Workplace productivity, employment issues, and resource utilization in patients with bipolar I disorder

Abstract

Objective: To collect workplace productivity and healthcare utilization data from subjects with bipolar I disorder and compare the results with those from normative subjects.

Methods: A cross sectional survey was administered to patients and recruiting physicians. Data collected included employment status, Endicott Workplace Productivity Scale (EWPS) results, healthcare resource utilization, and quality-of-life.

Results: In comparison with normative subjects, bipolar I subjects reported lower levels of work productivity (measured by the EWPS). Bipolar I subjects also reported more frequent outpatient visits and more prescribed pharmaceuticals. Bipolar I subjects were more likely to miss work, have worked reduced hours due to medical or mental health issues, receive disability payments, been involved in a crime, be uninsured or covered by Medicare, or have been fired or laid off. The study groups were age- and gender-matched to reduce the impact of selection bias associated with a non-randomized study design. Other potential limitations affecting the results of the study include recall bias and possibly an impact of different data collection methods (e.g. Internet versus telephone).

Conclusions: Bipolar I disorder is associated with a negative effect on work productivity and resource utilization and is an appropriate disease management target for employers and healthcare decision makers.

Key words::

• bipolar disorder
• employment
• productivity
• resource utilization

Introduction

Bipolar disorder, also referred to as manic-depressive illness, is a chronic and sometimes severe condition, characterized by episodes of mania and depression. There are two types: bipolar I disorder is defined by one or more episodes of mania or a mixed state with symptoms of mania and depression overlapping; bipolar II disorder is defined by hypomanic episodes¹. The World Health Organization's Global Burden of Disease Study (GBD) found that 29.5 million people worldwide suffered from bipolar disorder in 2004². The GBD reports that bipolar disorder is the 12th leading cause of years of healthy life lost—ahead of asthma, schizophrenia, and glaucoma. US population-based studies have estimated that lifetime prevalence for bipolar I disorder is 1.0% and for bipolar II disorder, 1.1%^3–5.

The majority of people with bipolar disorder are younger than age 60². In the US, the median age of onset for bipolar disorder is 25 years³. Bipolar disorder is also associated with a high relapse rate, and the majority of patients who relapse also experience significant mood pathology and reduced psychosocial functioning between episodes⁴.

The effect of bipolar I disorder on patients should be evaluated separately from other mood disorders⁵. A national survey of people with mood disorders found that workers with bipolar disorder, although fewer in number, experienced significantly more lost workdays than people with major depressive disorder⁵. A recently reported literature review concluded that bipolar disorder in a working population had a negative impact on absenteeism, productivity, and workplace relationships⁶.

Cost estimates for the total impact of bipolar disorder in the US have ranged from US$24 billion for lifetime direct and indirect costs (using incidence-based data projected for the lifetime of bipolar patients identified in 1998)⁷ to US$45.2 billion per year (using prevalence-based data from 1991)⁸. A retrospective healthcare claims study found that patients with bipolar disorder incurred healthcare costs four times higher than those of matched controls⁹. In the US, the indirect cost of lost productivity associated with bipolar disorder was estimated to exceed US$14 billion per year based on data from 2002⁵. Bipolar disorder was associated with a 20% reduction in annual workplace productivity based on the results of a retrospective evaluation of healthcare claims data¹⁰.

Although bipolar I disorder affects a relatively small percentage of the US population, the condition is likely to develop just as those affected are entering the workforce. The authors believe that the burden of bipolar I disorder on both direct healthcare costs and indirect costs associated with lost productivity should be a disease management target for employers and healthcare decision makers.

The goal of this study was to evaluate a group of subjects with bipolar I disorder by collecting data on workplace productivity, employment issues, and healthcare resource utilization and to compare the results with those from a group of subjects with no history of any type of bipolar disorder or serious mental illness.

