Longitudinal Relationships between Depression and Chronic Illness in Adolescents: An Integrative Review

Abstract

Introduction:

Depression is prevalent among adolescents with chronic illness. Little is known about how depression affects chronic illness over time. This review aimed to synthesize longitudinal relationships between depression and disease control, self-management behaviors, illness-related morbidity, and quality of life.

Method:

Four databases were searched: PubMed, CINAHL, PsycINFO, EMBASE. Inclusion criteria were cohort studies examining depression among adolescents ages 10–21 with a chronic illness and published in English. Study quality was appraised using the Newcastle-Ottawa scale and data was synthesized by outcome.

Results:

Of 3,463 articles identified, 11 were included. For adolescents with diabetes, increased depressive symptoms predicted decreased metabolic control/monitoring, medication adherence, quality of life, and increased hospitalization. Studies on cystic fibrosis, congenital heart disease, sickle cell disease, and juvenile idiopathic arthritis were limited but demonstrated that depressive symptoms affected quality of life, disability, pain, and hospitalization rates/costs.

Discussion:

Evidence supports the need for mental health care strategies suitable for adolescents with chronic illness. Future research is needed to examine effects of depressive symptoms across diversified chronic illness populations.

Introduction

Approximately one-third of adolescents in the United States (U.S.) live with a chronic medical illness (Park, 2013). Adolescents often have difficulty adjusting to the demands imposed by chronic illness during this developmentally complex stage in life (Compas et al., 2012), thereby making it common for chronic illness to be accompanied by poor mental health. The general prevalence of depression among adolescents has increased from 8.7% in 2005 to 11.3% in 2014 (Mojtabai et al., 2016), and those with chronic illness are shown to be at even higher risk for depression than their healthy peers (Kline-Simon et al., 2016; Pinquart & Shen, 2011). This is due in part to the burdens (e.g., unpredictable symptom exacerbations, daily care regimens) that illness can impose on social activities or relationships that are crucial for positive development (Spirito et al., 1991; Suris et al., 2004; Taylor et al., 2008). As chronic illness and depression independently contribute to the highest proportions of years lived with disability (Patton et al., 2016), adolescents with chronic illness and comorbid depression are doubly disadvantaged in that depressive symptoms can exacerbate issues (e.g., impeding social activities, symptom burden) secondary to a chronic illness.

In adults, the longitudinal relationships between chronic illness and comorbid depression have been described and demonstrate that depression commonly predicts disease-related morbidity and mortality over time (Voinov et al., 2013). However, less is known about the long-term, cumulative effects that depression has on chronic illness in adolescents. It is likely that depression may affect this population in unique ways, as pubertal timing is associated with onset of depressive symptoms (Copeland et al., 2019; Lewis et al., 2018). While depressive symptoms have commonly been associated with poor disease control (Kwong et al., 2016; Richardson et al., 2006), existing reviews examining such relationships among adolescents focus on singular illnesses, primarily using cross-sectional studies that do not describe relationships over time (Kongkaew et al., 2014).

The persistence of depression among adolescents (Brady et al., 2017) emphasizes the need for an increased healthcare focus on adolescents affected by chronic illness and comorbid depressive symptoms/depression in order to alleviate the challenges faced by this vulnerable yet understudied population. Therefore, we conducted an integrative review to critically evaluate and synthesize literature exploring the longitudinal relationships between comorbid depressive symptoms/depression and disease control, self-management behaviors, chronic illness-related morbidity, and quality of life (QOL).

Methods

Literature Search

A comprehensive literature search was conducted in February 2019 using four primary databases (PubMed, Cumulative Index of Nursing and Allied Health Literature (CINAHL), PSYCHinfo, and EMBASE). We consulted with a library informationist to facilitate the development of the search strategy. We also explored subject headings and keywords within each respective database to inform the search strategies. Appendix A lists the full search strategies with key terms and Boolean operators. Search strategies incorporated terms for each chronic illnesses of interest. Though definitions of chronic illness can vary, we included chronic illnesses meeting the following definition: a long-term, non-communicable medical illness (e.g., diabetes, cystic fibrosis) that still allows for participation in activities of daily living through proper self-management (Bernell & Howard, 2016). Across databases, we searched the following keywords: chronic disease, chronic illness, asthma, diabetes, cystic fibrosis, arthritis, epilepsy, heart condition, blood disorder, inflammatory bowel disease, depression, comorbid depression, depressive, adolescent, teen, youth, young people, functional impairment, functional limitation, quality of life, disease management, symptom. Consistent with recommended methodology for integrative reviews (Whittemore & Knafl, 2005), we did not place restriction on year of publication to ensure inclusion of all relevant sources of evidence. Discussions among the reviewers led to a subsequent revision in the search strategies, and the searches were updated in April 2019 to include sickle cell as a keyword in the event that the keyword blood disorder did not pick up relevant articles.

