Health Care Use Over 3 Years After Adolescent SBIRT

The study was conducted at Kaiser Permanente Northern California (KPNC), an integrated health care–delivery system serving a diverse population of >4 million members. The sample was drawn from a pragmatic, randomized clinical trial at KPNC’s Oakland Pediatrics Department of implementation and patient outcomes of 3 different modalities of delivering SBIRT for adolescents ages 12 to 18 years, including 2 brief-intervention arms: (1) a pediatrician-only arm consisting of pediatricians who were trained to screen and (when indicated) deliver brief interventions and refer to specialty substance use or mental health treatment; (2) an embedded behavioral clinician arm in which pediatricians administer the initial assessment and (if needed) refer patients to an embedded behavioral clinician for further assessment, brief intervention, and referral to treatment; and (3) a usual-care arm in which pediatricians had access to the EHR screening tools but no formal SBIRT training. SBIRT training of the pediatricians and behavioral clinicians in the original trial was similar and has been described previously in detail elsewhere.¹³ 

The sample consists of a subset of adolescents aged 12 to 18 years who endorsed either the substance use or mental health symptoms on the Teen Well Check Questionnaire between November 1, 2011, and October 31, 2013, or were deemed eligible for further assessment, brief intervention, and/or referral to treatment by their pediatricians (n = 1871). The index date was defined as the date on which the adolescent screened positive during this period. Additional details of the study protocol and the Teen Well Check Questionnaire are given elsewhere.^13,24,25 

To examine associations between comorbidity and health care use between patients who had access to brief interventions for problems (if needed) and those who did not, we created a dichotomous indicator (SBIRT group = 1; usual care = 0).

We obtained sex, age at index date, race and/or ethnicity (coded as white, African American, Hispanic, Asian American, and other), and length of enrollment in the health plan in the 3 years after the index date from the EHR.

The EHR was used to obtain all outpatient and inpatient (medical and psychiatric) service use for 1-year before and up to 3 years after the index date. Counts of visits made to the ED, primary care, substance use treatment, and psychiatric services and inpatient uses as well as a count of all outpatient visits were created for baseline and 1- and 3-year follow-up periods.

We examined 2 sets of comorbidities recorded in the EHR during each patient’s visits at baseline and over the 1- and 3-year postindex visits. These comorbidities included the prevalence of any behavioral health conditions (International Classification of Diseases, Ninth Revision [ICD-9] codes 290, 293–302, and 306–319; International Classification of Diseases, 10th Revision [ICD-10] codes F01–F09, and F20–F99) and any of the 7 most common or chronic medical conditions found in this age group, including asthma (ICD-9: 493; ICD-10: J45), arthritis (ICD-9: 710–719; ICD-10: M01, M02, M05–M08, and M11–M25), diabetes (ICD-9: 250; ICD-10: E08–E13), irritable bowel syndrome (ICD-9: 555, 560.89, and 5641; ICD-10: K50, K58, and K566), migraine (ICD-9: 346; ICD-10: G43), rhinitis (ICD-9: 477; ICD-10: J30 and J31), and sinusitis (ICD-9: 461 and 473; ICD-10: J01 and J32).²⁸ Two dichotomous indicators, 1 to denote the presence of any behavioral health condition and 1 for the presence of any of the medical conditions (any of the conditions was recorded = 1; otherwise = 0), were created for each time period. We also created indicators for the most common behavioral health conditions from the subset of behavioral health conditions noted above: depression (ICD-9: 296.2, 296.3, 296.82, 298.0, 300.4, 301.12, 309.0, 309.1, 309.28, and 311; ICD-10: F32, F33, F34.1, F43.21, and F43.23), anxiety (ICD-9: 300.00, 300.02, 300.09, 309.21, 309.24, 309.81, F41.1, F41.3, F41.8, F41.9, F43.22, and F43.1), attention-deficit/hyperactivity disorder (ICD-9: 314; ICD-10: F90), bipolar disorder (ICD-9: 296 [except 296.1 and 296.3], 301.13, and 301.11; ICD-10: F30, F31, F34.0, F34.8, F34.9, and F39), and substance use disorders (ICD-9: 291, 292, and 303–305; ICD-10: F10–F19).

