Scholarlink

Background <p>Estimating mortality risk in incarcerated adults is important for identifying individuals who may benefit from palliative care and compassionate release referrals.</p> Objective <p>To develop and internally validate a 2-year mortality prediction model in incarcerated adults.</p> Design <p>Cohort study (February 1, 2018–February 1, 2020).</p> Participants <p>Incarcerated adults aged ≥ 18&#xa0;years residing at a California Department of Corrections and Rehabilitation (CDCR) prison for ≥ 1&#xa0;year.</p> Main Measures <p>Model predictors included demographics (e.g., age, sex), housing status (general housing vs. higher acuity infirmary bed vs. lower acuity infirmary bed), functional assessment (level of mobility restriction), healthcare utilization (e.g., hospitalizations and intensive care unit admissions in the previous year), and chronic conditions. The primary outcome was natural death at 2&#xa0;years, defined as death due to causes other than suicide, homicide, accidental injury, or drug overdose. Cox proportional hazards regression with LASSO for variable selection was used to develop the model. Model performance was assessed by discrimination (area under the receiver operating characteristic curve (AUC) at 2&#xa0;years) and calibration (plots of predicted and observed mortality). Classification metrics were assessed at clinically relevant thresholds.</p> Key Results <p>The final cohort included 89,430 adults (median age 40&#xa0;years (interquartile range = 20), 10.2% ≥ 60&#xa0;years, 30.6% Black, 41.4% Hispanic). At 2&#xa0;years, 506 (0.6%) individuals experienced a natural death. The optimism-corrected AUC at 2&#xa0;years after bootstrap internal validation was 0.926 (95% confidence interval (CI) = 0.915–0.938). The calibration plot at 2&#xa0;years suggested good calibration. At a 2-year mortality risk threshold of 5%, sensitivity, specificity, and positive predictive value were 47.6% (95% CI = 42.3–51.6%), 98.4% (95% CI = 98.2–98.4%), and 16.7% (95% CI = 14.3–18.5%), respectively.</p> Conclusions <p>The mortality risk estimates from this model can help clinicians identify individuals who may most benefit from advance care planning discussions, palliative care services, and compassionate release referrals.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development of a Mortality Prediction Model for Incarcerated Adults to Identify Palliative Care Needs

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月18日

查看全文
  • W. James Deardorff,
  • Alexandra K. Lee,
  • Kaiwei Lu,
  • Bocheng Jing,
  • W. John Boscardin,
  • Michele DiTomas,
  • John Dunlap,
  • Brie A. Williams,
  • Sei J. Lee,
  • Alexander K. Smith

摘要

Background

Estimating mortality risk in incarcerated adults is important for identifying individuals who may benefit from palliative care and compassionate release referrals.

Objective

To develop and internally validate a 2-year mortality prediction model in incarcerated adults.

Design

Cohort study (February 1, 2018–February 1, 2020).

Participants

Incarcerated adults aged ≥ 18 years residing at a California Department of Corrections and Rehabilitation (CDCR) prison for ≥ 1 year.

Main Measures

Model predictors included demographics (e.g., age, sex), housing status (general housing vs. higher acuity infirmary bed vs. lower acuity infirmary bed), functional assessment (level of mobility restriction), healthcare utilization (e.g., hospitalizations and intensive care unit admissions in the previous year), and chronic conditions. The primary outcome was natural death at 2 years, defined as death due to causes other than suicide, homicide, accidental injury, or drug overdose. Cox proportional hazards regression with LASSO for variable selection was used to develop the model. Model performance was assessed by discrimination (area under the receiver operating characteristic curve (AUC) at 2 years) and calibration (plots of predicted and observed mortality). Classification metrics were assessed at clinically relevant thresholds.

Key Results

The final cohort included 89,430 adults (median age 40 years (interquartile range = 20), 10.2% ≥ 60 years, 30.6% Black, 41.4% Hispanic). At 2 years, 506 (0.6%) individuals experienced a natural death. The optimism-corrected AUC at 2 years after bootstrap internal validation was 0.926 (95% confidence interval (CI) = 0.915–0.938). The calibration plot at 2 years suggested good calibration. At a 2-year mortality risk threshold of 5%, sensitivity, specificity, and positive predictive value were 47.6% (95% CI = 42.3–51.6%), 98.4% (95% CI = 98.2–98.4%), and 16.7% (95% CI = 14.3–18.5%), respectively.

Conclusions

The mortality risk estimates from this model can help clinicians identify individuals who may most benefit from advance care planning discussions, palliative care services, and compassionate release referrals.