The analysis describes a mixed picture regarding the quality of care received by persons with bipolar-I depression. It is important to consider these results in the context that we used a liberal interpretation of the guidelines when constructing these measures of “best treatment” (i.e., receiving at least one psychotherapy session or at least one nondopamine blocking antimanic medication prescription).
Throughout the year only approximately half of the patients received treatment most compatible with guideline recommendations (i.e., both antimanic medications and psychotherapy). A quarter of the outpatient depressed phases ended due to a break in treatment for at least two months. Only approximately two-thirds received pharmacotherapy consistent with contemporary guideline recommendations; liberalizing the definition of acceptable antimanic medications to include all antipsychotics increased this number to nearly three-quarters.
Nearly three-quarters of the depressed phases included some psychotherapy (but few received family psychotherapy [<7%]). We cannot determine from the claims data if the psychotherapy received was consistent with the psychotherapies that have demonstrated efficacy in clinical trials. However, receiving at least some psychotherapy allows the possibility that these patients who have a chronic and disabling illness received needed educational and psychotherapeutic treatment. Prior research by this investigator group using similar Medstat data found that the proportion of persons with bipolar-I disorder receiving any psychotherapy has declined since 1991.36
Thus, while it appears to be a more positive finding that many of these depressed phases included at least some psychotherapy, further research is needed to determine if rates for bipolar-I depression psychotherapy are declining as well.
Approximately half of the depressed phases were treated with both an antidepressant and a non-dopamine blocking antimanic medication, whereas only 12% with only a non-dopamine blocking antimanic medication (and no antidepressant). Given the often persistence of depressive symptoms, more limited treatment options for persons with bipolar-I depression in 1999 and 2000, and uncertainty in the literature about prescribing antidepressants for patients with bipolar disorder, it is not surprising that a high proportion of those who are on antimanic medications also received an antidepressant. Possibly, with the availability of newer medication and psychotherapy options; and even newer evidence that antidepressants may not be efficacious in treating bipolar depression,9
this practice pattern is changing.
Twenty percent of these depressed phases ended due to a change in polarity that occurred in the outpatient setting. This is particularly important when one considers that often in mental health quality of care studies using claims data, illness instability is measured by looking at hospitalization rates. In this cohort, if hospitalization were the only marker of instability examined then a considerable amount of symptom instability would go unnoticed. Thus, this observation has important implications for bipolar disorder quality assessment using claims data analyses.
As much as 28% of the mental health costs for this population went towards care that did not meet professional guideline standards—and 12%−21% were costs for care that experts have deemed to be potentially harmful. We considered the possibility that these were patients possibly more likely to have entered our depressed cohort by inpatient claims and did not transition to outpatient care (thus their costs would be high but we would not observe outpatient treatments that met our quality standards). However, post-hoc analyses indicated that patients whose treatment was most concordant with guidelines were more—not less—likely to have been hospitalized (non-dopamine blocking antimanic medication χ2= 6.6, d.f. =2, p=.04; any antimanic medication: χ2= 12.4, d.f. =2, p=.002).
A limitation of claims data is that we cannot observe all elements of the clinical history that could dictate treatment and clarify the extent that patients received appropriate care. For example, some patients in our sample who did not receive an antimanic agent may have been stable from a mania perspective (i.e., no manic episode over past 20 years) and remained free of mania while receiving medications not considered traditional mood stabilizers, such as clonazepam or clonidine. Similarly, patients who did not receive psychotherapy during an episode of bipolar depression may have been those who were quickly responsive to medication changes to treat the depressive symptoms. Our observation that patients who received both psychotherapy and antimanic agents were more frequently hospitalized than patients who did not receive both is consistent with this hypothesis that deviations from guideline recommendations may be related, in part at least, to patients’ stability. However, it is also important to consider that bipolar depression is often treatment resistant37-40
and that we would expect poor outcomes in many bipolar-I patients who do not receive a mood stabilizer.41-44
Therefore, while some of these results may be consistent with tailoring treatment to patients needs, it is more likely that for an overwhelming proportion they represent poor care.
Our data reflect similar spending on mental health treatment compared to several published studies of bipolar disorder, including a study using data from care delivered in 200426, 32, 33
—a time in which more second generation antipsychotic medications were receiving FDA approval for acute and maintenance phase bipolar disorder treatment. These studies were similar to ours in that they used administrative data from multiple health plans, employers and regions of the United States, as well as similar case finding and treatment definitions. However, two prior studies found considerably lower costs but are not directly comparable to ours due to differing methodology, such as a more limited geographical distribution of enrollees, or not including all mental health treatment costs.29, 30 30
One potential limitation of this study is our use of administrative data to determine the cohort, which raises the concern of diagnostic accuracy in claims data. However, previous comparisons in a privately insured population found substantial agreement (94%) between claims diagnoses of bipolar disorder and chart review, and that claims data have demonstrated validity in analyses utilizing a bipolar diagnosis to establish the cohort and assess population-based quality of care.45, 46
Additionally, our analysis utilized a more stringent algorithm of the claims data than these studies to determine our bipolar cohort. While we did not determine the accuracy bipolar depression diagnoses specifically in the claims data (and to our knowledge there are no published reports on this either), studies examining the agreement between administrative data and either structured clinical interview or chart review have found fair agreement for depressive disorders.47, 48
Additionally, we cannot directly observe symptoms in administrative data. While typically our definition of depressed outpatient treatment phase ends were based on events: hospitalization, changes in polarity, a gap in treatment, or ICD-9 coding indicating remission, our estimated duration of bipolar-I depression specific treatment would be inflated if depressive symptoms had resolved but the diagnosis was not coded to a fifth digit indicating remission. Thus, our estimates of depressed phase duration should be considered an “upper bound”.
Our analysis took into consideration that since 1999 and 2000 (when this cohort received treatment), FDA indications and clinical guidelines have evolved such that antipsychotic medications (particularly second generation) are increasingly acceptable as maintenance monotherapy antimanic medications. In our cohort, most antipsychotic prescribing was in conjunction with a traditional mood stabilizer, not as stand-alone maintenance antimanic treatment. However, analyses conducted by members of this investigator group using data from the Systematic Treatment Enhancing Program for Bipolar Disorder (STEP-BD) study during years 2001 through 2004 indicate that for patients with bipolar-I disorder, there was no increased likelihood of receiving any antimanic agent over time, nor an increased likelihood of receiving an antipsychotic as antimanic monotherapy.49
Additionally, our antimanic medication rates, hospital rates and mean mental health costs were similar to a bipolar disorder usual care study examining care as recent as 2004.50
Thus, we have reason to believe that the practice patterns and costs we observed in this study have persisted at least through 2004. Still, an important area of future research would be to examine current treatment patterns relative to second generation antipsychotic use and maintenance antimanic monotherapy for bipolar-I disorder.