We used χ² and Student’s t tests to examine differences in demographic characteristics (age, sex, and race and/or ethnicity) and medical and behavioral health comorbidities between the SBIRT and usual-care groups. Multivariate logistic regression models were used to examine the dichotomous outcomes (any mental health or substance use comorbidities and any common or chronic medical comorbidities), and negative binomial regression models were used to examine visit counts (number of ED, primary care, substance use treatment, psychiatry, and all outpatient visits). In these models, the exponent of the coefficient for the treatment variable represents the odds ratio (OR) and the incidence rate ratio (iRR) for logistic and negative binomial regressions, respectively, for the SBIRT group relative to the usual-care group. Multivariate use models adjusted for age, sex, race and/or ethnicity, and previous use of similar services (eg, ED visits in the 12-month preindex period were included in the model for postindex ED visits). Similarly, we adjusted for corresponding preexisting diagnoses in the models for postindex comorbidity (eg, previous-12-month mental health diagnoses in the model for mental health comorbidity). Length of membership was accounted for by creating a variable whose coefficient is constant. All analyses were performed by using SAS 9.3 (SAS Institute, Inc, Cary, NC).

The SBIRT group had significantly more girls (59.5% vs 48.1%; P ≤ .05) and multiracial adolescents and adolescents of color (77.9% vs 68.3%; P ≤ .05) than the usual-care group (Table 1); there was no difference in the mean age at index date (15.8 years [SD = 1.5] vs 15.9 years [SD = 1.5]; P > .10). More than 93% of the sample had continuous membership for the 1-year postindex period, and >80% were continuous members up to 3 years postindex. At 1 year postindex, the usual-care group had more continuous members than the brief-intervention group (95.6% vs 92.3%; P ≤ .05); there were no differences in membership length at 3 years (data not shown).

Although this was a randomized trial, we examined differences in service use and comorbidities between the SBIRT and usual-care groups in the year before the index date (Fig 1). We found no baseline or postindex differences in any of the health service measures (Table 2).

The prevalence of mental health and substance use diagnoses was similar between the groups at baseline through 1 year postindex (Table 2). Among those with a behavioral health diagnosis 1 year postindex (6.8% of adolescents), depression (43.7%) and anxiety (36.7%) were the most common; there were no differences between the treatment groups. The prevalence of any behavioral health conditions in the 3 years postindex was 28.7%; fewer adolescents in the SBIRT group had a depression diagnosis than those in usual care (45% vs 54%; P < .05). All other diagnoses did not differ significantly. There was no difference in substance use rates at baseline through 1 year postindex between the 2 groups; by 3 years postindex, the SBIRT group was less likely to have a substance use diagnosis (2.1% vs 3.7%; P < .05). Eighteen percent of all adolescents had at least 1 of the 7 common or chronic medical conditions at baseline; there was no difference between the treatment groups. At 1 year postindex, the SBIRT group was less likely to have any chronic medical condition (5.4% vs 7.8%; P < .05); this trend was also observed at 3 years postindex, although the difference was narrower (31.6% vs 32.6%; P < .10).

In the 1-year postindex period, there were only 6 inpatient stays and 6 adolescents with a substance use treatment visit, 5 of whom were in the SBIRT group; therefore, these outcomes could not be modeled. There was no significant difference in the likelihood of ED visits, but there was a declining trend in the number of ED visits in the postindex period (iRR = 0.87; P > .05). Adolescents in the SBIRT group were also likely to have 24% fewer psychiatry department visits (iRR = 0.76; P = .05). There were no differences in primary care or total outpatient visits (Table 3).

Three years postindex, adolescents in the SBIRT group were likely to have had more substance use treatment visits (iRR = 2.04; P < .01) and fewer psychiatry department visits (iRR = 0.65; P < .05) and total outpatient visits (iRR = 0.85; P < .05) over the 3 years compared with those in usual care (Table 4).

Less than 1% of all adolescents had a substance use diagnosis during the 1-year postindex period; therefore, this outcome could not be modeled. In multivariable models adjusted for patient characteristics and the presence of previous diagnoses, adolescents in the SBIRT group were less likely to have a mental health diagnosis (OR = 0.69; 95% confidence interval [CI] = 0.48–1.01; P = .05) or a common or chronic medical condition (OR = 0.66; 95% CI = 0.45–0.98) in the 1 year postindex compared with patients in the usual-care group (Table 5).

In multivariate models adjusted for patient characteristics and previous diagnoses, we found that adolescents in the SBIRT group were less likely to have a substance use diagnosis (OR = 0.64; 95% CI = 0.45–0.91) than those in usual care over the 3-year postindex period. There were no statistically significant differences in the prevalence of mental health diagnoses or common or chronic medical conditions between groups over the 3 years (Table 6).

This study examined the association between access to SBIRT and health care use and comorbidities for adolescents at risk for substance use and other behavioral health problems. As in adult studies,²⁹ we observed that adolescents with access to brief interventions for behavioral health risks had fewer ED visits in the subsequent 12 months compared with those in usual care. This finding echoes those of Tait et al,²³ who also found lower ED use associated with the receipt of a brief intervention for substance use. In that study, however, the adolescents presented in the ED for substance-related problems and represented a more severe population than this primary care sample. That we found a trend toward lower ED use even in this lower-severity population underscores the potential of offering SBIRT in primary care to reduce ED use among adolescents.