Methods

Study design

This paper describes the results from a cross-sectional survey study of subjects with diagnosed bipolar I disorder and normative subjects without a history of bipolar disorder or other serious mental illness. Subjects were recruited by physicians during regularly scheduled office visits. Subjects in the bipolar I group were recruited by 24 psychiatrists, and normative subjects were recruited by 31 primary care physicians. All recruiting physicians were enrolled in the i3 Physician Investigator Database. Most bipolar subjects were recruited by psychiatrists in Florida (17.7%), Texas (19.3%), and Ohio (12.4%), and most normative subjects were recruited by primary care providers in Pennsylvania (30.9%), Florida (12.1%), and Missouri (10.2%).

Data were collected from physicians via an electronic case report form (eCRF) and from subjects using a paper screening questionnaire and another questionnaire completed via telephone or the Internet within 14 days of their enrollment office visit. Subjects were enrolled in the study from July 24, 2006, to November 8, 2006.

Independent institutional review board (IRB) approval was obtained before the study commenced. All study materials were handled in compliance with the Health Insurance Portability and Accountability Act.

Nonrandomized subject selection

For both groups, recruitment eligibility was limited to adult subjects in the US who were age 18–64 years at their screening visit and could read and understand English. All subjects were employed for pay for a minimum of 20 hours per week or were receiving workers' compensation or disability insurance payments.

For the bipolar group, patients of the recruiting psychiatrists were eligible for the study if they met the criteria described above and had a diagnosis of bipolar I disorder with at least two episodes of mania in their lifetime and at least one episode of mania within the previous 2 years. Patients were excluded if they also had a diagnosis of bipolar II disorder, schizophrenia, schizoaffective disorder, or unipolar depression. No exclusions were made for medical conditions.

For the normative group, patients of the recruiting primary care physicians were eligible for the study if they met the criteria above but were excluded if they had a diagnosis of bipolar I disorder, bipolar II disorder, schizophrenia, schizoaffective disorder, or unipolar depression. No exclusions were made for medical conditions.

To match the group by age and gender, subjects were recruited beginning with the bipolar group and with overlap in recruitment of the normative group. Primary care physicians received pre-assigned age-gender categories to recruit subjects into the normative group.

For both groups, the recruiting physicians provided data about each subject's demographic information, treatment history, and comorbid conditions via eCRF. While in the office, the subjects completed a questionnaire that included the Mood Disorder Questionnaire (MDQ)¹¹. The office staff provided subjects with the telephone and Internet access information to complete the survey portion of the study.

The survey included questions about employment status, the Endicott Work Productivity Scale (EWPS), and measures of healthcare resource utilization. Also included was the Quality of Life Enjoyment and Satisfaction Questionnaire-Short Form (Q-LES-Q-SF), the results of which are reported elsewhere (unpublished manuscript submitted to patient reported outcomes focused journal). The MDQ was administered to compare the percentages of mood disorder diagnoses in each group of patients. The inclusion of a subject in the study was not dependent on the results of the MDQ.

Internet responses were collected via a secure Internet site. Telephone interviews were conducted using a computer-assisted telephone interviewing system. Subjects who completed the survey either by the Internet or by telephone received US$20 to compensate them for their time. This amount was approved by the external institutional review board.

Data collection

Workplace productivity data were collected using the EWPS¹², a standard subject-completed instrument that measures absenteeism and presenteeism using 25 items. The instrument was designed to capture lost productivity data in clinical trials. Each item in the EWPS captures the frequency of productivity-related behaviors during the previous week, using a 5-point Likert scale. A sum of scores is then computed, with total EWPS scores ranging from zero (best score) to 100 (worst score). In addition to the items used to calculate the total EWPS score, individual items on this instrument include whether subjects missed work during the previous week and the reason, whether subjects are working a reduced schedule, and whether subjects have ever been laid off or fired and the reason.

Other employment issues were measured by the EWPS and individual questions developed by the authors. Subjects were asked whether they were receiving workers' compensation or short-term or long-term disability benefits, and whether they had ever been jailed, arrested, or convicted of a crime other than drunk driving.

Resource utilization information was collected by asking subjects to recall the number of visits in the previous 30 days to each of the following: psychiatrist, other physicians, psychotherapist, and emergency department. Information regarding medications and previous treatment was collected by the enrolling physician on the eCRF.