Eligibility Criteria and Study Selection

We exported all articles yielded from the initial database search to Covidence online management software (Veritas Health Innovation, 2019) for duplicate removal. Two reviewers then independently screened for relevance based on title and abstract and reviewed the full-texts of the remaining articles based on predetermined eligibility criteria. To be included in the review, articles must have met all of the following criteria: 1) prospective or retrospective cohort study published in English; 2) included a sample of adolescents, defined as individuals ages 10 to 21 (Sawyer et al., 2018); 3) examined a chronic illness meeting our definition as described above 4) examined depression or depressive symptoms comorbid to a chronic illness; and 5) explored the impact of depression or depressive symptoms on disease control, chronic illness-related morbidity, self-management behaviors such as medication adherence, or quality of life. As depression is often underdiagnosed in adolescents (Sigalas et al., 2014), we included studies that identified depressive symptoms either through self-report measures or a diagnosis confirmed through medical records. In addition, we included studies examining youth below the 10-year threshold of our age criteria were included if the mean sample age fell within the targeted age range or if statistical analyses were adjusted for age. Studies were excluded if they: 1) used a cross-sectional study design that did not examine long-term associations; 2) focused on adolescents with physical/developmental/neurological disabilities (e.g., autism, cerebral palsy, Turner syndrome) or illnesses that could be either acute or chronic (e.g., cancer) due to the substantial differences in care considerations and social implications of these conditions; 3) focused on psychiatric conditions (e.g., anorexia, bipolar disorder) other than depression.

Quality Appraisal

Two reviewers independently appraised and rated the quality of each study using the Newcastle-Ottawa Assessment Form for Cohort Studies, a validated tool developed by a panel of healthcare experts (Wells et al., 2009). The reviewers met to discuss discrepancies in ratings until consensus was reached. If consensus could not be reached between the two reviewers, a third reviewer independently assessed the quality of the study. The Newcastle-Ottawa appraises items across three study domains: 1) selection, which has four criteria addressing representativeness, recruitment, exposure assessment, and ascertainment of outcome; 2) comparability, which has one criterion addressing methods to control for confounding variables; and 3) outcome, which has three criteria addressing follow-up, outcome assessment, and attrition rates. Each criterion that is met is awarded a star, with the comparability criterion being awarded a maximum of two stars. In total, studies can earn a maximum of nine stars.

The reviewers then used the Newcastle-Ottawa appraisal ratings to determine overall study quality (i.e., good, fair, poor) based on reporting standards published by the Agency for Healthcare Research and Quality (Viswanathan et al., 2018). Good quality studies are defined as 3 – 4 stars awarded in the selection domain, 1 – 2 stars awarded in the comparability domain, and 2 – 3 stars awarded in the outcome domain. Fair quality studies are defined as 2 stars awarded in the selection domain, 1 – 2 stars awarded in the comparability domain, and 2 – 3 stars awarded in the outcome domain. Poor quality studies are defined as 0 – 1 star awarded in the selection domain, or 0 stars awarded in the comparability domain, or 0 – 1 star awarded in the outcome domain.

Data Extraction

Two reviewers systematically extracted data from each article by summarizing the author, year, country, study design, sample characteristics, depression measure, study outcomes, and key findings. The reviewers classified associations into the following categories: disease control; self-management behaviors; illness-related morbidity; and quality of life. If an investigator conducted both cross-sectional and longitudinal analyses, only the longitudinal outcomes were extracted and synthesized. Studies were grouped by illness type to allow for comparisons across studies reporting similar outcomes.

Results

Data Sources

Figure 1 depicts a flow diagram detailing the article selection process. The initial search results yielded a total of 3,463 studies. After removal of duplicates, the reviewers initially screened 2,608 studies. Among these, the reviewers deemed 25 articles relevant by title and abstract and reviewed them in full. In total, we included 11 cohort studies that met eligibility criteria, 10 of which were prospective (Baucom et al., 2018; Guo et al., 2015; Hanns et al., 2018; Helgeson et al., 2009; Hoff et al., 2006; Hood et al., 2011; Katz et al., 2016; Luyckx et al., 2014; Stewart et al., 2005; Van Buren et al., 2018) and one of which was retrospective (Snell et al., 2014).

Figure 1.

Flow diagram depicting article search and selection

Study Characteristics

Table 1 provides a summary of the study characteristics. Study settings spanned across several countries, including the United States (n = 8), China (n = 1), England (n = 1), and Belgium (n = 1). Five studies were multi-site studies (Baucom et al., 2018; Guo et al., 2015; Hanns et al., 2018; Katz et al., 2016; Van Buren et al., 2018). Six studies focused on adolescents with type 1 diabetes (Baucom et al., 2018; Guo et al., 2015; Helgeson et al., 2009; Hood et al., 2011; Stewart et al., 2005). The remaining studies examined type 2 diabetes (n = 2) (Katz et al., 2016; Van Buren et al., 2018), cystic fibrosis (n = 1) (Snell et al., 2014), congenital heart disease (n = 1) (Luyckx et al., 2014), and juvenile idiopathic arthritis (n = 1) (Hanns et al., 2018). One study followed adolescents with juvenile idiopathic arthritis and sickle cell disease (Hoff et al., 2006). Samples were followed over a period ranging from 6 months (Hood et al., 2011) to 4 years (Hanns et al., 2018). The retrospective study examined medical records from 2002 to 2003 (Stewart et al., 2005). Two studies analyzed data from the same sample of adolescents with type 2 diabetes (Katz et al., 2016; Van Buren et al., 2018), but we included both because different outcomes were assessed. Ages of the participants in the samples ranged from 8 to 21 years. In total, the current review reflects data from 2,287 adolescents with a chronic illness (70% diabetes, 1% cystic fibrosis, 19% congenital heart disease, 7% juvenile idiopathic arthritis, 3% sickle cell disease).

Table 1.

