Introduction
Justice health research, or epidemiological criminology, sits at the critical intersection of public health and the legal system. To develop effective interventions for justice-involved populations, it is essential to understand who is being studied. However, the sheer volume of literature makes manual review nearly impossible, and inconsistent reporting standards often obscure whether research is equitable or representative. This study addresses this "knowledge gap" by automating the extraction of population characteristics from over 34,000 PubMed abstracts published between 1963 and 2023.
Methods
The researchers employed a rule-based text mining approach using the General Architecture for Text Engineering (GATE) and its Java Annotations Pattern Engine (JAPE). Unlike black-box machine learning models, this method used 140 specific rules to identify lexical patterns—anchored by specific verbs, nouns, and prepositions—to extract data on sex, age, offender type, and nationality.
The primary corpus consisted of 34,481 PubMed abstracts. To ensure accuracy, raw extractions were processed via Python scripts to map varied terms (e.g., "adolescent," "teen," "juvenile") into standardized bins. The system was evaluated against a manually annotated "gold standard" of 100 random abstracts, achieving a micro-F1 score of 89.9%, with high precision (94.3%) and respectable recall (85.9%).
Results
The analysis revealed a significant reporting crisis: 49.4% of abstracts mentioned none of the four key characteristics, and only 1.3% reported all four. This suggests that the "front door" of most papers—the abstract—often fails to inform readers about the study's target population.
Shifting Demographics
A major finding was the "Gender Paradox." While men represent roughly 95% of the global incarcerated population, research focus has shifted dramatically. Since 1990, female-only studies have surged, accounting for nearly half (48.6%) of research outputs in 2023. While studying women is vital, the authors suggest that the overwhelming majority of prisoners (men) may now be under-researched.
Regarding age, 69.3% of abstracts reporting age focused on minors (<18), reflecting a global research priority on youth justice. In terms of offender types, mentally ill offenders were the most studied subgroup (17.4%), reflecting a 30-year trend where prisons have increasingly become de facto mental health facilities.
Geographic Disparities
The United States leads in sheer volume (31.4% of articles), but when research output is scaled per 1,000 prisoners, Nordic countries (Sweden, Norway, Finland, and Denmark) produce the most research. This likely reflects their progressive, data-driven approach to offender rehabilitation.
Discussion and Conclusion
While the study is limited by its focus on abstracts rather than full-text bodies (due to paywalls) and its restriction to English-language articles, the findings are clear. There is a critical need for better reporting. The authors call for journals to enforce stricter adherence to reporting guidelines (like STROBE or CONSORT) specifically within abstracts to ensure justice health research remains transparent and actionable. By promoting diversity in participant selection and improving reporting clarity, the field can better address the needs of all justice-involved individuals.
Summary updated