Adapting evidence-based interventions (EBIs) guided by implementation science frameworks is a promising way to bridge the gap between research and practice and advance effective sexual violence prevention in schools.
Keywords: Diabetes, Depression, Screening, Patient Health Questionnaire (PHQ), Pediatrics
Abstract
Psychosocial guidelines recommend routine screening of depressive symptoms in adolescents and young adults (AYA) with diabetes. Best practices for screening in routine care and patient characteristics associated with depressive symptoms require further investigation. The purpose of this study was to examine psychometric properties of the Patient Health Questionnaire (PHQ-2 and PHQ-9); document rates of depressive symptoms and related clinical actions; and evaluate associations with patient characteristics. The Patient Health Questionnaire (PHQ-2 or PHQ-9) was administered at five pediatric academic medical centers with 2,138 youth with type 1 diabetes. Screening was part of routine clinical care; retrospective data from electronic health records were collected for the first screening date as well as 12 months prior. The PHQ demonstrated good psychometric properties. Evaluation of item-level PHQ-9 data identified 5.0% of AYA with at least moderate depressive symptoms who would not have been flagged for further screening using the PHQ-2 only. On the PHQ-9, 10.0% of AYA with type 1 diabetes endorsed elevated depressive symptoms and 7.0% endorsed thoughts of self-harm. Patients with moderate or greater depressive symptoms had a 43.9% documented referral rate for mental health treatment. Higher BMI, older age, public insurance, shorter diabetes duration, higher HbA1C, and a diabetic ketoacidosis (DKA) event in the past year were associated with depressive symptoms. The PHQ-9 identified AYA with elevated depressive symptoms that would not have been identified using the PHQ-2. Depressive symptoms were associated with negative diabetes indicators. To improve referral rates, standardized methods for provision and documentation of referrals are needed.
Implications.
Practice: In the United States, screening for depression in pediatric diabetes care should utilize a consistent measure, cut points, and focus on improvement of referrals and documentation of clinical actions for those in need.
Policy: Policy makers and hospital administrators who want to support youth with depressive symptoms should provide electronic health records with the ability to readily document and track clinical actions associated with screening.
Research: Additional multisite studies are needed. In addition, studies should use prospective designs and interventional research informed by implementation science to generate generalizable and sustainable recommendations.
Adolescents and young adults (AYA) with type 1 diabetes are at increased risk for experiencing symptoms of depression compared to the general population [1]. A recent meta-analysis found that up to 30% of youth with diabetes report elevated depressive symptoms on self-report questionnaires [2]. Depressive symptoms in youth with diabetes are relatively stable over time and are associated with negative health outcomes, suggesting that early identification and intervention are necessary [3, 4]. The American Diabetes Association (ADA) and the International Society of Pediatric and Adolescent Diabetes (ISPAD) recommend routine screening for depressive symptoms in clinical care [5, 6]. It is recommended that screening be conducted at least annually using an age-appropriate, valid self-report measure [3]. Further, systems must be in place to ensure reliable recommendations and resources for treatment and follow-up [7].
A number of academic health systems have implemented depression screening as part of clinical care for youth with diabetes, reporting that 7%–14% of AYA with type 1 diabetes endorse elevated depressive symptoms and 3%–7% of AYA endorse suicidal ideation [8–11]. Various depression screening measures have been used with youth with diabetes, including the Children’s Depression Inventory (CDI), the Patient Health Questionnaire-2 (PHQ-2), and the Patient Health Questionnaire-9 (PHQ-9) [4, 8–14]. These efforts have focused largely on quality improvement and have detailed the processes and procedures necessary to initiate routine screening for depressive symptoms within clinical care. Results suggest that routine depression screening with youth with diabetes is feasible for health care systems, valued by health care providers and patients, and efficacious in identifying youth in need of mental health resources and support [8, 9]. However, these reports also shed light on current gaps in what is known about best practices for depression screening in routine care. For example, there has not been consistency in the measures used for depression screening [8, 9, 13], and some programs utilize a tiered screening system in which all patients complete the brief PHQ-2 and only those with elevated scores complete the PHQ-9 [11]. While the PHQ-2 is a valid measure, further evaluation of sensitivity and use in a type 1 diabetes population is needed [15]. Further, clinical actions following routine depression screening and how these actions are documented require additional investigation. Evaluation of depression screening outcomes across a variety of academic health settings representing racial, ethnic, and geographic diversity can inform best practices and translate targeted quality improvement findings into broad clinical care.
Increased understanding of associations among clinical and demographic characteristics and depressive symptoms can inform how to best identify patients at risk for elevated depressive symptoms, but these relationships have not been sufficiently explored within clinic-based populations. For example, some research suggests that depressive symptoms are more prevalent in females and youth with lower reported family income [2, 10, 11]. Other studies in youth with diabetes have not found significant differences in depressive symptoms by sex, and findings have been inconclusive on differences in depressive symptoms by age, race, and disease duration [3, 4, 10]. Research also suggests that elevated depressive symptoms are associated with increased risk for diabetic ketoacidosis (DKA), emergency department visits, and poorer diabetes self-management [3, 4, 11, 16, 17]. Therefore, information gleaned from the electronic health record (EHR) across multiple clinic-based samples can provide a comprehensive look at clinical characteristics in a large population of youth with diabetes.
