Impact of initiation timing of SSRI or SNRI on depressed adolescent healthcare utilization and costs

Abstract

Background:

Adolescents with newly diagnosed depression may not receive timely antidepressant therapy. Clinical and economic effects of early versus late treatment initiation are unclear.

Objective:

To compare effects of early versus late initiation of second-generation (SSRI/SNRI) antidepressants on emergency room (ER) visits, hospitalizations and healthcare costs in adolescents with depression.

Methods:

Patients aged 12–17 with a diagnosis of depression were identified in a claims database (1999–2007). Patients initiating antidepressants within 1 month of initial diagnosis were considered early initiators; patients initiating within 2–12 months were late initiators. Clinical resource use and healthcare costs were measured during the 6-month pre-index and 12-month post-index (study) periods and compared descriptively between groups. Logistic regression compared healthcare services utilization; a generalized linear model compared costs. All models were adjusted for baseline characteristics, including demographics, comorbidities, and healthcare services utilization.

Results:

A total of 7344 adolescents met study criteria. 4415 (60%) initiated antidepressant treatment within 1 month of diagnosis. At baseline, early initiators had more all-cause inpatient visits (14 vs. 7%) and all-cause ER visits than late initiators (25 vs. 21%, both p < 0.01). They had higher medical ($1434 vs. $1160) and total costs ($1565 vs. $1290) (both p < 0.01). In the study period, late initiators had higher risk of ER visits (OR = 1.13, p = 0.03). They incurred higher medical costs ($5415 vs. $4061) and higher total healthcare costs ($6001 vs. $4907), but lower adjusted drug costs ($767 vs. $888) (all p < 0.01).

Limitations:

Clinical data are scarce in the claims database, and the ability to observe disease severity and reasons for delayed treatment is limited. The definition of early and late initiation was based on empirical analysis, and no clear cutoff was identified beyond what was observed in the data.

Conclusions:

Adolescents who initiated SSRI/SNRI therapy earlier experienced lower risk of ER visits and had lower total costs compared to late initiators.

Key words::

• Adolescent
• Depression
• Early treatment
• Second-generation antidepressants
• SNRI
• SSRI

Introduction

Depressive disorders have a frequent incidence in adolescents, and impose a heavy burden on their development. For instance, the prevalence of major depressive disorder (MDD) is estimated to be approximately 3–5% in children and 4–8% in adolescents, with a cumulative incidence of up to 20% by the age of 181–3. The disorder can cause substantial setbacks in adolescent patients’ emotional and physical development and impede their social interactions and academic performance at a particularly crucial and fragile stage in life. Epidemiological evidence has shown that youths with MDD are at increased risk for recurrent MDD episodes and/or anxiety disorders into adulthood, and suicide4–6.

For many patients, depression is a chronic, recurrent illness. Research has shown that approximately 85% of patients with a history of MDD suffer a recurrent depressive episode within 15 years, and that longer duration of a depressive episode and a greater number of prior depressive episodes significantly predict a higher risk of a future episode7. Longitudinal studies estimated that depressed youths have a 70% probability of recurrence within 5 years of remission from a depressive episode3. Thus, clinicians are urged to treat a depressive episode with the goals of achieving symptom remission and minimizing young patients’ suffering and impairment3.

Psychotherapy or antidepressant medications have demonstrated comparable efficacy in treating adolescents with moderate-to-severe depression8–10. The controversies and regulatory warnings about a small but significant increase in the risk of suicidal thoughts and behaviors in children, adolescents, and young adults treated with selective serotonin reuptake inhibitors (SSRIs), as well as other classes of antidepressants including serotonin norepinephrine reuptake inhibitors (SNRIs), tricyclic antidepressants (TCAs), and monoamine oxidase inhibitors (MAOIs) have raised concerns about the use of these drugs. Consequently, warning labels have been recommended for antidepressants with respect to their use in adolescents11. However, untreated depression carries its own risk of suicidal behavior2,12,13, and SSRIs and SNRIs remain staples in treating adolescent depressive disorders because of their advantages in efficacy and tolerability over TCAs3,14–19. Only two of the SSRIs or SNRIs, escitalopram and fluoxetine, are currently approved by the FDA for acute and maintenance treatment of depression in pediatric and/or adolescent patients (ages 8–18 for fluoxetine, and 12–17 for escitalopram), but other SSRIs and SNRIs are also widely prescribed14.

