Staying Positive: Positive Affect as a Predictor of Resilience in Adolescents With Type 1 Diabetes

Abstract

Background Adolescents with Type 1 Diabetes (T1D) are at increased risk for diminished quality of life, deteriorating glycemic control, and psychological symptoms, yet some adolescents are able to adapt to the challenges associated with having diabetes exceptionally well. We sought to examine positive affect as a protective process predicting resilience over time in youth with T1D. Method Adolescents and their mothers completed questionnaire data, and HbA1c was obtained from adolescents’ medical records at baseline and after 6 months. Adolescents were coded for observed positive mood during a videotaped interaction with their mothers. Results Positive mood, including both self-report and observed mood, was associated with glycemic control, psychological symptoms, and quality of life. In addition, positive mood predicted improvements in glycemic control and externalizing problems over 6 months. Conclusions Positive affect emerged as a protective process for resilient outcomes in adolescents with T1D, suggesting novel targets for intervention in this high-risk population.

Type 1 diabetes (T1D) is one of the most prevalent chronic illnesses diagnosed in childhood. Approximately 167,000 youth are currently living with T1D, and >15,000 new cases are diagnosed annually in youth aged <20 years (Liese, 2006). Although pharmaceutical and technological advances have improved patient care options, T1D still requires lifelong monitoring and management. The recommended treatment regimen includes multiple daily injections, monitoring carbohydrate intake, and frequently checking blood glucose (American Diabetes Association, 2014). Glycemic control typically deteriorates in adolescence; only 20% of adolescents meet the target for HbA1c, which is a proxy for disease control (Wood et al., 2013). In addition, adolescents with T1D report declining quality of life (Hood et al., 2014). Finally, youth with T1D experience higher levels of internalizing (e.g., anxiety, depression) and externalizing (e.g., behavioral problems) disorders than their peers without a chronic illness (Reynolds & Helgeson, 2011). A recent population-based study found increased risk for both internalizing and externalizing disorders in youth with T1D as compared with siblings, owing to the biological and/or psychological effects of the disease (Butwicka, Frisén, Almqvist, Zethelius, & Lichtenstein, 2015). Given the risk for poor psychosocial and physiological outcomes, adolescents with T1D face significant adversity. Yet, some adolescents are able to do exceptionally well when faced with diabetes-related challenges maintaining both physical and mental health (Helgeson et al., 2010).

Studying those who thrive despite chronic illness may help those who struggle in similar circumstances. Resilience is the idea that some people fare better than others when exposed to adversity (Rutter, 2012), and it is described as a dynamic process surrounding positive adaptation when one has been exposed to significant adversity (Luthar, Cicchetti, & Becker, 2000). Hilliard, Harris, and Weissberg-Benchell (2012) have proposed a model of diabetes resilience, which outlines specific individual, family, and social/contextual processes that lead to two domains of resilience. The first domain is behavioral resilience, or the ability to competently carry out diabetes management tasks and experience high health-related quality of life. The second domain is health resilience, or physiologic indicators of diabetes, such as maintaining good glycemic control. This model includes protective processes that mediate risks and assets, and allow youth to attain behavioral and health resilience. For example, prosocial characteristics, greater general well-being, and stronger cognitive abilities have been identified as assets related to positive outcomes in youth with type 1 diabetes (Hoey et al., 2001; Mackey et al., 2011; Northam, Lin, Finch, Werther, & Cameron, 2010; Soutor, Chen, Streisand, Kaplowitz, & Holmes, 2004). Protective processes such as emotional processing and adaptive coping are thought to improve adolescents’ self-efficacy and lead to resilient outcomes (Jaser & White, 2011; Yi, Vitaliano, Smith, Yi, & Weinger, 2008). These findings provide evidence for the buffering effect of protective factors and processes in youth with T1D; however, the majority of these studies relied on individuals’ self-reports of these factors and processes. Therefore, observational methods may provide more objective evidence of protective processes that predict resilience (Kerig, 2001).

