The Clinical History and Costs Associated With Delayed Diagnosis of Bipolar Disorder

Paul E. Stang, PhD; Cathy Frank, MD; Anupama Kalsekar, MS; Marianne Ulcickas Yood, DSc, MPH; Karen Wells, BS; Steven Burch, PhD

In This Article


Data Source

We conducted this study in a large, vertically integrated health system serving the primary and specialty healthcare needs of Midwestern residents. This system is affiliated with a multi-specialty, salaried, physician group that provides most of the care for health system patients. The health system also owns a large, nonprofit, mixed-model health maintenance organization (HMO). To optimize the computerized data available for this study, the population was limited to HMO members assigned to the medical group physicians.

Computerized Data Sources and Medical Records. The health system maintains an extensive, centralized system of computerized databases. Data for this study came from electronic and paper medical record review and 3 computerized databases: (1) the HMO membership file; (2) the outpatient pharmacy database; and (3) the encounter and claims databases.

The HMO membership database includes information about historical and current coverage dates, covered benefits, and the amount of copayments for which the enrollee is responsible. The pharmacy database contains information on outpatient prescriptions filled by members. To receive payment, HMO-contracted pharmacies, which include several major chains, must file a claim for each prescription filled. The HMO collects the claims information and stores the data in the HMO drug claims database. This database includes information on the date a member filled a prescription, drug name, national drug code, number of pills, dosage dispensed, and number of days supplied.

The encounter and claims databases store comprehensive patient data for outpatient, emergency department, and inpatient care delivered by the medical group. For each outpatient encounter, information about date of visit, diagnoses, physician delivering care, procedures delivered, and clinic in which the care was delivered are compiled. Likewise, for each inpatient stay, information is collected about admission and discharge dates, diagnoses, and procedures. In addition to these encounter-specific fields, data on the patient's age, sex, and race are available.

Study Population and Bipolar Cohort Definition

Using the data sources described, we identified all patients with newly diagnosed bipolar disorder from January 1, 2000, through August 31, 2002, by obtaining records of all patients who had an inpatient, outpatient, or emergency department encounter with an ICD-9 coded diagnosis indicative of bipolar disorder (296.00-296.06, 296.40-.46, 296.50-296.56, 296.60-296.66, 296.7, 296.80, 296.89). All coded diagnoses were reviewed in this process (ie, this identification was not limited to primary diagnoses). To distinguish newly diagnosed cases, we required that all patients be continuously enrolled in the health plan for at least 1 year before bipolar disorder diagnosis, and excluded from the study population all patients who had any previous encounter with the health system that was coded as bipolar disorder. The date of bipolar diagnosis was labeled the index date.

For each patient in the cohort, we downloaded all data on member enrollment (ie, dates that patient enrolled and disenrolled in the health plan), demographics (age, race, sex), and all available utilization (inpatient, outpatient, emergency department, and pharmacy encounters) for the period January 1, 1990, through August 31, 2003. Unless otherwise stated, the results presented here are limited to those patients with full pharmacy benefit and linkage.

For each patient in the cohort, we identified the first encounter with a mental health-related diagnosis that occurred from January 1, 1990, through the index date. The first mental health-related diagnosis was defined using the following ICD-9 codes: depression (296.2x, 296.3x, 311); anxiety (293.89 300.00, 300.01, 300.02, 300.20-300.23, 300.29, 300.3, 300.4, 300.9, 308.3, 308.4, 308.9, 309.81); substance abuse (303.00-305.90); schizoaffective disorder (2975.7x); and schizophrenia (295.xx, excluding 295.7x). An underlying assumption in our study is that among those who eventually received a bipolar diagnosis, previous mental health encounters may have represented early presentations of bipolar disorder that went unrecognized. We further assumed that when patients finally did receive a bipolar diagnosis, they did in fact have bipolar disorder.

The cohort of patients with bipolar disorder was then stratified according to the following categories: (1) patients diagnosed with bipolar disorder fewer than 6 months after first mental health-related visit; (2) patients diagnosed with bipolar disorder at least 6 months but less than 1 year after the first mental health-related visit; (3) patients diagnosed with bipolar disorder at least 1 year but less than 2 years after the first mental health-related visit; (4) patients diagnosed with bipolar disorder at least 2 years but less than 4 years after the first mental health-related visit; and (5) patients diagnosed with bipolar 4 years or more after the first mental health-related visit.

Identification and Follow-up of Comparison Cohort

For each patient in the bipolar cohort, we randomly selected 5 comparison patients from the general population of health system members and matched them with the patient with bipolar disorder by sex, race, and age (± 5 years). The index date for the matched patient with bipolar disorder (ie, the date of bipolar diagnosis) served as the index date for the comparison patients. As with members of the bipolar cohort, we required that comparison cohort patients be enrolled in the HMO for at least 1 year before the index date. Because patients with bipolar disorder were identified at the time of a visit (ie, the time the ICD-9 code was recorded), we required that the comparison patient have a visit within 60 days of the index date. This requirement most likely resulted in an analysis with the most conservative comparison of cost because, by definition, we selected a comparison population that may have higher healthcare utilization than the general population in the healthcare system.

Healthcare Utilization Data

For each patient in the cohort, we tabulated, using data as far back as 1990, all inpatient stays, outpatient visits, emergency department visits, and prescriptions filled beginning at the first mental health-related visit through the index date (ie, date of bipolar diagnosis). In addition, these same data were tallied for the 1-year period post-index date.

Cost Data

Data on costs are available beginning in 1995. Therefore, analyses of costs are limited to patients with a first mental health diagnosis recorded January 1, 1995 or later. For each patient in the study, we downloaded all encounters and prescriptions filled from the time of the first mental health-related visit through the index date. As with the health care utilization data, we also tallied costs incurred during the year following the index date. Encounters were categorized as follows: inpatient, outpatient, emergency department, pharmacy, and other.

Pharmacy data are stored such that actual costs paid are recorded in the database. The databases recording all other utilization that occurred from January 1, 1995 through December 31, 2000, store only charges. For utilization occurring during this time, we downloaded charges and applied corresponding cost-to-charge ratios stored in the databases to obtain costs. For January 1, 2001 through August 30, 2003, actual costs are recorded in the databases and were used for this analysis.

Data Analysis

We calculated the mean number of visits overall and per month of follow-up time (time between first mental health-related visit and index date) and the mean costs overall and per month of follow-up time, with corresponding standard deviations. T-tests were used to compare mean monthly costs before and after the diagnosis of bipolar disorder. Differences in mean total costs 1 year after bipolar diagnosis were assessed by time from initial mental health encounter using ANOVA. Utilization data were categorized as follows: outpatient (primary care, specialty, mental health), laboratory tests, emergency department, inpatient stays, psychiatric inpatient stays, ambulatory short stay, and others. Cost data were categorized as follows: outpatient pharmacy, inpatient stays, outpatient visits, and emergency department encounters. Finally, to evaluate potential depression misdiagnoses, we tabulated all antidepressant prescriptions filled by the study population and calculated the total number of different antidepressants filled, the mean number of days exposed, and the proportion of follow-up time exposed overall and stratified by time between first mental health-related encounter and bipolar diagnosis.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.