Data analysis

Descriptive and inferential statistics, including t-tests and analysis of variance, were used to compare the demographic and subject characteristics for each group and to estimate the effect of bipolar I disorder on productivity, employment issues, and resource utilization. Characteristics of subjects who enrolled but failed to complete the final survey were compared using t-tests and Fisher's exact tests from data provided in the eCRF (results not shown but available on request). For the EWPS, an analysis of internal consistency was conducted by calculating Cronbach's alpha.

Ordinary least squares regression analyses were conducted to compare the results from subjects in the bipolar I disorder group with the results from the normative group on measures of work productivity, while adjusting for demographic and medical history variables. A stepwise regression analysis was used to identify significant determinants of work productivity in this population. The primary predictor of interest was the regression coefficient used to compare subjects in the bipolar I disorder group and those in the normative group by predicting work productivity scores. Results were tested for significant interactions between the dichotomous variable (bipolar vs. normative) and other variables to determine any significant moderator effects.

All statistical analyses for this paper were generated using either SAS software [Version 9.1 for Windows, SAS Institute Inc., Cary, NC, USA] or STATA software [Version 10, StataCorp LP, College Station, TX, USA].

Results

Noncompletion of the productivity and employment surveys within the required 14-day period was the primary reason for dropout for both groups (65 of 88 subjects in both the bipolar I disorder group and the normative group did not complete the survey during the time allowed; 74% of all dropouts). Fewer than ten subjects per study group dropped out for the following stated reasons: refusal, nonworking or wrong phone number, and language barrier.

No significant differences were noted between completers and noncompleters for age, gender, highest level of education, or insurance coverage. The only significant difference noted between responders and nonresponders was that nonwhite subjects were more likely to complete the survey than white subjects (18 vs. 9%; p=0.045).

Of the 249 bipolar I subjects enrolled in the study, 219 completed the follow-up survey within 14 days (Figure 1; 88% response rate). For the normative group, 198 of the 256 recruited subjects completed the follow-up survey (77% response rate).

As noted previously, the two groups were matched by age and gender during data collection and, as a result, the mean ages were similar for both groups (Table 1). However, a higher percentage of normative subjects were observed in the 30- to 39-year age category than in the bipolar I group, and the difference approached statistical significance depending on rounding of the p-value (33 vs. 25%; p=0.051).

Table 1. Demographic and clinical characteristics of the study sample.

Characteristic	Bipolar group	Normative group	p-value
	n=219	n=198
Age at enrollment
Mean±SD, years	39.6±10.9	39.4±10.6	0.877
Median age (range), years	40 (19–63)	38 (18–64)
Age category, n (%)
18–29 years	54 (24.7)	37 (18.7)	0.140
30–39 years	54 (24.7)	66 (33.3)	0.051
40–49 years	64 (29.2)	59 (29.8)	0.898
≥50+ years	47 (21.5)	36 (18.2)	0.402
Gender, n (%)
Male	69 (31.5)	60 (30.3)	0.791
Female	150 (68.5)	138 (69.7)
Race, n (%)
White	190 (86.8)	153 (77.3)	0.011
Black	16 (7.3)	30 (15.2)	0.011
American Indian/Native American	3 (1.4)	1 (0.51)	0.366
Asian	3 (1.4)	2 (1.01)	0.736
Native Hawaiian/Pacific Islander	0 (0)	1 (0.51)	0.292
Other	9 (4.1)	11 (5.6)	0.490
Mixed race	1 (0.5)	0 (0)	0.341
Highest level of education
Some high school	9 (4.1)	7 (3.5)	0.760
High school graduate	45 (20.6)	42 (21.2)	0.868
Some college, no degree	68 (31.1)	57 (28.8)	0.615
Associate degree	33 (15.1)	28 (14.1)	0789
Bachelor's degree	45 (20.6)	43 (21.7)	0.770
Master's degree or higher	19 (8.7)	21 (10.6)	0.504
Employer-sponsored health care coverage, n (%)
Yes	108 (49.3)	137 (69.2)	< 0.001
No	111(50.7)	61 (30.8)
History of medical condition, n (%)
Cancer	7 (3.2)	6 (3.0)	0.922
Hypertension	44 (20.1)	35 (17.7)	0.530
Hypercholesterolemia	26 (11.9)	25 (12.6)	0.814
Cardiovascular disease	3 (1.4)	5 (2.5)	0.390
Diabetes	15 (6.9)	5 (2.5)	0.039
Arthritis	21 (9.6)	16 (8.1)	0.589
Autoimmune disorders	2 (0.9)	5 (2.5)	0.201
Neurologic disorders	7 (3.2)	1 (0.5)	0.045
Epilepsy or other seizure	2 (0.9)	1 (0.5)	0.622
Other	33 (15.1)	24 (12.1)	0.382
At least one medical condition	96 (43.8)	81 (40.9)	0.546
History of psychiatric condition, n (%)*
Any psychiatric condition	139 (63.5)	20 (10.1)	< 0.001
No psychiatric condition	79 (36.1)	178 (89.9)	< 0.001
Receiving medication for any psychiatric condition other than bipolar disorder	22 (10.1)	5 (2.5)	0.002