As noted above, previous research in resilience in T1D has examined coping strategies, self-esteem, and self-efficacy to explain why some adolescents fare better than others, but few researchers have examined positive affect as a protective process in adolescents with chronic illness. Positive affect, defined as pleasurable engagement with the environment (e.g., feeling happy, cheerful, or proud) may help people to manage negative thoughts and feelings during stressful times (Watson, Clark, & Tellegen, 1988). Generally, prospective and longitudinal studies show that positive emotions may precede and predict positive outcomes, including better mental and physical health, rather than just resulting from them (Lyubomirsky, King, & Diener, 2005). Studies indicate that positive affect may buffer the risks for adverse physical and mental health outcomes for people living with a chronic illness such as T1D (Tugade, Fredrickson, & Feldman Barrett, 2004). The experience of positive affect is thought to provide respite from chronic stress (Moskowitz et al., 2012). Additionally, induced positive affect has been shown to enhance intrinsic motivation (Isen & Reeve, 2005), which is associated with improved glycemic control (Viner, Christie, Taylor, & Hey, 2003). One of the only studies to examine positive affect in youth with T1D found that positive affect was associated with lower (better) blood glucose levels (Fortenberry et al., 2009). In this same sample, youth who reported using benefit finding to deal with diabetes-related stress had higher levels of positive affect (Tran, Wiebe, Fortenberry, Butler, & Berg, 2011). In addition, a study of older women (without diabetes) found that positive affect predicted positive changes in glycemic control over time (Tsenkova, Love, Singer, & Ryff, 2008). Taken together, these findings suggest that positive affect may serve as a protective process for adverse outcomes in youth with T1D. However, to our knowledge, no studies have examined positive affect in relation to resilience in this population.

Current Study

Given that adolescents with T1D are at high risk for internalizing and externalizing symptoms, poor health-related quality of life, and deteriorating glycemic control (Fogel & Weissberg-Benchell, 2010), further research is needed to identify protective factors and processes in this population. The present study aimed to describe positive affect (using both observations and self-report) as a protective process in adolescents with T1D and to examine the associations between positive affect, glycemic control, quality of life, and psychological symptoms. In line with Hilliard’s model of diabetes resilience, our study examines internalizing and externalizing symptoms as outcomes related to behavioral resilience. We also examined Hba1c as an indication of good glycemic control, an outcome related to health resilience. Positive mood was conceptualized as a protective process that would predict resilient outcomes above and beyond preexisting demographic related risks and assets. Based on the literature, we hypothesized that higher levels of positive affect (as measured with both observed and self-report) would predict better outcomes (i.e., better quality of life, better glycemic control, fewer internalizing and externalizing behavior problems) in adolescents with T1D 6 months later.

Methods

Participants

All participants were part of a prospective, observational study of adolescents with T1D and their mothers. Adolescents aged 10–16 years were eligible if they were diagnosed with T1D for at least 6 months with no other major health problems and were able to speak and read English. We chose this age range to capture the high-risk transition to adolescence. Of the 295 eligible families approached, 118 families declined to participate (time and distance being the most common reasons), 60 expressed interest but were unable to schedule a visit, and 117 families completed data (40%). Although all families completed questionnaire data, some of the videotaped observations were unable to be coded owing to technical difficulties (e.g., problems with audio quality), limiting our current sample to 93 adolescents.¹ Of these 93 participants, 12 were lost to follow-up 6 months later owing to failure to return for a clinic appointment within the study’s time frame. There were no significant differences between those who completed follow-up information and those who did not related to age, sex, family income, duration of diabetes, therapy type (pump vs. injections), or glycemic control.

Procedures

Adolescents and their mothers were recruited from an outpatient pediatric diabetes clinic during their quarterly clinic visits. If they expressed interest in participating, a separate study visit to the laboratory was scheduled. At this baseline visit, procedures approved by the university’s institutional review board were used to obtain assent from the adolescent and consent from the mother for them to participate in the study. The participants then completed the questionnaires and participated in a 15-min videotaped discussion of a stressful topic related to the child’s diabetes. The topic was chosen using both parent and adolescent responses to a list of diabetes-specific stressors on the Response to Stress Questionnaire (Connor-Smith, Compas, Wadsworth, Thomsen, & Saltzman, 2000). The first 10 items ask about topics that adolescents with T1D may find stressful (e.g., Diabetes got in the way of my [my child’s] personal goals, I felt guilty or upset about my [my child’s] “bad” numbers) (Davidson, Penney, Muller, & Grey, 2004). Mothers and adolescents reported on the frequency of these stressors on a 4-point Likert scale (0 = never, 1 = a few times, 2 = many times, and 3 = almost every day), and their ratings were summed across reports to determine a common stressor. A cue card with questions about this stressor (e.g., What happened the last time [diabetes got in the way of your personal goals], what kind of emotions do you have when [diabetes gets in the way of your personal goals?]) was given to each dyad to guide their discussion during the interaction. Despite these prompts, most dyads discussed topics other than diabetes stress during part of the 15 min. Following the interaction, participants were compensated for their time ($20 each). Participants were met again at their regular diabetes clinic visit, approximately 6 months later, during which they completed questionnaire data only and received $10 for their time.

