Youth Study has found larger increases in the annual incidence rates among Hispanic youth (4.2%) compared to White Non-Hispanic (WNH) youth (1.2%; Mayer-Davis et al., 2017). In addition, a recent meta-analysis reported that ethnic minority youth, including Hispanic youth, display higher HbA1c levels, with more acute and long- term complications than WNH youth (Borschuk & Everhart, 2015). Studies have often found that Hispanic youth with T1D have poorer glycemic control and lower rates of adherence compared to WNH youth (Gallegos‐Macias, Macias, Kaufman, Skipper, & Kalishman, 2003; Hilliard, Wu, Rausch, Dolan, & Hood, 2013; Naranjo, Schwartz, & Delamater, 2015). However, findings have been inconsistent, with a few studies reporting no differences in glycemic control between Hispanic and WNH youth after adjusting for socioeconomic status (Springer et al., 2006; Willi et al., 2015).

Hispanic youth with T1D may be at greater risk for complications than WNH youth if they experience poorer glycemic control and lower rates of adherence, especially during adolescence. It has been reported that nearly two-thirds of adolescents engage in suboptimal diabetes management, defined as “the amount, precision and regularity of behaviors carried out by patients or family members to establish, maintain, or monitor diabetic control” (Harris et al., 2000). Hispanic youth are even more likely to fall into a high-risk group of adolescents than their WNH counterparts with poor diabetes management and glycemic control (Hilliard et al., 2013). Research indicates that changes in autonomy support, youth regimen responsibility, and family conflict are important developmental considerations that influence disease care and diabetes management over the course of adolescence (Drotar et al., 2013; Miller & DiMatteo, 2013; Rausch et al., 2012; Sood et al. 2012). These changes may be most amenable to intervention during “early adolescence,” defined in the current study to include children ranging from ages 9 to 14.

Recent research has found that parental autonomy support is associated with improved diabetes management, including maintenance of frequent blood glucose monitoring (Wu et al., 2014). Parent autonomy support has been defined as “being responsive to youths’ perspectives and needs, providing choices, encouraging initiative, and reasoning about expectations or requirements” (Hanna, Dashiff, Stump, & Weaver, 2013). Parent and adolescent report of autonomy support have been associated with perceptions of shared diabetes care responsibility among family members. Studies suggest that shared regimen responsibility is associated with improved psychosocial outcomes and improved diabetes management (Hanna et al., 2013; Helgeson et al., 2017). Additionally, there is consistent evidence of a relationship between increased disease-related family conflict in adolescence and reduced diabetes management, as well as suboptimal glycemic control (Anderson et al., 2002). Increases in family conflict and perceived parental criticism may interfere with the quality of collective problem-solving between parents and adolescents leading to poorer disease outcomes (Wysocki et al., 2000).

While a developmentally sensitive understanding of diabetes management in adolescence is emerging, the research in this area is understudied with regard to ethnic minority youth. As in the research described above for largely WNH populations, it is suspected that characteristics of the parent-adolescent relationship are salient in diabetes management for Hispanic adolescents due to the centrality of family (Lopez, 2006; Main et al., 2014). Family connectedness is a predominant aspect of family functioning and characteristic of family cohesion within collectivistic family orientations, like traditional Hispanic families (Olson, 2000). Literature suggests that Hispanic youth may have increased parental supervision, support, and control, rooted in cultural values, like familismo (familism), respeto (respect), and educación (moral education) in an effort to promote communication and familial support (Halgunseth, Ispa, & Rudy, 2006). These family values have been associated with positive wellbeing and psychological functioning in Hispanic youth (Schwartz et al., 2010). In at least one study, higher levels of family connectedness were associated with lower levels of asthma-related functional limitation in Hispanic youth (Koinis-Mitchell et al., 2012). Findings like this suggest that family connectedness may be used to help support the management of chronic illness, like asthma and T1D. Literature suggests that similar to asthma management, family dynamics are salient to T1D management (Lewin et al., 2006; McQuaid, Walders, Kopel, Fritz, & Klinnert, 2005). Hispanic values and family connectedness may have implications for Hispanic families of youth with T1D as increased disease-specific parent involvement and support may be protective factors that can be further amplified in intervention with families of Hispanic youth with T1D.

