Misdemeanor Charges Among Individuals With Serious Mental Illnesses: A Statewide Analysis of More Than Two Million Arrests

Individuals with mental illnesses are overrepresented in all stages of the criminal legal system in the United States. Approximately 1–2 million people with serious mental illnesses are jailed each year (1, 2); detention can exacerbate symptoms, disrupt treatment and connections to services and social supports, and result in myriad collateral consequences (3, 4). Reducing the number of persons with serious mental illnesses in jails through interventions at the point of police contact has become an important focus of practice and policy (5). Most of what is known about the relationship between mental illness and involvement with the criminal legal system has focused on individuals who are incarcerated; less is known about law enforcement officers’ charging practices during encounters (6). Wide variation in methodologies (7) makes it difficult to determine the true arrest rate among people with mental illnesses compared with the general population. Research examining complete state-level arrest data would be informative.

Misdemeanor offenses account for approximately 80% of the 10.5 million arrests in the United States annually (5, 8); thus, it is unsurprising that studies report much higher numbers of misdemeanor than felony arrests among those with mental illnesses (6, 9–11). In Constantine and colleagues’ analysis of a jail cohort (9), for example, detained individuals had an average of 2.5 misdemeanor arrests and 1.4 felony arrests. In a study by Fisher et al. in Massachusetts (11), mental health service recipients were 3.7 times and 2.5 times more likely than the general Massachusetts population to be arrested for felony crimes against persons and property, respectively, and 4.2 times more likely to be arrested for misdemeanor crimes. The need for alternatives to arrest and incarceration is underscored by a growing literature on the unmet health and social needs of people who experience repeated contact with the criminal legal and mental health systems (12, 13).

Although the problem of overrepresentation of individuals with mental illnesses in the criminal legal system is well established, few studies have examined the specific charges among those with mental illnesses in comparison with the general population. We addressed this gap by exploring specific types of charges that are most common in arrests involving individuals with serious mental illnesses. We used New York State arrest data spanning 4 years. In a previous report (6), we found that an indicator of serious mental illness was associated with a >50% increase in the odds of a jail sentence for misdemeanor arrestees, after controlling for other case characteristics. Conversely, the indicator was unrelated to likelihood of a prison sentence in cases with a felony arrest. Here, we compared charges (examining the most serious charge within each arrest event) involving individuals with and those without the serious mental illness indicator in terms of percentage of arrests for a violent crime, across penal code classifications (classes of felonies and misdemeanors) and across Uniform Crime Reporting (UCR) codes; we also ranked the most common specific charges.

Methods

Arrest Data

The data set included all arrests for a misdemeanor or felony in New York State in 2010–2013 (6). In the current, event-based analysis, all arrests were included, meaning unique individuals could contribute multiple events to the data. Data came from the state’s Division of Criminal Justice Services’ Computerized Criminal History database. For arrests involving multiple charges (e.g., third-degree criminal trespass, a class B misdemeanor, along with third-degree assault, a class A misdemeanor), only the most serious charge (the assault charge, in this example) is reported on here.

Arrests were categorized and examined in four ways. First, the most serious charge within each arrest was categorized as violent or nonviolent, with violence defined as offenses related to assault, strangulation, homicide, sex offenses, kidnapping or coercion, robbery, firearms and dangerous weapons, and terrorism. This categorization overlapped substantially with the statutorily defined “violent felony offenses,” although with some differences (e.g., violent felony offenses include some types of burglaries, which were not included in our measure, and our measure included many lower-level crimes of violence that are not statutorily defined violent felonies).

Second, the most serious charge within each arrest was categorized into a felony or misdemeanor penal code class. Misdemeanors are offenses other than traffic infractions or nontraffic violations for which any imposed imprisonment sentence is 15–364 days. A felony is an offense for which a prison sentence of ≥1 year may be imposed. Table 1 defines felony and misdemeanor classes.

