Migration

The pre-unified exports (query_event, filter_event, cohort helpers, …) remain importable from encodebox and keep their signatures. Each call emits a DeprecationWarning naming the replacement.

import warnings
warnings.filterwarnings("default", category=DeprecationWarning, module="encodebox")

Quick map

Legacy	Replacement
`query_event(...)`	`db.diagnosis.matching(...)`
`filter_event(..., occurrence, gap, limit)`	`EventFrame.occurring(...)` / `.first_per_patient()`
`query_enrollment(...)`	`Database(...).enrollment`
`compute_followup(...)`	`EventFrame.with_followup(...)`
`compute_age(...)`	`EventFrame.with_age(...)`
`derive_continuous_enrollment(...)`	`Database(...).enrollment` + EventFrame verbs
`inclusion_criteria_*`	`Cohort(...).include(...)` / `.exclude(...)`
`generate_final_cohort` / `cohort_attrition*`	`Cohort(...).to_df()` / `.attrition()`
`rwe_study(...)`	`Cohort(...)`
`compute_prevalence` / `compute_incidence`	Still `eb.compute_prevalence` / `eb.compute_incidence` (deprecated; no EventFrame/Cohort replacement yet)

Semantic traps

Occurrence counting

	Legacy `filter_event`	New `EventFrame.occurring`
Default count	Rows	Distinct dates
Same-day duplicates	Count as 2	Count as 1

Opt into legacy-style row counting with occurring(..., count_distinct_on=None).

Gap definition

	Legacy `gap=`	New `gap_days=`
Meaning	Span: `max(date) − min(date)`	Consecutive distinct-date gaps (`lead()`)
SPAN_TRAP (Jan 1/11/21/31)	kept at `gap=30`	dropped at `gap_days=30`

Do not assume gap=30 and gap_days=30 select the same patients.

Index date after a gap filter

After occurring(gap_days=...), first_per_patient() / with_index_date() use the first date of the first qualifying consecutive pair, not the earliest event on the patient.

Side-by-side examples

Query + filter → EventFrame

# Legacy
events = eb.query_event(
    connector="postgres_local_sentinel",
    code=[["E11"]], code_type=["10"], domain=["diagnosis"],
    match=["start with"], engine=conn,
)
filtered = eb.filter_event(events, occurrence="[2,]", gap=30, limit="first")

# New
db = eb.connect(..., connector="postgres_local_sentinel")
filtered = (
    db.diagnosis
    .matching("E11", code_type="10", match="starts_with")
    .occurring(at_least=2, gap_days=30)
    .with_index_date()
    .first_per_patient()
)

Inclusion ladder → Cohort

cohort = (
    eb.Cohort("study", database=db)
    .entry(filtered, index="first")
    .include(db.demographic.age_at_index(18, 89))
    .include(db.enrollment.covering_index(days_before=365, days_after=0))
)
attrition = cohort.attrition()
final = cohort.to_df()

What stays

Connector registry (register_connector, list_connectors, …)
connect without connector= still returns an engine
domain_code_type_map
Sentinel CDM YAML connectors unchanged

Getting started

Data model

Guides

API reference

Quick map

Semantic traps

Occurrence counting

Gap definition

Index date after a gap filter

Side-by-side examples

Query + filter → EventFrame

Inclusion ladder → Cohort

What stays

​Quick map

​Semantic traps

​Occurrence counting

​Gap definition

​Index date after a gap filter

​Side-by-side examples

​Query + filter → EventFrame

​Inclusion ladder → Cohort

​What stays

Quick map

Semantic traps

Occurrence counting

Gap definition

Index date after a gap filter

Side-by-side examples

Query + filter → EventFrame

Inclusion ladder → Cohort

What stays