Criminogenic or criminalized? Testing an assumption for expanding criminogenic risk assessment.

Proponents of criminogenic risk assessment and algorithmic prediction for criminal recidivism have called for its widespread expansion throughout the criminal justice system, including for setting bail, pre-trial detention decisions, sentencing, and even policing. Its success in predicting recidivism is taken as evidence that criminogenic risks tap into the causes of criminal behavior even for first offenses, and that targeting these factors can reduce correctional supervision rates and even prevent crime.  However, recent work by Seth Prins suggests that these algorithms for predicting recidivism after release from incarceration cannot be applied wholesale across the various decision points in the criminal justice system.  His findings suggest that current risk assessments for recidivism cannot fully distinguish between individuals’ propensities for committing crime and the fact that they have already been criminalized by a runaway criminal justice system.  In the era of mass incarceration, the idea that risk factors for staying trapped in the criminal justice system are the same as the risk factors for initial exposure to the system ignores all the social, economic, and policy-related factors that have nothing to do with individual characteristics. He argues that we need to focus on what puts people at risk of criminogenic risk, and one of those things, arguably, is current criminal justice policy.

In related news, Seth and his colleague Brett Story also did an interview with WBAI on  how many of the policies and programs proposed in the Green New Deal can also put us on the path to decarceration and healthequity.

Seth also created the Mass Incarceration Mix on Social Epi Radio and the Mass Incarceration Info-Graphix hosted on this web-site.


Posted in Environmental Justice, Mass Incarceration, Racism, Social Epi Radio | Leave a comment

Body Mass Index across the Life-Course: Emergence of Race by Sex Disparities in Early Childhood.

In the U.S. 35% of adults aged 20 years or older are obese and the obesity epidemic represents a critical public health issue.  There are marked disparities in body mass index (BMI) and obesity prevalence by race/ethnicity and sex.  Among men the age adjusted prevalence of obesity is modestly higher among Non-Hispanic Blacks and Hispanics than among non-Hispanic Whites, while among women the prevalence of obesity is substantially higher among Non-Hispanic Blacks and Hispanics than among non-Hispanic Whites.  Non-Hispanic Black women have the highest prevalence of obesity of any racial, ethnic, sex group, a disparity that has been in place for several decades, with an age adjusted obesity prevalence of 57% in 2011-2012.

Work by Andrew Rundle and colleagues recently published in the Annals of Epidemiology used data from the Child Health and Development Studies (CHDS) to assess when in the life course the race by gender disparity in BMI for Blacks and Whites begins. CHDS participants were born in the early 1960s and height and weight data were collected at ages 5, 9-11 and 15-17.  Six hundred and five CHDS participants were recently follow-up again at age ~50 and height and weight were measured. Analyses of these data showed that the race by gender disparity in BMI was present by age 9-11 years and continued at ages 15-17 and 50 years, with Black women having the highest BMI scores.  A large proportion of the race by sex disparity in BMI at age 50 could be accounted for by the participant’s BMI at age 9-11.

1. Socio-demographic covariates were: paternal and maternal education, maternal pre-pregnancy BMI, participant obtained a college degree, participant’s smoker/non-smoker status at age 50 and participant age of assessment

Continue reading

Posted in Uncategorized | Leave a comment

“You probably can’t feel as safe as normal women”: Hispanic women’s reactions to breast density notification

A new qualitative study by Alsacia Pacsi-Sepulveda, Rachel Shelton, Carmen Rodriguez, Arielle Coq, and Parisa Tehranifar that explored the understanding of, and reactions to, New York State’s breast density notification language was recently published in the journal Cancer. In 2013, New York State enacted legislation requiring that women receive a written letter if a mammogram reveals they have heterogeneously or extremely dense breasts. The letter is intended to notify them of their density status, increased risk of breast cancer, and lower sensitivity of mammography screening. A paragraph of mandated text containing this information is specified in the legislation.

The researchers asked 24 self-identified Hispanic women who had a history of dense breasts about their understanding of the NYS mandated breast density information, directly after reading them the mandated notification text. All respondents had received screening mammograms since the law went into effect, but most did not recall receiving notification and showed low levels of understanding of breast density.

Using inductive content analysis, the researchers identified five overarching themes that arose in their interviews, including: confusion about, and lack of understanding of, what ‘dense breasts’ means; the perception that dense breasts are an abnormal and serious condition; and worries about breast cancer risk and the need for additional tests for screening. Additional themes found were a “reliance on faith and acceptance of destiny” – that dense breasts are something predetermined, and that the information learned in the notification was important and actionable – that more diligent screening or additional testing was important for them.