Cross-sectional studies have suggested some family related differences in adolescent diabetes management for Hispanic youth including increased parent supervision of diabetes management tasks and increased diabetes-related family conflict compared to WNH youth (Gallegos-Macia et al., 2003; Main et al., 2014). In addition, Hsin, Greca, Valenzuela, Moine, & Delamater (2010) found that family support mediated a significant relationship between adolescent responsibility and improved adherence in a sample of Hispanic youth with T1D. Differences in Hispanic culture and family values may result in differences in the experience of autonomy support, shared regimen responsibility, and family conflict over the course of adolescence for Hispanic youth and their families. Understanding culture-specific differences that exist for Hispanic youth and families may be an important consideration for pediatric psychologists treating youth with T1D.

Given findings that Hispanic youth with T1D have disparate health outcomes (Lotstein et al., 2013), the current study uses a longitudinal, multi-site database to compare diabetes management trajectories of Hispanic and WNH youth with T1D over the course of late childhood and early adolescence. In addition, the study compares youths’ trajectories on a variety of variables that have been previously shown to impact diabetes management during adolescence (e.g., regimen responsibility, family support, and family conflict). It is hypothesized that the subset of Hispanic youth will have different longitudinal trajectories on a variety of predictors and outcomes (e.g., regimen responsibility, family support, blood glucose monitoring frequency, HbA1c) in comparison to the larger WNH group of youth. Specifically, the limited pediatric diabetes research suggests that family variables may act as a protective factor for Hispanic youth. We hypothesize that Hispanic families may remain more involved in regimen responsibility, have fewer changes in autonomy support, and a lower rate of family conflict compared to WNH youth (Hsin et al., 2010).

Methods

Two-hundred and thirty-nine youth (M age = 10.5 years, SD = .94) with T1D were recruited from three university-affiliated pediatric diabetes clinics within the United States as part of a larger longitudinal study of diabetes regimen and management throughout early adolescence (Rohan et al., 2014). Parents identified youth as WNH (74.9%), Hispanic (13.8%), and African-American (4.6%). Reports of longitudinal data from this study have been previously described in Drotar et al. (2013), Rausch et al. (2012), Rohan et al. (2013), Rohan et al. (2014), and Wu et al. (2014). This is the first report from this study focusing on the comparison of diabetes management trajectories across ethnicities.

Youth were recruited during routine outpatient clinic visits. Inclusion criteria included duration of T1D for at least 1 year, between 9 and 11 years of age at recruitment, English speaking, no plans to relocate out of the area during the study duration, and absence of secondary causes for T1D diagnosis (e.g., cystic fibrosis). Exclusionary criteria included active involvement in foster care, presence of severe psychiatric disorders or comorbid chronic conditions that required intensive treatment regimens (e.g., renal disease), or diagnosis of intellectual and/or developmental disabilities.

Participants and their maternal caregivers completed study measures at four time points (i.e., baseline, 1 year, 2 years, 3 years). Youth received $35 in gift cards and caregivers received $20 cash for completing measures at each visit. Additionally, participants received $5 for providing blood glucose records from their blood glucose meter or logbook at each visit. Each of the three university-affiliated Institutional Review Boards approved the study. The overall recruitment rate was 66.5%. Reasons for declining participation in the study included being too busy (n = 54), difficulties maintaining transportation (n = 3), and other (n = 64; not interested, did not return recruitment phone calls, did not attend clinic regularly, etc.). Caregivers provided signed informed consent and youth aged 11 years and older provided written assent, while youth under 11 years of age provided verbal assent for participation.

This report includes all WNH and Hispanic participants who completed measures of interest during at least one time point. Other minority groups (e.g., African-American) were excluded in the present study due to small sample sizes and the focus on Hispanic youth. Overall, at baseline, the study yielded 211 participants; 178 WNH (M age = 10.5 years, SD = 1.03) and 33 Hispanic youth participants (M age = 10.76 years, SD = 1.00) as well as their maternal caregivers (WNH n = 174; Hispanic n = 32 maternal caregivers). Participant attrition during data collection has been described previously as low, approximately 4% from baseline to 3 years (n = 10). Reasons for attrition include transfer of medical care to a provider non-affiliated with the study (n = 1), family too busy to participate (n = 4), family relocation from the geographic area (n = 1), and family lost to follow-up (n = 4). Wu and colleagues (2014) found no significant demographic differences between participants who completed assessments at all time points versus those who did not in the study.