Characteristics of Longitudinal Cohort Studies Included in Review

Author, Year, Country	Study Length & Attrition	Chronic Illness Sample	Age, Sex, Race/Ethnicity	Depression Measure	Comparison Groups	Outcomes Measured
Baucom et al., 2018, United States*	2 years (Prospective) 21% Attrition	N = 247 with type 1 diabetes	17 – 18 years Sex NR Race/Ethnicity NR	Center for Epidemiologic Studies Depression Scale (Range 0 – 60, cut-off ≥ 16)	No comparison groups (41% above CES-D cut-off score at baseline)	Self-management Medication adherence Disease control HbA1c levels
Guo et al., 2015, China*	1 year (Prospective) 36.8% Attrition	N = 136 with type 1 diabetes	8 – 19 years 62.8% female 100% Asian	Depression Self-Rating Scale for Children (Range 0–36, cut-off ≥ 15)	No comparison groups (Mean DSRS score 9.43 at baseline, 11.43 at follow-up)	Disease control HbA1c levels Quality of life Satisfaction
Helgeson et al., 2009, United States	3 years (Prospective) 5% Attrition	N = 132 with type 1 diabetes	11 – 13 years 53% female 93% white, 2% black, 1% Asian, 1% American Indian, 3% mixed	Children’s Depression Inventory- Short Form (Range 0–20, cut-off ≥ 3)	No comparison groups (Mean CDI-S score at baseline: 1.15)	Disease control HbA1c levels
Hood et al., 2011, United States	6 months (Prospective) Attrition NR	N = 145 with type 1 diabetes	13 – 18 years 52.4% female 86.9% non-Hispanic white, other NR	Children’s Depression Inventory (Range 0–54, cut-off ≥ 13)	No comparison groups (22.8% above cut-off score at baseline, 15.2% at 6 months)	Disease control HbA1c levels Self-management Blood glucose monitoring
Stewart et al., 2005, United States	2 years (Prospective) Attrition NR	N = 231 with type 1 diabetes	11 – 18 years 60% female Race/Ethnicity NR	Center for Epidemiologic Studies Depression Scale (Range 0 – 60, cut-off score ≥ 12 for boys, ≥ 22 for girls)	Above CES-D cut-off: N = 76 (33% Below CES-D cut-off: N = 155 (67%)	Illness-Related Morbidity Hospitalization
Katz et al., 2016, United States*	2 years (Prospective) 50.4% Attrition	N = 699 with type 2 diabetes	10 – 17 years 63.8% female 21.9% non-Hispanic white, 34.4% non-Hispanic black, 43.7% Hispanic	Children’s Depression Inventory if age <16 (Range 0–54, cut-off score ≥ 13) Beck’s Depression Inventory if age ≥16 (Range 0–63, cut-off score ≥ 14)	Above CDI/BDI cut-off: N = 105 (15%) Below CDI/BDI cut-off: N = 594 (85%)	Self-management Medication adherence
Van Buren et al., 2018, United States*	2 years (Prospective) 16% Attrition	N = 699 with type 2 diabetes	10 – 17 years 65.5% female 20.2% non-Hispanic white, other NR	Children’s Depression Inventory if age <16 (Range 0–54, cut-off score ≥ 13) Beck’s Depression Inventory if age ≥16 (Range 0–63, cut off score ≥ 14)	No comparison groups (20% above CDI or BDI cut-off score during at least 1/3 time points)	Disease control HbA1c levels
Snell et al., 2014, United States	2 years (Retrospective) Attrition NA	N = 40 with cystic fibrosis	12 – 21 years 40% female 95% white, other race/ethnicity NR	Depression diagnosis determined from medical records	Depression: 20 (50% No depression: 20 (50%)	Disease Control FEV1 Self-management Medication adherence Illness-Related Morbidity Hospitalization Care costs
Luyckx et al., 2014, Belgium	1.5 years (Prospective) 18.8% Attrition	N = 429 with congenital heart disease	14 – 18 years Sex NR Race/Ethnicity NR	Center for Epidemiologic Studies Depression Scale (Range 0–60, cut-off ≥ 16)	No comparison groups (% above CES-D cut-off or mean score NR)	Illness-Related Morbidity Loneliness Quality of life Satisfaction Perceived health status
Hanns et al., 2018, England*	4 years (Prospective) 0% Attrition	N = 102 with juvenile idiopathic arthritis	11 – 16 years 58% female Race/Ethnicity NR	Mood and Feelings Questionnaire (Range 0–66, cut-off ≥ 27)	Above MFQ cut-off score: N = 15 (15%) Below MFQ cut-off score: N = 87 (85%)	Illness-Related Morbidity Pain Disability
Hoff et al., 2006, United States	1 year (Prospective) 6% Attrition	N = 66 with juvenile idiopathic arthritis and 61 with sickle cell disease	8 – 17 years 81% female 87% Caucasian, 10% black, 3% other	Revised Child Anxiety and Depression Scale (Range 0–18, cut-off ≥ 11)	No comparison groups (14% with JIA above RCADS cut-off, 27% with SCD at baseline)	Quality of Life Pain Disability

Note.

^*Indicates multi-site study;

FEV₁ = Forced Expiratory Volume; NR = Not Reported; NA = Not Applicable; BMI = Body Mass Index; CES-D = Center for Epidemiologic Studies Depression Scale; CDI = Children’s Depression Inventory; DSRS = Depression Self Rating Scale.