Currently, there is very little published evidence to clarify the optimal timing of prescribing antidepressants in either adults or younger patients and whether early treatment after a diagnosis makes a difference in patient outcomes. Several small studies conducted in adult patients suggest that a prolonged duration of untreated depressive illness (time from the first MDD emergence to the first adequate antidepressant treatment) may be associated with worse outcomes, such as prolonged duration of the disorder, persistence of symptoms, and a higher number of recurrent episodes20–23. The implications of delaying pharmacotherapy in adolescents remain unknown, for both patients and healthcare providers.

The objective of this retrospective observational study was to compare, from a third-party payer’s perspective, the effect on indirect clinical indicators, such as the rates of hospitalization and emergency room visits, and the cost implications of early versus late initiation of treatment with SSRIs or SNRIs in adolescent depressive disorders.

Methods

Data source

An employer claims database, including 9 million insured individuals, was used for this study. The de-identified healthcare claims spanned from January 1999 through December 2007 and contained claims records from 40 self-insured Fortune 500 companies located across the United States. Medical claims data include type of service, diagnosis, current procedural terminology (CPT) code, and payments to providers. Prescription drug data include national drug code (NDC), days of supply, dose strength, and payments.

Study design

A retrospective claims data analysis was conducted to assess the impact of the timing of prescribing SSRIs and SNRIs on the healthcare service utilization and costs in adolescents diagnosed with depression.

The study sample was designed to include adolescent patients who demonstrated signs of depression severe enough to warrant treatment with antidepressants. Included patients were required to have a diagnosis of depressive disorder as identified by any of the International Classification of Diseases, 9th Revision (ICD-9) codes of 296.2, 296.3, 298.0, 300.4, 309.1, and 311, and who received at least one prescription of SSRI or SNRI in the 12 months after the diagnosis. The date on which the patient received the first diagnosis of a depressive disorder was defined as the index date. The 6 months prior to the index date were defined as the baseline period; the 12 months after the index date were defined as the study period. Patients were 12–17 years of age on the index date and were continuously eligible during the baseline and study periods. Patients who were enrolled in health maintenance organization (HMO) plans were excluded from this analysis, as these organizations tend to be more restrictive than self-insured or preferred provider plans, and may restrict the treatment options available to patients directly after diagnosis. Patients were also required to have received no prescription for any antidepressant or antipsychotic drug during the baseline period. Moreover, patients whose first antidepressant use after the index date was not an SSRI or SNRI were excluded, because older antidepressants (i.e., monoamine oxidase inhibitors, tricyclic antidepressants) are considered to lack evidence of clinical efficacy in adolescent populations, according to clinical guidelines24.

Included study subjects were grouped into an early and a late initiator cohort according to the timing of the first SSRI or SNRI prescribed after the index diagnosis of a depressive disorder: early initiators were patients initiating antidepressants within 1 month of the diagnosis while late initiators were patients initiating at least 1 month after but no more than 12 months after the diagnosis. This cut-off between early and late initiators was chosen empirically as a preliminary review of the data indicated that most patients who were initiated on antidepressants clustered within the first month of diagnosis.

Outcomes were analyzed over the 6-month baseline period and the 12-month study period and included indirect clinical indicators that could be identified in a healthcare claims database and healthcare costs estimated from a third-party payer’s perspective. The indirect clinical indicators included rates of hospitalizations (inpatient care), outpatient hospital visits, and emergency room (ER) visits. Healthcare costs included prescription drug cost, medical cost, and total healthcare cost. Medical cost was divided into subcategories of inpatient cost, outpatient hospital cost, ER cost, and cost of other professional services. All costs were inflation-adjusted to 2008 US dollars. In addition, mental health-related hospitalization (i.e., inpatient visits), ER visits, and other visits and healthcare costs were analyzed, and were identified by medical claims associated with ICD-9 codes for depression, schizophrenia, other psychotic disorders, dementia, generalized anxiety disorder, phobia, panic disorder, obsessive-compulsive disorders, post-traumatic stress disorder, alcohol and drug abuse, sleeping disorder and eating disorder. Finally, depression-related outcomes were identified by ICD-9 codes for depression.