Measures

Demographic Variables

Caregivers provided information on youth’s date of birth, gender, race/ethnicity, duration of diagnosis, family income, maternal education, and household composition (i.e., one vs. two parent). Youth age was calculated from birth date and date of study assessments.

Blood Glucose Monitoring Frequency

Blood Glucose Monitoring Frequency (BGMF) was evaluated using data downloaded from youths’ blood glucose meter(s) for the previous two weeks starting with the day prior to the assessment visit. If one or more of the blood glucose meters were not accessible at the time of the visit (e.g., a school meter), information from youths’ blood glucose monitoring logbooks was used. Frequency of blood glucose monitoring (i.e., number of BG checks per day) were recorded and the average daily blood glucose monitoring frequency over the last two weeks was calculated.

Glycemic Control

Hemoglobin HbA1c levels were obtained at six month intervals from baseline to 3 years post baseline. Blood samples, obtained by finger stick at study visits, were processed and analyzed by one central laboratory, using the TOSOH-G7 method (reference range 4.0–6.0%).

Family Conflict

Family conflict regarding diabetes management was assessed with the Diabetes Family Conflict Scale (Hood, Butler, Anderson, & Laffel, 2007). The measure evaluates the amount of conflict around tasks of diabetes management (e.g., conflict about remembering to check blood sugars, what to eat when away from home). It consists of 19 items on a 3-point Likert scale ranging from 1 (never argue) to 3 (always argue). Total scores may range from 19 to 57 with higher scores indicative of higher family conflict. The measure has demonstrated good reliability and validity as well as good internal consistency for caregivers (α = .81) and for youth (α = .85; Hood et al., 2007). Youth as well as their maternal caregiver completed this measure at each time point. Internal consistency across study time points ranged from .85 to .87 (youth) and .81 to .87 (maternal caregivers) in the current sample.

Regimen Responsibility

Primary responsibility for diabetes care (caregiver, youth, or shared responsibility) was assessed through caregiver report using modified versions of the Continuous Subcutaneous Insulin Infusion (CSII)-Use Survey (Weissberg-Benchell, Goodman, Antisdel Lomaglio, & Zebracki, 2007) and Diabetes Independence Survey (DIS; Wysocki et al., 1996), depending on type of insulin therapy (insulin pump vs. injection) used. The measures have demonstrated both adequate reliability and validity (Weissberg-Benchell et al., 2007; Wysocki et al., 1996). Youth report was not obtained as the original measures were not validated for use with an adolescent population. The DIS was modified for the current study to allow for items to address modern diabetes management recommendations (e.g., rechecking blood glucose levels when high). The DIS was also modified to allow for assessment of caregiver perception of which family member is primarily responsible for monitoring and carrying out tasks associated with injections to increase its comparability to the modified CSII-Use Survey. Overall, the modified DIS included 38 items, while the modified CSII-Use Survey included 28 items. Both the modified DIS and CSII-Use Survey were scored based on “Who is responsible?,” with response options of Parent (1), Shared (2), Child (3), or Not Applicable. Cronbach’s alphas for the two measures ranged from .88 to .95 across time points in the current sample. For both respective measures, an average responsibility score was calculated by averaging scores across the responsibility items. Participants were classified as having a caregiver primarily responsible if the average responsibility score was between 1 and 1.49, shared responsibility of diabetes tasks if the score was between 1.5 and 2.49, and youth primarily responsible if the score was between 2.5 and 3.