Importantly, respondents also provided recommendations for communicating breast density information, stating that health care providers were the best source of this information, but that written informational materials, videos, ads, or community programs, were also recommended.

The researchers suggest that revising the notification text to lower the literacy level and include a definition of breast density or other clarifying information could assist women in their understanding of this notification.

In Feb 2019, a US federal law was passed requiring the FDA to develop mandatory reporting language that must be included in patient and provider mammography reports across the US. [source]. The minimum information that is required includes:

  • The effect of breast density in masking the presence of breast cancer on a mammogram
  • The qualitative assessment of [breast density by] the provider who interprets the mammogram, and
  • A reminder to patients that individuals with dense breast tissue should talk with their providers if they have any questions or concerns about their summary.

Though the mandated information is specified, specific language and text to be used is not. The authors’ findings and recommendations have implications for informing the specific language used in notification letters directed towards patients to ensure that the information is understood and actionable.

Posted in Health Communication, Mixed Methods, Qualitative Research | Leave a comment

Neighborhood Health Effects: Does The Way We Define “Neighborhood” Alter the Effect?

There are many different ways that aspects of the social and physical environment can affect a person’s health. For example, body mass index and chronic disease are associated with the walkability of the area where a person lives. Spending more time near fast food outlets is associated with greater saturated fat intake. Being present in neighborhoods with more decay and disorder is related to increased risks for assault. The studies that identified these associations all link individuals to their neighborhood environments in some way. But if the researchers had used different approaches to make this link, would they have arrived at the same findings?

Christopher Morrison and his colleagues from the Prevention Research Center in Berkeley, CA, published a paper in Epidemiology that addressed this question. They used one month of GPS data for 231 adolescents aged 14 to 16, and tested whether exposure to retail alcohol outlets was associated with increased alcohol consumption. The group used three different approaches to link the environmental condition (alcohol outlets) to the individuals. They measured exposure around the person’s home, around the places they attended most frequently, and around the full GPS route path.

The researchers found that the different measures of exposure to alcohol outlets were, at best, moderately correlated, which means they probably all measure different constructs. Perhaps more importantly, the different measures produced different associations with alcohol outlets—in some cases the exposure was positively related to the outcome, and in others there was no association. It seems the way we measure individuals’ exposures to environmental conditions could very substantially affect the results of a given study.

Activity space captured by GPS depicted in space and time. Residence-based measures, activity location-based measures, and activity path-based measures of exposure to alcohol outlets can be derived from the data.

Posted in Alcohol Consumption, Methods, Neighborhood Environments, Urban Health | Leave a comment

Intimate partner violence severity and depression in rural Bangladesh—a high prevalence setting

Precious Esie, Lisa Bates, and colleagues recently published their work examining the relationship between the severity of intimate partner violence (IPV) and the risk for a major depressive episode (MDE) in the journal Social Science and Medicine –Population Health. The authors applied a novel approach to measuring IPV by operationalizing IPV as the frequency of various acts of physical, psychological, and sexual IPV separately, as well as experiencing injury due to physical or sexual IPV — in contrast to using standard dichotomous indicators. The study population consisted of women who were between 16-37 years old and were married for 4-12 years to their current husband. Experiences of IPV were recent, occurring within 10 months (on average) prior to interviews.

The figure below illustrates the ubiquity and multi-dimensionality of IPV in rural Bangladesh. Nearly 83% of women experienced some form of recent IPV, and nearly all women who experienced physical or sexual IPV also experienced psychological IPV.

Euler diagram illustrating the prevalence and co-occurrence of physical, psychological, and sexual IPV, as well as injury due to physical or sexual IPV

The authors found that incorporating IPV severity consistently revealed a substantially elevated risk of recent MDE for all forms of IPV. Alternatively, the standard dichotomous measures of IPV appeared to underestimate the elevated risk of MDE associated with IPV exposure. For example, in the figure below, although the most severe exposure to sexual IPV was significantly associated with a 65% increased risk for MDE, the corresponding dichotomous measure of sexual IPV (experiencing at least one act versus none) indicated no significant increase for MDE (RR=1.15; 95% CI: 0.90–1.48). This suggests dichotomous indicators of IPV may fail to reveal an association with MDE, but also obscure the much more elevated risks associated with severe levels of IPV exposure.

Risk ratio for sexual IPV severity (left) and as a dichotomous indicator (right).