The current study advances the literature by reporting on real-world experiences with and results of depression screening using the PHQ-2 or the PHQ-9 implemented in five pediatric diabetes clinics across the United States. In each clinic, screening for depressive symptoms was conducted as part of routine clinical diabetes care with youth with type 1 diabetes, aged 12–24 years. The objectives of the current study were to: (i) assess psychometric properties of the PHQ-2 and PHQ-9 in youth with type 1 diabetes; (ii) compare rates of elevated depressive symptoms in youth with type 1 diabetes between the PHQ-2 and PHQ-9; (iii) document depression treatment rates and referrals associated with elevated scores; and (iv) evaluate associations among elevated depressive symptoms and relevant clinical and demographic factors.
RESEARCH DESIGN AND METHODS
Participating adolescents and young adults
Inclusion criteria were diagnosis of type 1 diabetes for at least 6 months, 12–24 years of age, completion of one version of the PHQ during routine clinical care (i.e., PHQ-2 or PHQ-9), EHR documentation of an HbA1C value in the previous 3 months as a marker of overall glycemic control, and no more than 8 medical clinic visits in the prior 12 months to represent typical experiences with diabetes care. Screening data were included for clinic visits that occurred between April 1, 2016 and March 31, 2017 for all sites. Exclusion criteria were a diagnosis of type 2 diabetes, MODY, or unknown diabetes diagnosis at the time of EHR data extraction.
Data collection sites
Six pediatric diabetes clinics provided data from EHRs for the study period: (a) Vanderbilt University, Nashville, TN, (b) Cincinnati Children’s Hospital, Cincinnati, OH, (c) Barbara Davis Diabetes Center, Aurora, CO, (d) University of Florida, Gainesville, FL, (e) Stanford University, Palo Alto, CA, and (f) Children’s National Medical Center, Washington, DC.
Patient Health Questionnaire
The PHQ is a validated self-report measure of depressive symptoms and is commonly used in clinical settings. The PHQ asks respondents to report the extent to which symptoms have been bothersome during the previous 2 weeks using four response options ranging from “Not at All” (0) to “Nearly Every Day” (3). The PHQ-9 is comprised of nine items that reflect diagnostic criteria for major depressive disorder. The total score categories are minimal (0–4), mild (5–9), moderate (10–14), moderately severe (15–19), and severe (20–27). One item assesses thoughts of suicidal ideation and/or harm-to-self (hereafter, “harm-to-self”). The PHQ-9 has been validated with adolescents in primary care [18–20] and has documented concurrent validity with the Diagnostic Interview Schedule for Children (DISC-IV) [19, 20]. A score of 10 on the PHQ-9 (indicating moderate depressive symptoms) or greater has demonstrated adequate discrimination between adults diagnosed with and without major depressive disorder (AUC = 0.95) [21], and has established sensitivity to change in adults [22]. The PHQ also is available in a brief two-item screener [23]. The two-item version of the PHQ is used to indicate the need for administration of the additional seven items of the PHQ-9. PHQ-2 thresholds of ≥3 and ≥2 have been recommended for administration of the full PHQ when screening adults [21, 24].
Screening and referral procedures
As expected with a retrospective study utilizing EHR data derived from real-world clinical practices, PHQ procedures and documentation varied by site. Two sites administered the PHQ-2 followed by the PHQ-9 if warranted, and three sites administered the PHQ-9 only. Of the two sites using the PHQ-2, one site used ≥3 and one site used ≥2 as the thresholds for administration of the full PHQ-9 [11, 20]. Item-level PHQ data were available from three sites. All sites used the same threshold on the PHQ-9 to identify moderate depressive symptoms (PHQ-9 score 10–19) and severe depressive symptoms (PHQ-9 score ≥ 20) [19].
The frequency of PHQ administration varied by site, ranging from administration every diabetes clinic visit (two sites) to once every 6 months (one site) to once every 12 months (two sites). For the purposes of this study, the AYA’s first PHQ completion during the study period was used and additional administrations of the PHQ for individual patients over the 12-month period were not analyzed. PHQ measures were administered via completion in REDCap via iPad (two sites), using MyChart through the Epic electronic medical record system (one site), or paper/pencil only (two sites). At three sites, the primary medical provider (physician or nurse practitioner) was responsible for communicating the PHQ results to the patient and family at the time of screening and assessing for risk of self-harm. One site utilized both the medical team and the behavioral team to communicate screening results and assess for risk of self-harm, while the final site primarily utilized psychologists or psychology trainees to review the screening results and assess for risk of self-harm.
Follow-up actions based on PHQ scores were documented in the medical record at all sites by diabetes team members based on the level of consultation needed. The team members involved in this documentation included the primary medical provider (MD or NP), diabetes educator, or behavioral health specialist (social worker, psychologist, or psychology trainee). Four sites utilized handouts with mental health resources and referrals for patients with elevated PHQ scores (i.e., PHQ-9 score ≥ 10); one site provided in-person consultation with the psychology team for all patients with elevated PHQ-9 scores. All sites worked with behavioral health teams comprised of psychologists and/or social workers, and in-person consultation was conducted as available with AYA endorsing thoughts of self-harm. Patients with an immediate risk for self-harm were taken to the emergency department for further evaluation as needed.
Electronic health record data
Data were retrospectively collected from EHRs for the 12 months before completion of the first PHQ clinic screen. Data were screened for accuracy and anonymized at each site, and then were sent to the data coordinating site, cleaned, and aggregated.
