Examining All-cause and Cause-specific Mortality among Former Prisoners in Pennsylvania

Abstract

Former prisoners have a higher than expected risk of death following release from incarceration. However, little is known about the specific risk factors for post-release mortality among former prisoners. The current study uses a unique set of measures obtained from administrative records from Pennsylvania to examine demographic, custodial, behavioral, and criminal history factors that impact mortality risk following release from incarceration. Moreover, this study is the first to assess whether risk factors for post-release mortality are consistent or variable across race and ethnicity. Using data from the Pennsylvania Department of Corrections and mortality records from the Pennsylvania Department of Health we find several demographic, custodial, behavioral, and criminal history measures are related to post-release mortality risk. Moreover, while most risk factors for mortality are generally consistent across race and ethnicity, we find evidence that some custodial and criminal history factors vary by race and ethnicity.

Keywords:

• incarceration
• mortality
• race and ethnicity
• health

Notes

¹ Notably, Dirkzwager and colleagues (Dirkzwager, Nieuwbeerta, & Blokland, 2012) are among the only studies comparing differences in postprison mortality between ex-prisoners and both the general population and offenders who received a non-custodial sentence. The results of this study reveal that while former prisoners are at a heightened risk of mortality relative to those in the general population, there is no significant difference in mortality risk between former prisoners and non-imprisoned offenders (matched on age, gender, and propensity score).

² The finding of high rates of post release mortality among former prisoners appear across several countries including the United States, Russia, United Kingdom, Australia, and France (Binswanger et al., 2007, 2011a; Farrell & Marsden, 2008; Hobbs et al., 2006; Krazlan, Ridout, Mai, Knuiman, & Chapman, 2006; Pridemore, 2014; Stewart, Henderson, Hobbs, Ridout, & Knuiman, 2004; Verger, Rotily, Prudhomme, & Bird, 2003).

³ The standardized mortality ratio (SMR) is a ratio of the number of deaths in a population over a selected time period to the expected number of deaths in that population if it was composed of the same age, sex, and race specific death rates in a “standard” population (Julious, Nicholl, & George, 2001). The SMR is calculated using the following formula:

⁴ However, other studies have found longer periods of incarceration increase risk for post-release mortality (see Kariminia et al., 2007a; Patterson, 2013).

⁵ The linking process of records between the Pennsylvania Department of Corrections and Department of Health was achieved using 4 identifiers: last name, first name, date of birth, and social security number.

⁶ Overall, the current study includes 21,417 of the 22,307 individuals release from a Pennsylvania state prison in 2006 or 2007.Observations were removed from the data-set because of missing data. The primary sources of missing data include removal of observations for nonsensical reports on administrative records such as reporting negative initial sentence length (541 observations). The remainder of the observations are removed as a result of listwise deletion from missing data on behavioral and criminal history covariates included in the analytic sample.

⁷ Because the current study examines mortality following release from prison, a total of 49 individuals who died while incarcerated were removed from the analysis sample. A sensitivity analysis retaining these observations does not alter the results.

⁸ American Indian, Asian, and “Other Race” make up less than 1 percent of the sample. In total, there are 22 American Indians, 62 Asians, and 32 other race included in the original records.

⁹ BMI was calculated using information on height in inches and weight in pounds using the following formula: BMI = weight (lb) / [height (in)]² x 703. For more information see: http://www.cdc.gov/healthyweight/assessing/bmi/adult_bmi/. In coding BMI, we follow prior research and bottom coded BMI at 14 and top-coded BMI at 50 (Besen & Pranksy, 2014). In total, 3 observations (.02%) were bottom coded and 65 observations (.31%) were top-coded.

¹⁰ Because misconducts have a strong positive skew we top-code misconducts at the 95th percentile. A sensitivity analysis including the misconduct measure without the top-coding does not alter the results.

¹¹ LSI-R score was top-coded at 54, which is the maximum score possible on the instrument. The sample contained 20 observations (.07%) had scores above 54. Subsequent analysis was conducted omitting these cases, which does meaningfully alter the results.

¹² A dichotomous variable is included for whether mental health status is missing or unknown. This retains a total of 648 observations. An analysis removing these observations from the analytic sample found no difference from the main findings.

¹³ According to the guidelines from the Center for Disease Control and Prevention, the BMI ranges include: underweight (below 18.5), healthy weight (18.5–24.9), overweight (25.0–29.9) and obese (30.0 and above).

¹⁴ We estimated a logistic regression using mortality from homicide as the dependent variable and including the full set of variables reported in Table 4. The results demonstrate that relative to whites, blacks are approximately 5 times more likely to die from homicide (OR = 5.05; p < .01). Additionally, there is not statistically significant difference in mortality risk between Hispanics and whites (OR = 1.23; p > .10).

¹⁵ We performed additional analyses using a measure of percentage of the minimum sentence and percentage of maximum sentence served calculated using minimum and maximum sentence release date and the actual release date. These variables did not have a statistically significant association with post-release mortality among either the full sample and the race-specific samples.

¹⁶ Legislative changes to the sentencing practices in Pennsylvania may also have impacted the composition of the sample. In 2005, the Pennsylvania State Legislature enacted the state intermediate punishment program (SIP Program Act 112 of 2004). Under this program, judges were given the discretion to sentence convicted offenders with a maximum sentence of 2–5 years to county jails instead of state prisons. Accordingly, if judges systematically sentenced certain types of offenders to country jails rather than state prisons, the implementation of this law may have impacted the results of the current study. However, as only 4 percent of eligible offenders received sentences under this program by 2012, it is unlikely that the program substantially altered the results (Pennsylvania Commission on Sentencing, 2013; Ulmer & Laskorunsky, 2016).

Additional information

Notes on contributors

Alexander Testa

Alexander Testa is a PhD candidate in the Department of Criminology and Criminal Justice at the University of Maryland. His research interests include the consequences of criminal justice contact across the life course, criminal justice decision-making, cross-national causes and consequences of violence, and criminal justice policy.

Lauren C. Porter

Lauren C. Porter is an assistant professor of Criminology and Criminal Justice at University of Maryland. Her research revolves around issues of punishment, health, and the influence of place and space on criminal behavior.

Kiminori Nakamura

Kiminori Nakamura is an assistant professor of Criminology and Criminal Justice at the University of Maryland. His research focuses on corrections and reentry, and life-course and criminal-career issues, including redemption and desistance.