Notably, the authors also found that for all forms of IPV, the lowest levels of violence severity were not associated with an elevated risk for MDE, relative to no IPV. These results deviate from what has been observed in high-income countries, where as little as one episode of IPV may be associated with more than twice the risk for depression (e.g. Miszkurka et al., 2012). In sum, the use of dichotomous measures, at least in high-prevalence settings such as rural Bangladesh, may mask important nuances in the relationship between IPV and depression and more work is necessary to help explain the IPV-depression gradient.



Miszkurka, M., Zunzunegui, M.V., & Goulet, L. (2012). Immigrant status, antenatal depressive symptoms, and frequency and source of violence: what’s the relationship? Archives of Women’s Mental Health, 15, 387-396.

Posted in Uncategorized | Leave a comment

Mental illness, drinking, and the social division and structure of labor in the United States: 2003-2015

New research by Seth Prins, Sarah McKetta, Jonathan Platt, Carles Muntaner, Kerry Keyes, and Lisa Bates shows the ways that the social division and structure of labor are associated with mental illness and drinking.  Their work was published online in the American Journal of Industrial Medicine.

Occupations involving manual labor and customer interaction, entertainment, sales, or other service‐oriented labor were associated with increased odds of mental illness and drinking outcomes. Physical/risky work was associated with binge and heavy drinking and serious mental illness; technical/craft work and automation were associated with binge drinking. Work characterized by higher authority, autonomy, and expertise was associated with lower odds of mental illness and drinking outcomes.

Odds Ratio for heavy drinking for a 1 standard deviation difference in each dimension of work. The dimensions of work were created from analyses of Department of Labor’s Occupational Information Network (O*NET) Database

The most compelling finding–which verifies common sense but is rarely, if ever, demonstrated in mainstream quantitative research–is that the productivity-to-pay gap (a crude indicator of economic exploitation) seems to have real consequences for mental health. For full-time workers below the top 1% of wage earners, every unpaid hour of labor was associated with higher odds of moderate and serious mental illness.

Predicted probability of moderate and serious mental illness in relation to the productivity to pay gap. Mental illness was defined using the Kessler 6 instrument.

Also important is that inequity in the division of domestic labor places women at risk of moderate and serious mental illness. This disparity represents persistent processes of oppression and economic exploitation, given that women’s gains in workforce participation by the turn of the century did not reduce their disproportionate burden of unpaid domestic work.


Posted in Alcohol, Gender, Health Disparities, Mental Health, Occupation | Leave a comment

Mortality and Work-Family Trajectories for U.S. Women, 1968–2013

Cluster member Sarah McKetta, working with Seth Prins, Jonathan Platt, Lisa Bates, and Katherine Keyes, recently published research examining social roles of US women and how the patterning of these roles impacts mortality.

Changes in employment patterns in the 20th century have led to the majority of US women participating in the labor force. However, as women enter the workforce and become laborers, their traditional domestic roles have not been supplanted; rather, the role of laborer is an additional social role to women who already frequently occupy the roles of parent and spouse. It turns out that having more social roles – parent, spouse, laborer – is associated with better health than having fewer. Yet, measurement of roles generally does not account for the dynamics of women’s lifecourses: women’s roles change throughout the lifecourse, and it may be that the timing, onset, and duration of these roles is as meaningful for health as the number or types of roles themselves.

These researchers partially replicated a study by Sabbath et al. using the Panel Study on Income Dynamics, a nationally-representative, longitudinal survey of households in the United States. The current study examined the social roles of women ages 18-50 every survey wave from 1968-2013; each participant’s pattern of social roles – composed of marriage status, labor status, and parent status – was compiled, and then using sequence analysis and hierarchical clustering these patterns were clustered so that women with similar lifecourse patterns were grouped together into categories.

They found 5 empirically-derived lifecourse types in this sample:

Type 1, “Non-working, married, later-mothers” (N=2,340): These women were consistently married, had children later than other women (e.g., mid-to-late twenties compared to early twenties), and were unlikely to work during their reproductive years.
Type 2, “Working, divorced mothers” (N=2,264): These women had a high probability of marriage early in life and to be unmarried later. They were more likely to work as they got older (e.g., after age 35). Nearly all had children.
Type 3, “Working and non-working never-married mothers” (N=581): These women were much less likely to be married at any point in time than other women with children. They were about equally likely to be working as not during these years. Nearly all had children.
Type 4, “Working, never-married non-mothers” (N=345): These women worked consistently, and were less likely than other women to marry or have children.
Type 5, “Non-working, married, earlier-mothers” (N=767): These women were less likely to work, were consistently married, and had children earlier than the “Non-working, married later mothers.”