*Other psychiatric conditions include unipolar depression, generalized anxiety disorder, obsessive compulsive disorder, attention deficit disorder, alcohol abuse or alcoholism, substance abuse or addiction, borderline personality disorder, posttraumatic stress disorder, panic disorder, and anorexia, bulimia, or other eating disorder.

Approximately two-thirds of both groups were female (p=0.791). Similar proportions of subjects in both groups had a history of at least one medical condition (43.8% bipolar vs. 40.9% normative; p=0.546). Nearly 7% of bipolar subjects had a history of diabetes mellitus compared with 2.5% of the normative group (p=0.039). Neurologic disorders appeared in the history of 3% of bipolar subjects, compared with less than 1% of normative subjects (p=0.045). Nearly two-thirds of bipolar I subjects (63.5%) had a history of at least one additional psychiatric condition, compared with 10% of the normative group with a diagnosis of any psychiatric condition (p< 0.001), including unipolar depression, generalized anxiety disorder, obsessive compulsive disorder, attention deficit disorder, alcohol abuse or alcoholism, substance abuse or addiction, borderline personality disorder, posttraumatic stress disorder, panic disorder, and anorexia, bulimia, or other eating disorder.

As measured by the MDQ, 140 bipolar I disorder subjects (66.7%) had a positive screen for a mood disorder compared with two normative subjects (1.0%; p< 0.001). This finding confirms that the normative group did not include undiagnosed bipolar I subjects.

Almost 70% of normative subjects received healthcare insurance through their employer compared with 49% of bipolar I subjects (p< 0.001). More than three-fourths (77%) of normative subjects had health maintenance or preferred provider organization insurance coverage compared with 55% of bipolar subjects (p< 0.001). Significantly higher percentages of bipolar I subjects versus normative subjects received Medicare coverage (7.3 vs. 1.5%; p=0.005) or were uninsured (22.4 vs. 6.6%; p< 0.001).

Workplace productivity

A significantly higher percentage of normative subjects versus bipolar I subjects reported that they were currently receiving pay for work (86.9 vs. 73.5%; p< 0.001). In the bipolar I group, 39 (17.8%) subjects reported that they did not receive pay for work or participate in volunteer work in the previous 2 weeks, compared with 18 (9.1%) subjects in the normative group (p=0.01). Normative subjects were significantly more likely than bipolar I subjects to report that they had not missed any work in the previous week (49.4 vs. 36.0%; p=0.01). Normative subjects who did report missing work were more likely than bipolar I subjects to report that the reason was a holiday or vacation (33.1 vs. 21.7%; p=0.017). Bipolar subjects who reported missing work during the previous week were significantly more likely than normative subjects to report that they missed work because they were too upset, depressed, or nervous (30.6 vs. 1.7%; p< 0.001) or physically ill (19.0 vs. 10.2%; p=0.021). In terms of the overall score from the EWPS, a note on scale creation is necessary. According to scaling conventions, EWPS scores were not computed for 42 bipolar subjects and 20 normative subjects because these subjects either failed to answer at least one-third of the required EWPS questions or self-reported not working or receiving work-related insurance payments the week before the survey.