Measures

Demographics

Mothers provided information on demographic and clinical variables, including adolescents’ date of birth, gender, and race/ethnicity, date of diagnosis, and treatment type (insulin pump or injections).

Diabetes-Related Stress

As described above, adolescents and their mothers completed the Type 1 Diabetes version of the Responses to Stress Questionnaire (Connor-Smith et al., 2000) to report on diabetes-related stress. These first 10 items were used to select a topic for the videotaped interaction task. The stressor rated highest by the adolescent was used in the case of two stressors being rated equally high.

Self-Reported Positive Affect

Adolescents reported on their positive affect using the Positive subscale of the Positive and Negative Affect Schedule (PANAS, Watson et al., 1988). This measure consists of 15 positive emotions (e.g., happy, interested, excited, calm) and 15 negative emotions (e.g., sad, frightened, gloomy, lonely) rated on a 1 (slightly) to 5 (extremely) scale to describe to what extent each adolescent felt that way over the past week. In the current sample, internal consistency was .80.

Observed Positive Affect

We used the Iowa Family Interaction Rating Scales (IFIRS), a global coding system, to code the videotaped interactions. The Positive Mood code was used to measure observed positive affect. This code measures the degree to which the focal appears content, happy, optimistic (e.g., “[My diabetes] doesn’t really get in the way of anything”), and/or demonstrates positive behavior toward self (e.g., “I can control [my diabetes] now; I am independent”), others, or things in general. In our study, we observed positive statements related to both diabetes (e.g., “I did a good job checking my blood sugar this week”) and life in general (e.g., “life is great”) as well as nonverbal indicators of positive mood (e.g., smiling, laughing). Two trained bachelor’s level coders independently viewed each video interaction and rated adolescents on a scale from 1 (the focal never or rarely displays this behavior) to 9 (the focal frequently or consistently displays this behavior) based on the behaviors frequency, context, intensity, and proportion of the observed behaviors. The mean of the two coders’ scores were used in analyses. The IFIRS coding system is one of five observational coding systems designed to measure family functioning deemed “well established” for use in pediatric populations (Alderfer et al., 2008) and has been validated in use with parent–child dyads for a wide variety of ages (Melby & Conger, 2001). Interrater reliability was .78 for the Positive Mood code.

Glycemic Control

Glycosylated hemoglobin (HbA1c) was obtained from adolescents’ medical records, as a measure of glycemic control over the previous 8–12 weeks. The Bayer Diagnostics DCA2000® machine was used to conduct analyses of HbA1c; a normal range is considered 4.2–6.3%. The American Diabetes Association recommends a target HbA1c of ≤7.5% for children and adolescents (American Diabetes Association, 2014).

Internalizing and Externalizing Problems

Mothers reported on adolescents’ behavior problems on the Child Behavior Checklist (CBCL; Achenbach & Rescorla, 2001). For each of the 113 items, mothers rated behaviors on a 3-point Likert scale (0 = absent, 1 = occurs sometimes, 2 = occurs often). The Externalizing (composed of the aggressive and rule-breaking behavior subscales) and Internalizing (containing the somatic complaints and withdrawn or anxious depressed subscales) scales were used for this study. The CBCL provides T-scores based on age and sex from a nationally reported sample.

Quality of Life

The Type 1 Diabetes module of the Pediatric Quality of Life (PedsQL) was used to measure diabetes-specific quality of life (Varni et al., 2003). We used the total diabetes quality of life score, as recommended in the recent literature (Nansel, Weisberg-Benchell, Wysocki, Laffel, & Anderson, 2008). Scores range from 0 to 100, and higher values indicate better quality of life. Cronbach’s α was .88 at Time 1 and .92 at Time 2.