Across prospective studies, researchers measured depressive symptoms using self-report instruments validated for use in child and adolescent populations: The Center for Epidemiologic Studies Depression Scale (CES-D) (Baucom et al., 2018; Luyckx et al., 2014; Stewart et al., 2005); Depression Self-Rating Scale (DSRS) for Children (Guo et al., 2015); Children’s Depression Inventory (CDI) (Helgeson et al., 2009; Hood et al., 2011; Katz et al., 2016; Van Buren et al., 2018); Beck’s Depression Inventory (BDI);(Katz et al., 2016; Van Buren et al., 2018) Revised Child and Anxiety Depression Scale (RCADS) (Hoff et al., 2006); and the Mood and Feelings Questionnaire (MFQ) (Hanns et al., 2018). Though a diagnosis cannot be determined from the use of these instruments alone, each instrument identifies a cut-off score that signifies clinically elevated depressive symptoms suggestive of a depressive disorder. Table 1 summarizes the cut-off scores by instrument. Three studies compared groups above and below a clinical cut-off score (Hanns et al., 2018; Katz et al., 2016; Stewart et al., 2005). The remaining studies examined depressive symptom scores over time (Baucom et al., 2018; Guo et al., 2015; Helgeson et al., 2009; Hoff et al., 2006; Hood et al., 2011; Van Buren et al., 2018). For the retrospective study, researchers compared adolescents with and without a formal diagnosis of depression and medical charts were reviewed to validate a diagnosis of depression as documented by a psychiatrist (Snell et al., 2014).

Quality of Studies

Table 2 summarizes the quality of studies as determined from the Newcastle-Ottawa ratings. Four studies were good quality (Hanns et al., 2018; Katz et al., 2016; Snell et al., 2014; Stewart et al., 2005), six were fair quality (Baucom et al., 2018; Guo et al., 2015; Helgeson et al., 2009; Hoff et al., 2006; Hood et al., 2011; Luyckx et al., 2014), and one was poor quality (Van Buren et al., 2018). As illustrated in table 2, the majority of studies did not fully meet criteria within the selection domain. Six lacked a comparison group of adolescents without depression/clinically elevated depressive symptoms, and 10 studies (Baucom et al., 2018; Guo et al., 2015; Hanns et al., 2018; Helgeson et al., 2009; Hoff et al., 2006; Hood et al., 2011; Katz et al., 2016; Luyckx et al., 2014; Stewart et al., 2005; Van Buren et al., 2018) relied on self-report measures of depressive symptoms rather than a diagnosis of depression. Within the comparability domain, three studies (Hanns et al., 2018; Katz et al., 2016; Van Buren et al., 2018) lacked statistical methods that controlled for both demographic factors (e.g., age, sex, socioeconomic status) and study-related factors (e.g., peer/family support, duration of diagnosis, study site, treatment regimens, disease severity). Within the outcome domain, one study (Luyckx et al., 2014) used self-reported outcome measures (i.e., quality of life satisfaction, perceived health status, and loneliness) without indication that surveys were administered by blinded study personnel, and one study (Katz et al., 2016) reported high attrition (> 20% drop-out) with no description of the lost participants.

Table 2.

Summary of Quality Ratings Across Studies

Study	Selection				Comparability	Outcome			Quality
	Representative-ness of the exposed cohort	Selection of the non-exposed cohort	Ascertainment of exposure	Demonstration that outcome of interest was not present at start of study	Comparability of cohorts or analysis controlled for confounders	Assessment of outcome	Follow-up long enough for outcomes to occur	Adequacy of follow-up cohorts
Baucom et al. (2018)	★	-	-	★	★★	★	★	★	Fair
Van Buren et al. (2018)	★	★	-	★	- -	★	★	★	Poor
Guo et al. (2015)	★	-	-	★	★★	★	★	★	Fair
Hanns et al. (2018)	★	★	-	★	★ -	★	★	★	Good
Helgeson et al. (2009)	★	-	-	★	★★	★	★	★	Fair
Hoff et al. (2006)	★	-	-	★	★★	★	★	★	Fair
Hood et al. (2011)	★	-	-	★	★★	★	★	★	Fair
Katz et al. (2016)	★	★	-	★	★ -	★	★	-	Good
Luyckx et al. (2014)	★	-	-	★	★★	-	★	★	Fair
Stewart et al. (2005)	★	★	-	★	★★	★	★	★	Good
Snell et al. (2014)	★	★	★	★	★★	★	★	★	Good

Above is a summary of the quality ratings across studies using the Newcastle-Ottawa Assessment Form for Cohort Studies. Good quality equates to 3 or 4 stars in selection domain AND 1 or 2 stars in comparability domain AND 2 or 3 stars in outcome domain. Fair quality equates to 2 stars in selection domain AND 1 or 2 stars in comparability domain AND 2 or 3 stars in outcome domain. Poor quality equates to 0 or 1 star in selection domain OR 0 stars in comparability domain OR 0 or 1 stars in outcome domain.