Bipolar I subjects reported higher scores (i.e., greater negative effect on workplace productivity) on the EWPS than normative subjects (Table 2). The mean EWPS score for the bipolar I group was 37.2 compared with 15.8 for the normative group (p< 0.001). The EWPS had high internal consistency for both subject groups, as indicated by Cronbach's alpha (0.94 for bipolar I subjects; 0.91 for normative subjects).

Table 2. Results from Endicott Workplace Productivity Scale.

Productivity	Bipolar group n=219	Normative group n=198	p-value
Overall score
n	177	178
Mean±SD	37.2±19.2	15.8±11.1	< 0.001
Range	2.1–84.0	0.0–70.0
Cronbach's alpha	0.94	0.91
Hours of work
Usual number of hours per week
n	180	180
Mean±SD	39.3±12.5	41.8±11.6	0.051
Range	4.0–90.0	0.0–95.0
Number of hours worked during the previous week
n	180	180
Mean±SD	35.9±15.4	40.4±15.3	0.005
Range	0.0–85.0	0.0–95.0

As stated earlier, subjects in both groups held jobs associated with a minimum of 20 hours per week. When asked to estimate the usual number of hours worked, normative subjects reported an average of 41.8 hours per week and bipolar I subjects reported an average of 39.3 hours per week (p=0.051). When asked the number of hours worked in the previous week, bipolar I subjects reported working an average of 35.9 hours, compared with normative subjects' reported average of 40.4 hours (p=0.005).

Employment issues

Bipolar I subjects were significantly more likely than normative subjects to report working a reduced schedule for a medical reason (12.4 vs. 2.4%; p< 0.001) (Table 3). None of the normative subjects reported working a reduced schedule because of mental health issues compared with 23% of bipolar I subjects (p< 0.001). A significantly higher percentage of bipolar I subjects reported having been fired or laid off from a job (66.2%) than normative subjects (38.9%; p< 0.001). Of subjects who reported being laid off or fired, significantly more bipolar I subjects said the reason for the situation was that their supervisor was not happy with their work, behavior, or mental attitude (70.4 vs. 18.9%; p< 0.001).

Table 3. Results from employment questions.

Question asked of participants	Bipolar group n=219		Normative group n=198		p-value
	Valid n	n answering ‘yes’ (%)	Valid n	n answering ‘yes’ (%)
Are you working a reduced schedule for pay for a medical reason?	170	21 (12.4)	170	4 (2.4)	< 0.001
Are you working a reduced schedule for pay for a mental health reason?	174	40 (23.0)	173	0 (0)	< 0.001
Have you ever been fired from a job or laid off?	219	145 (66.2)	198	77 (38.9)	< 0.001
If yes, was this because your supervisor was not happy with your work, behavior, or mental attitude?	142	100 (70.4)	74	14 (18.9)	< 0.001
Have you received short- or long-term disability benefits for a medical reason?	218	73 (33.5)	198	42 (21.2)	0.005
Have you received workers' compensation benefits for any reason?	218	44 (20.2)	197	29 (14.7)	0.144
Have you been in jail, arrested, or convicted of a crime other than drunk driving?	219	64 (29.2)	198	14 (7.1)	< 0.001

No significant difference was observed between the percentage of subjects in each group who reported having received workers' compensation benefits for any reason (20.2% bipolar group vs. 14.7% normative subjects; p=0.144) (Table 3). However, a significantly higher percentage of bipolar I subjects than normative subjects reported receiving short-term or long-term disability benefits for a medical reason (33.5 vs. 21.2; p=0.005).

In the bipolar I group, 54 (29.2%) subjects reported that they had been jailed, arrested, or convicted of a crime other than drunk driving, compared with 14 (7.1%) normative subjects (p< 0.001) (Table 3).