Autonomy Support

Parental support for youth autonomy in diabetes management tasks was assessed using a modified form of the Diabetes-Specific Parental Support for Adolescents’ Autonomy Scale. Specifically, the measure evaluates parental behaviors that promote youths’ autonomy in the management of their diabetes regimen (e.g., “In the past three months, how often have your parents answered your questions about figuring your insulin dose?”). This measure has demonstrated adequate reliability and validity (Hanna, Dimeglio, & Fortenberry, 2005). Maternal caregivers and youths were each asked to respond to six items on a 5-point Likert scale ranging from 0 (none of the time) to 4 (all of the time) in regard to the frequency of autonomy-supporting behaviors. Two items were added to the original measure to assess for parental support of youth autonomy in regard to blood glucose monitoring and carbohydrate counting. Item responses were summed for a total score ranging from 0 to 24 with higher scores indicating higher levels of parental support for youth autonomy. Cronbach’s alphas for maternal caregivers and youth at baseline were 0.68, 0.67; at 1-year were 0.67, 0.72; at 2-year were 0.78, 0.70; and at 3-year were 0.75, 0.72, respectively.

Diabetes Self-Management

Behaviors related to diabetes self-management were measured using The Diabetes Self-Management Profile (DSMP), a 25-item structured interview administered to both youth and their caregivers (Harris et al., 2000). The profile assesses diabetes management for the previous 3 months through open-ended questions addressing domains of: blood glucose monitoring, diet, exercise, hypoglycemia management, and insulin administration. The DSMP includes items with a variety of response scales (i.e., yes/no questions and 3- to 5- point Likert scale items). A total self-management score is calculated by summing all items, and domain scores are computed by summing items within that domain. Higher scores are indicative of more meticulous self-management behaviors. Overall, the DSMP has demonstrated good internal consistency (r = .76), moderate agreement between reporters (parent and youth; r = .61), and strong interrater agreement (r = .94; Harris et al., 2000). Additionally, the measure has demonstrated good predictive validity between caregiver and youth report of self-management behaviors and glycemic control (Harris et al., 2000). Internal consistency across time points ranged from .66 to .71 for maternal caregiver report and .60 to .69 for youth report.

Analytic Technique

All analyses were conducted using IBM SPSS v.20. Differences in participant characteristics between WNH and Hispanic youth were examined through independent t-tests for age and chi-square tests for all other demographic variables. Similarly, baseline group differences on BGMF, HbA1c, autonomy support, diabetes management, and regimen responsibility were assessed by conducting independent t-tests with baseline (i.e., time one) observations. When appropriate, Welch’s t-tests were utilized rather than student’s t-tests to address violations of the assumption of homogeneity of variance. Significant baseline differences were further examined using ANCOVA models to control for the effects of demographic variables.

In order to examine participant trajectories over time on these same variables, we utilized an individual growth curve modeling approach (Singer & Willett, 2003). This approach, which is functionally equivalent to multilevel modeling (Snijders & Bosker, 1999), allows for examination of growth trajectories at both the aggregate and individual levels (Delucia & Pitts, 2006). Of note, while growth curve models typically have sample sizes of at least 100, models have been successfully fit to samples as small as 22 and allow for comparisons among unequal groups (Curran, Obeidat, & Losardo, 2010). Our unconditional models were fit to samples well over 100. The same analytic process was conducted with each dependent variable separately. First, we determined the best fitting unconditional growth model by comparing the −2 log likelihood statistics of a series of increasingly sophisticated, nested models until no further significant improvement in fit was detected (α = .05). This series began with a simple model including only a fixed intercept parameter, and one additional parameter was added in each subsequent model in the following order: (1) random intercept, (2) fixed linear slope, (3) random linear slope, and (4) fixed quadratic slope. Time was represented in all models by chronological age, centered at the sample mean age at time one. Second, between group differences between WNH and Hispanic youth on growth curve intercepts and slopes were assessed by adding ethnicity and an ethnicity-by-time interaction to the best fitting unconditional growth curve model for each dependent variable. Finally, if a statistically significant ethnicity-by-time interaction was detected, we added to the model the income, maternal education, and household composition variables, as well as the interaction of each with time, in order to control for the potential confounding effects of these demographic variables. The SPSS mixed procedure with full maximum likelihood estimation was used for all growth curve models. This procedure addresses the potential influence of missing data by weighting observations without missing data points more heavily than observations with missing data points. Additionally, sensitivity analyses of missing data were conducted and indicated that data were missing completely at random.