Outcomes Across Studies

Table 3 summarizes the results reported across studies. Six studies examined outcomes related to disease control (i.e., metabolic control) (Baucom et al., 2018; Guo et al., 2015; Helgeson et al., 2009; Hood et al., 2011; Snell et al., 2014; Van Buren et al., 2018), four studies examined outcomes related to self-management behaviors (i.e., blood glucose monitoring and medication adherence) (Baucom et al., 2018; Hood et al., 2011; Katz et al., 2016; Snell et al., 2014), five studies examined illness-related morbidity (i.e., hospitalization, care costs, pain, disability, loneliness) (Hanns et al., 2018; Hoff et al., 2006; Luyckx et al., 2014; Snell et al., 2014; Stewart et al., 2005), and two studies examined quality of life (i.e., satisfaction with quality of life and perceived health status) (Guo et al., 2015; Luyckx et al., 2014).

Table 3.

Effects of Depressive Symptoms/Depression across Studies

Study	Main Predictor Variable(s)	Outcome Variable(s)	Follow-up Periods	Effect
Effects on Disease Control
Baucom et al., 2018	Mean CES-D score	HbA1c change	Baseline, 1-year, 2-year	Between-person: β = .028* Within-person: β = .009
Guo et al., 2015	DSRS score change	HbA1c change	Baseline, and 6 to 12-mo	β = .03
Helgeson et al., 2009	CDI score change CDI change × time	HbA1c change	Baseline, 1-year, 2-year, 3-year	Score change: β = 1.18** Interaction: β = −.71***
Hood et al., 2011	CDI score change	HbA1c change	Baseline, 6-mo	β = .11*
Snell et al., 2014	Comparison groups by depression diagnosis	FEV1 Body mass index	2 years of medical records	Mean FEV1 in depressed vs. non-depressed groups: 67% vs. 71%
Van Buren et al., 2018	Baseline CDI score	HbA1c change Glycemic control	Baseline, 6-mo, 24-mo	Explanatory results NR
Effects on Self-Management Behaviors
Baucom et al., 2018	Mean CES-D score	Medication adherence	Baseline, 1-year, 2-year	Between-person: β = −.004* Within-person: β = −.002*
Hood et al., 2011	CDI score change	Blood glucose monitoring	Baseline, 6-mo	β = .024***
Katz et al., 2016	Comparison groups by CDI cut-off score	Medication adherence	Every 2-mo for year 1, quarterly for year 2	Proportion of adolescents above CDI cut-off in high vs. low med. adherence group: 17% vs. 12%***
Snell et al., 2014	Comparison groups by depression diagnosis	Medication adherence	2 years of medical records	Mean adherence concerns documented by clinician in depressed vs. non-depressed groups: 3.42 v. 2.3
Effects on Illness-Related Morbidity
Stewart et al., 2005	Comparison groups by CES-D cut-off score	Hospitalization	2 years of medical records	Below CES-D (reference) vs. above CES-D: HR = 2.58, 95% CI 1.12 – 5.96***
Snell et al., 2014	Comparison groups by depression diagnosis	Hospitalization Care costs	2 years of medical records	Mean number of annual hospitalizations in depressed vs. non-depressed group: 4 vs. 1.2* Total hospitalization costs in depressed vs. non-depressed group: $280,000 vs. $60,116*
Luyckx et al., 2014	CES-D score at baseline CES-D score at 9 months	Loneliness	Baseline, 9-mo, 18-mo	Baseline to 9 months: r = .19* 9 months to 18 months: r = .18*
Effects on Illness-Related Morbidity (Continued)
Hanns et al., 2018	Comparison groups by MFQ cut-off score	Pain, Disability	Baseline, 6-mo, 12-mo, 24-mo, 36-mo, 48-mo	Group difference in pain score changes: Z = −2.79** Change in group below MFQ cut-off: 1.206 Change in group above MFQ cut-off: 3.143 Group difference in disability score changes: Z = −3.31* Change in group below MFQ cut-off: 0.19 Change in group above MFQ cut-off: 0.72
Hoff et al., 2006	Baseline RCADS score × baseline pain	Pain, Disability	Baseline, 6-mo, 12-mo	JIA group: Pain: β = −.017** Disability: β = −.026** SCD group: no significant results, NR
Effects on Quality of Life
Guo et al., 2015	DSRS score change	Quality of life satisfaction change	Baseline, and 6 to 12-mo	β = −.17**
Luyckx et al., 2014	CES-D score at baseline CES-D score at 9 months	Quality of life satisfaction, Perceived health status	Baseline, 9-mo, 18-mo	Baseline to 9 months: QOL: r = −.12** PHS: r = −.08*** 9 months to 18 months: QOL: r = −.12** PHS: r = −.08***

Note.

^*P < 0.001;

^**P < 0.01;

^***P < 0.05;

FEV₁ = Forced Expiratory Volume, NR = Not Reported, CES-D = Center for Epidemiologic Studies Depression Scale, DSRS = Depression Self Rating Scale, CDI = Children’s Depression Inventory, HADS= Hospital and Anxiety Depression Scale, MFQ = Mood and Feelings Questionnaire, RCADS= Revised Children’s Anxiety and Depression Scale, RR = Rate Ratio

Disease Control

Researchers most extensively examined the relationship between depressive symptoms and disease control among adolescents with diabetes. Of these six studies, one study examined whether baseline CDI scores predicted longer-term HbA1c levels among a sample of adolescents with type 2 diabetes and found no significant relationship (Van Buren et al., 2018).

