Follow-Up Mental Health Care in Youth and Young Adults With Type 1 Diabetes After Positive Depression Screen and/or Suicidal Ideation

Depression screening has been clinically performed annually in patients ≥10 years of age since January 2016 at the Barbara Davis Center for Diabetes via the PHQ-9 (29), a standardized screen for depressive symptoms and SI. A retrospective chart review was conducted of patients at the center who scored positive for depressive symptoms and/or endorsed SI between January 2016 and July 2018. The University of Colorado’s Multiple Institutional Review Board approved this study.

The PHQ-9 is a validated self-administered depression screen that directly assesses the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, symptoms of depression (29,30). Respondents rate whether various symptoms have occurred in the previous 2 weeks, from 0 (not at all) to 3 (nearly every day). The severity level of each depressive symptom is categorized on a numerical scale as minimal (0–4), mild (5–9), moderate (5,10–13), moderately severe (14–18), or severe (19–26). A PHQ-9 score ≥10 has 88% specificity and 88% sensitivity to detect major depression (29). The PHQ-9 includes one SI item, which asks whether the individual has had “thoughts that you would be better off dead or of hurting yourself in some way.” Any answer other than 0 (not at all) on this question suggests positivity for SI.

Eligibility for screening was determined on a weekly basis for patients with scheduled appointments the following week. Eligibility criteria included individuals who are ≥10 years of age and English-speaking who had not had a depression screen in the previous year and were not developmentally delayed, resulting in the inability to understand the screening questions. Screening was only provided to patients who spoke and read English and who had a caregiver who spoke English if the patient was <18 years of age, as the clinic did not have a Spanish-speaking professional on the suicide risk assessment team.

For those eligible, the PHQ-9 was given to patients at their clinic visit by a patient navigator before they met with their provider. The PHQ-9 was scored by the patient navigator, and providers were notified if a patient scored positive for depressive symptoms and/or SI. When patients scored positive for SI, a member of the suicide risk assessment team was also notified.

Figure 1 provides details regarding the clinical flow for depression screening. Because of staffing constraints, clinic flow, and time limitations, not every eligible patient on a clinic day was able to have a depression screening conducted. For example, if a provider entered the room before the patient navigator could administer the PHQ-9, the patient was followed to receive a depression screening at a future diabetes appointment.

If the PHQ-9 score was clinically elevated (score ≥10) without SI, the provider reviewed depressive symptoms with the individual and provided a referral list of mental health providers in the community if the individual was not already actively seeing a mental health specialist. The mental health referral list included social workers, counselors, and psychologists in different cities and counties throughout the state. The referral list also contained information explaining how patients should contact their insurance company to find out which mental health providers in their area were covered by their insurance, as well as information about which providers offered payment via a sliding scale. If a mental health provider was known to have additional training in chronic illnesses such as diabetes, this was noted on the referral list as well.

If an individual endorsed SI on the PHQ-9 (with or without additional positive depressive symptoms), the provider would discuss the results with the individual, and a member of the suicide risk assessment team (comprising three social workers, a psychologist, and a physician) would also meet with the individual before the patient left clinic and conduct a suicide risk assessment. The mental health referral list was then provided after the suicide risk assessment if the individual was not already actively seeing a mental health specialist.

The suicide risk assessment team assessed individuals’ current risk of suicide using a standardized suicide risk assessment protocol recommended by Joiner et al. (34) and Chu et al. (35). Specifically, the risk assessment included questions pertaining to desire and ideations, plans and preparation, and other risk factors, which have been described elsewhere (8). Based on the information obtained from the risk assessment, patients were classified as being at either low, moderate, severe, or extreme risk for suicide, based on the decision tree described in Chu et al. (35). Recommendations provided were based on risk level and included creating a safety plan, providing a suicide hotline number, and providing the mental health referral list.

Individuals were categorized based on whether they had obtained any type of mental health follow-up (therapy and/or starting on psychiatric medication) within 1 year after the initial positive depression screen. Data on mental health follow-up were obtained via review of each participant’s chart, which included all encounters, medications administered, and notes at hospitals and clinics in Colorado that link or share with the Barbara Davis Center/Children’s Hospital Colorado electronic medical record (EMR) system. The following search terms were used: “psychiatric,” “psychiatry,” “psychology,” “psychologist,” “psychiatrist,” “mental health,” “behavioral health,” “counselor,” “counseling,” “therapy,” “therapist,” “depression” (which also populated any antidepressant medications), “suicide,” “suicidal,” “ideation,” and “psychotic” (which also populated any antipsychotic medications). Any medication lists, encounters, free-written text, or diagnoses in the EMR that contained any of these search terms were individually reviewed and recorded. Data on prior mental health care use was obtained by using the same search term methodology above and reviewing the visit notes from the day of screening. Those who had active (at the time of screening) or previous mental health treatment were combined in a group designated as having had previous mental health treatment.

Sex, age, race, ethnicity, and insurance status at the time of depression screening were collected from all individuals from the EMR. Diabetes-related characteristics, including diabetes duration, pump use, continuous glucose monitoring (CGM) use, and A1C at the time of screening, were also obtained from the EMR. All A1C measurements for visits between 13 months before and 13 months after the depression screening were also collected for each patient. One-year follow-up PHQ-9 scores were collected for those who had been rescreened by study end date (n = 27).