Demographic and clinical data
The demographic and clinical data closest in time before the depression screen were used if multiple data options existed, along with other clinical data from the prior 12 months. The following data were extracted from EHRs: age; sex; race; insurance status; diabetes duration; most recent HbA1C and date; height/weight for body mass index (BMI) calculation; continuous glucose monitoring (CGM) use; insulin pump use; diabetes oral medications; mean number of blood glucose (BG) checks per day; mean BG level; % hyperglycemia, % hypoglycemia, and % time in range calculated from meter glucose data in EHR; diabetes ketoacidosis (DKA) episodes; hospitalizations for diabetes; number of clinic visits; and clinical interventions taken after depression screening.
Mental health diagnoses
Depression, anxiety, and externalizing behavioral disorders (e.g., conduct disorder, hyperactivity) were coded if noted by an International Classification of Diseases (ICD) code in the EHR.
Mental health treatment
Prior mental health treatment was coded if: (i) a caregiver indicated that the AYA had obtained psychological or behavioral treatment or medication for depressive symptoms; or (ii) the EHR included medications used to treat depression, anxiety, or behavioral symptoms such as hyperactivity (e.g., SSRIs, SNRIs, central alpha agonists, methylphenidate).
Clinical actions taken after PHQ screening
All sites identified clinical actions taken after a screening resulted in a moderate or higher (≥10) total score or any endorsement of the harm-to-self item. The following interventions were coded as: (i) no need for response; (ii) provision of referral information for mental health treatment; or (iii) immediate transfer to the emergency department.
Blood glucose meter data
The definition of hyperglycemia was ≥180 mg/dL and for hypoglycemia ≤70 mg/dL. The time in range consisted of values between hyperglycemia and hypoglycemia (71–179 mg/dL). Time frames associated with meter data were limited to 7–90 days.
Statistical approach
Confirmatory factor analyses (CFA) were conducted to verify the unidimensional structure of the PHQ-9. Goodness of fit was based on empirically supported indices [25]: root mean square error of approximation (RMSEA) values <0.05, comparative fit index (CFI) and Tucker–Lewis Index (TLI) values >0.95 and standardized root mean square residual (SRMR) values <0.08. Analyses used Mplus version 8.2 and its weighted least squares with mean and variance adjustment estimator and theta parameterization [26]. Cronbach’s alpha was calculated to assess reliability.
Generalized linear regression models (GLM) with maximum likelihood estimation were conducted in Stata v15, accounting for the clustering of individuals within study sites, to examine predictors of PHQ-9 scores and endorsement of the harm-to-self item which was dichotomized and coded for those who responded a 0—“Not at all” and coded as a 1 (positive response/endorsed) for those who responded 1, 2, or 3 on this item. A linear model was used to predict total PHQ-9 scores and a logistic model was used to predict harm-to-self item endorsement. Predictors included both demographic and clinical variables.
RESULTS
Sample characteristics
Table 1 contains descriptive statistics for demographic, clinical, and health care utilization variables (Total n = 2,138). The mean time difference between the most recent HbA1C and depression screening administration was 0.83 days (SD = 7.11, range = 0–90). The mean number of days of BG meter data available was 40.40 (SD = 31.34, range = 7–90). In the previous year, 11.5% individuals had at least one DKA event.
Table 1.
Sample characteristics of adolescents and young adults with type 1 diabetes
| Characteristic | Mean (SD) | Median | Range | N |
|---|---|---|---|---|
| Age at screening (years) | 16.45 (2.38) | 16 | 12.01–23.90 | 2,138 |
| Sex (% male) | 49.44 | – | – | 2,138 |
| Race (% white) | 70.14 | – | – | 2,120 |
| Insurance (% private) | 67.32 | – | – | 2,102 |
| Diabetes duration (years) | 6.98 (50.69) | 4.22 | 0.5–20.47 | 2,128 |
| HbA1c (%) HbA1c (mmol/mol) | 9.29 (2.12) 78 | 8.80 73 | 5.1–16.2 32–154 | 2,120 |
| BMI (kg/m2) | 24.02 (4.98) | 23.24 | 14.07–69.5 | 1,388 |
| CGM use (%) | 18.01 | – | – | 2,017 |
| Pump use (%) | 40.06 | – | – | 1,949 |
| Average BG level (mg/dL) | 218.14 (65.29) | 208 | 0–601 | 1,816 |
| Hyperglycemia (%) | 61.74 (21.61) | 64 | 0–100 | 1,174 |
| Time in range (%) | 31.62 (19.06) | 30 | 0–100 | 1,184 |
| Hypoglycemia (%) | 6.32 (7.82) | 4 | 0–100 | 1,184 |
| DKA events (n/year) | 0.16 (0.51) | 0 | 0–4 | 2,138 |
| Diabetes hospitalizations (n/year) | 0.21 (0.61) | 0 | 0–6 | 2,137 |
| Clinic visits (n/year) | 2.98 (1.35) | 3 | 0–8 | 2,138 |
| BG checks per day | 3.19 (2.10) | 3 | 0–14 | 1,823 |
| PHQ-9 only total | 3.85 (4.68) | 2 | 0–27 | 1,901 |
PHQ-2 and PHQ-9 scale analyses
Cronbach’s alpha for the PHQ-2 was 0.68 and for the PHQ-9 was 0.85. Confirmatory factor analysis for the PHQ-9 indicated that a unidimensional model was a good fit to the data, χ 2 (27) = 250.34, p < .001, RMSEA = 0.07 (90% confidence interval = 0.06, 0.08), CFI = 0.97, TLI = 0.96, SRMR = 0.04.