Third, the most serious charge within each arrest was classified with UCR codes from the U.S. Federal Bureau of Investigation UCR program. Arrests were categorized into 48 UCR codes (e.g., arson, burglary, drug offenses, larceny-theft, simple assault, and weapons offenses). Finally, we examined specific charges to identify the most common charges; however, as noted above, our data included only the most serious charge within each arrest.

Indicator of Serious Mental Illness

We used the same two sources of mental health data as our previous report (6): Medicaid billing records for mental health services and state hospital admission and discharge records from the state Office of Mental Health’s Mental Health Automated Record System. The time frame for the mental health data was 2009–2013. This sample of approximately 1 million mental health service recipients was matched to the arrestee sample by using probabilistic matching software that matched multiple identifiers, including name, date of birth, and Social Security number. Serious mental illness was indicated by any of the following diagnoses in a treatment episode in the 2 years before the individual’s arrest: schizophrenia, schizoaffective disorder, other psychotic disorders, bipolar disorder, and major depressive disorder. Data were originally compiled for operational purposes and evaluation. Personal identifiers were deleted before this analysis, so the institutional review board of the New York State Psychiatric Institute determined that the project did not meet human subject research criteria.

Data Analysis

The current analysis utilized event-based data. As mentioned above, only the most serious charge within each arrest was represented, and that charge could have entered the record at arrest, arraignment, indictment, or disposition. Individuals involved in offenses classified only as violations in New York State (e.g., loitering, disorderly conduct, or harassment) are not usually fingerprinted, so these cases were not included in the data. The analysis included 2,224,847 arrests: 91,363 (4.1%) involving individuals with the indicator of serious mental illness and 2,133,484 (95.9%) involving those without the indicator. The provided deidentified data set included only a frequency run of all charges among those with the indicator and a frequency run among those without the indicator. As such, we report raw frequencies and percentages as well as simple comparisons of proportions of charge types, analyzed by using chi-square tests. Multivariable models—and multilevel analyses that would account for the clustered nature of the data (e.g., within individuals, within jurisdictions, or within counties)—were not possible because these variables were not included within the data set provided. Therefore, the figures reported are purely descriptive.

The first analysis compared the proportion of arrests for a violent crime among those with and those without the indicator. Second, we analyzed differences in penal code classifications among arrests of individuals with and of those without the indicator. In the third analysis, we examined UCR codes. Among 48 UCR codes, those with a very low prevalence (<2% in the overall sample of arrests, e.g., arson, embezzlement, extortion, manslaughter, and prostitution) were excluded; accordingly, we focused on 14 UCR codes. Fourth, the most serious specific charges were ranked separately for arrests of individuals with and of those without the indicator of serious mental illness. All charges with a frequency of >1% in each group were included.

Results

Violent Offenses and Penal Code Classifications

Among arrests involving individuals with the indicator of serious mental illness, 7.3% (N=6,715) were for a violent offense, compared with 7.6% (N=163,069) involving individuals without the indicator. As shown in Table 2, among 10 classes of felonies and misdemeanors, class B felonies and class A misdemeanors were more likely in arrests involving those with the indicator than in arrests involving those without the indicator, whereas the other classes of offenses (except class A-I felony, reducible) were more frequent in arrests involving individuals without the indicator. The differences in percentages were largest for class A misdemeanors, unclassified misdemeanors, and class B felonies. Class A misdemeanors, comprising >50% of arrests in both groups, was the offense class with the largest disparity between arrests involving those with and those without the serious mental illness indicator.

UCR Codes and Most Common Charges

Table 3 shows the 14 UCR codes that occurred at a frequency of ≥2% in the overall sample. Percentages of each code were statistically significantly different between the two groups (which was unsurprising given their large sample sizes), except for robbery and burglary. Seven codes were more common among those with the serious mental illness indicator, including larceny-theft, fraud, criminal mischief, and aggravated assault. Codes that were more common in arrests involving individuals without the indicator included possession of marijuana, driving under the influence of alcohol, and those pertaining to dangerous weapons.

