This study estimated changes during the 1990s in the quality of usual-care treatment among persons diagnosed as having bipolar I disorder in a privately insured population. Retrospective private insurance administrative data were analyzed for enrollees aged 18 to 64 who were diagnosed as having bipolar I disorder during 1991 (431 person-years), 1994 (598 person-years), and 1999 (600 person-years). Medication and psychotherapy quality indicators were derived from bipolar disorder expert guidelines published in 1994, which were consistent with guidelines published until year 2002. The unadjusted prevalence of receiving any lithium, valproate, or carbamazepine improved over the study period (68 percent in 1991, 64 percent in 1994, and 77 percent in 1999), whereas, compared with 1991, receiving any antidepressant in the absence of lithium, valproate, or carbamazepine increased in 1994 and then declined in 1999 (13 percent in 1991, 23 percent in 1994, and 14 percent in 1999). The unadjusted prevalence of receiving any psychotherapy declined steadily and sharply (94 percent in 1991, 89 percent in 1994, and 69 percent in 1999). The unadjusted prevalence of receiving any lithium, valproate, or carbamazepine and therapy together declined over time (65 percent in 1991, 58 percent in 1994, and 54 percent in 1999). After the analyses adjusted for patient characteristics, these changes were significant from p<.01 to p<.001. The prevalence of receiving the pharmacotherapy recommended in the guidelines improved after guideline publication in 1994, whereas other quality measures that included receiving psychotherapy declined throughout the study period. These results suggest different psychotherapeutic modalities are under differing constraints under managed care, constraints that overpower consensus in the literature of quality practice. Policy makers should measure a variety of key therapeutic modalities when measuring quality in order to capture these differences.