Healthcare resource utilization

Bipolar I subjects reported an average of 1.7 visits to the psychiatrist during the previous month compared with 0.1 visits for normative subjects (p< 0.001) (Table 4). Subjects were not asked to exclude their screening visit for this study when estimating their resource utilization. Bipolar I subjects reported an average of 1.2 visits to other physicians or doctors in the previous month. Although this number was not significantly different from the mean number of visits for the normative group (1.3; p=0.41), the screening visits for this study are included in the resource utilization estimates for the normative group but not the bipolar group because primary care physicians recruited the subjects during a regularly scheduled visit. Bipolar I subjects reported significantly more frequent visits to a psychotherapist (0.67 vs. 0.05; p< 0.001) compared with subjects in the normative group. Although few emergency department visits were reported by either group, the average number of visits by bipolar I subjects (0.14) was significantly higher than the average reported by normative subjects (0.06; p=0.03).

Table 4. Results from patient-reported resource utilization questions.

Resource utilization	Bipolar group n=219	Normative group n=198	p-value
Number of visits to a psychiatrist in the previous 30 days
n	218	198
Mean±SD	1.7±1.5	0.08±0.40	< 0.001
Range	0–11	0–3
Number of visits to other physicians in the previous 30 days
n	215	198
Mean±SD	1.2±1.6	1.3±1.1	0.413
Range	0–9	0–9
Number of visits to a psychotherapist in the previous 30 days
n	218	198
Mean±SD	0.7±1.6	0.1±0.3	< 0.001
Range	0–14	0–2
Number of visits to an emergency department in the previous 30 days
n	218	198
Mean±SD	0.1±0.5	0.1±0.3	0.027
Range	0–3	0–2

Based on results from the eCRF, 68% of bipolar I subjects had taken at least one medication, including antidepressants, benzodiazepines or other antianxiety medications, and hypnotics, compared with 14% of normative subjects (p< 0.001). Seventy-six percent of bipolar I subjects had a history of antimania medications compared with 1% of normative subjects (p< 0.001). Antimania medications taken by bipolar I subjects included anticonvulsants (46%) and atypical antipsychotics (38%). Frequently reported medications taken by bipolar I subjects included antidepressants (52%) and benzodiazepines and other antianxiety medications (27%).

Evaluation of administration mode differences

Because the 14-day questionnaire was offered for completion by telephone or the Internet, an evaluation of potential administration mode effects is warranted. Significant differences in the preferred mode of survey completion were noted, with bipolar I subjects preferring the telephone over the Internet (Table 5). Specifically, 58% of the bipolar I group completed a telephone survey versus 42% of the normative group (p=0.012). No statistically significant differences in administration mode were noted for gender, age, or education level. However, subjects who completed the survey using the Internet had significantly higher mean total scores on the EWPS (mean score = 28.6) than those who completed the survey via telephone (mean score = 24.3; p=0.032). Because of these administration mode differences, a dichotomous variable (telephone vs. Internet) was included in the multivariate analysis to control for selection effects.

Table 5. Comparison of survey administration modes.

Measure	Statistic	Telephone n=204	Internet n=213	p-value
Study cohort
Bipolar I disorder	n (%)	120 (58.82)	99 (46.48)	0.012
Normative	n (%)	84 (41.18)	114 (53.52)
Gender
Male	n (%)	72 (35.29)	57 (26.76)	0.059
Female	n (%)	132 (64.71)	156 (73.24)
Age	Mean±SD	40.3±10.85	38.6±10.55	0.061
Education
High school or less	n (%)	53 (25.98)	50 (23.47)	0.553
Some college	n (%)	151 (74.02)	163 (76.53)
EWPS score	Valid n	176	179
	Mean score±SD	24.3±15.8	28.6±21.5	0.032

EWPS indicates Endicott Workplace Productivity Scale.