Results

Baseline Comparisons

Between-group comparisons of baseline participant characteristics are displayed in Table I. The groups differed on household composition, with the Hispanic group composed of a significantly greater percentage of participants from single-parent households (39.4%) compared to the WNH group (14.6%), Χ²(1) = 11.4, p = .001. Between-group differences were also detected on the income variable, Χ²(5) = 26.2, p = .001, with the Hispanic group including more individuals in the lower income levels compared to the WNH. In terms of insulin administration method, 69.7% of Hispanic participants used injection rather than a pump or pod at baseline, whereas only 37.6% of WNH participant used injection, Χ²(1) = 11.7, p = .001. The groups did not significantly differ on age, gender, or maternal education.

Table I.

Sample Characteristics

Variable	White, Non-Hispanic (n = 178)	Hispanic (n = 33)	Statistic	p-value
Age (M, SD)	10.5 (1.03)	10.76 (1.06)	t(203) = −1.3	.19
Gender (n, %)			Χ2(1) = 2.3	.09
Female	93 (52.2%)	22 (66.7%)
Male	85 (47.8%)	11 (33.3%)
Maternal Education (n, %)			Χ2(2) = 4.0	.14
No high school diploma/equivalent	7 (4.0%)	1 (3.0%)
High school diploma/equivalent	45 (25.5%)	14 (42.4%)
Some college or college degree	125 (70.6%)	18 (54.5%)
Household Composition (n, %)			Χ2(1) = 11.4	.001
Single parent	26 (14.6%)	13 (39.4%)
Two parent	152 (85.4%)	20 (60.6%)
Insulin Administration (n, %)			Χ2(1) = 11.7	.001
Injection	67 (37.6%)	23 (69.7%)
Pump/pod	111 (62.4%)	10 (30.3%)
Income (n, %)			Χ2(5) = 26.2	<.001
<$18,745	6 (3.4%)	7 (21.2%)
$18,745–$32,874	13 (7.3%)	3 (9.1%)
$32,875–$48,999	15 (8.4%)	7 (21.2%)
$49,000–$72,999	38 (21.3%)	9 (27.3%)
$73,000–$126,500	66 (37.1%)	5 (15.2%)
>$126,500	37 (20.8%)	2 (6.1%)

Baseline between-group comparisons on the outcome variables of interest are displayed in Table II. On average, Hispanic participants checked their blood glucose (i.e., BGMF) less frequently (M = 4.4 times per day, SD = 1.98) than WNH participants (M = 5.2 times per day, SD = 1.74), d = 0.43, p = .02. Hispanic participants had higher HbA1c levels at baseline (M = 8.8%, SD = 1.48) compared to WNH participants (M = 8.0%, SD = 1.08), d = 0.62, p = .002. The groups also differed on both child and maternal reports of family conflict, with Hispanic families generally reporting higher levels of conflict, d = .49, p = .01 and d = 0.68, p = .005 for child and maternal reports, respectively. When controlling for household composition, income, and insulin administration method, differences remained significant for HbA1c, F(1, 201) = 3.76, p = .05, and both child and maternal reports of family conflict, F(1, 203) = 4.02, p = .05, F(1, 197) = 13.002, p < .001, respectively. However, BGMF did not differ after controlling for these variables, F(1, 201) = 0.65, p = .42.

Table II.

Baseline Comparisons

	White Non-Hispanic			Hispanic
Variable	n	Mean	SD	n	Mean	SD	t	p-value	d
BGMF	176	5.2	1.74	33	4.4	1.98	2.4	.02*	0.43
HbA1c	176	8.0	1.08	33	8.8	1.48	−3.2	.002**	0.62
Autonomy (ch)	178	12.8	5.32	33	14.7	4.34	−1.9	.06	0.39
Autonomy (m)	174	13.2	4.93	32	13.7	4.80	−0.6	.55	0.10
Responsibility (m)	174	1.7	0.27	32	1.7	0.31	0.05	.95	<0.01
Conflict (ch)	178	25.7	5.35	33	28.4	5.68	−2.5	.01*	0.49
Conflict (m)	173	24.1	3.75	32	28.1	7.43	−3.0	.005**	0.68
DSMP (ch)	178	61.2	8.06	33	59.1	7.97	1.4	.16	0.26
DSMP (m)	175	65.8	7.83	32	62.7	11.33	1.5	.15	0.32

^* p < .05, **p < .01; ch = child report; m = maternal report.

