_pkgdown.yml

url: https://andrew-farrey.github.io/sysPrep/

template:
  bootstrap: 5
  bootswatch: flatly
  bslib:
    primary: "#2766A9"
    secondary: "#5F805C"
    font_scale: 1.0

navbar:
  structure:
    left:  [intro, reference, articles, news]
    right: [search, github]
  components:
    home:
      icon: ~
      image: man/figures/logo.png
      href: /
      aria-label: sysPrep logo
    github:
      icon: fab fa-github
      href: https://github.com/andrew-farrey/sysPrep
      aria-label: GitHub

home:
  title: "sysPrep -- NSSP ESSENCE Preprocessing Methods"
  description: >
    An R package providing formalized preprocessing methods for NSSP ESSENCE
    syndromic surveillance data: deduplication, care setting filtering,
    encounter linkage, and geographic attribution. These steps are optional
    -- most valuable for small-count, high-impact case definitions where
    minimizing false-positive clusters/anomalies matters most.
  links:
    - text: "Report a bug"
      href: https://github.com/andrew-farrey/sysPrep/issues

reference:
  - title: "Deduplication"
    desc: >
      ESSENCE data frequently contains multiple rows for the same visit due to
      data feed retransmissions, midnight-crossing date changes, patient
      identifier corrections, and patient class transitions. These functions
      identify, characterize, and remove duplicate records that would
      otherwise inflate case counts or create false-positive clusters.
    contents:
      - dedupe
      - summarize_duplicates
      - classify_duplicates

  - title: "Care Setting Quality Assurance"
    desc: >
      The ESSENCE API returns all facilities that submitted matching records,
      including non-emergency providers (primary care clinics, specialty
      practices). Free-standing emergency departments (FSEDs) may be onboarded
      with a FacilityType of "Urgent Care" -- one of two valid ESSENCE facility
      type assignments for FSEDs -- which prevents front-end API filtering by
      facility type from correctly identifying them as emergency providers.
      These functions filter to valid care settings, correct known facility
      type assignments before filtering, and flag facilities with anomalous
      visit volumes.
    contents:
      - filter_care_setting
      - review_facility_ed_visits

  - title: "Encounter Linkage"
    desc: >
      A row showing HasBeenAdmitted = 1 and HasBeenE = 0 can mean either a
      genuine direct admission or a mis-submitted ED-to-inpatient continuity
      break reported as two unrelated records -- nothing in a single row
      tells you which. The former is invisible to HasBeenE = 1 queries; the
      latter double-counts a single encounter if left unlinked.
      link_encounters() links ED and inpatient records into unified care
      episodes, correcting both problems at once.
    contents:
      - link_encounters

  - title: "Geographic Attribution"
    desc: >
      ESSENCE records include out-of-state residents and visits with unknown
      residential geography (OTHER_REGION). Discarding these visits
      systematically understates burden at facilities serving cross-border
      populations and can exclude real visits from cluster/anomaly detection
      scoped to the treating facility's county.
      assign_treating_geography() implements a hybrid approach --
      preserving residential geography for in-state patients while reassigning
      only out-of-state and OTHER_REGION visits to the treating facility's
      county. assign_facility_geography() applies full incidence-at-location
      reassignment, attributing all visits to the treating facility's geography
      regardless of patient origin. Both write to new columns by default
      (region_hybrid/region_facility), leaving the original Region/ZipCode
      untouched, so calling either -- or both, chained -- is always safe to
      try and safe to re-run.
    contents:
      - assign_treating_geography
      - assign_facility_geography

  - title: "Data"
    desc: >
      Synthetic example datasets demonstrating common ESSENCE data quality
      issues. All data are fully fabricated with no real patient, facility,
      or geographic identifiers.
    contents:
      - essence_raw
      - essence_clean
      - essence_ed_raw
      - essence_inp_raw

articles:
  - title: "Getting Started"
    desc: "Core concepts and the recommended preprocessing pipeline"
    contents:
      - getting-started

  - title: "Deep Dives"
    desc: >
      Detailed methodology for each preprocessing stage, including the
      underlying data quality mechanisms, when each step is worth applying,
      and real-world consequences observed when it was omitted.
    contents:
      - deduplication-and-classification
      - encounter-linkage
      - geography-assignment

news:
  - one_page: true