PHQ-2 and PHQ-9 score comparison
One goal of this study was to identify potential differences between the PHQ-2 and PHQ-9. We examined the scores of the first two items (the PHQ-2 items) for the AYA who were administered the PHQ-9 and had item-level data. Sixty-one AYA (5.0%) who would have been categorized as “minimal” by the PHQ-2 (scores of less than 3) scored in the “moderate” or “moderately severe” range on the PHQ-9. Based on these results, only AYA who were administered the full PHQ-9 were included in regression models. The majority of AYA in the study (88.6%) were administered the PHQ-9.
PHQ-9 scores
Table 2 shows distribution of the PHQ-9 scores and rates of endorsement to the harm-to-self item by score category. Aggregating the moderate and above risk categories (i.e., total score ≥ 10), 10.0% (n = 214/2,138) of the sample endorsed elevated depressive symptoms. The PHQ item related to harm-to-self was endorsed in 7.0% (n = 80/1,141 participants with item-level data).
Table 2.
Portion of PHQ scores and endorsement of harm-to-self in each category
| PHQ Scores | Minimal (%) | Mild (%) | Moderate (%) | Moderately severe (%) | Severe (%) |
|---|---|---|---|---|---|
| Overall rates | 72.6 | 17.3 | 6.4 | 2.5 | 1.2 |
| Harm-to-selfa | 0.5 | 10.1 | 26.1 | 57.7 | 93.8 |
aBased on n = 1,141 patients with item-level data available.
Diagnosis, referral and treatment rates
Depression diagnoses were documented in the EHR for 5.6% (n = 120/2,137) of AYA with type 1 diabetes. Related mental health diagnoses for individuals with type 1 diabetes were as follows: 3.7% mood disorders (not depression), 4.0% anxiety disorders, and 2.8% externalizing behavioral diagnoses. Of those individuals with a prior diagnosis of depressive disorder, 76.1% (n = 89/120) had indication of treatment within the prior 12 months (hereafter, “in treatment”). Of those individuals endorsing the harm-to-self item, 31.3% (n = 25/80) had indication of mental health treatment within the prior 12 months. However, 68.8% (55/80) of those individuals endorsing the harm-to-self item did not have indication of mental health treatment. In addition, 65.4% (n = 36/55) of those who expressed harm-to-self without indication of mental health treatment in the EHR also did not have a documented referral for future treatment as a result of the screening.
Table 3 shows rates of referrals and treatment status within each PHQ-9 score category. Because all sites used a cutoff total score of ≥10 for referral, the moderate, moderately severe, and severe PHQ-9 score categories were combined for further examination. For this higher score range, 43.9% (n = 94/214) received a referral and 56.1% (n = 120/214) did not have a documented referral for mental health treatment resulting from screening . Of those referred, 35.1% (n = 33/94) had indication of treatment in the previous year, whereas 64.9% (n = 61/94) had no indication of treatment. Thus, within this higher score range, 56.7% (n = 148/261) did not have a documented referral for treatment. Of those who were not referred, 10.0% (n = 12/120) were already in treatment and 90.0% (n = 108/120) were not in treatment. As indicated by Table 3 rates for previous depression treatment increased with severity of depressive symptoms (see “total in treatment” category).
Table 3.
Proportions of adolescents and young adults with type 1 diabetes by PHQ score category, referral status, and treatment status (Tx)
| PHQ score category (score range) Category % (n) | Referral | No referral | Total in treatment | ||
|---|---|---|---|---|---|
| Minimal (0–4) 72.60% (1,536) | Referral 4.20% (65) | No referral 95.80% (1,471) | 2.79% (43) | ||
| In Tx 10.77% (7) | No Tx 89.23% (58) | In Tx 2.45% (36) | No Tx 97.55% (1,434) | ||
| Mild (5–9) 17.30% (367) | Referral 11.40% (42) | No referral 88.60% (325) | 7.90% (29) | ||
| In Tx 19.05% (8) | No Tx 80.95% (34) | In Tx 6.46% (21) | No Tx 93.54% (304) | ||
| Moderate (10–14) 6.40% (136) | Referral 33.10% (45) | No referral 66.90% (91) | 15.44% (21) | ||
| In Tx 35.56% (16) | No Tx 64.44% (29) | In Tx 5.49% (5) | No Tx 94.51% (86) | ||
| Moderately severe (15–19) 2.50% (52) | Referral 59.60% (31) | No referral 40.40% (21) | 25.00% (13) | ||
| In Tx 25.81% (8) | No Tx 74.19% (23) | In Tx 23.81% (5) | No Tx 76.19% (16) | ||
| Severe (≥20) 1.20% (26) | Referral 69.20% (18) | No referral 30.80% (8) | 42.31% (11) | ||
| In Tx 50.00% (9) | No Tx 50.00% (9) | In Tx 25.00% (2) | No Tx 75.00% (6) | ||
Regression models
Table 4 shows results from generalized linear regression analyses with AYA with type 1 diabetes. Model 1 results show regression coefficients using PHQ-9 scores as the dependent variable. Higher BMI, greater age at screening, public insurance, shorter diabetes duration, higher HbA1C, and having at least one DKA event in the previous 12 months were related to higher PHQ-9 scores. Model 2 results show odds ratios using endorsement of the PHQ-9 harm-to-self item (no/yes) as the dependent variable. Higher BMI, lower age at screening, female sex, private insurance, greater % hypoglycemia, and having at least one DKA event were associated with increased likelihood of endorsing the harm-to-self item.