Analytical Plan

First, we conducted Little’s Missing Completely at Random (MCAR) test (Little, 1988) to determine whether data were missing completely at random. Because the test was not significant (χ² = 98.21, df = 99, p = .554), we used multiple imputation methods to replace missing data. Next, we conducted descriptive analyses to better understand positive affect in this sample. We tested for differences related to demographic (i.e., age, gender, race/ethnicity) and diabetes-related (i.e., treatment type, duration) variables using Wilcoxon rank sum tests (for categorical variables) and bivariate correlations (for continuous variables). To reduce the shared method variance and improve predictive power (Holmbeck, Li, Schurman, Friedman, & Coakley, 2002), we combined observed positive mood and self-reported positive affect (PANAS) into a composite variable of positive affect by summing standardized scores (z-scores). Next, bivariate correlations were calculated to examine associations between key study variables (i.e., positive affect, glycemic control, internalizing and externalizing problems, and quality of life) at each time point. Finally, a series of multiple regression analyses were conducted to test positive affect as a predictor of resilience. In line with previous longitudinal research in T1D (Hood, Rausch, & Dolan, 2011), we used change scores as the dependent variables, rather than examining a later outcome (e.g., internalizing problems at 6 months). Change scores were calculated for internalizing and externalizing problems (on the CBCL), glycemic control (HbA1c), and quality of life (PedsQL) by subtracting the baseline value from the value at 6 months. Adolescent age, gender, and race were added as covariates to the regression models, to account for their potential contribution to outcomes. Treatment type (pump or injection) was added to the model predicting change in HbA1c, as it has been shown to affect glycemic control (Sherr & Tamborlane, 2008). Each model was tested with individual indices of positive affect (observed and self-reported), and again with the composite positive affect variable.

Results

Descriptive Analyses

The mean age of the sample was 12.6 years old (SD = 2.1) with a mean duration of diabetes of 5.0 years (SD = 3.5) (Table I). At baseline, the average HbA1c was 7.6% (SD = 1.2), with 55.7% of participants meeting the target for glycemic control (HbA1c < 7.5%). Six months later, the average HbA1c was 7.9% (SD = 1.1) with 43.4% of participants meeting the target. Adolescents in our sample reported fairly high diabetes-related quality of life; mean PedsQL was 83.8 (SD = 11.7) at baseline and 85.8 (SD = 10.8) at Time 2 (compared to a mean of 71–75 in a national sample; Hood et al., 2014). At baseline, 24% of adolescents in our sample met criteria for borderline or clinical levels of internalizing behavior problems (19% at Time 2), consistent with previous research findings indicating increased levels internalizing problems and depressive symptoms in adolescents with T1D (Dantzer, Swendsen, Maurice-Tison, & Salaman, 2003). Finally, while 13% of our sample met borderline or clinical levels of externalizing behavior problems at baseline, 6% met the cutoff at Time 2.

Table I.

Demographic Characteristics (n = 93)

Adolescent variable
Age (years)
M(SD)	12.6 (2.1)
Range	10–16
Sex (%)
Female	44
Male	56
Race (%)
White	83.3
Black	6.7
Asian	1.1
Biracial	6.7
Other	2.2
Ethnicity (%)
Hispanic	9.8
Non-Hispanic	90.2
Duration of diabetes
M(SD) (years)	5.0 (3.5)
Range	0.5–14
Insulin therapy type (%)
Pump	83.9
Injections	16.1

Preliminary Analyses

Data were investigated for multivariate outliers and none was found. We conducted preliminary analyses to describe positive affect in this sample, by testing for demographic differences related to child age, gender, and race/ethnicity (White/non-White). We did not find any significant differences in observed positive mood related to age (r = −.08, p = .460). In addition, Wilcoxon rank sum tests indicated that there were not significant differences in observed positive mood related to sex (U = 1,049.00, p = .814) or race/ethnicity (U = 825.00, p = .368). Likewise, we did not find any significant differences in adolescents’ reports of positive affect on the PANAS related to age (r = .04, p = .685), sex (U = 1,057.00, p = .497), or race/ethnicity (U = 801.50, p = .251). Next, we tested for differences in positive affect related to diabetes variables. We did not find a significant association between observed positive mood and duration of diabetes (r = .03, p = .785), or treatment type (U = 461.50, p = .191). Similarly, we did not find a significant association between self-reported positive affect on the PANAS and duration of diabetes (r = .15, p = .165) or treatment type (U = 428.00, p = .128).