Statistical methods

Baseline characteristics, including demographics, comorbidities, healthcare services utilization, and healthcare costs in the 6-month baseline period, were compared descriptively between the early and late initiator cohorts. Healthcare services utilization rates and comorbidity rates were compared using chi-square tests. Continuous variables, such as healthcare costs, age and Charlson Comorbidity Index (CCI) were compared using two-sided Wilcoxon rank sum tests.

All study period outcomes were compared descriptively between the early and late initiators. The healthcare services utilization and major healthcare cost categories (prescription drug, medical, and total healthcare cost) in the 12-month study period were further analyzed using multivariate analysis. Logistic regression was used for healthcare services utilization, and Generalized Linear Models (GLM) with gamma distribution and log link were used to estimate healthcare costs. Regression analyses controlled for insurance type (PPO vs. other), gender, age, CCI, index year, comorbidities, asthma, injuries, MDD diagnosis, and the rates of inpatient visits, outpatient visits, ER visits, and other professional services visits as measured during the 6-month baseline period.

Results

Baseline characteristics

A total of 7344 adolescent patients met the study’s inclusion and exclusion criteria and were included in the analyses. Table 1 summarizes sample counts by selection criteria. Among these patients, 4415 (60%) were included in the early initiator cohort versus 2929 patients in the late initiator cohort. Among late initiators, most patients initiated during month 2 (36.5%), and over 70% had initiated by month 5.

Table 1.  Sample selection for adolescent depression patients with early vs. late treatment initiation.

Criteria	Total number of patients
All patients with at least one diagnosis of general depression*	642 902
All patients between 12 and 17 years old on the date of first depression diagnosis (index date)	44245
All patients continuously enrolled during the 6 months prior to the index date (baseline period)	31441
All patients continuously enrolled during the 12 months following the index date (study period)	24445
All patients with no antidepressant use during the baseline period	20100
All patients having at least one prescription of second-generation antidepressants within the study period	7344
	Early initiator cohort	Late initiator cohort
All patients^†	4415	2929

*Having at least 1 claim with an ICD-9 code of: 296.2, 296.3, 298.0, 300.4, 309.1, or 311.

^†Early initiators begin treatment with a second-generation antidepressant in the first month after the index date. Late initiators begin treatment with a second-generation antidepressant within 2–12 months of the index date.

Table 2 summarizes the baseline characteristics of both groups. The study patients had a mean age of 15 years, and 64% of the early initiators and 66% of the late initiators were female. The comorbidity profiles of the two cohorts presented some differences. Early initiators were more likely to have a diagnosis of panic disorder (relative risk [RR] = 2.08, p < 0.05), post-traumatic stress disorder (RR = 1.64, p = 0.051), and drug abuse (RR = 1.41, p < 0.05). They were also less likely to have a diagnosis of eating disorder (RR = 0.69, p < 0.05). During the 6-month baseline period, early initiators were more likely to have all-cause inpatient visits (RR = 2.22, p < 0.01) and all-cause ER visits (RR = 1.17, p < 0.01). Additionally, they had higher rates of depression and mental health-related rates of inpatient visits, outpatient hospital visits, and ER visits (all p < 0.05). Finally, early initiators had higher total medical cost during the baseline period, in particular, higher depression and mental health-related costs, while prescription drug costs were similar between the two cohorts. Overall, early initiators appeared to be more severely ill than late initiators.

Table 2.  Baseline characteristics for adolescent depression patients with early vs. late treatment initiation.