Table 4.
Regression analyses for adolescents and young adults with type 1 diabetes predicting PHQ-9 scores and endorsement of harm-to-self (n = 1,901)
| PHQ-9 scores | Coefficient | Robust SE | p Value | [95% CI] |
|---|---|---|---|---|
| BMI | 0.101 | 0.029 | < .001 | [0.046, 0.156] |
| Age at screening | 0.013 | 0.003 | < .001 | [0.008, 0.018] |
| Sex | 0.182 | 0.935 | .846 | [−1.651, 2.014] |
| Race (white) | 0.119 | 0.502 | .812 | [−0.856, 1.104] |
| Insurance | −0.680 | 0.266 | .011 | [−1.229, −0.146] |
| Diabetes duration | −0.004 | 0.001 | <.001 | [−0.005, −0.002] |
| HbA1c | 0.179 | 0.061 | .003 | [0.060, 0.299] |
| Average BG level | 0.004 | 0.002 | .059 | [0.002, 0.009] |
| % Hyperglycemia | 0.004 | 0.029 | .884 | [−0.054, 0.063] |
| % in range | −0.005 | 0.019 | .786 | [−0.042, 0.032] |
| % hypoglycemia | 0.001 | 0.008 | .615 | [−0.011, 0.019] |
| DKA events | 1.230 | 0.217 | < .001 | [0.808, 1.662] |
| No. of BG checks per day | 0.007 | 0.028 | .887 | [−0.095, 0.111] |
| Harm-to-self endorsement | Odds ratio | Robust SE | p Value | [95% CI] |
| BMI | 1.022 | 0.005 | <.001 | [1.011, 1.031] |
| Age at screening | 0.999 | 0.001 | <.001 | [0.998, 0.999] |
| Sex | 0.510 | 0.139 | .013 | [0.299, 0.870] |
| Race (white) | 1.192 | 0.623 | .737 | [0.428, 3.318] |
| Insurance | 1.266 | 0.144 | .039 | [1.012, 1.584] |
| Diabetes duration | 1.000 | 0.001 | .729 | [0.998, 1.002] |
| HbA1c | 1.111 | 0.129 | .368 | [0.884, 1.397] |
| Average BG level | 0.998 | 0.002 | .186 | [0.994, 1.001] |
| % hyperglycemia | 1.005 | 0.004 | .239 | [0.997, 1.012] |
| % in range | 1.010 | 0.007 | .154 | [0.996, 1.024] |
| % hypoglycemia | 1.011 | 0.005 | .043 | [1.000, 1.022] |
| DKA events | 2.148 | 0.478 | .001 | [1.389, 3.322] |
| No. of BG checks per day | 0.921 | 0.005 | .429 | [0.752, 1.128] |
Conclusions
This is the first multisite study of pediatric depression screening using the PHQ. The study uniquely integrated retrospective EHR data from five geographically diverse pediatric diabetes clinics to evaluate depressive symptoms in AYA presenting for routine diabetes care. Findings inform measurement of depressive symptoms in AYA with diabetes and offer implications for screening practices, including identifying subpopulations at greater risk for depressive symptoms and highlighting the need to improve referral rates for mental health services and related documentation in the EHR.
A broad range of levels of depressive symptoms have been identified in AYA with type 1 diabetes due to differences in assessment instruments, study age ranges, single-site samples, and varying cut points. In this present study of AYA presenting for routine diabetes care over a 1-year period, 10% reported moderate or greater symptoms on the depression screening measure. Direct comparison of symptom rates across studies is challenging. Nonetheless, symptom rates were lower compared with research samples of AYA with type 1 diabetes using measures such as the CDI (13–21%) [8, 10, 27–29] or the CES-D (25%) [17], but slightly higher compared with results from a large multi-site study using the CES-D (6.6%) [30]. This variability in prevalence of depression screening scores heightens the need to utilize a common measure for clinical screening of depression in this population. The PHQ-9 is increasingly used in primary care and patient-reported outcomes research and may provide the best option for diabetes care.
Demographic and clinical variables were related to higher PHQ-9 scores and endorsement of the harm-to-self item in AYA with type 1 diabetes in the present study. Prior research has been inconclusive on differences in depressive symptoms in youth with type 1 diabetes by sex, age, race, and disease duration [3, 4, 10]. The present findings underscore the complexity of the associations between demographic characteristics and depression risk. In the current sample, older age, shorter disease duration, public insurance, and higher BMI were associated with elevated depressive symptoms, offering new insights into risk factors for depression in a diverse, clinic-based population. AYA sex was not associated with depressive symptoms, suggesting that males and females with type 1 diabetes demonstrate similar risk for depression. In contract, female sex, private insurance, and younger age were associated with increased risk for endorsement of harm-to-self, along with higher BMI. Higher BMI has consistently emerged as a key risk factor associated with depressive symptoms and suicidal ideation, and the current study extends this literature in a population of youth with type 1 diabetes [31]. As found in prior studies, elevated depressive symptoms and endorsement of harm-to-self also were associated with more DKA events and elevated depressive symptoms were also associated with worse glycemic control [8, 17, 32, 33]. Given the variability of demographic characteristics associated with depressive symptoms and harm-to-self, results support enacting screening in a broader population of youth with diabetes rather than targeting specific subgroups of youth based on demographic characteristics.