We also conducted preliminary analyses to detect any relationships between the target outcome variables (i.e., glycemic control, internalizing and externalizing problems, and quality of life) and demographic variables that have been recognized as possible covariates in previous research (i.e., child’s age, race, gender and duration of diabetes, treatment type). Consistent with previous research, we found a significant difference in glycemic control related to race/ethnicity at baseline (U = 3,178.00, p = .018) and at 6 months (U = 2,594.00, p = .010), indicating that non-White participants had higher HbA1c values than White participants at both time points. Quality of life at baseline was related to duration of diabetes (r = −.24, p = .024), indicating that adolescents with a longer duration of diabetes had poorer quality of life. At Time 2, poorer quality of life was associated with greater child age (r = −.27) and duration (r = −.22, both p < .05). Child age and duration of diabetes were not significantly related to internalizing or externalizing problems. Child race/ethnicity was related to externalizing problems at 6 months (U = 282.00, p = .004), indicating that non-White participants had more externalizing problems than White participants. There were no significant differences in key variables related to child sex or treatment type (pump vs. injections).

Correlational Analyses

We conducted bivariate correlations to examine the relationship between positive affect (as measured with self-report on the PANAS, observed positive mood, and the composite score) in relation to glycemic control, internalizing and externalizing problems, and quality of life. Consistent with our hypotheses, higher levels of positive affect were associated with lower levels of internalizing problems and externalizing problems at baseline (Table II). Similarly, higher levels of positive affect were associated with better glycemic control, lower internalizing problems and externalizing problems and better quality of life at Time 2 (see Table II).

Table II.

Descriptive Statistics and Bivariate Correlations Among Positive Affect and Outcomes

	1	2	3	4	5	6	7	8	9	10	11
1. Observed Positive Mood	–
M = 5.65 (1.64)
2. PANAS Positive	.37***	–
M = 35.54 (6.30)
3. Positive Affect	.83***	.83***	–
M = −.01 (1.66)
4. HbA1c T1	−.17	.02	−.09	–
M = 7.62 (1.15)
5. HbA1c T2	−.28**	−.14	−.26*	.64***	–
M = 7.89 (1.13)
6. PedsQL T1	.15	.17	.20	−.11	−.17	–
M = 83.79 (11.67)
7. PedsQL T2	.21*	.15	.22*	−.23*	−.31**	.65***	–
M = 85.84 (10.81)
8. Internalizing T1	−.15	−.24*	−.24*	.21*	.17	−.22*	−.29**	–
M = 50.95 (11.49)
9. InternalizingT2	−.18	−.24*	−.25*	.17	.10	−.25*	−.44***	.40***	–
M = 49.52 (10.65)
10. Externalizing T1	−.13	−.27**	−.24*	.18	.18	−.07	−.11	.84***	.18	–
M = 47.66 (9.73)
11. Externalizing T2	−.24*	−.33***	−.35***	.19	.28*	−.17	−.18	.43***	.52***	.55***	–
M = 46.14 (9.01)

Note. PANAS = Positive and Negative Affect Scale. Positive Affect represents summed z-scores of observed positive mood and adolescent self-reported positive affect. PedsQL = Diabetes-Specific Pediatric Quality of Life. Internalizing and Externalizing Symptoms were reported by mothers on the Child Behavior Checklist. T1 = Baseline data, T2 = 6 month data.

Linear Regressions

A series of multiple linear regressions were conducted to test our hypothesis, that positive affect would predict improvements in glycemic control, internalizing and externalizing symptoms, and quality of life. For each model, we adjusted for baseline values of the outcome variable, followed by demographic covariates found to be significant in preliminary analyses, and observed and self-reported positive mood. We then conducted regression analyses using the composite positive affect variable. In this way, we could determine the relative contribution of each variable. Change scores were used as the dependent variables.

While the overall model predicting change in glycemic control was significant (F(4, 88) = 6.70, p ≤ .001), explaining 20% of the variance, baseline HbA1c was the only significant predictor; the individual indices of positive affect were not significant. However, when we included the composite positive affect variable in the model, the overall model predicting change in glycemic control remained significant (F(3, 89) = 9.04, p ≤ .001), and the composite positive affect variable emerged as a significant predictor of change in glycemic control (β = −.13, p = .026), even after adjusting for baseline glycemic control and the demographic covariate (see Table III).

Table III.