The remaining five studies examined whether changes in depressive symptoms predicted changes in HbA1c levels among adolescents with type 1 diabetes (Baucom et al., 2018; Guo et al., 2015; Helgeson et al., 2009; Hood et al., 2011), and, of these, all but one (Guo et al., 2015) found a significant relationship. Over a period of six months, Hood and colleagues found that a one-unit increase in CDI score predicted an increase of 0.11% in HbA1c (p < .001) (Hood et al., 2011). Using a longer follow-up period of three years, Helgeson and colleagues also found that a one-unit increase of CDI scores predicted an overall increase of 1.18% in HbA1c (p < .01) (Helgeson et al., 2009). However, the effect of depressive symptoms on HbA1c levels diminished over time (β = −0.71, p < .01). In a sample of 195 adolescents with type 1 diabetes, Baucom and colleagues examined the longitudinal relationship between CES-D scores at baseline, one year, and two years (Baucom et al., 2018). When examining within-person effects, a one-unit increase in an individual’s CES-D score predicted an increase of 0.28% in their HbA1c over time (p = .009). No significant between-person effects were found.

In one retrospective study of adolescents with cystic fibrosis, researchers reviewed medical records to examine the effect of a depression diagnosis on lung function over a one-year period following a diagnosis of depression. Mean forced expiratory volume (FEV₁) was compared in a sample of 40 adolescents with (n = 20) and without depression (n = 20) (Snell et al., 2014). There were no significant differences in FEV₁ between groups (67% vs. 71%).

Self-Management Behaviors

Of the five studies that examined the relationship between depressive symptoms and self-management behaviors, one study examined the effect of depressive symptoms on blood glucose monitoring (BGM) frequency among adolescents with type 1 diabetes (Hood et al., 2011). This study demonstrated that changes in depressive symptoms moderated the relationship between BGM frequency and HbA1c levels, such that increases in depressive symptoms worsened BGM frequency, which subsequently accelerated an increase in HbA1c levels over time (β = .024, p < .05) (Hood et al., 2011).

Three studies examined the relationship between depressive symptoms/depression and medication adherence (Baucom et al., 2018; Katz et al., 2016; Snell et al., 2014). Baucom and colleagues found significant between-person and within-person effects among adolescents with type 1 diabetes using the Diabetes Behavior Rating Scale (range 0.06 – 1.00) (Baucom et al., 2018). An individual’s adherence score decreased by .004 as their CES-D score increased one-unit above the sample mean score (p < .001) over two years. Additionally, an individual’s adherence score decreased by .002 as their CES-D score increased one-unit above their individual mean score (p < .001). Another study followed a sample of adolescents with type 2 diabetes and found that adolescents above the CDI cut-off score at baseline were more likely to demonstrate low medication adherence (defined as < 80% medication taken) than high medication adherence at 48-month follow up (17% vs. 12%, p < .05) (Katz et al., 2016). Among adolescents with cystic fibrosis, Snell and colleagues compared the number of adherence concerns documented in medical records for adolescents with and without depression throughout the one-year period and found no differences by depression status (3.42 vs. 2.30) (Snell et al., 2014).

Illness-Related Morbidity

Of the five studies that examined illness-related morbidity, two examined pain and disability (Hanns et al., 2018; Hoff et al., 2006). Using a Visual Analogue Pain Scale, Hanns and colleagues (2018) found that adolescents with juvenile idiopathic arthritis above the MFQ cut-off score at baseline had significantly worsened pain scores than those below the MFQ cut-off score at both 12 months (3.14 vs. 1.21, Z = −2.79, p < .01) and at 48 months (1.72 vs. 0.19, Z = −3.31, p < .01). Adolescents with scores above the MFQ cut-off had significantly worsened disability at both 12 months and 48 months relative to adolescents below the cut-off score, using the Child Health Assessment Questionnaire (1.72 vs. 0.19, Z = −3.31, p < .001). Using RCADS scores, one study examined the relationship between depressive symptoms and pain intensity (Faces Pain Scale, range 0 – 6) in a mixed sample of adolescents with juvenile idiopathic arthritis and sickle cell disease. For adolescents with juvenile idiopathic arthritis, the lower pain intensity was at baseline, the greater the effect that depressive symptoms had on pain at 6 and 12 months (β = −.017, p < .005) (Hoff et al., 2006). Using the Functional Disability Index, it was also found that the lower disability was at baseline, the greater the effect that depressive symptoms had on disability at 6 and 12 months (β = −.026, p < .009) (Hoff et al., 2006). For adolescents with sickle cell disease, no significant relationships were found when examining pain or disability.

Two studies examined hospitalization likelihood and frequency/care costs, respectively (Snell et al., 2014; Stewart et al., 2005). Stewart and colleagues compared hospitalizations secondary to diabetes complications between adolescents with type 1 diabetes above and below the CES-D cut-off score over a two-year period (Stewart et al., 2005). Adolescents above the CES-D cut-off score were 2.5 times more likely to be hospitalized than adolescents below the cut-off score (95% CI 1.12 – 5.96, p = .03). Among adolescents with cystic fibrosis, Snell and colleagues compared both hospitalization frequency and costs within a cystic fibrosis center between those with and without a depression diagnosis throughout a one-year period after depression had been diagnosed (Snell et al., 2014). Those with depression were hospitalized at a three times higher rate (4 mean annual hospitalizations vs. 1.2 mean annual hospitalizations, p < .01) with hospital costs that were four times higher ($280,000 vs. $60,116, p < .01) than those without depression.