Baseline characteristics*	Early initiators		Late initiators		Difference^†	p-value^‡
	[A]		[B]		[A]–[B] or [A]/[B]	[A] vs. [B]
Sample (n)	4415		2929
Demographic characteristics
Age on index date (mean, SD)	15.32	(1.48)	15.23	(1.51)	0.09	0.0195^§
Female (n, %)	2840	64%	1946	66%	0.97	0.0630
Insurance type (n, %)
PPO	2487	56.3%	1624	55.5%	1.02
Other	1928	43.7%	1305	44.6%	0.98	0.4540
Charlson Comorbidity Index (mean, SD)	0.08	(0.35)	0.09	(0.36)	–0.01	0.9824
Mental Comorbidity Profile (n, %)
Generalized anxiety disorder	139	3.2%	79	2.7%	1.17	0.2650
Phobia	40	0.9%	29	1.0%	0.92	0.7150
Panic disorder	35	0.8%	11	0.4%	2.08	0.0260^§
Obsessive-compulsive disorder	45	1.0%	21	0.7%	1.42	0.1790
Post-traumatic stress disorder	52	1.2%	21	0.7%	1.64	0.0510
Alcohol abuse	59	1.3%	27	0.9%	1.46	0.1060
Drug abuse	149	3.4%	70	2.4%	1.41	0.0150^§
Eating disorder	63	1.4%	61	2.1%	0.69	0.0330^§
Bipolar	55	1.3%	30	1.0%	1.23	0.3850
Healthcare service utilization (n, %)
Inpatient visit	633	14.3%	189	6.5%	2.22	<0.0001^§
Depression-related^¶	543	12.3%	124	4.2%	2.91	<0.0001^§
MHD-related**	565	12.8%	134	4.6%	2.80	<0.0001^§
Outpatient hospital visit	1507	34.1%	951	32.5%	1.05	0.1390
Depression-related	365	8.3%	208	7.1%	1.16	0.0680
MHD-related	426	9.7%	241	8.2%	1.17	0.0380^§
ER visit	1104	25.0%	627	21.4%	1.17	<0.0001^§
Depression-related	301	6.8%	131	4.5%	1.53	<0.0001^§
MHD-related	349	7.9%	157	5.4%	1.47	<0.0001^§
Healthcare costs, $ (mean, SD)^††
All prescription drug costs	$131	(437)	$129	(445)	$1	0.0774
Medical costs
Emergency room	$131	(439)	$119	(475)	$11	0.0006^§
Inpatient visit	$672	(5294)	$385	(3923)	$287	<0.0001^§
Outpatient hospital visit	$273	(1457)	$286	(1776)	−$12	0.1263
Other professional services	$358	(627)	$367	(604)	−$10	0.2108
Total medical costs	$1434	(5939)	$1160	(4791)	$274	<0.0001^§
Depression-related	$517	(1392)	$256	(1933)	$260	<0.0001^§
MHD-related	$605	(1612)	$328	(2124)	$277	<0.0001^§
Total costs	$1565	(6009)	$1290	(4860)	$275	<0.0001^§

*Baseline characteristics are measured over the 6-month pre-index period.

^†Relative risk is calculated for proportions; difference is calculated for continuous variables.

^‡Wilcoxon rank sum test for continuous variables; chi-square test for categorical variable.

^§p-value ≤0.05.

^¶Related to a claim with an ICD-9 code of: 296.2, 296.3, 298.0, 300.4, 309.1, or 311.

**Mental Health Diagnosis related.

^††Inflated to 2008 US dollars.

Indirect clinical indicators

During the 12 months after the diagnosis of depressive disorder, adolescents who were initiated early on SSRIs or SNRIs continued to have a higher unadjusted risk of all-cause hospitalizations (RR = 1.21, p < 0.01) as well as depression-related and mental health-related inpatient stays (RR = 1.32 and 1.27, respectively, p < 0.01) compared with adolescents who initiated pharmacotherapy later (Table 3). Unadjusted all-cause, depression-related, and mental health-related outpatient hospital visits were not statistically significantly different between the two cohorts. However, the early initiator cohort had a lower risk of all-cause, depression-related and mental health-related ER visits (all p < 0.05).

Table 3.  12-month rates of healthcare services utilization for adolescent depression patients with early vs. late treatment initiation.