Regression Analyses Predicting Change in Glycemic Control

	Change in glycemic control
Variable	Β	p value
Step 1: R2 change = .20***
HbA1c T1	−.38	<.001
Step 2: R2 change = .02
HbA1c T1	−.41	<.001
Child race	−.31	.208
Step 3: R2 change = .05
HbA1c T1	−.43	<.001
Child race	−.26	.283
Observed positive mood	−.10	.136
PANAS	−.02	.287
Full Model: Adjusted R2 = .20, F(88, 4) = 6.70***
Step 1: R2 change = .20***
HbA1c T1	−.38	<.001
Step 2: R2 change = .02
HbA1c T1	−.41	<.001
Child race	−.31	.208
Step 3: R2 change = .05*
HbA1c T1	−.42	<.001
Child race	−.26	.289
Composite positive affect	−.13	.026
Full Model: Adjusted R2 = .24, F(89, 3) = 9.04***

Note. β = standardized beta. PANAS = Positive and Negative Affect Scale.

*p < .05; **p < .01; ***p < .001.

Similarly, the overall model predicating change in externalizing behavior problems was also significant (F(4, 88) = 75.28, p < .001), explaining 76% of the variance (Table IV), but baseline symptoms of externalizing problems were the only significant predictor. The individual indices of positive affect were not significant. However, when the model was analyzed with the composite variable, positive affect emerged as a significant predictor of change in externalizing behavior problems (β = −.62, p = .016), even after adjusting for baseline externalizing behavior problems and demographic covariates.

Table IV.

Regression Analyses Predicting Change in Externalizing Symptoms

	Change in externalizing symptoms
Variable	Β	p value
Step 1: R2 change = .75***
Ext T1	−.72	<.001
Step 2: R2 change = .01
Ext T1	−.73	<.001
Child race	−1.49	.147
Step 3: R2 change = .02
Ext T1	−.76	<.001
Child race	−1.30	.2001
Observed positive mood	−.33	.219
PANAS	−.11	.128
Full Model: Adjusted R2 = .76, F(88, 4) = 72.58***
Step 1: R2 change = .75***
Ext T1	−.72	<.001
Step 2: R2 change = .01
Ext T1	−.73	<.001
Child race	−1.49	.147
Step 3: R2 change = .02*
Ext T1	−.76	<.001
Child race	−1.30	.199
Composite positive affect	−.62	.016
Full Model: Adjusted R2 = .76, F(89, 3) = 97.86***

Note. β = standardized beta. PANAS = Positive and Negative Affect Scale. Ext = Externalizing Symptoms on the Child Behavior Checklist.

*p < .05; **p < .01; ***p < .001.

Although the overall model predicting change in internalizing symptoms was significant (F(3, 89) = 73.08, p < .001), explaining 70% of the variance, positive affect was not a significant predictor after adjusting for baseline levels of internalizing symptoms (with either the individual indices of positive affect or the composite variable). Finally, the overall model predicting change in quality of life was significant (F(5, 83) = 8.54, p < .001), predicting 30% of the variance, but baseline quality of life was the only significant predictor (positive affect was not significant).

Discussion

The current study is one of the first to examine both observed and self-reported positive affect as a predictor of resilience in youth with T1D. Strengths of the study include the use of observational methods and the longitudinal design. Further, our sample was in relatively good glycemic control, with 43% meeting recommended targets—as compared with only 20% in a national study (Wood et al., 2013)—offering the opportunity to examine predictors of resilience in a relatively well-controlled sample. We found significant associations between observed and self-reported positive affect with glycemic control, quality of life, and psychological symptoms, and notably, higher levels of positive affect predicted improvements in glycemic control and internalizing symptoms over 6 months.

Findings from the current study are in line with Hilliard and colleagues’ (2012) proposed model of diabetes resilience, such that positive affect acted as an individual protective process that predicted better outcomes, in the face of diabetes-related stress. Notably, positive affect predicted improvements in glycemic control, similar to findings in an adult population (Tsenkova et al., 2008). The broaden-and-build hypothesis (Fredrickson, 2001) suggests that positive affect leads to a broadening of attention and cognition, enhancing creativity and flexibility in coping strategies. Thus, increased positive affect may help adolescents with T1D use more adaptive coping strategies to manage stress related to the daily demands of diabetes care (Tran et al., 2011). Additionally, positive affect may provide respite from chronic stress, giving adolescents the opportunity to renew social, intellectual, and physical reserves (Fredrickson, 1998). Finally, the fact that the individual measures of positive affect (observed and self-report) were not significant predictors, but the composite variable was significant, supports that the use of multisource and multimethod data improves predictive power (Holmbeck et al., 2002).