Interestingly, none of these lifecourse types was characterized by the “triple role” of laborer, parent, and spouse. However, the authors found that lifecourse types had differential mortality, with Type 1 (non-working, married, later-mothers) having the lowest age-standardized mortality rates. Adjusting for race, birth year, number of children, and educational attainment, Type 3 (working and non-working never-married mothers) and Type 4 women (working, never-married, non-mothers) had significantly higher mortality risks than Type 1 women (RR =1.57 and 1.81, respectively). These risks were somewhat attenuated with the addition of household income to the model, suggesting an incomplete mediating role of income, though the risks were persistently elevated in these groups.

Overall, the authors concluded that timing of childbearing did not significantly predicted poor health in this population, likely because no empirically-derived lifecourse type consistently occupied the “triple role” of laborer, partner, and parent in this sample. Delaying childbearing is thought to be beneficial for women because it allows for gains in the workforce; in this sample where few married mothers also worked, it made sense to not see protective effects of delaying childbearing. In addition, lifecourse type conferred an increased risk of mortality among women who did not have partners, regardless of parenting status, which is a common finding in social science research. Marriage is thought to provide social benefits and economies of scale, which are overall very beneficial for long-term health outcomes.

It may be that social roles impact health, or that women select into social roles because of preexisting health problems or disability. Because the women in the study were followed starting at age 18, these factors were not highly concerning to the authors. Overall, this study contributes to a large corpus of work examining how women’s social realities contribute to health disparities.

Below are a couple of key references for the sequence analysis methods used in the paper.

Gauthier, J., et al., Multichannel sequence analysis applied to social science data. Sociological Methodology, 2010. 40(1 ): p. 1 – 38.

Gabadinho, A., et al. Analyzing and Visualizing State Sequences in R with TraMineR. Journal of Statisitcal Software, 2011. 40(4): p. 1 – 37.


Posted in Gender, Health Disparities, Life Course, Occupation | Leave a comment

Agent Based Model of Alcohol Taxation and Rates of Violent Victimization

Kerry Keyes and colleagues recently published the results of an agent-based modeling simulation of the effects of alcohol taxation on alcohol consumption and non-fatal violent victimization and homicide in New York City.  The team simulated six examples of taxation interventions and the heterogeneous effects of alcohol price elasticities by income, level of consumption and beverage preferences, and examined whether taxation can reduce alcohol-related violence and income-related inequalities in this form of violence. The figure below shows the relations between agent, social network and neighborhood characteristics that were simulated in their agent‐based model. The simulations suggest that reductions in alcohol consumption in New York City can be sustained with modest increases in alcohol taxation (10% tax) and that these tax increases would have modest effects on alcohol-related violent crime.

A primer on agent-based model simulations was published here and an another example on how weight stigma may influence depression in the obese is here.

Posted in Agent Based Model, Alcohol, Violence | Leave a comment

Interdisciplinary Association for Population Health Science Annual Conference.

The program for the 4th annual conference of the Interdisciplinary Association for Population Health Science (IAPHS) is now available online. “Pushing the Boundaries of Population Health Science: Social Inequalities, Biological Processes, and Policy Implications,” to be held October 3-5 in Washington DC, features the latest in interdisciplinary research on population health and promotes exchange between scientists and stakeholders from policy and practice fields. You can register here; Early Bird rates have been extended through August 1.

Posted in Conference, Population Health Sciences | Leave a comment

Analyzing Mixtures of Environmental Contaminants

It has long been understood that the burden of environmental pollution is disproportionately felt in certain neighborhoods, particularly low-income or minority neighborhoods.  In the 1980’s the US Environmental Protection Agency (EPA) described Environmental Justice as the fair treatment and meaningful involvement of all people regardless of race, color, national origin, or income with respect to the development, implementation, and enforcement of environmental laws, regulations, and policies.  Community advocates argue that racial and ethnic minority neighborhoods and lower income neighborhoods are burdened by mixtures of multiple environmental pollutants and socioeconomic stressors, contributing to a “double jeopardy”.

Understanding the health effects of exposure to multiple, or complex mixtures of, pollutants is particularly challenging.  The Mailman School of Public Health’s Department of Environmental Health Sciences is hosting a 2 day training workshop (Aug 23-24) on statistical analyses of complex environmental contaminant mixture data.  Workshop information is here.

Posted in Environmental Justice, Event, Health Disparities, Neighborhood Environments, Teaching Tools | Leave a comment