Comparisons of Longitudinal Trajectories

Growth-specific curve estimates, ethnicity main effects, and interaction effects for all dependent variables are displayed in Tables III and IV. Except where noted, all models include random effects for linear slope. For most variables, the data were best fit by linear growth models. However, for maternal reports of regimen responsibility and family conflict, quadratic growth was observed. Estimates of intercepts for both groups were statistically significant for all dependent variables, indicating non-zero estimated average values for each group at age 10.54 (i.e., age centered at time 1). Similarly, slope estimates for each group were significant on most, but not all, dependent variables, indicating that most variables changed significantly over time. Specifically, youth in both groups had less BG monitoring, greater HbA1c, and less autonomy support (by youth report). Slope estimates for maternal reports of autonomy support and regimen responsibility were not significant for the Hispanic group, suggesting that there was relatively no to little change over time, though this may have been a function of the relatively small sample size of that group. Slope estimates for family conflict by maternal report were only significant for the Hispanic group and suggested decreasing family conflict over time. As seen in Table IV, significant main effects of ethnicity were found for HbA1c (d = 0.56, p < .01), child reports of autonomy support (d = 0.49, p = .01) and family conflict (d = 0.92, p = .02), and maternal reports of conflict (d = 0.78, p < .01). These main effects are indicative of significant differences in the groups at baseline and are qualitatively consistent with the t-test results in Table II.

Table III.

Growth Curve Estimates as a Function of Ethnicity

	WNH			Hispanic
Outcome	Est	SE	p	Est	SE	p
BGMF
Intercept	5.19	0.12	<.01*	4.62	0.30	<.01*
Linear	−0.17	0.05	<.01*	−0.27	0.12	.03*
HbA1c
Intercept	7.93	0.08	<.01*	8.66	0.20	<.01*
Linear	0.23	0.03	<.01*	0.25	0.08	<.01*
Autonomy Support (ch)#
Intercept	12.52	0.35	<.01*	14.89	0.86	<.01*
Linear	−0.58	0.13	<.01*	−0.90	0.31	<.01*
Autonomy Support (m)
Intercept	13.27	0.31	<.01*	14.19	0.76	<.01*
Linear	−0.52	0.13	<.01*	−0.28	0.32	.38
Response (m)
Intercept	1.73	0.02	<.01*	1.71	0.02	<.01*
Linear	0.09	0.01	<.01*	0.10	0.03	<.01*
Quadratic	0.01	0.00	<.01*	0.01	0.01	.49
Conflict (ch)
Intercept	25.55	0.35	<.01*	27.81	0.85	<.01*
Linear	−0.23	0.13	.09	−0.30	0.32	.36
Conflict (m)
Intercept	24.21	0.29	<.01*	28.35	0.72	<.01*
Linear	0.68	0.17	<.01*	1.24	0.43	<.01*
Quadratic	−0.07	0.05	.18	−0.25	0.11	.03*
DSMP (ch)
Intercept	61.41	0.57	<.01*	58.49	1.40	<.01*
Linear	−0.99	0.20	<.01*	−0.08	0.48	.86
DSMP (m)++
Intercept	58.95	3.39	<.01*	57.44	3.47	<.01*
Linear	−2.63	0.22	<.01*	−0.79	0.53	.14

Note: # = model did not include random slope parameter; ++ = controlling for income, maternal education, and household composition; BGMF= Blood Glucose Monitoring Frequency; Est = Estimated; ch = child report; m = maternal report; *p < .05.

Table IV.

Estimates of Ethnicity Effects on Intercept and Slope

	Ethnicity				Ethnicity*Time (Ethnicity*Time2)
Outcome	Est	SE	p	d	Est	SE	p	d
BGMF	−0.56	0.32	.08	0.30	−0.10	0.13	.44	0.05
HbA1c	0.73	0.22	<.01*	0.56	0.02	0.09	.80	0.02
Autonomy Support (ch)#	2.36	0.93	.01*	0.49	−0.32	0.33	.34	0.07
Autonomy Support (m)	0.92	0.82	.26	0.19	0.24	0.35	.50	0.05
Response (m)	−0.01	0.05	.81	0.03	0.01	0.03	.78	0.03
					−0.01	0.01	.52	0.02
Conflict (ch)	2.27	0.92	.02*	0.41	−0.07	0.35	.85	0.01
Conflict (m)	4.14	0.78	<.01*	0.70	0.56	0.47	.23	0.09
					−0.19	0.12	.13	0.03
DSMP (ch)	−2.92	1.51	.06	0.30	0.90	0.52	.08	0.09
DSMP (m)++	−1.51	1.63	.36	0.19	1.83	0.57	<.01*	0.23

Note: # = model did not include random slope parameter; ++ = controlling for income, maternal education, and household composition; BGMF= Blood Glucose Monitoring Frequency; Est = Estimated; ch = child report; m = maternal report; *= p < .05.