One study examined the relationship between depressive symptoms and loneliness among adolescents with congenital heart disease (Luyckx et al., 2014). Loneliness was examined using the UCLA Loneliness Scale (range 1 – 5) and depressive symptoms were measured using the CES-D. Increases in CES-D scores predicted increased feelings of loneliness (r = .19, p < .001) from baseline to 9 months as well as 9 to 18 months.

Quality of Life

Four studies examined outcomes related to QOL (Guo et al., 2015; Hanns et al., 2018; Hoff et al., 2006; Luyckx et al., 2014). Two studies measured general satisfaction with QOL (Guo et al., 2015; Luyckx et al., 2014). Guo and colleagues administered the Global Satisfaction with QOL scale (range 6 – 24), which asks about satisfaction with sleep, energy, life, health status, friends and happiness, to a cohort of Chinese adolescents with type 1 diabetes. A one-unit increase in DSRS scores predicted a 0.17 decrease in QOL satisfaction scores at 12 months (p < .01) (Guo et al., 2015). A study of adolescents with congenital heart disease demonstrated similar results (Luyckx et al., 2014). In this study, a Linear Analogue Scale was used to measure both satisfaction in quality of life (0 being worst imaginable quality of life to 100 being best imaginable quality of life) and perceived health status (0 being worst imaginable health to 100 being best imaginable health). From baseline to 9 months, increases in CES-D scores predicted decreases in both QOL satisfaction scores (r = −.12, p < .01) and perceived health status (r = − .08, p < .05). These results found from baseline to 9 months were confirmed using the same measures from 9 to 18 months.

Discussion

Our findings from this integrative review demonstrate that depressive symptoms/depression can have long-term effects on disease control, self-management behaviors, illness-related morbidity, and QOL. As we deemed most studies fair to good quality, we can be reasonably confident in these findings. The majority of included studies followed adolescents with diabetes, demonstrating that depressive symptoms negatively influence glycemic control, blood glucose monitoring frequency, medication adherence, satisfaction in QOL, and odds of hospitalization. Two studies (Guo et al., 2015; Van Buren et al., 2018) found no relationship between depressive symptoms and glycemic control, which was contradictory to the other three studies (Baucom et al., 2018; Helgeson et al., 2009; Hood et al., 2011) examining glycemic control among adolescents with diabetes. One of these studies (Van Buren et al., 2018) focused on adolescents with type 2 diabetes, while the remaining focused on type 1 diabetes. The other study (Guo et al., 2015) followed a sample of Chinese adolescents, reflecting potential cultural differences given that depressive symptoms significantly decreased QOL among this population despite having no association with glycemic control. Depression is heavily stigmatized and not expected to compromise daily activities or self-care within Chinese communities (Hsu et al., 2008; Yen et al., 2000; Yeung et al., 2004). Understanding an adolescent’s respective culture and lifestyle preferences can help healthcare professionals identify appropriate mental health resources and treatments that suit the individual needs of an adolescent.

Longitudinal studies examining the effect of comorbid depressive symptoms on other illnesses (i.e., cystic fibrosis, congenital heart disease, sickle cell disease, and juvenile idiopathic arthritis) are limited in number, but present similar implications in that depressive symptoms/depression can decrease satisfaction in QOL and increase disability/pain, hospitalization, and health care costs. However, findings were inconsistent between different illness types. Depressive symptoms demonstrated a long-term effect on pain and disability among adolescents with juvenile idiopathic arthritis, but not among adolescents with sickle cell disease (Hoff et al., 2006). This may be attributed to differences in pathology, as pain in juvenile idiopathic arthritis is acutely caused by inflammation whereas pain in sickle cell disease is more persistent due to vascular occlusion. Likewise, disease control (i.e., FEV₁) and medication adherence among adolescents with cystic fibrosis did not appear to differ between those with and without depression (Snell et al., 2014). However, samples across these studies were relatively small and the number of longitudinal studies conducted within these respective illness populations is scarce. More robust and comprehensive research is needed to confirm the generalizability of these results across these particular illnesses.

Overall, there is a paucity of longitudinal research on long-term illnesses other than type 1 diabetes. Comorbid depressive symptoms/depression is a pertinent problem regardless of the chronic illness that is affecting an adolescent. The prevalence of depressive symptoms and/or depression among adolescents with varying illnesses, including epilepsy and asthma, has been repeatedly highlighted within the literature (Kline-Simon et al., 2016; Pinquart & Shen, 2011). While a number of cross-sectional studies have examined illness-related correlates of depressive symptoms/depression among adolescents with asthma and epilepsy (Delmas et al., 2011; Kwong et al., 2016; McCauley et al., 2007; Salcedo & Rios, 2009), we did not identify any longitudinal studies conducted among either of these subpopulations. This is surprising, given that asthma and epilepsy are among the most common chronic conditions in adolescents (Jin et al., 2017). Future research is needed to elucidate how depressive symptoms can be alleviated across adolescents with varying types of chronic illnesses.

To date, only two studies have examined the effects of comorbid depressive symptoms/depression on healthcare utilization and healthcare costs over time (Snell et al., 2014; Stewart et al., 2005). However, there is extensive cross-sectional evidence demonstrating that illness exacerbations (e.g., sickle cell crises, recurrent diabetic ketoacidosis, asthma exacerbation) are more frequently reported in adolescents with depressive symptoms/depression (Benton et al., 2007; Lawrence et al., 2006; Richardson et al., 2006). As these events can lead to avoidable hospitalizations and increased heath care costs, more research is needed to explore these relationships over time. Similarly, there is a lack of longitudinal studies examining the effect of depressive symptoms/depression on self-management behaviors for illnesses other than diabetes. This warrants greater attention, given that poor self-management is a key driver of symptom exacerbations and hospitalizations in adolescents with various chronic illnesses (Eakin & Riekert, 2013; Engelkes et al., 2015; McGrady & Hommel, 2013).

