The Role of Incarceration as a Risk Factor for Cognitive Impairment—Comprehensive Study Notes

Publication & Context

  • Article: “The Role of Incarceration as a Risk Factor for Cognitive Impairment” (Journals of Gerontology: Social Sciences, Vol. 77, No. 12, 2022)
  • Part of a special section on Interdisciplinary Aging Research addressing health disparities in Alzheimer’s Disease & Related Dementias (ADRD)
  • Open-access; emphasizes structural/institutional contributors to cognitive health disparities
  • Authors: Robynn J. A. Cox, PhD (Public Policy, UC-Riverside) & Robert B. Wallace, MD (Public Health, Univ. of Iowa)

Objectives

  • Quantify disparities in early-midlife cognitive impairment between formerly incarcerated (FI) and never-incarcerated (NI) adults
  • Test whether incarceration status independently predicts:
    • Continuous cognitive performance (TICS-m)
    • Categorical outcomes: Cognitive Impairment-Not Dementia (CIND) & Dementia
  • Identify mediators (education, premorbid cognition, health, socioeconomic status, adverse childhood experiences, etc.)
  • Situate findings within cognitive-reserve & life-course frameworks

Data Source: NLSY79 Cohort

  • National Longitudinal Survey of Youth, 1979 cohort (NLSY79)
  • Baseline ages 14–2214\text{–}22 (1979); cognition module administered 2006–20162006\text{–}2016 when respondents were 46–6046\text{–}60
  • Sample sizes after imputation:
    • Total observations: N=8,021N = 8{,}021
    • Formerly/currently incarcerated (FI/I): N=573N = 573
      • Prior incarceration (before cognition survey): N=567N = 567
      • Incarcerated at interview year: N=60N = 60 (6 had no prior spell)
    • Never-incarcerated (NI): N=7,448N = 7{,}448
  • Multiple Imputation by Chained Equations (MICE) for missing data

Key Measures

Cognitive Function (TICS-m)
  • Telephone Interview for Cognitive Status-modified (range 0–270\text{–}27)
    • Immediate 10-word recall (episodic memory)
    • Delayed 10-word recall
    • Serial 7s subtraction (working memory)
    • Backward count from 20 (mental status)
  • Cut-points (literature based):
    • Dementia: ≤6\le 6
    • CIND: 7≤score≤117 \le \text{score} \le 11
    • No impairment: ≥12\ge 12
  • Overall cognitive impairment indicator: Score≤11\text{Score} \le 11
Incarceration Variables
  • “Type of Residence R Is Living In” identifies “jail” in each wave
  • Binary indicators:
    • Prior incarceration (any time before cognition module)
    • Current incarceration (year of cognition module)
  • Limitation: short jail spells between waves may be missed; jail vs. prison not differentiated

Covariate Blocks (Added Sequentially in 10 Models)

  1. Demographics: Age, sex, race/ethnicity (Non-Hispanic Black, Hispanic, White/Other)
  2. Health: Body-Mass-Index categories, vascular diagnoses (heart problems, high blood pressure, diabetes, stroke)
  3. Socio-economic: Net-worth quartiles
  4. Education: Highest Grade Completed (HGC) (<1212, 1212, 13–1513\text{–}15, ≥16\ge16 years)
  5. Augmented factors:
    • Premorbid cognition: 1980 Armed Forces Qualification Test (AFQT, range 0–1000\text{–}100)
    • Adverse Childhood Experiences (ACEs): poor childhood health, long hospitalization, prolonged confinement to bed, household mental illness, household alcoholism, repeated physical harm, low parental affection
    • Lifetime smoking (>100 cigarettes)
    • Emotional disorder/depression diagnosis
  6. Fixed effects: Urbanicity & Census region at baseline and interview year

Statistical Strategy

  • Continuous outcome: OLS regression for TICS-m score
  • Binary outcome: Logistic regression for overall cognitive impairment
  • Multicategory outcome: Generalized Ordered Logit (partial-proportional odds) for Normal vs CIND vs Dementia
  • Ten nested models (Models 1–101\text{–}10); paper focuses on 1,5,6,101,5,6,10 for brevity

Descriptive Highlights (Weighted Means)

  • FI/I group older (mean 48.5448.54 vs 48.3148.31 years)
  • Racial composition: Blacks 51.3%51.3\% FI/I vs 29.6%29.6\% NI; Hispanics 21.8%21.8\% vs 19.2%19.2\%
  • Sex: Women 12.2%12.2\% FI/I vs 53.9%53.9\% NI
  • Health burdens higher among FI/I: heart problems 12.1%12.1\% (vs 8.0%8.0\%), high BP 45.3%45.3\% (vs 37.7%37.7\%), emotional disorder 26.4%26.4\% (vs 19.2%19.2\%)
  • Socio-economic disadvantage: 48.9%48.9\% FI/I in bottom wealth quartile (vs 20.2%20.2\%)
  • Education: <HS diploma 45.2%45.2\% FI/I (vs 13.6%13.6\%)
  • AFQT mean: FI/I 21.521.5 vs NI 42.342.3 (gap ≈21\approx 21 points)
  • Cognition:
    • Mean TICS-m: FI/I 14.1014.10 vs NI 16.5616.56 (difference −2.46-2.46)
    • Cognitive impairment prevalence: FI/I 26.0%26.0\% vs NI 12.5%12.5\%
    • Dementia prevalence: FI/I 5.0%5.0\% vs NI 1.8%1.8\%

