Results of the Brief Jail Mental Health Screen Across Repeated Jail Bookings

The mean±SD number of bookings per person in our analytic sample was 3.35±2.53 (range 2–65) during the study period, the mean length of stay was 21.42±50.59 days (range 0–1,049), and the mean time between bookings was 846.15±288.37 days (range 0–1,276). The sample was 34.8% (N=4,362) Caucasian, 65.2% (N=8,162) non-Caucasian, 65.1% (N=8,151) African American, and 0.1% (N=11) Asian. The mean age at first booking during the study period was 31.50±11.57 years (range 16–79). Most detainees (81.7%, N=10,232) were men. About 8.0% (N=1,002) of detainees reported homelessness, and 40.0% (N=4,994) screened positive for problematic substance use. At their initial booking, 62.9% (N=6,600) of the sample were charged with more serious offenses (typically felonies punishable by imprisonment for 1 year or more, such as assault, burglary, and sexual offenses), whereas 37.1% (N=3,886) were charged with less serious offenses (typically misdemeanors punishable by imprisonment for less than 1 year, for example, stolen property and disorderly conduct). The most common charges at initial booking included assault (21.2%, N=2,654), drugs or narcotics related (15.3%, N=1,916), and larceny or theft (11.5%, N=1,440).

In total, 23.3% (N=2,900) of detainees screened positive overall on the BJMHS at initial booking during the study period. Rates of positive screens for each decision rule at initial booking ranged from 2.7% (N=340) for current psychiatric symptoms to 12.7% (N=1,446) for current psychiatric medication. Visual exploration of the descriptive statistics suggested that the frequency of positive screens may be changing over time; however, rates and directions of change seemed to differ across items and decision rules (Figure 1). A breakdown of positive responses to individual BJMHS items and to BJMHS decision rules by booking number can be seen in Table 1.

Table 2 presents the estimates for each of the five multilevel logistic models. Results of the first multilevel logistic model confirmed that the odds of a positive screen overall (i.e., meeting at least one of the decision rules) increased with each additional booking into jail. Specifically, each additional booking was associated with an increase in the odds of a positive screen (odds ratio [OR]=1.02, 95% confidence interval [CI]=1.01–1.04). Results of the model including covariates showed the same pattern [see online supplement].

Results of the multilevel logistic models examining each of the decision rules showed that the odds changed across repeated bookings for three of the four rules (Table 2). The odds of screening positive for prior hospitalization and for reporting two or more current psychiatric symptoms decreased over time. With each additional booking, the odds of screening positive for prior hospitalization decreased significantly (OR=0.96, 95% CI=0.95–0.98) as did the odds of screening positive for current symptoms (OR=0.95, 95% CI=0.93–0.98). In contrast, detainees were more likely to be referred for any other reason (at the assessor’s discretion) with each additional booking into jail. For this decision rule, each additional booking was associated with an increase in the odds of referral (OR=1.03, 95% CI=1.02–1.04). Finally, the odds of screening positive for current use of psychiatric medication did not change across bookings. As before, results of the models including covariates showed the same patterns of results [see online supplement].

Detainees were less likely to screen positive for current psychiatric symptoms across repeated bookings. On the one hand, it is possible that a decrease in the reporting of symptoms reflects improvements in illness trajectories. Indeed, as discussed earlier, jails may represent a critical point of identification, referral, and initiation of psychiatric treatment for many adults with serious mental illness. On the other hand, changes in responses to the symptom items across bookings may not necessarily represent improved illness trajectories. Instead, different responses may reflect natural fluctuation, but not necessarily marked improvement, in clinical presentation (31). Furthermore, symptoms not queried by the BJMHS, such as symptoms associated with anxiety disorders, may have worsened over time and gone undetected by the BJMHS (32).

Our findings also may be attributable to problems associated with the self-report nature of the items, such as differences in insight or willingness to disclose information. Indeed, the question about hospitalization on the BJMHS queries lifetime hospitalization. The fact that detainees were less likely to screen positive for lifetime hospitalization across repeated bookings cannot indicate that lifetime hospitalization rates decreased. Rather, this indicates that reporting of hospitalization decreased among detainees who had previously reported being hospitalized; in other words, a small proportion of detainees (0.5%, N=63) recanted their prior hospitalization. This calls into question the reliability of this item across repeated administrations. Alternatively, it is possible that jail staff skipped this question at some bookings or forgot to record detainees’ responses. It is also possible that changes in this item are due to data-entry errors. These possibilities are important areas for further investigation.