The portion of AYA already in treatment for depression appropriately increased with PHQ scores, indicating there was enhanced access to mental health services for those who potentially needed them. Five percent of AYA with type 1 diabetes in this study had an already documented depression diagnosis in the EHR, and a portion of the sample was reportedly accessing care for depression even prior to the depression screening completion in the diabetes clinic. Importantly, although having been in treatment in the previous year, in-clinic screening identified one-third of these AYA with type 1 diabetes who were in prior mental health treatment still scored in the moderate or above range. This finding underscores the importance of routine screening for depressive symptoms in diabetes care even for AYA who had been previously screened and/or treated for depression outside the diabetes clinic.
Referrals for mental health services also increased with PHQ scores in the present study. However, a significant portion of AYAs with scores in the moderate or severe PHQ categories did not have a documented referral for subsequent mental health treatment; further, the majority of those individuals who were not referred had no indication of prior mental health treatment in the EHR. This gap represents a critical opportunity for improvement in depression screening and follow-up procedures in routine diabetes care. A necessary component of the screening process is the provision of appropriate support to any patients who endorse elevated depressive symptoms. It is essential that diabetes providers are equipped to provide mental health resources and referrals to patients with identified mental health needs. Additionally, the use of standardized documentation or templates for mental health history and interventions taken as a result of the depression screening scores can improve documentation in the EHR. For example, the use of simple input formats such as check-boxes to document common or recommended actions and automatic triggers to place a mental health referral for elevated scores could improve the submission and tracking of these referrals.
The rates of endorsement of harm-to-self found here are comparable to two previous studies using the CDI [8, 13]. In the PHQ-9 scoring system, the item related to harm-to-self is weighted equally with other items. In practice, all five sites had procedures in place for further evaluation of endorsement of harm-to-self regardless of the PHQ-9 total score. Yet, two-thirds of the AYA who endorsed any level of harm-to-self were not in mental health treatment, and the majority of those who were not in treatment did not have a documented referral in the EHR for such treatment. Results highlight the importance of collecting data on harm-to-self and using that information in addition to an overall score to guide clinical interventions. It is recommended that the harm-to-self item score is highlighted for providers clearly in assessment results, with well-established plans for specific clinical interventions, including transfer to the emergency room, if needed and readily accessible referral sources for future mental health treatment.
One strength of this study is inclusion of both the PHQ-2 and PHQ-9. We found evidence for the reliability of both versions and the PHQ-9 reflected a single construct in confirmatory factor analyses. Through examination of the PHQ-2 items for those individuals who completed the PHQ-9, we found that a small portion of individuals taking the brief two-item version could have been misclassified as minimal or mild when they were moderate on the PHQ-9. Sites in this study implementing the PHQ-2 used a cut score of 2 or 3 [24, 34]. Measurement choice may drive some of the differences in rates of elevated depressive symptoms across published samples. For example, the current study found 10.0% of AYA with type 1 diabetes endorsed a PHQ-9 score ≥10, indicating at least moderate depressive symptoms. Yet a recent article detailing the depression screening process at a large pediatric diabetes center utilized the PHQ-2 and found only 7.6% of their sample of AYA endorsed PHQ-2 elevations. Further, only 6.7% of this elevated subsample then endorsed a PHQ-9 total score ≥5, which also is lower than the clinical cutoff of PHQ-9 total score ≥10 used in the current study [11]. Given the current study results and the lack of sensitivity and specificity data for applications of the PHQ-2 versus PHQ-9 in AYA with diabetes, the use of the PHQ-9 is the most cautious option. In addition, it will be important to determine what total cutoff score (i.e., 5 vs. 10) on the PHQ-9 is the most appropriate to identify risk early and proactively refer AYA to mental health treatment.
Given challenges with data from real-world settings, this study also had several limitations. Data associated with the PHQ were derived from EHR data within the previous 12 months. This time frame improved feasibility, but may have resulted in under-estimations of those who were previously identified as high-risk or in mental health treatment. While the PHQ total score was readily obtained, referrals for mental health treatment were often documented in free text in the EHR in a clinic note and this documentation was not consistently completed. A recent study of four clinics within the same health care system found 10% of consultations with social workers due to elevated depressive symptoms were not documented in the EHR [11]. Monitoring these systematic errors on a routine basis is necessary for diabetes clinics as a part of constantly improving depression screening procedures over time. Finally, we were not able to document overall diabetes clinic depression screening rates, as clinical informatics systems did not track those eligible versus screened in the context of missed appointments.