Multivariate analyses

A regression analysis demonstrated that being in the bipolar I group was associated with a significantly higher score (greater effect on productivity) on the EWPS. Approximately 32% of the variance in productivity scores was explained by this variable alone (results not shown but available on request). In the adjusted model (Table 6), the results for age, gender, and race did not have a statistically significant effect on work productivity scores. Survey administration mode was significantly related to EWPS scores when controlling for other variables in the model. Neither the presence of at least one significant medical condition nor a history of medication use had a significant effect on EWPS score. Significant interaction terms included being in the bipolar I group and choosing to complete the survey via the Internet (b=12.41; p< 0.001) and being in the bipolar I group and reporting at least some college when asked about education level (b=–13.91; p< 0.001). Overall, the variables in Table 6 explained approximately 42% of the variance in EWPS scores.

Table 6. Regression analysis: EWPS results including significant interaction terms (n=417)

Variable	Coefficient (β)	p-value
Intercept	15.64
Bipolar I group (vs. normative)	12.45	0.028
Age	–0.10	0.20
Male (vs. female)	2.20	0.21
Nonwhite race (vs. white race)	3.18	0.13
Some college (vs. high school or less)	1.85	0.479
Internet survey (vs. telephone survey)	5.99	0.94
At least one medical condition (vs. no medical condition)*	3.16	0.06
History of medication use (vs. no medication use)^†	3.21	0.11
Interaction: bipolar I group and some college	–13.91	< 0.001
Interaction: bipolar I group and Internet survey	12.41	< 0.001

*Medical conditions include cancer, hypertension, hypercholesterolemia, cardiovascular disease, diabetes, arthritis, autoimmune disorder, neurological disorder, epilepsy, or other seizure disorders.

^†Medications include antidepressants, benzodiazepine or other antianxiety medications, hypnotics, and other medications not including antimanics. EWPS indicates Endicott Workplace Productivity Scale.

Discussion

A group of bipolar I subjects who were receiving psychiatric treatment reported significantly lower workplace productivity and poorer employment issues, as well as a significantly greater consumption of healthcare resources, than an age- and gender-matched group of control subjects. When multivariate analyses were conducted, being in the bipolar I disorder group was associated with significant lower work productivity scores, even after adjusting for demographic and medical history measures. Bipolar I subjects also reported more problems maintaining employment. Nearly one-third of subjects in the bipolar I group who reported missing work during the previous week did so because they felt upset, depressed, or nervous, compared with less than 2% of subjects in the normative group. Bipolar I subjects were likely to have difficulty obtaining and maintaining private, employer-sponsored healthcare coverage. The results of this study suggest that bipolar I disorder is associated with difficult interpersonal relationships at work, particularly with superiors.

The two subject groups were matched by age and gender. Few significant differences were noted between subject groups in terms of their demographic characteristics. However, blacks were significantly more numerous in the normative group. Bipolar I subjects were more likely than normative subjects to complete the survey via telephone; bipolar I subjects were also significantly less likely than normative subjects to have employer-sponsored healthcare insurance.

Global and national estimates have shown that unlike with unipolar depression, which is more prevalent among females, the prevalence of bipolar I disorder does not differ by gender¹³. The GBD reports that number of years lived with disability from bipolar disorder are similar for men and for women, whereas the disability burden of unipolar depression is 50% greater for women than for men². For this study, psychiatrists were not provided with recruitment targets for bipolar I disorder subjects. The study population was nearly two-thirds female. A retrospective study using administrative healthcare claims data from insured and treated bipolar subjects also identified a population consisting of 66% females, and the authors of that study speculated that women may be more likely than men to seek healthcare services for bipolar disorder¹⁴. This assumption is consistent with findings from the National Ambulatory Medical Care Survey in the 1990s. Results from that study found that the number of psychiatric visits by women with bipolar disorder was more than double those for men¹⁵. Evidence suggests that women with bipolar disorder experience more frequent and longer depressive episodes than men experience¹⁶. Because subjects are more likely to seek treatment for depressive episodes¹⁷, more severe depressive episodes associated with female bipolar subjects may have contributed to the oversampling of women in this study population of bipolar I disorder subjects who are under the care of a recruiting psychiatrist.