	Descriptive results						Multivariate results
	Early initiators		Late initiators		Relative risk	p-value^†	Odds ratio^‡	p-value
	[A]		[B]		[A]/[B]	[A] vs. [B]
Healthcare service utilization* (n,%)
Inpatient visit	1081	24.5%	595	20.3%	1.21	<0.0001**	0.97	0.6105
Depression-related^§	868	19.7%	438	15.0%	1.32	<0.0001**	0.89	0.1326
MHD-related^¶	937	21.2%	490	16.7%	1.27	<0.0001**	0.93	0.3206
Outpatient hospital visit	2374	53.8%	1603	54.7%	0.98	0.4200	1.09	0.1096
Depression-related	793	18.0%	476	16.3%	1.11	0.0580	0.97	0.6137
MHD-related	973	22.0%	603	20.6%	1.07	0.1380	1.01	0.9248
ER visit	1552	35.2%	1104	37.7%	0.93	0.0270**	1.13	0.0272**
Depression-related	280	6.3%	231	7.9%	0.80	0.0110**	1.30	0.0074**
MHD-related	369	8.4%	285	9.7%	0.86	0.0430**	1.21	0.0307**

*Rates are measured over the 12-month post-index period.

^†Chi-square test for categorical variables.

^‡Logistic regression was used; specified baseline covariates include cohort, insurance type, gender, age, Charlson Comorbidity Index, index year, asthma, injuries, MDD, inpatient visits, outpatient visits, ER visits, and other professional services visits.

^§Related to a claim with an ICD-9 code of: 296.2, 296.3, 298.0, 300.4, 309.1, or 311.

^¶Mental Health Diagnosis related.

**P-value ≤0.05.

Results of the logistic regression analysis indicate that, after adjusting for differences in baseline characteristics, rates of inpatient visits and outpatient hospital visits were not statistically significantly different between the two cohorts. However, late initiators had higher odds of all-cause ER visits than early initiators (OR = 1.13, p < 0.05) (Table 3). The risks for depression-related and mental health-related ER visits for late initiators were also higher compared to early initiators (OR = 1.30 and 1.21, respectively, p < 0.05).

Healthcare costs

Figure 1 reports descriptive results of the cost analysis. Early initiators had higher unadjusted prescription drug costs during the 12-month study period compared with late initiators (+$84 per patient, p < 0.01). The unadjusted medical cost incurred by early initiators was $719 lower than that incurred by late initiators (p < 0.01). Overall, early initiators incurred $636 less total unadjusted healthcare cost per person than late initiators.

Figure 1.  12-month medical and prescription drug costs for adolescent depression patients with early vs. late treatment initiation. *P-value ≤0.05.

Results of the multivariate analysis were consistent with findings from the descriptive analysis. Prescription drug costs remained higher for early initiators (+$121, p < 0.01) (Figure 1). After adjusting for baseline differences, medical costs incurred by early initiators were significantly lower than those of late initiators (−$1354, p < 0.01). Overall, early initiators had $1094 lower total healthcare cost than late initiators (p < 0.01).

Discussion

Although early initiators of SSRI and SNRI treatment used more healthcare services and had higher healthcare costs during the 6-month baseline period before the index diagnosis of a depressive disorder, these adolescents used fewer healthcare services and incurred lower total healthcare cost than the late initiators during the 12 months after the depression diagnosis. Some of the early initiators may have been recognized for depressive symptoms during hospitalization or ER visit in the baseline period and received a diagnosis of depression and an immediate antidepressant prescription as follow-up care. Overall, early initiators appear to be in worse general health, but the data available do not have sufficient details for a specific assessment. After adjusting for differences in baseline demographics and comorbidities and baseline healthcare services use, early initiators were significantly less likely to have ER visits during the 12-month study period. Most notably, these patients had considerably lower risk for mental health-related ER visits, indicating that they required fewer urgent medical interventions for depressive disorders. None of the other outcomes, including all-cause, depression-related, and mental health-related inpatient stays and outpatient hospital visits, suggested an advantage of late initiators over the early initiators during the study period.

These findings echo previous clinical, observational studies, which suggest an association between the duration of untreated illness in adult depression patients and poorer outcomes, including persistent symptoms, higher number of recurrent episodes, and prolonged course of disease20–23. Conversely, it is reasonable to expect that timely intervention is likely to reduce the severity, duration, and risk of recurrence of depression, resulting in better outcomes and lower healthcare service utilization in the long term.