To further illustrate the susceptibility of BJMHS items to reporting issues, research demonstrates that some people find repeated screening or assessment to be therapeutic, possibly increasing the likelihood that they will self-report symptoms, whereas others find it frustrating and suspicious (33). The jail mental health screening process is unlikely to confer positive therapeutic effects. Instead, it seems more realistic to conclude that having to answer the same questions across multiple bookings led detainees to feel that staff are not paying attention to their answers or that staff are trying to catch them giving conflicting answers.

Detainees may grow tired of repeating the same responses across multiple screens and may become less likely to report current symptoms or prior hospitalization. Furthermore, detainees may be fearful that disclosing symptoms will deny them pretrial release, lead to involuntary hospitalization, or affect their case in some way (33). Finally, detainees may be especially less likely to disclose mental health information if they feel that the disclosure of this information is not being used in a meaningful way (33), a strong possibility given the limited availability of treatment in jails generally (34) and the study jail, specifically.

In spite of the decreases in self-reported symptoms and hospitalization, the odds of a positive screen overall increased across bookings. This increase appears to be driven by increases in discretionary referral for “any other reason.” This may indicate that jail staff do not rely on the results of the other three decision rules when making referral decisions. In this study, the BJMHS detected current symptoms only in about 2.0% of detainees, whereas other studies have placed estimates of serious mental illness at 14.5% for men and 31.0% for women (2). Staff may not trust the screening results because of the incredibly low detection rate, which may explain why referral rates for staff overrides increased over time even though positive screens for the other three decision rules decreased or held steady. At the same time, however, this discretion is built into the BJMHS itself because one of the decision rules dictates that detainees should be referred “if [staff] feel it is necessary for any other reason” (8).

Thus, it may not be that staff do not trust the screening results but rather that they do not trust the detainee self-report. Such a “professional override” is included in many instruments designed for use in correctional settings (35). In this way, assessors are using the BJMHS as designed. Whether this rate of professional override is acceptable is questionable; no guidance or research exists to support the acceptable rate of professional override on this or any other instrument. This is an important avenue of investigation in relation to the BJMHS and more generally.

It is also possible that symptoms not included on the BJMHS increased across repeated bookings and were identified by jail staff, resulting in referral. Alternatively, staff may become familiar with a detainee across repeated bookings and refer on the basis of past rather than current clinical presentation. The familiarity hypothesis suggests that an evaluator will gain additional information with each subsequent administration of a test, which leads to improved assessment accuracy (28). However, this hypothesis has not always held when tested within the context of risk assessments (28, 36) and, thus, warrants further examination in the context of jail mental health screening. Indeed, confirmation bias suggests that such prior knowledge, or familiarity, may result in decreased screening accuracy over time.

It is also possible that changes in screening results across bookings reflect the limited training that jail staff receive before implementation of the BJMHS. There is no formal training required for the BJMHS, which may affect the fidelity with which the screen is administered and scored. Establishing a standardized training protocol, including booster training, should improve adherence to items and decision rules. Finally, changes in screening results may be due to poor reliability of the BJMHS items. A review of BJMHS studies demonstrated highly variable findings across studies (37), calling into question the tool’s accuracy and consistency in detecting mental health problems. Jails may be better served by other screening measures such as the Correctional Mental Health Screen or the Jail Screening Assessment Tool, which have been demonstrated in other studies to outperform the BJMHS in terms of detecting mental health problems among jail inmates (14, 38).

A few limitations of the study design should be noted. First, we do not have information on the number of times a person was booked into this jail before our study period. As such, our first BJMHS results may not be the first time they were screened. Second, because the BJMHS is the only mental health screen administered in this jail and follow-up with an in-depth mental health evaluation was inconsistent, we were unable to compare results with another screening tool or against a criterion measure, such as the SCID-IV or clinical diagnosis. For these reasons, we cannot determine whether the observed changes reflect actual changes in mental health status, issues with fidelity of administration, issues of disclosure and reporting of symptoms, characteristics of the BJMHS itself, or otherwise. Third, we did not have access to treatment data and cannot speak to the treatment-related or other mechanisms through which BJMHS results changed. Fourth, we investigated changes in mental health screening results across repeated jail bookings using one tool in one jurisdiction; generalizability of findings must be examined in future research. Despite these limitations, the study’s strength lies in our longitudinal analysis of results of a widely used mental health screening tool in a large jail population. This study is the first of its kind.