Table 4 shows the rank order of specific charges (the most serious charge within each arrest event) with a frequency of >1% in arrests involving those with and those without the serious mental illness indicator. Although many charges appeared for both groups, the two groups showed some differences. Four charges (i.e., operating a vehicle with ≥0.08% blood-alcohol content, first offense; driving while intoxicated, first offense; criminal possession of a weapon, fourth degree; and criminal possession of a controlled substance, third degree, narcotic, intent to sell) were among the most common charges in arrests involving individuals without the indicator but were not among the most common charges in arrests for those with the indicator. On the other hand, two criminal trespass charges (second and third degree) that were among the most common charges in arrests involving individuals with the serious mental illness indicator did not appear on the list of most common charges in arrests involving individuals without the indicator.

Discussion

We highlight four noteworthy findings from this study. First, arrests involving individuals with the indicator of serious mental illness were less likely than arrests involving individuals without a serious mental illness to be categorized as a violent offense, and most arrests among those with serious mental illness were for nonviolent crimes. As Neusteter and colleagues (5) suggest, law enforcement officers’ discretion in applying minor charges for nonviolent behavior can have compounding effects in shaping people’s trajectories through the criminal legal system; such discretion is less consequential in violent felonies. For example, misdemeanor arrestees with serious mental illnesses, but not felony arrestees, face an increased risk for receiving a jail sentence (6), and those with serious mental illnesses represent most jail entrants who are chronic misdemeanants (14).

Second, in terms of penal code classifications, class A misdemeanors were substantially more common in arrests involving individuals with serious mental illnesses than in arrests involving those without the indicator of serious mental illness. On the other hand, unclassified misdemeanors, including driving while intoxicated, driving with a suspended license, and reckless driving, were less common among individuals with serious mental illnesses, probably because they are less likely to own and drive vehicles. Third, some UCR codes were more likely to occur in arrests involving those with serious mental illnesses, and the magnitude of the overrepresentation of some UCR codes was striking—for example, larceny-theft (i.e., the unlawful taking, carrying, leading, or riding away of property from the possession of another, such as theft of bicycles or shoplifting) (15) was 20% more common. Fourth, rank-ordering the most common charges among the most serious charge within each arrest event revealed that charges related to driving under the influence of alcohol were among the most common charges in arrests of those without the indicator (6.5% [N=138,377] of arrests, compared with 1.4% [N=1,307] among those with the indicator), and criminal trespass charges were among the most common charges in arrest involving individuals with serious mental illnesses (2.5% [N=2,265] of arrests, compared with 1.7% [N=35,284] among those without serious mental illnesses).

These findings should be viewed in light of literature seeking to explain why individuals with serious mental illnesses are overrepresented within the criminal legal system. The “criminalization hypothesis” argues that deinstitutionalization and failure to provide adequate mental health services led to arrest and criminal legal involvement as a default response to managing individuals with serious mental illnesses (16). However, the “criminogenic risk” perspective posits that myriad risk factors beyond symptomatic mental illness, such as low socioeconomic status and substance use, drive criminal involvement (17). These two conceptual frameworks provide complementary lenses for viewing our findings. On the one hand, criminal legal records describe the behavior of the legal system: the official labels attached (often with discretion) to alleged lawbreaking, the determination of guilt, and the sanctions meted out in particular cases. If the criminal legal system is using its processes to control behavior (i.e., behavior that is unacceptable to a community or to specific complainants) because the mental health system (or other social service systems) cannot sufficiently control that behavior, this process could be deemed one of “criminalization.” On the other hand, criminal legal records can be seen as providing a rough proxy measure of actual offending behavior in the community (a rough proxy because many people who break the law never get arrested, and many who get arrested are never convicted for the crime they were initially charged with because of plea bargaining, etc.). In this case, the burden of criminogenic risks carried by individuals with serious mental illnesses (e.g., unstable family and peer relationships, inadequate education, unemployment and underemployment, housing instability, substance use, and neighborhood instability) puts them at high risk for behaviors meeting the threshold of offenses for which the criminal legal system can legitimately use and process misdemeanor charges. As such, the criminogenic risk perspective and the criminalization hypothesis likely go hand in hand: individuals with serious mental illnesses are, by virtue of their psychosocial circumstances, at high risk for behaviors qualifying as criminal offenses, and simultaneously, the criminal legal system readily uses its processes to control those behaviors that the community or complainants insist be controlled.