The results of this study suggest that people with bipolar I disorder who are both employed and under the supervision of a psychiatrist experience severe limitations in the workplace that are likely related to their mental health condition, Additionally these patients consume more healthcare resources than age- and gender-matched primary care patients. These results support previous research demonstrating an association between bipolar disorder and significant functional disability and a negative impact of this disease on workplace productivity. Kessler et al ⁵ found that bipolar disorder was associated with an average of 65.5 lost workdays per year compared with 27.2 lost workdays per year associated with unipolar depression. A prospective employer study found that although employees with bipolar I disorder made up only 0.3% of their workforce, each patient cost nearly US$7000 per year more in health benefits and health-related absences than the average employee without this condition¹⁸. Gitlin et al ⁴ found that bipolar I disorder was associated with significant mood pathology and reduced psychosocial function during episodes of relapse as well as periods between relapses. Simon et al ¹⁹ report that depressive episodes of bipolar disorder are associated with unemployment and absenteeism, but mania and hypomania have more variable effects on productivity.

Having bipolar I disorder translates into greater medical resource utilization in the form of more office visits to psychiatrists and psychotherapists, as well as more trips to the emergency department. Although outpatient visits are an important component of bipolar treatment, economic models have indicated that the primary cost driver of direct costs for bipolar disorder is hospitalisations^7,10. Bipolar disorder associated with relatively frequent episodes can lead to hospitalizations that could potentially disrupt a patient's employment status. One study reported that subjects with bipolar I disorder had a higher average total cost per subject for all medical encounters and a higher number of prescriptions for central nervous system drugs than both the all-bipolar group and the non-bipolar group²⁰.

The results of this study should be considered in the context of its limitations. First, selection bias may have influenced the results. Although the study groups were matched by age and gender, other covariates may have influenced the results. This study is observational, and recruiting physicians and subjects were selected using a nonrandomized sampling procedure. No adjustments were made for state or regional variations in the availability of insurance or employee unionization. As noted earlier in this section, female subjects were probably overrepresented in this study. No data were collected from patients describing their tenure, occupation, or industry. All subjects included in the bipolar I group were being treated by psychiatrists and employed; therefore, results are not representative of the total population of bipolar I subjects. The authors were concerned that the results could be affected by undiagnosed bipolar I patients in the normative group; however, the results of the MDQ demonstrated that the majority of patients in the bipolar I disorder group screened positively for a mood disorder diagnosis compared with only 1% of normative patients.

Second, recall bias is also a limitation inherent in questionnaire-based studies. Subjects were asked to consider the previous week when answering questions about workplace productivity and the previous 30 days when asked to consider resource utilization.

Third, the effect of different technologies for data collection is unclear. Other studies have found no difference in the results of surveys administered via telephone or the Internet^21,22. In contrast, the authors found that bipolar I subjects were significantly more likely to complete the survey using a telephone and that, across both groups, subjects who completed the survey using the Internet were more likely to report greater productivity loss than were those who completed the survey using the telephone. Multivariate analyses were conducted to control for the effect of survey administration mode on productivity results.

The results of this study suggest that bipolar I disorder was associated with significant losses in work productivity and significant employment issues as well as significantly greater resource utilization. Subjects with bipolar I disorder were less likely to be in a position to obtain healthcare coverage through their employer than a matched group of primary care subjects without a major psychiatric disorder. Further studies are needed to provide a link between effective treatment and outcomes.

Physicians, employers, and healthcare decision makers may be able to improve workplace productivity and employment issues and reduce healthcare resource utilization in patients with bipolar I disorder by optimizing treatment and targeting bipolar disorder with disease management programs.

Acknowledgments

Declaration of interest: This study was funded by Ortho-McNeil Janssen Scientific Affairs, LLC, Titusville, NJ, USA. At the time of data collection, B.J.M. has disclosed that she was an employee of i3 Innovus, Ingenix, a company that received funding from Ortho-McNeil Janssen to conduct this research. K.E.D. has disclosed that she is an employee of Relevant Health Outcomes, who received funding from Ortho-McNeil Janssen to conduct this study. J.M.P. and R.D. have disclosed that they are stockholders and employees of Ortho-McNeil Janssen, which is part of Johnson & Johnson, Inc. The authors wish to acknowledge the technical and editorial support provided by Matthew Grzywacz, PhD, Helix Medical Communications.