Regression Results

Continuous Scores (TICS-m)
  • Model-1 (unadjusted): prior incarceration coefficient −2.46-2.46 (p<.001)
  • Model-5 (adds demographics, health, wealth, education): coefficient −0.37-0.37 (marginal p<.10)
  • Model-6 (adds ACEs): coefficient −0.36-0.36 (marginal p<.10)
  • Model-10 (full augmented): coefficient −0.12-0.12 (ns)
  • Key independent predictors in full model:
    • Age: β=−0.067\beta = -0.067 per year
    • Black β=−0.345\beta = -0.345; Hispanic β=−0.539\beta = -0.539
    • Female β=+0.894\beta = +0.894
    • Overweight +0.447+0.447; Obese +0.610+0.610 ("obesity paradox")
    • High BP −0.432-0.432
    • Bottom-wealth quartile −0.885-0.885
    • <HS education −1.303-1.303; HS −0.585-0.585
    • ACEs: poor childhood health −0.946-0.946; household mental illness −0.533-0.533; repeated physical harm −0.279-0.279
    • Emotional disorder −0.542-0.542
    • AFQT: +0.055+0.055 per point
Overall Cognitive Impairment (Binary)
  • Model-1 OR for prior incarceration OR=2.402OR = 2.402 (p<.001)
  • Model-5 OR =1.001=1.001 (ns) once education entered
  • Model-10 OR =0.887=0.887 (ns)
  • Significant risk factors (full model):
    • Black OR=2.495→1.288OR = 2.495\rightarrow1.288; Hispanic OR=2.135→1.323OR = 2.135\rightarrow1.323 after full adjustment
    • High BP OR=1.278OR = 1.278; Diabetes OR=1.202OR = 1.202 (marginal)
    • <HS education OR=1.420OR = 1.420; Female OR=0.683OR = 0.683 (protective)
    • Overweight OR=0.655OR = 0.655; Obese OR=0.619OR = 0.619 (protective)
    • Emotional disorder OR=1.296OR = 1.296
    • AFQT OR=0.960OR = 0.960 per point (protective)
Dementia vs CIND/Normal
  • Model-1 OR prior incarceration OR=2.683OR = 2.683 (p<.001)
  • Difference erased once education or AFQT entered; full Model-10 OR =0.881=0.881 (ns)
  • Independent correlates (Model-10):
    • Age: OR=1.148OR = 1.148 per year
    • Black OR=2.103OR = 2.103; Hispanic OR=2.268OR = 2.268
    • Overweight OR=0.573OR = 0.573; Obese OR=0.478OR = 0.478
    • AFQT OR=0.961OR = 0.961 per point
    • Education effects attenuated after AFQT (supporting cognitive-reserve hypothesis)
  • Findings align with Cognitive Reserve Hypothesis:
    • Higher premorbid cognition (AFQT) & education buffer against neuropathology → lower dementia odds
  • Life-course & Cumulative Disadvantage frameworks:
    • Incarceration compounds early disadvantages (ACEs, poor schools, poverty) → reduced cognitive reserve
    • Structural racism increases both incarceration risk & cognitive decline risk among Blacks/Hispanics
  • Stress pathway: chronic stress of incarceration & stigma may damage cognition; emotional-disorder variable partially captures this
  • Obesity/overweight showing protective associations mirrors “obesity paradox” in aging literature

Practical, Ethical & Policy Implications

  • Reentry planning must consider high prevalence of cognitive impairment among FI (≈11 in 44)
    • Cognitive screening pre-release; tailored employment, housing, and health services
  • Educational & cognitive-skill interventions (both in youth & during incarceration) could build reserve and mitigate later ADRD risk
  • Address structural factors (school quality, discrimination, poverty, policing) to reduce upstream disparities
  • Health-care systems & corrections should coordinate to manage chronic diseases (HTN, diabetes) that elevate cognitive risk
  • Implications for state budgets: ADRD costs $355\$355 billion; if incarceration adds to prevalence, burden will increase in disproportionately impacted communities

Limitations Noted by Authors

  • Incarceration measure may miss short spells → attenuation bias
  • No timing for chronic disease diagnoses; temporality ambiguous
  • Survivorship bias: higher mortality among FI could understate effects
  • TICS-m influenced by education quality; may conflate educational disadvantage with cognitive decline
  • Observational design → cannot claim causality; need richer causal data

Numerical & Statistical References (Key)

  • FI mean TICS-m =14.10=14.10; NI =16.56=16.56 (gap −2.46-2.46)
  • Unadjusted odds: Cognitive impairment OR=2.4OR = 2.4; Dementia OR=2.7OR = 2.7 for FI vs NI
  • Education <HS prevalence: FI 45.2%45.2\% vs NI 13.6%13.6\%
  • AFQT means: FI 21.521.5; NI 42.342.3 (scale 0–1000\text{–}100)
  • Over 600,000600{,}000 releases from prison annually; prisoners aged ≥55\ge55 grew 400%400\% since 19931993
  • ADRD treatment cost (2021): $355\$355 billion

Connections to Prior Literature

  • Confirms micro-level evidence of cognitive decline during incarceration (Ezenwa 2020; Umbach 2018)
  • Extends work on racial disparities in early cognitive impairment (Lines 2014; Mayeda 2016) by inserting incarceration as mediator
  • Supports studies linking education/IQ to dementia delay (Schmand 1997; Stern 2012)
  • Aligns with findings that emotional distress predicts cognitive decline (Wilson 2007)

Study Take-Home Messages

  • Apparent incarceration–cognition link is largely explained by educational attainment & premorbid cognitive ability → cognitive reserve matters
  • Nonetheless, FI exhibit double the raw prevalence of impairment; actionable need for screening & support
  • Black & Hispanic disparities persist even after extensive controls, underscoring role of structural racism across the life course