Between July 2003 and November 2004, a total of 597 individuals were detained at the Ottawa County Jail. The MDQ was completed by 526 detainees (350 male, 80 female, and 96 gender not indicated) as part of jail standard operating procedures. The remaining 71 detainees either refused or provided an incomplete questionnaire. Of 526 detainees who completed the MDQ, 37 (7%) screened positive for bipolar disorder. After the booking process was completed, 164 subjects (142 male, 22 female) volunteered to participate in a research evaluation where the MINI was administered. The mean age (SD) of the research subjects was 33.4 (10.9) years old. The ethnicity of the sample was 87.8% (n = 144) Caucasian, 9.1% (n = 15) African-American, 3.0% Hispanic (n = 5), and 0.6% (n = 1) other, which was reflective of the respective county.
A MINI primary diagnosis of bipolar disorder was assigned to 55 (33.5%) inmates (bipolar I: N=39, bipolar II: N=13, bipolar disorder not otherwise specified (NOS): N=3). Of the inmates screening positive on the MDQ, 6 subjects were not found to have bipolar disorder according to the MINI. Among these subjects, 4 were diagnosed with schizoaffective disorder and 2 were not found to have a diagnosable Axis I disorder.
Limited socio-demographic data regarding age, ethnicity, and gender was available on the larger sample of 597 jail detainees who completed the MDQ as part of jail standard operating procedure as presented in . Compared to the 164 subjects who volunteered to participate in the diagnostic interview, no differences were observed in age or gender. A greater number of African-American subjects agreed to participate in the diagnostic interview, though the statistical significance of this association should be interpreted with caution given the small number of subjects available for analysis in this subgroup. The complete socio-demographic characteristics of the volunteer sample in relation to MDQ scores are presented in . No differences in age, educational level, or marital status were observed between those screening positive on the MDQ and those screening negative. Anxiety disorders, both lifetime and current, were more prevalent in those screening positive (p < .001). Similarly, rates of current substance use disorders (p = .003) and rates of lifetime substance abuse (p = .024) and dependence (p = .003) were more prevalent in the MDQ positive cohort. Comparisons of anxiety and substance use disorders are shown in . Arrests for crimes against persons, including domestic violence, rape, assault, and violation of a protection order were more common in the MDQ positive group (p = .017). No differences were found between groups in the rates of property crimes (e.g. robbery, breaking and entering, or trespassing), disorderly conduct, or drug violations. A history of physical (p = .003) and emotional (p = .023) abuse, but not sexual abuse (p = .167), was more common among subjects screening positive for bipolar disorder.
Sociodemographic Characteristics of Inmates Refusing Versus Consenting to Participate in a Research Diagnostic Evaluation
Sociodemographic Characteristics Versus MDQ Results
Substance Use and Anxiety Disorder Co-Morbidity Versus MDQ Results
Sensitivity, specificity, and predictive values of the MDQ were calculated. The MDQ correctly screened out 103 of 109 individuals not meeting the MINI criteria for lifetime bipolar disorder, resulting in a specificity of 0.94. However, the MDQ correctly identified only 26 of 55 individuals meeting the MINI criteria for bipolar disorder, resulting in a sensitivity of 0.47. The positive predictive value of the MDQ was 0.81 and the negative predictive value was 0.78. shows the ROC curve of the MDQ in this sample and shows the sensitivity and specificity of the MDQ plotted across different scoring thresholds. Sensitivity indices were also calculated based upon the bipolar disorder subtypes I or II. The MDQ did not identify any of the 3 patients who were found to have BP NOS. Among the 39 individuals with BP I, 23 were correctly identified by the MDQ, while only 3 of 13 subjects with BP II were correctly identified. The difference in sensitivity for detecting BP I (0.59) versus BP II (0.23) trended toward significance (p = 0.052). The sensitivity of the MDQ for detecting BP I (0.59) was significantly better than for detecting BP II/NOS combined (0.19; p = .008).
a A score of ≥ 8 (vertical line) was identified as the optimal cutoff.
In an attempt to improve the sensitivity for detecting BP II, we modified the scoring algorithm by adapting the MDQ from a 13-item to a 15-item questionnaire. The items measuring symptom severity and symptom co-occurrence were converted into dichotomous measures. In this manner, any patient endorsing 7/15 items would be considered to have a positive MDQ test. The scoring modification allowed for the detection of 13 additional cases of bipolar disorder. The sensitivity for detecting BP II more than doubled, increasing from 0.23 to 0.54, and the overall sensitivity for detecting bipolar disorder improved modestly from 0.47 to 0.55, with minimal decrease in specificity (0.94 to 0.84).
Among the 29 subjects identified as false negatives (i.e. did not screen positive for bipolar disorder on the MDQ but were subsequently found to have bipolar disorder on the MINI), comparison of responses on the MDQ and the MINI revealed that 28/29 subjects failed to endorse that manic or hypomanic symptoms co-occurred or resulted in moderate to severe problems. This suggests that requesting patients with bipolar disorder to recall lifetime symptom co-occurrence or functional impairment on a screening questionnaire may be a challenging task. Thus, we undertook a separate analysis to determine if eliminating these 2 items substantially altered the operating characteristics of the MDQ. Despite this modification, the sensitivity remained unchanged at 0.47 but the specificity decreased from 0.94 to 0.88. An ROC curve was constructed to determine whether a different scoring threshold might improve the detection of bipolar disorder (see and ). Interestingly, the optimum combination of sensitivity (0.71) and specificity (0.68) was obtained with a positive response to ≥ 3 of the 13 core mood questions, resulting in an Area under the Curve value of 0.75.
a ROC curve generated using a modified scoring algorithm for the MDQ that did not require subjects to indicate that the symptoms occurred during the same time period and caused moderate or serious problems.
a A score of ≥ 3 (vertical line) was identified as the optimal cutoff.