The SBIRT group also had lower psychiatric service use in both the 1- and 3-year follow-up periods. There were no differences in use or comorbidities between the SBIRT and usual-care groups in the year before participants’ index screening. Therefore, it appears that the lower ED- and psychiatric service–visit rates we found could be attributable to having access to SBIRT in primary care. This premise is supported by our finding that the SBIRT group had lower likelihood of medical and mental health comorbidity during the 1-year follow-up period relative to the usual-care group.

Over the longer 3-year period, the SBIRT group had lower use of psychiatric services and total outpatient services, a lower likelihood of a substance use diagnosis, and a slightly lower likelihood of a mental health diagnosis. The SBIRT group was more likely to use substance use treatment services than the usual-care group. This suggests that although fewer patients in this group over time develop substance use problems that are serious enough to warrant specialty substance use treatment, the referral-to-treatment component of SBIRT available to them might be operative in getting those who do to treatment.

We also observed that the prevalence of mental health, substance use, and medical conditions increased over time among the adolescents in both groups; almost one-third of all adolescents had a common or chronic medical condition in the 3 years postindex. Together with the findings on service use, this underscores the importance of screening for substance use problems and brief intervention or referral to treatment as necessary, occurring at every visit as for other health conditions, particularly in light of the developmental changes and environmental exposures experienced by adolescents. This also suggests the need for patients to be given brief interventions in a timely manner and, if needed, more than once so that their substance use and other psychiatric problems do not escalate, resulting in worsening morbidity and critical life events that lead to costly ED visits or hospital stays. Access to SBIRT, by helping to reduce substance use and initiation and/or emotional distress, may also bolster physical health through the encouragement of healthier behavior choices (eg, not smoking, not drinking, and stress reduction), reducing the prevalence of medical comorbidities and ultimately contributing to lower health care use.

Consistent with the adolescent mental health literature,²⁸ we found that depression and anxiety were the most common mental health diagnoses recorded in the EHR, emphasizing the need for early screening and timely intervention. Although we acknowledge that comorbidities impact service use, we did not control for these in our analyses of use because the prevalence of comorbidities over time are in and of themselves important outcomes of interest. Instead, we included corresponding health care use, which serves as both a proxy for comorbidity as well as accounting for regression to the mean.

This study has several limitations. It was conducted in an integrated health care system with an insured population and may not be generalizable to uninsured populations. This was a pragmatic trial that did not recruit patients but relied on EHR-based clinical information collected during regular care. This strengthened the study by allowing us to examine the population base of adolescents with pediatric visits in the clinic, increasing its generalizability to pediatric primary care populations. This was an intent-to-treat analysis in which outcomes were examined among all eligible patients in both groups, regardless of whether they received a brief intervention. However, post hoc analyses showed that the receipt of brief intervention was associated with significantly lower rates of ED services at 1 year, which speaks to the potential value of SBIRT in helping to avert problems necessitating emergency medical care.

This study is among the first to examine associations between access to brief intervention and health care service use. Among those in the SBIRT group, we found lower use of psychiatry services at both 1 and 3 years and lower overall outpatient use at 3 years. These lower use rates are particularly notable because they serve as important proxy indicators of health and well-being. Importantly, we also found that the adolescents in the SBIRT group were less likely to have a mental health or a medical diagnosis at 1 year and less likely to have a substance use or depression diagnosis at 3 years, suggesting that offering SBIRT in pediatric primary care may have an enduring impact on both health and health care use during this critical developmental period. Future research is needed on the effects of SBIRT for adolescents on these important outcomes.

We thank the KPNC Adolescent Chemical Dependency Coordinating Committee, the KPNC Adolescent Medicine Specialists Committee, and Thekla Brumder Ross, PsyD, Derek Satre, PhD, and Jennifer Mertens, PhD, for their guidance. We also thank David Bacchus, MD, and all the physicians, medical assistants, nurses, receptionists, managers, and especially the patients and parents of KPNC’s Oakland Pediatrics clinic for their participation in the activities related to this study. All contributing authors have been included above.

CI

confidence interval

ED

emergency department

EHR

electronic health record

ICD-9

International Classification of Diseases, Ninth Revision

ICD-10

International Classification of Diseases, 10th Revision

iRR

incidence rate ratio

KPNC

Kaiser Permanente Northern California

OR

odds ratio

SBIRT

screening, brief intervention, and referral to treatment