Data analysis was performed using R, v. 4.0.4, statistical software (R Foundation for Statistical Computing, Vienna, Austria). Demographic and diabetes-related characteristics for those who obtained follow-up mental health care versus those who did not receive follow-up mental health care were compared using t tests for continuous variables or Pearson χ² tests or Fisher exact tests for categorical variables. Logistic regression was used to examine predictors of obtaining follow-up mental health care. Piecewise linear mixed models were used to compare changes in A1C after screening in those who obtained mental health care versus those who did not. All A1C values were plotted, and a line of best fit was created for both groups (mental health follow-up vs. no mental health follow-up). The slope of A1C change over time before and after the positive screen was calculated for each group.

Depression screening was conducted in 1,376 youth and young adults (mean age 15.2 ± 3.2 years; 52.8% male; 64.1% identified race as Caucasian; 10.1% identified ethnicity as Hispanic; 68.3% privately insured). Two hundred patients (14.5%) were identified as having a positive depressive symptom (PHQ-9 score ≥10) and/or SI. Ninety-nine (49.5%) had positive depressive symptoms without SI, 70 (35.0%) had positive depressive symptoms and SI, and 31 (15.5%) had SI without positive depressive symptoms. Of those identified as positive for depressive symptoms and/or SI, 106 (53%) obtained mental health follow-up. Follow-up consisted of therapy (n = 62 [58.5%]), starting psychiatric medications (n = 10 [9.4%]), or both therapy and medication (n = 34 [32.1%]).

Sixteen (15%) of those with mental health follow-up after screening had a history of having had mental health care before the screening. It was not clear from chart review whether or when mental health treatment had stopped before the depression screening.

Those who had mental health follow-up after screening were more often female (67.0%, P <0.001) and had a higher initial PHQ-9 score (13.7 vs. 11.9, P = 0.005) (Table 1). The presence of SI at screening was not associated with obtaining mental health follow-up care (P = 0.47).

There were no differences between mental health follow-up groups in type 1 diabetes duration, insulin pump use, CGM use, or average A1C at time of screening (Table 1). Of the 200 individuals who screened positive, 10 (5%) did not have any subsequent diabetes visit after their screening visit and no follow-up A1C. Among those who had mental health follow-up, A1C post-screening decreased by an average of 0.25% per 3-month period, and by 1% per year (P = 0.008) when adjusting for the presence of mental health treatment before screening. Among those with no mental health follow-up, A1C post-screening increased by an average of 0.08% per 3-month period and 0.34% per year when adjusting for previous mental health treatment before screening, which was significantly different from those with evidence of mental health follow-up (P = 0.03) (Figure 2).

In models testing predictive factors for obtaining mental health follow-up, males were 64.8% less likely to do so (P = 0.003). Those with no documented previous mental health treatment were also 66.9% less likely (P = 0.002) to obtain mental health follow-up after a positive screen. For each 1-unit increase in PHQ-9 score, those who screened positive were on average 8.0% more likely to obtain follow-up mental health care (P = 0.039). Individuals of Hispanic ethnicity were 61.8% less likely to obtain mental health follow-up after screening positive, although this difference did not reach statistical significance (P = 0.068) (Table 2).

Although depressive symptoms and SI are more common in the pediatric and young adult population with type 1 diabetes, only half of those who have a positive screen obtain follow-up mental health care. This finding is higher than that noted in the studies by Vassilopoulos et al. (31) and Shapira et al. (32) but similar to that in Watson et al. (33). The fact that our study accessed medical records that were outside of the diabetes clinic but integrated within or linked to the clinic’s EMR system may have contributed to the higher rate of mental health follow-up seen in our study and the one by Watson et al., who also accessed records outside of their clinic (33). Specifically, in those who had mental health follow-up, we found follow-up treatment via medication lists and encounters in 67% of the patients identified, and documentation of medication and/or therapy in clinic notes for the other 33%. The findings in all four studies indicate an important gap in care between individuals screening positive and then receiving follow-up mental health care.

This study also found that higher PHQ-9 scores were predictive of following up with mental health care after referral was provided in clinic, suggesting that those with greater depressive symptoms are more likely to seek follow-up care. Our study is consistent with others that have shown that males and those of Hispanic ethnicity are less likely to seek mental health care compared with females and individuals of non-Hispanic ethnicity (36,37). Studies have shown that there may be a greater stigma associated with seeking mental health care among males and individuals of Hispanic descent (38–40), although it is not clear whether that is the reason for decreased follow-up in this study. Further studies to understand the roles of sex and race/ethnicity in adherence to mental health referral would be beneficial to improve rates of participation. Additionally, developing care coordination resources would be beneficial to ensure that patients with depression and/or SI and type 1 diabetes access behavioral care in a timely manner.