Internalizing and externalizing symptoms in adolescents have been associated with the use of less adaptive coping strategies (i.e., avoidance) and less frequent blood glucose monitoring (Northam et al., 2010), resulting in deteriorating glycemic control (Herzer & Hood, 2010; Holmes et al., 2006; Luyckx, Seiffge-Krenke, & Hampson, 2010). Our finding that positive affect predicted improvements in externalizing symptoms suggests that positive affect may have benefits both physiologically and psychologically. Lower quality of life is also important to consider as a meaningful outcome in itself, as well as a risk factor for later problems with glycemic control (Hood et al., 2014). Although we found associations between positive affect and quality of life, positive affect was not a significant predictor of changes in quality of life. This may be owing to the fact that the current sample was relatively high functioning, and therefore may experience fewer problems with quality of life than those typically seen in adolescents with T1D.

While we have conceptualized positive affect as a protective process in this study, it could also be considered an indicator of resilience in itself. Davidson defines resilience as “the maintenance of high levels of positive affect and well-being in the face of significant adversity” (Davidson, 2000, p. 1198). Future studies, therefore, could examine predictors of positive affect, such as coping. For example, the ability to maintain and generate positive affect within a stressful context (e.g., talking about diabetes stress) may be the result of adaptive coping strategies (Folkman & Moskowitz, 2000). Further, including measures of adherence or self-management would help to further explain the relationship between positive affect and glycemic control.

Findings from the current study highlight positive affect as a novel target for interventions to improve outcomes in this high-risk population. Laboratory studies and clinical trials demonstrate that positive affect can be induced, increasing effort and motivation (Erez & Isen, 2002; Isen & Reeve, 2005). While interventions to promote positive affect, which include exercises in gratitude, self-affirmations, and small gifts, have been tested in adults (Moskowitz et al., 2012) and have successfully improved adherence to medication and physical activity recommendations in adults (Charlson et al., 2009), to our knowledge, this type of intervention has only been pilot tested in pediatric population (Jaser, Patel, Rothman, Choi, & Whittemore, 2014). Given that we did not find any differences in positive affect (either observed or self-reported) related to demographic or clinical variables, interventions to promote positive affect may have potential to improve outcomes in both boys and girls across a range of ages, racial/ethnic backgrounds, duration of diabetes, and treatment type.

Limitations

Some limitations of the current study must be acknowledged. First, the sample size limited power to detect smaller effects, and a larger sample may reveal differences in positive affect related to age or gender. The relatively low rate of participation, although it is similar to other studies that include observational measures (e.g., Dunn et al., 2011; Janicke, Mitchell, & Stark, 2005), may limit the generalizability of findings. Future studies should consider collecting data during regularly scheduled clinic visits or in participants’ homes to increase participation rates. Further, the current sample had relatively high socioeconomic status, which may be an important personal asset to consider in the context of resilience. In addition, an older sample or one in poorer glycemic control may have exhibited lower levels of positive affect. Finally, although the current study used a longitudinal design, it is possible that preexisting psychological problems or suboptimal glycemic control influenced our observations of positive affect.

Conclusion and Implications

Despite these limitations, findings from the current study extend previous findings and establish a connection between positive affect and resilient outcomes. Our results suggest that providers should consider asking about sources of positive affect when working with adolescent patients with T1D. For example, for many adolescents, every blood glucose check is a test that they could fail, and many report feeling guilty about “bad” numbers (Davidson et al., 2004). Given this negative association, it is not surprising that adolescents often avoid checking blood glucose. An intervention study by Jaser and colleagues (2014) found a significant association between positive affect and multiple measures of adherence (self-report, meter download) in teens with type 1 diabetes. Similarly, an intervention for adults with chronic cardiopulmonary disease using positive affect and self-affirmation for behavior change found that inducing positive affect not only buffered stress, but increased self-efficacy and promoted success in behavior change (Charlson et al., 2014). These findings suggest that providing positive affirmations or asking youth to recall something that they are proud of may provide additional motivation for diabetes management. Protective factors and processes may be easier to influence and reinforce than risk factors (Hilliard et al., 2012) and therefore deserve attention from both researchers and providers. Future studies are needed to replicate these findings in other samples and to understand how positive affect relates to adherence.

Funding

This research was supported by grants from the National Institute of Diabetes and Digestive and Kidney Diseases [K23 DK088454] and the National Center for Research Resources [UL1 RR024139]. The contents of this manuscript are solely the responsibility of the authors and do not necessarily represent the official view of NIH.

Conflicts of interest: None declared.