Given our focus on ethnicity-based differences in growth trajectories, we were most interested in examining ethnicity-by-time interaction effects. A significant interaction was detected for maternal report of diabetes self-management, even after controlling for income, maternal education, and household composition (i.e., DSMP; Estimate = 1.83, d = 0.23, p <.01). This interaction is displayed graphically in Figure 1. Regarding self-management as measured by DSMP, WNH maternal caregivers reported significantly decreasing scores over time, whereas Hispanic maternal caregivers reported smaller and non-significant decreases over time.

Figure 1.

Trajectory for caregiver report of disease management.

Discussion

This study aimed to understand differences between Hispanic and WNH youth in regard to diabetes management trajectories over early adolescence. In addition, longitudinal data was used to compare Hispanic and WNH family trajectories on a variety of disease-specific family variables (i.e., autonomy support, regimen responsibility, and family conflict) that have been previously shown to impact disease management during this developmental period. Results indicated that, similar to WNH youth, Hispanic youth reported decreasing autonomy support. The study also examined the developmental trajectory of diabetes management behaviors and health outcomes in this sample and found, as expected, worsening blood glucose monitoring frequency (BGMF) and glycemic control (as measured by HbA1c) for all youth during early adolescence. While both Hispanic and WNH youth had worsening self-management and health outcomes during this period, this study highlights that there may be some differences in the acceleration of these processes. Hispanic caregivers in this study reported a more gradual change in youth’s diabetes management over early adolescence.

The present study cannot clearly connect maternal report of a more gradual change in diabetes management to the disease-related family variables studied here. There were no other differences over time by ethnicity on the variables assessed. It may be that there are differences in how Hispanic caregivers perceive diabetes-related behaviors and family interactions. As suggested by some of our data (e.g., Hispanic mothers reported significantly less family conflict and did not report significant changes in autonomy support over the study period, but Hispanic youth did report decreases in autonomy support during this period). Future studies should measure differences in acculturation and assess how acculturative status may relate to differences in child and mother perception of these important diabetes-related variables (Hsin et al., 2010). However, it is also important to note that Spanish translations of the Diabetes Self-Management Profile, the disease management measure used in the present study, have demonstrated strong parent-child concordance and a significant relationship between Hispanic caregiver-report and both A1c and physician perceptions of self-management (Valenzuela et al., 2010). If Hispanic mother-report of a slower trajectory of decline in disease-management is accurate, future studies that randomize Hispanic youth to various groups and manipulate variables such as autonomy support, family conflict, and regimen responsibility through intervention may be helpful in further understanding the role of these interactions in the disease management behaviors of Hispanic youth.

It is important to note that despite finding that mothers report a more gradual change in diabetes management and decreasing family conflict over time for Hispanic youth, there was clear evidence of increased risks for this group. At baseline, Hispanic youth had significantly poorer glycemic control, less blood glucose monitoring behavior, and more family conflict compared to WNH youth. These findings ranged from moderate to large effect sizes. While the literature has at times been inconsistent about these disparities for Hispanic youth with diabetes, evidence appears to be accumulating regarding clear disparities and poorer outcomes (Gallegos‐Macias et al., 2003; Hilliard et al., 2013; Naranjo et al., 2015). In the current study, there were clear differences between Hispanic and WNH youth on a variety of variables including household composition, annual household income, and insulin regimen. Baseline comparisons suggest that ethnic group differences exist after controlling for these variables. The existing literature remains mixed on the effects of SES versus ethnicity on health-related disparities (Gallegos‐Macias et al., 2003; Willi et al., 2015). However, it is important to consider the role of variables outside of the family system that may disparately affect Hispanic youth with diabetes. Hispanic youth functioning within problematic school, community, and healthcare systems may have inadequate supports for optimal disease management despite having similar family support to WNH youth studied (Delamater et al., 1999; Ellis et al., 2005; Valenzuela et al., 2014). Future studies should also examine the role of other systems variables, beyond the family, to better understand these existing disparities.