Results from this study inform future directions for screening in pediatric diabetes practices. Depressive symptoms are only one indicator of psychosocial functioning, and psychosocial guidelines advocate for the assessment of general (e.g., depression, anxiety, disordered eating) and diabetes-specific (e.g., diabetes distress) domains [5, 6]. Further, it can be challenging to disentangle symptoms of depression from symptoms of diabetes distress. Approximately 10% of AYA with type 1 diabetes endorsed moderate depressive symptoms in the current study. Yet research suggests more than 33% of AYAs experience significant diabetes distress, and diabetes-specific distress may be a more specific predictor of glycemic control than depressive symptoms [35]. Future research could implement tailored psychosocial screening based on clinical and demographic indicators of risk (e.g., shorter disease duration, higher HbA1C) and expand screening to include both general and diabetes-specific psychosocial domains associated with diabetes-related outcomes [15].
Despite advances in screening, there remain broad challenges in ensuring access to appropriate mental health care once risk is identified. All sites had integrated mental health professionals, but multiple challenges faced providers once elevated risk for depression was identified: (i) mental health resources were not readily available for treatment or were not easily identified, (ii) diabetes providers are not typically trained in health communications surrounding depression, and (iii) many mental health professionals are not trained to support persons with diabetes. Given the rates of AYA with depressive symptoms in the present sample and potential gaps found in clinical actions, advancing efforts to address these challenges is paramount to improving health and wellbeing for youth with diabetes.
Acknowledgments
The authors would like to thank Douglas Conway, Cindy Lybarger, and the Vanderbilt Institute for Clinical and Translational Research (VICTR) for pilot funding and data coordination (Vanderbilt); Cheyenne Reynolds and Savannah Summy for data extraction and collection (University of Florida). Research reported in this publication was supported by the National Center For Advancing Translational Sciences of the National Institutes of Health under Award Numbers UL1TR002243 (Vanderbilt University) and UL1TR001427 (University of Florida). Content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Compliance with Ethical Standards
Conflict of Interest: All authors declare that they have no conflicts of interest.
Authors’ contributions: S.A.M., M.M., S.M. and J.C.K. conceived the study; S.A.M. obtained funding for and carried out data harmonization and cleaning; all authors planned analyses and carried out EHR data extraction; C.A.M. performed statistical analyses; all authors contributed to interpreting results, writing, and editing the manuscript.
Ethical Approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Informed Consent: For this type of study, formal consent is not required.
Welfare of Animals: This article does not contain any studies with animals performed by any of the authors.
References
- 1. Hedden S, Blau G, Lipari R, Rubenstein L, Forman-Hoffman V, Glasheen C. Comparison of Physical Health Conditions Among Adolescents Aged 12 to 17 With and Without Major Depressive Episode. Bethesda, MD: Substance Abuse and Mental Health Services Agency; 2017. [Google Scholar]
- 2. Buchberger B, Huppertz H, Krabbe L, Lux B, Mattivi JT, Siafarikas A. Symptoms of depression and anxiety in youth with type 1 diabetes: a systematic review and meta-analysis. Psychoneuroendocrinology. 2016;70:70–84. [DOI] [PubMed] [Google Scholar]
- 3. Hood KK, Beavers DP, Yi-Frazier J, et al. Psychosocial burden and glycemic control during the first 6 years of diabetes: results from the SEARCH for Diabetes in Youth study. J Adolesc Health. 2014;55(4):498–504. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. McGill DE, Volkening LK, Pober DM, Muir AB, Young-Hyman DL, Laffel LM. Depressive symptoms at critical times in youth with type 1 diabetes: following type 1 diabetes diagnosis and insulin pump initiation. J Adolesc Health. 2018;62(2):219–225. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Delamater AM, de Wit M, McDarby V, et al. ISPAD clinical practice consensus guidelines 2018: psychological care of children and adolescents with type 1 diabetes. Pediatr Diabetes. 2018;19(suppl 27):237–249. [DOI] [PubMed] [Google Scholar]
- 6. Young-Hyman D, de Groot M, Hill-Briggs F, Gonzalez JS, Hood K, Peyrot M. Psychosocial care for people with diabetes: a position statement of the American Diabetes Association. Diabetes Care. 2016;39(12):2126–2140. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Kichler JC, Harris MA, Weissberg-Benchell J. Contemporary roles of the pediatric psychologist in diabetes care. Curr Diabetes Rev. 2015;11(4):210–221. [DOI] [PubMed] [Google Scholar]
- 8. Corathers SD, Kichler J, Jones NH, et al. Improving depression screening for adolescents with type 1 diabetes. Pediatrics. 2013;132(5):e1395–e1402. [DOI] [PubMed] [Google Scholar]
- 9. Iturralde E, Adams RN, Barley RC, et al. Implementation of depression screening and global health assessment in pediatric subspecialty clinics. J Adolesc Health. 2017;61(5):591–598. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Silverstein J, Cheng P, Ruedy KJ, et al. ; Pediatric Diabetes Consortium . Depressive symptoms in youth with type 1 or type 2 diabetes: results of the pediatric diabetes consortium screening assessment of depression in diabetes study. Diabetes Care. 2015;38(12):2341–2343. [DOI] [PubMed] [Google Scholar]