As a result of the reduced utilization of healthcare services, early initiators incurred lower medical and total healthcare costs during the 12-month study period than late initiators. Despite a slightly higher total drug cost due to greater antidepressant use, the savings in total healthcare cost amounted to 7 times the difference in drug cost between cohorts. From the third-party payer’s perspective, the findings suggest that there is little benefit in restricting young patients’ access to antidepressant medications as it may lead to worse outcomes and higher overall costs.

Limitations

This study has several limitations. First, as a common limitation in claims studies, clinical data are scarce, thereby limiting the extent to which one can observe differences in disease severity between the two cohorts. Therefore, healthcare services utilization was used as a proxy for clinical outcomes. Baseline characteristics, particularly the baseline healthcare services utilization and comorbidities, suggested that the early initiators may have been in poorer health compared to late initiators. Every effort was made to control for these differences using multivariate analyses, but inter-cohort differences in unobserved clinical characteristics might exist.

The reasons for delayed therapy were not observable in the data, but could have included counseling and other therapies that have been found to be effective in some patients. Patients who were late initiators may not have been severe enough to receive therapy or may not have responded to other forms of therapy. Another reason for delayed therapy is that patients may have initiated therapy on professional samples (which would not appear in claims) prior to their filling their prescriptions, but there is no reason to believe this sampling is different among groups. Moreover, although we excluded HMO patients from our analysis in order to limit the impact of treatment restrictions on the analysis, it is possible that some preferred provider organization plans might have restrictions in place, such as prior authorization, which could influence the timing of treatment initiation. Details of plan restrictions other than plan type were not available in the data, limiting the extent to which plans with such designs could be filtered out.

While some late-initiating patients might have benefited from earlier initiation with antidepressants, identifying those who are not likely to respond to other forms of therapy may be difficult. Because of the post-hoc nature of this analysis, it is not possible to identify patients who would have benefited from early initiation. Further research is needed to identify the characteristics at time of diagnosis of patients who would benefit from early initiation of antidepressants. Moreover, this study does not assess the resource utilization and costs of patients who were not initiated on antidepressants. An across-the-board early initiation of antidepressants for adolescent patients could threaten the potential cost savings from initiation among those who would benefit from it by adding treatment costs for patients who would have recovered without treatment. Further, if late antidepressant therapy initiation timing is a result of informal clinical practice patterns, it may be difficult to change physicians’ prescription practices, which may result from individual preferences and personal experiences in treating particular patient populations.

Finally, the definition of early and late initiation was based on an empirical analysis of the timing of treatment initiation, and no clear cutoff was identified beyond what was observed in the data. While this may limit the clinical inference it is possible to draw from this study, indicators still show a significant and persistent difference in both severity and costs between the two cohorts. Long-term outcomes studies are warranted to further confirm the benefits of early treatment in adolescent populations with depression, as well as to determine the characteristics of patients who would benefit the most from early treatment.

Conclusion

Adolescents who received SSRI or SNRI treatment within 1 month of being diagnosed with depressive disorders had a lower risk for all-cause, mental health-related, and depression-related ER visits and incurred lower total healthcare cost in the 12 months after the diagnosis, compared with those who received delayed SSRI or SNRI treatment. Future research is warranted to assess the reasons for delayed antidepressant treatment and systematically evaluate the clinical and economic consequences of the delay.

Transparency

Declaration of funding

This study was funded by Forest Research Institute.

Declaration of financial/other relationships:

A.P.Y., A.K., R.B., J.X., R.B.-H., and E.Q.W. are employees of Analysis Group, Inc. S.B. is an employee of Forest Research Institute. M.H.E. was an employee of Forest Research Institute at the time the study was conducted.

Acknowledgments

Ms Jun Yan, an independent medical writer, contributed to drafting this manuscript.

This study was funded by Forest Research Institute and was presented at the Academy of Managed Care Pharmacy’s Educational Conference in San Antonio, TX, USA, October 7–10, 2009.