The present study is not without limitations. The focus on Hispanic families treats individuals of different national origins the same despite important differences in Hispanic subgroups based on many factors including cultural beliefs, socioeconomic status, and acculturation status (Hsin et al., 2010). These variables were not assessed in the current study. Additionally, the majority of the Hispanic youth in the current sample came from one geographic region. There are likely important differences among Hispanic groups that suggest the critical need for more research in this area, especially given the growing population of Hispanic youth with T1D in the United States (Naranjo et al., 2015). Furthermore, the limited sample size of Hispanic participants in this secondary data analysis from a large longitudinal, multi-site dataset limits our confidence and suggests the need for continued efforts to recruit larger samples of ethnic minority youth in longitudinal research of this kind. Future research should directly assess these important variables in a larger, representative sample of Hispanic families of youth with diabetes.

In the study, three self-report measures (i.e., CSII- Use Survey, DIS, and Diabetes-Specific Parental Support for Adolescents’ Autonomy Scale) were modified from their original forms in order to more accurately assess regimen responsibility and autonomy support, given changes in diabetes management (e.g., technology, medical recommendations) and to account for all participant perspectives (i.e., caregiver and youth response). Cronbach’s alphas were calculated for all assessment tools utilized in this study, including the updated measures. While the data from these measures provide initial evidence regarding the trajectories of Hispanic youth with T1D, further psychometric properties on the consistency, validity, and reliability of these adapted measures and their cross-cultural use is needed.

A final limitation of the present study is that these findings focus on early to mid-adolescence and may not extend to later adolescence and early adulthood, a critical developmental transitional period. It would be important to investigate what happens over the course of late adolescence and transition into adulthood for Hispanic youth and their families managing diabetes. Young adults are at increased risk for diabetes-related complications (i.e., 50% of young adults develop diabetes-related complications in their 20s) and are the least likely group to maintain consistent medical care for disease management (Monaghan, Helgeson, &Wiebe, 2015). In at least one study, parental support was found to decrease psychological distress and improve glycemic control in emerging adults with T1D (Helgeson et al., 2017). More research might suggest that parental support and transition readiness interventions could target different trajectories in Hispanic youth and families compared to WNH families.

Despite the limited sample size, this study presents important movement towards exploring change in family and youth behaviors over a vulnerable developmental period, and considering how these changes may differ for ethnic minority populations. Because disease-specific family variables have been shown to be critical to diabetes outcomes (Lewin et al., 2006), they are an important part of effective intervention in this population. Therefore, understanding common trajectories for Hispanic youth has the potential to influence treatment in this population. This study found decreases evidence of decreasing autonomy support and diabetes self-management in both WNH and Hispanic families of youth with diabetes during this developmental period. It suggests that addressing these factors in treatment for Hispanic youth and families may be as effective as it is when studied in largely WNH intervention studies (Wysocki et al., 2000). Further, differences in youth and mother report of autonomy support in Hispanic families may speak to the importance of considering factors impacting perception of these variables intergenerationally in treatment settings (e.g., acculturative differences). If further research more robustly finds that a slower trajectory of worsening disease self-management for Hispanic youth, this protective factor could be bolstered in culturally competent ways by clinicians in pediatric diabetes centers. There is evidence that focusing on strengths, such as limited erosion of family involvement in diabetes management, is an effective tool for helping improve outcomes in pediatric diabetes (Hilliard et al., 2013). Furthermore, clinical interventions should also consider individual characteristics of the identified patient in order to adequately assess transition readiness for diabetes self-management and potential unique individual barriers, like patient beliefs and values (Delamater, 2006). Future studies are needed to further identify culturally competent strategies for supporting families of youth with diabetes during adolescence and into the transition to adulthood. Longitudinal studies like this one can help us learn how disparities grow and change over time, and may be essential in working to close the gap for ethnic minority youth with T1D.