- 11. Marker A, Patton SR, McDonough RJ, Feingold H, Simon L, Clements MA. Implementing clinic-wide depression screening for pediatric diabetes: an initiative to improve healthcare processes. Pediatr Diabetes. 2019;20(7):1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Hood KK. The influence of caregiver depressive symptoms on proxy report of youth depressive symptoms: a test of the depression-distortion hypothesis in pediatric type 1 diabetes. J Pediatr Psychol. 2009;34(3):294–303. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Matlock KA, Yayah Jones NH, Corathers SD, Kichler JC. Clinical and psychosocial factors associated with suicidal ideation in adolescents with type 1 diabetes. J Adolesc Health. 2017;61(4):471–477. [DOI] [PubMed] [Google Scholar]
- 14. Corathers S, Mara CA, Chundi PK, Kichler JC. Depression screening of adolescents with diabetes: 5-years of implementation and outcomes. J Am Acad Child Adolesc Psychiatry. 2019;58(6):628–632. [DOI] [PubMed] [Google Scholar]
- 15. Hilliard ME, De Wit M, Wasserman RM, et al. Screening and support for emotional burdens of youth with type 1 diabetes: strategies for diabetes care providers. Pediatr Diabetes. 2018;19(3):534–543. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Lawrence JM, Standiford DA, Loots B, et al. ; SEARCH for Diabetes in Youth Study . Prevalence and correlates of depressed mood among youth with diabetes: the SEARCH for Diabetes in Youth study. Pediatrics. 2006;117(4):1348–1358. [DOI] [PubMed] [Google Scholar]
- 17. Baucom KJW, Turner SL, Tracy EL, Berg CA, Wiebe DJ. Depressive symptoms and diabetes management from late adolescence to emerging adulthood. Health Psychol. 2018;37(8):716–724. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Johnson JG, Harris ES, Spitzer RL, Williams JB. The Patient Health Questionnaire for adolescents: validation of an instrument for the assessment of mental disorders among adolescent primary care patients. J Adolesc Health. 2002;30(3):196–204. [DOI] [PubMed] [Google Scholar]
- 19. Richardson LP, McCauley E, Grossman DC, et al. Evaluation of the Patient Health Questionnaire-9 item for detecting major depression among adolescents. Pediatrics. 2010;126(6):1117–1123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Richardson LP, Rockhill C, Russo JE, et al. Evaluation of the PHQ-2 as a brief screen for detecting major depression among adolescents. Pediatrics. 2010;125(5):e1097–e1103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Kroenke K, Spitzer RL, Williams JBW. The PHQ‐9. J Gen Intern Med. 2001;16(9):606–613. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Löwe B, Schenkel I, Carney-Doebbeling C, Göbel C. Responsiveness of the PHQ-9 to psychopharmacological depression treatment. Psychosomatics. 2006;47(1):62–67. [DOI] [PubMed] [Google Scholar]
- 23. Kroenke K, Spitzer RL, Williams JB. The Patient Health Questionnaire-2: validity of a two-item depression screener. Med Care. 2003;41(11):1284–1292. [DOI] [PubMed] [Google Scholar]
- 24. Carey M, Boyes A, Noble N, Waller A, Inder K. Validation of the PHQ-2 against the PHQ-9 for detecting depression in a large sample of Australian general practice patients. Aust J Prim Health. 2016;22(3):262–266. [DOI] [PubMed] [Google Scholar]
- 25. Hu LT, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equ Modeling. 1999;6(1):1–55. [Google Scholar]
- 26. Muthen LK, Muthén BO, Muthén M. Mplus version 8 user’s guide. Los Angeles, CA: Muthen & Muthen; 2017. [Google Scholar]
- 27. Herzer M, Hood KK. Anxiety symptoms in adolescents with type 1 diabetes: association with blood glucose monitoring and glycemic control. J Pediatr Psychol. 2010;35(4):415–425. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Hood KK, Huestis S, Maher A, Butler D, Volkening L, Laffel LM. Depressive symptoms in children and adolescents with type 1 diabetes: association with diabetes-specific characteristics. Diabetes Care. 2006;29(6):1389–1391. [DOI] [PubMed] [Google Scholar]
- 29. McGrady ME, Hood KK. Depressive symptoms in adolescents with type 1 diabetes: associations with longitudinal outcomes. Diabetes Res Clin Pract. 2010;88(3):e35–e37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Hood KK, Lawrence JM, Anderson A, et al. ; SEARCH for Diabetes in Youth Study Group . Metabolic and inflammatory links to depression in youth with diabetes. Diabetes Care. 2012;35(12):2443–2446. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Quek YH, Tam WWS, Zhang MWB, Ho RCM. Exploring the association between childhood and adolescent obesity and depression: a meta-analysis. Obes Rev. 2017;18(7):742–754. [DOI] [PubMed] [Google Scholar]
- 32. Kongkaew C, Jampachaisri K, Chaturongkul CA, Scholfield CN. Depression and adherence to treatment in diabetic children and adolescents: a systematic review and meta-analysis of observational studies. Eur J Pediatr. 2014;173(2):203–212. [DOI] [PubMed] [Google Scholar]
- 33. Baucom KJ, Queen TL, Wiebe DJ, et al. Depressive symptoms, daily stress, and adherence in late adolescents with type 1 diabetes. Health Psychol. 2015;34(5):522–530. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Arroll B, Goodyear-Smith F, Crengle S, et al. Validation of PHQ-2 and PHQ-9 to screen for major depression in the primary care population. Ann Fam Med. 2010;8(4):348–353. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Hagger V, Hendrieckx C, Sturt J, Skinner TC, Speight J. Diabetes distress among adolescents with type 1 diabetes: a systematic review. Curr Diab Rep. 2016;16(1):9. [DOI] [PubMed] [Google Scholar]