System Exit and Longitudinal Follow-Up (Tracking Outcomes Over Time)

Pipeline 09E

Transparency note: post-exit outcomes are derived from linked administrative datasets. Attrition due to unmatched records or migration is quantified, and confidence cues are applied when coverage falls below 85%.

Overview

When youth leave supervision or placement, longitudinal follow-up helps measure stability and change. Outcomes are tracked using cross-agency data on education, employment, and recidivism. A stable cohort design—anchored to an exit date—ensures comparability across time and programs. This page describes how those follow-ups are constructed and interpreted.

What We Track

Exit Definitions

End of formal involvement recorded with an exit date
- Probation closure
- Release from placement
- End of court jurisdiction
Exit reason coded (completion, time-served, transfer, other)

Observation Windows

Fixed follow-ups: 6 / 12 / 24 months from exit date
Window completeness checked (left/right censoring flags)

Education & Employment

Enrollment & attendance (term-based) post-exit
Diploma/GED completion verified (date, credential type)
Employment/wage records matched to UI/workforce data (coverage noted)

Recidivism & Re-engagement

New referrals / petitions / arrests within each window
Charge severity & time-to-event captured (survival-friendly)

Attrition & Censoring

Unmatched records & migration coded (out-of-state, unknown)
Sealed/expunged cases excluded with censor timestamp

Typical Flow

Define exit cohort by closure date/program/jurisdiction
- Verify follow-up windows are feasible for each exit month
Link datasets (justice, education, workforce, health)
- Use common IDs or probabilistic rules; record match score & review status
Calculate outcomes for each window
- Education: enrollment/attendance, diploma/GED
- Employment: any UI-covered earnings; wages/hours where available
- Justice: new referral/petition/arrest; time-to-event
Adjust denominators for attrition & censoring
- Flag incomplete windows; consider weights for representativeness
Publish indicators with transparency notes
- Document coverage, definitions, series breaks, and small-n handling

Fields

Field	Type	Required	Codeset	Description
pipeline_place_id	uuid	✅	—	Unique identifier for this exit-window record.
pipeline_stage_id	enum	✅	stages.yml#stage_key(8) intake detention adjudication disposition supervision commitment_and_placement reentry closure	One of the canonical stages; here = reentry_aftercare.
pipeline_place_key	enum	✅	pipeline_places.yml#place_key(45) intake pre_petition_diversion_and_deflection diversion youth_assessment_tools prosecutorial_screening status_offenses adjudication case_planning family_engagement case_timeliness language_access restorative_justice detention_screening community_atds electronic_monitoring court_appearance_and_fta initial_hearing pretrial_supervision probation intensive_supervision …and 25 more	Canonical key for this place (maps to route/slug).
youth_id	string	✅	—	Salted/hashed or state UID. No plaintext PII.
staff_id	string		—	Staff responsible for follow-up coordination (ID only).
occurred_datetime	datetime	✅	—	Anchor timestamp for row creation/update (e.g., window closeout).
jurisdiction_code	string	✅	—	County/parish/circuit or standardized local code.
legal_case_id	string		—	Case/docket identifier if exit linked to court jurisdiction.
source_system	string	✅	—	Origin system name.
source_file	string		—	Source batch/file id if flatfile.
extract_run_id	string		—	ETL run id for lineage.
series_break_flag	boolean		—	Comparability break applies to this row.
series_break_reason	enum		series_breaks.yml#reason(4) definition_change vendor_change coverage_change policy_change	Reason for break (policy/process/tool change).
exit_date	date	✅	—	Date of exit from formal involvement (cohort anchor).
exit_type_code	enum	✅	exit_types.yml#type(3) probation_closure placement_release end_of_court_jurisdiction	probation_closure, placement_release, end_of_court_jurisdiction.
exit_reason_code	enum		exit_reasons.yml#reason(5) completion time_served transfer administrative other	completion, time_served, transfer, administrative, other.
followup_window_code	enum	✅	followup_windows.yml#window(4) followup_30d followup_90d followup_180d followup_365d	6m, 12m, 24m (fixed windows from exit_date).
window_start_date	date	✅	—	Inclusive start of observation window (typically = exit_date).
window_end_date	date	✅	—	Exclusive end date of the observation window.
window_complete_flag	boolean	✅	—	True if full window observed (no right-censoring).
left_censor_flag	boolean		—	True if pre-window data missing (rare for exit-based cohorts).
right_censor_flag	boolean		—	True if observation ended before window_end_date.
censor_timestamp	datetime		—	Timestamp when censoring occurred (sealed/expunged/moved).
migration_status_code	enum		migration_statuses.yml#status(3) in_state out_of_state unknown	in_state, out_of_state, unknown.
post_exit_enrollment_status_code	enum		education_status.yml#status(6) in_school alt_edu withdrawn graduated ged unknown	Enrollment status during the window.
post_exit_attendance_days_present	integer		—	Days present accrued within the window (if available).
post_exit_attendance_days_enrolled	integer		—	Days enrolled within the window.
diploma_credential_type_code	enum		credential_types.yml#type(6) hs_diploma ged industry_cert short_term_training associate_degree other	HS diploma, GED/HiSET, industry_cert, etc., if completed within window.
diploma_credential_awarded_date	date		—	Award date of diploma/GED within the window.
workforce_match_method_code	enum		matching_methods.yml#method(3) deterministic probabilistic clerical_review	deterministic, probabilistic, clerical_review.
workforce_match_score	number		—	Probabilistic match score (if applicable).
workforce_match_outcome_code	enum		match_outcomes.yml#outcome(6) auto_match auto_no_match clerical_queue confirmed_match confirmed_no_match merged_duplicate	auto_match, clerical_queue, confirmed_no_match, confirmed_match.
ui_coverage_status_code	enum		employment_coverage_statuses.yml#status(4) full partial none unknown	full, partial, none, unknown (UI data coverage for the youth/window).
any_ui_earnings_flag	boolean		—	True if any UI-covered wages observed within window.
earnings_observation_quarter_count	integer		—	Count of quarters with UI-covered wages in window (0–4).
recidivism_event_type_codes	array<string>		recidivism_event_types.yml#type(5) referral petition arrest adjudication violation_return	New referral/petition/arrest/adjudication within window; semicolon-delimited.
first_event_datetime	datetime		—	Timestamp of first qualifying event within the window.
first_event_severity_code	enum		event_severity_levels.yml#level(4) low moderate high unknown	Severity category of first event (local mapping).
survival_days_to_first_event	integer		—	Days from window_start_date to first_event_datetime (structural store; non-negative).
coverage_note_code	enum		coverage_notes.yml#note(6) sealed_case_excluded moved_out_of_state unmatched_record incomplete_window data_hold other	Notes about data coverage (e.g., sealed_case_excluded, moved_out_of_state).
Download CSVwhat_we_track.csv

Data & Methods

Longitudinal follow-up uses cohort-based denominators and fixed observation windows (e.g., 12 months post-exit). Each dataset’s coverage is validated by match rates and duplication checks. Attrition is handled through right-censoring or weighting adjustments per Attrition & Censoring. Cross-agency data integration follows Data Linkage & Integration and Data Interoperability & Architecture. Confidence intervals and suppression rules follow Suppression & Small-n, and time trends are interpreted using Trend Interpretation & Baseline Selection.

EDORA • Learn