In this review, few studies examined relationships between more than two types of outcomes (i.e., disease control, self-management, morbidity, and quality of life). However, it is likely that depressive symptoms have a rippling effect across several types of outcomes. One study found that depressive symptoms affect BGM frequency, which subsequently leads to worsened metabolic control (Hood et al., 2011). This exemplifies that depressive symptoms can have both a direct and indirect impact on multiple aspects of illness-related care, control, morbidity, or quality of life. Hence, more research is needed to examine the moderating and mediating relationships between depressive symptoms/depression and chronic illness-related care and outcomes.

Current findings suggest clinically relevant relationships between depressive symptoms and the progression of a chronic illness. For example, depressive symptoms have a stronger effect on long-term disability/pain when disability/pain at baseline is lower (Hoff et al., 2006). Alongside the indications that disease severity and depressive symptoms fluctuate in relation to one another, this supports literature urging more frequent depression screenings among adolescents with chronic illness (Corathers et al., 2013; Guilfoyle et al., 2015; Luyckx et al., 2016), especially before the condition of an adolescent begins to decline. Helgeson and colleagues also found that the effect of depressive symptoms on metabolic control decreases as years pass (Helgeson et al., 2009). This may be due to various reasons, such as improved coping/self-management as a result of adolescents growing older, or increased utilization of mental health services. However, measurement of mental health service utilization was limited to only one study included in this review, and it is unclear whether adolescents across other studies were in treatment or therapy (Hood et al., 2011). Relatedly, all studies using self-report depressive symptom screening instruments identified a subset of adolescents exceeding clinical cut-off scores, supporting the strong recommendations of primary care guidelines for annual screening of depression in adolescents (Zuckerbrot et al., 2018). However, it is important to acknowledge the communicative barriers that create challenges in determining a diagnosis among this population. For example, adolescents often worry about the stigma around depression and are less willing to disclose emotions or symptoms (Gulliver et al., 2010). In addition, adolescents are often wary about confidentiality and trust in respect to their potential sources of help. While guidelines also recommend primary care training for depression screening, it is important to ensure that such strategies are centered on fostering comfortable patient-provider communication around sensitive topics such as depression.

Finally, this review shows that depressive symptoms/depression can have negative effects on disease-related morbidity in adolescents, which is consistent with literature focusing on adults with chronic illness (Voinov et al., 2013). However, there may be more specific ways in which depressive symptoms/depression impact adolescents with chronic illness differently than adults with chronic illness. While there are studies that examine quality of life satisfaction and adjust their analysis using peer support measures, current literature lacks exploration on the effects that depressive symptoms/depression may have on developmental issues that are particular to adolescents with chronic illness. Chronic illness has been shown to impede upon important developmental activities, such as forming peer relationships and identity formation (Luyckx et al., 2008; Santos et al., 2016). Thus, it is plausible that depressive symptoms may either contribute to or exacerbate these issues, given their relationship with disease control, morbidity, and self-management behaviors. More research is needed to investigate the multifaceted relationships between depressive symptoms/depression, chronic illness, and developmental factors to increase understanding of how care can be better tailored towards the needs of the adolescent population.

Limitations

We acknowledge that there are limitations to this review. All but one study examined depressive symptoms using self-report instruments instead of determining a diagnosis of depression. It is possible that adolescents living with diagnosed depression may exhibit different outcomes, but we were not able to discern the effects of having a diagnosis of depression versus clinically elevated levels of depressive symptoms. Although the literature search was conducted systematically and comprehensively across three databases, the exclusion of non-English language articles also makes it is possible that relevant studies were missed. In addition, we excluded conference proceedings, which may have highlighted additional key findings of new and upcoming longitudinal research. For the scope of this review, we chose to focus on a selective set of common illnesses that can be fully managed through ongoing care regimens, but we acknowledge that there are other types of conditions that should also be studied in future work. Finally, our suggestions are generalized across adolescents with chronic illness, but with the limited diversity in illness types, we were not able to comprehensively identify or explain inconsistencies between adolescents with differing chronic illnesses. Despite these limitations, this review is among the first to synthesize the longitudinal effects of depressive symptoms across varying chronic illnesses.

Conclusions

In this review, we synthesized literature examining ways in which depressive symptoms/depression can affect various measures of disease control, self-management behaviors, morbidity, and quality of life among adolescents with chronic illness over time. To date, longitudinal research remains limited, both in the types of outcomes assessed and among different chronic illness populations. Further investigation is needed to build upon the findings in this review and develop evidence on the underlying mechanisms by which depressive symptoms/depression may affect this vulnerable population. Current findings support the need to develop strategies for long-term monitoring and treatment of depressive symptoms alongside of chronic illness-related care. In addition, we encourage healthcare providers to consider how bio-psychosocial or behavioral factors can influence the effect of depressive symptoms long-term. Finally, more research is needed to investigate the effects of depressive symptoms/depression on factors that are developmentally unique to adolescents.