In this study, those found to be positive for SI had similar rates of mental health follow-up as those without SI. There is no other known study in the literature comparing mental health follow-up in those with versus without SI and diabetes. Matlock et al. (9) discussed mental health follow-up after SI, with 60% (23 of 38) seeing a mental health provider for follow-up after endorsing SI, whereas the rate of follow-up in this study was 50% (51 of 101). It is possible that the increased interaction in clinic between the suicide risk assessment team and those who endorsed SI played a role in the rate of mental health follow-up found, although whether this interaction led to an increased or decreased rate of follow-up is unclear. Further study is needed to determine why approximately half of those with SI did not obtain mental health follow-up care and to investigate how to overcome barriers to seeking mental health care in this group.

Previous studies have found that, often, depressive symptoms are associated with higher A1C levels and decreased adherence over time (4,10). The improvement in A1C trend slope seen in our study (A1C decrease by an average of 1% per year) in the group that used mental health resources compared with those who did not have mental health follow-up (increase in A1C by 0.34% per year) may suggest benefits in diabetes management for those who obtain mental health care. Given that this occurred in an age-group typically associated with increasing A1C, the improvement seen in those who obtained mental health follow-up care may indicate the importance of mental health integration into diabetes care, particularly in youth and young adults. Alternatively, those who are motivated to seek mental health follow-up care may also be motivated to improve diabetes management. Greater longitudinal studies are needed to further understand the long-term effects of mental health care on subsequent type 1 diabetes–related outcomes.

Some steps are currently underway to improve access to care, although further study and steps are needed to increase participation in mental health care after referral. The American Diabetes Association and American Psychological Association have started a diabetes education program for mental health professionals who are interested in providing care to people with diabetes (41), which could increase access to care. The results of our study underscore the benefits of having integrated mental health professionals in diabetes clinics, which has been successful in other studies (42–44). Providing integrated mental health care within a diabetes center can decrease barriers for patients in accessing follow-up mental health care and normalize mental health care as part of routine diabetes care. In diabetes centers where behavioral health care specialists are not present, further investigation is needed to determine how best to ensure that follow-up mental health care is obtained.

Although chart review was conducted in every individual and the EMR system pulls in clinical information from a variety of clinics and hospitals throughout the state, it is possible that this study underreported mental health utilization if it was not documented in the EMR or captured in facilities or practices that do not use a connecting record system. Additionally, the frequency of mental health follow-up visits, compliance, and length of psychiatric medication use was not always known in this study. Because of the differences in treatment for various forms of depression, type and duration of therapy may be important to note in future studies and be aware of in clinical practice. Although the PHQ-9 is an effective screening tool for depressive symptoms, it does not always indicate diagnosis, as it may overlap with diabetes distress, anxiety, and even emotions related to significant life events. This shortcoming might cause an overestimation of depression.

Documentation of patient refusal to complete a depression screening did not start until April 2017, so the refusal rates during the study period are not complete. However, the current overall clinical depression screening refusal rate is ∼3%. Although not all patients eligible for screening on a given clinic day were screened, we followed up with the missed patients at a later diabetes appointment to complete screening.

Depression screening in the study was also done only in English-speaking patients, as there were no Spanish-speaking members of the suicide risk assessment team to perform follow-up when there was a positive screen. Therefore, the results may not be generalizable to non-English–speaking patients and families. The lack of a Spanish-speaking team member excludes many of the most vulnerable patients facing barriers to care resulting from language and health literacy barriers, who may also be at high risk for depression and SI. For study participants of Hispanic/Latino ethnicity, English fluency was not recorded and may have influenced communication about mental health support to these families. It is essential that future studies include a Spanish-speaking member of the suicidal risk assessment team or a certified translator to avoid excluding this population from screening and research.

Although insurance status was not different between the groups, we could not determine whether socioeconomic status or affordability of mental health care played a role in low rates of accessing follow-up mental health care. It is important to investigate health disparities in mental health care to ensure that equitable care is being provided. Identifying specific barriers to obtaining follow-up mental health care in all patients is important to determine how best to connect patients with mental health care services. This study did not formally assess barriers such as cost and affordability, transportation, resource availability in rural communities, or stigma and preexisting beliefs pertaining to mental health problems. This is an important next step to create a more patient-centered approach to mental health referrals in the type 1 diabetes clinic and improve follow-up care.

Overall, this study found that almost 50% of those who screen positive for depressive symptoms and/or SI lack mental health care follow-up, particularly among males. This study highlights the need for further investigation into barriers to obtaining mental health care services both within and outside of the diabetes clinical setting to provide more complete diabetes care.

The authors thank Emma Mason, University of Denver, for help with preparing the manuscript.

This work was supported by National Institutes of Health (NIH)/National Center for Advancing Translational Sciences Colorado Clinical and Translational Sciences Institute grant UL1 TR002535. The authors are solely responsible for the contents of this article, which do not necessarily represent official NIH views.

No potential conflicts of interest relevant to this article were reported.

K.R.S.W. and S.M. performed the research, wrote the manuscript, and reviewed/edited the manuscript. T.V. and L.P. conducted data analysis and reviewed/edited the manuscript. E.M.Y., E.F.-I., and J.T. performed the research and reviewed/edited the manuscript. J.K.R., J.S.-B., and A.S. reviewed/edited the manuscript. S.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.