In the future, deeper explorations of specific class A misdemeanor charges, in terms of exactly where the behaviors are occurring, could illuminate dynamics that are unique to encounters with police among individuals with serious mental illnesses and concurrent socioeconomic disadvantage. Given links between mental illness and housing instability (18), for instance, more quantitative and qualitative work is needed to understand the use and processing of criminal trespass.

This study had several methodological limitations. First, our indicator of a serious mental illness was imperfect because it did not include individuals with a serious mental illness who had not received billable treatment services within the 2 years before arrest, those with private health insurance, or those with less severe mental illnesses, such as posttraumatic stress disorder. Therefore, our findings were specific to people with serious mental illnesses who were socioeconomically disadvantaged and receiving at least some treatment services. Second, although they were relevant to our research question, we could not capture minor violations for which offenders are not fingerprinted, such as disorderly conduct and loitering, even though such charges are misdemeanors in other states and might be germane to the problem of criminalization in those settings. Future research should examine violations in addition to misdemeanors and felonies. Third, we examined only the most serious charge for each arrest event; additional charges might be informative and could be issued more commonly (as secondary and tertiary charges) among individuals with serious mental illnesses. Fourth, although charges within an arrest event have a lifecycle and may change from arrest, to arraignment, to disposition, our data represented only a snapshot; when the data set was created, more distant arrest records were more likely to have a disposition, whereas recent ones might not. As such, criminal charges from different stages in the legal process were treated as equivalent, even though the role of discretion and standards of evidence differ at each stage (e.g., probable cause for arrest vs. beyond reasonable doubt for conviction). Results could differ if the same analyses focused only on charges at arrest or charges at disposition. Fifth, the study period was 2010–2013; internal and external validity might have been limited given the rapidly shifting criminal legal landscape in New York State and in the United States. For example, the New York State Bail Elimination Act of 2019 eliminated money bail and thus mandated pretrial release (i.e., disallowed pretrial jail detention) for most misdemeanors and nonviolent felonies.

Finally, using charges as a research measure inherently limits interpretation. For example, a higher arrest rate for trespassing among those with serious mental illness than among those without serious mental illness could reflect psychiatric symptoms impairing normal social functioning in ways that cause these individuals to break community norms. However, it is also possible that the mental illness creates social and economic disadvantages—the criminogenic risks described above—that put these people in situations where they are more likely to breach private property boundaries. Conversely, criminal trespass could be a discretionary, and potentially discriminatory, charge that officers (and ultimately the larger misdemeanor processing system) rely on (e.g., criminalization to resolve a situation and pacify a complainant, such as a shop owner). Examining charges alone cannot decipher the extent to which these interacting forces are at play.

Conclusions

Research that delves into law enforcement officers’ misdemeanor charging decisions would inform specific diversion strategies. Research is also needed on pathways for individuals with serious mental illnesses charged with minor misdemeanors, namely, their pathways across the various layers or intercepts of the criminal legal system, which can include technical violations, such as failure to appear to court and probation violations. Associated costs should also be studied. For example, Swanson et al. (19) examined administrative records for >25,000 adults with schizophrenia or bipolar disorder served by the Connecticut Department of Mental Health and Addiction Services in 2006–2007 and found that the group with criminal legal system involvement (about one in four) incurred costs approximately double those of the group with no involvement: $48,980 compared with $24,728 per person. Such work is especially vital in an era of rapidly evolving criminal legal reform.