Emergency Department Data Overview

Emergency Department Data Sources

This vignette provides a comprehensive overview of the emergency department data sources used in the SEM cohort study. The data comes from two primary systems:

1. Patientliggaren (Patient Registry): A logistic system used by healthcare staff to track patient location and processes
2. SVAR (Svenska Akutvårdsregistret): The Swedish Emergency Care Register, which contains clinical data including vital signs

Each dataset follows standardized variable naming conventions and processing procedures as outlined in the Variable Naming Standards.

Comparison of Data Sources

When planning analyses involving emergency department data, it’s important to understand the relative strengths and limitations of these complementary data sources:

Feature	Patientliggaren	SVAR
Primary strength	Complete linkage to other datasets (100%)	Contains vital parameters
Linkage rate	Complete patient_id and contact_id	Incomplete (76.7% linked to patient_id)
Triage data quality	Good, directly from triage systems	Poor, especially for walk-in patients (95%+ labeled as Green)
Vital signs	Not available	Available but with substantial missingness
Chief complaint accuracy	Good	Known accuracy issues
Temporal coverage	2017-2019	2017-2018
Integration with other data	Seamless via contact_id and patient_id	Limited by incomplete linkage

Recommendation: For most analyses, the Patientliggaren dataset should be used as the primary source for emergency department visits. SVAR should be used only when vital signs data is specifically required, with appropriate caution regarding data quality.

Patientliggaren (Patient Registry) Data

The Patientliggaren system provides comprehensive tracking of emergency department visits and patient flow.

Dataset	Description	Key Variables	Timeframe
SEM Emergency Dataset	Emergency department visits during SEM cohort enrollment	chief_complaint, triage_priority, ed_care_time_mins, discharged_to	2017-2018
Post-SEM 1-Month Dataset	Emergency visits in January 2019 after SEM period	Same as SEM Emergency Dataset	January 2019
Post-SEM All Returns Dataset	All return ED visits through end of 2019	chief_complaint, arrival_time, admitted_to_hospital	2017-2019

Key Features

• Complete linkage to other SEM cohort datasets via patient_id and contact_id
• Accurate triage information with triage_priority and triage_time
• Reliable tracking of patient flow with arrival_time, discharge_time, and discharged_to
• Precise timing metrics including ed_care_time_mins and time_to_physician_mins
• Complete capture of hospital admissions with admitted_to_hospital and admission_ward

Structure and Temporal Coverage

The Patientliggaren data is organized into three datasets:

1. SEM Emergency Dataset (2017-2018): Primary dataset covering the main SEM cohort period
2. Post-SEM 1-Month Dataset (January 2019): Detailed follow-up immediately after the SEM period
3. Post-SEM All Returns Dataset (through December 2019): Extended follow-up tracking return visits

These datasets can be linked chronologically to construct longitudinal patient trajectories.

SVAR (Svenska Akutvårdsregistret) Data

The SVAR registry provides clinical data for emergency department visits, with a particular focus on vital signs.

Dataset	Description	Key Variables	Timeframe
SVAR Emergency Registry	Emergency department visits with vital parameters	vital signs (systolic_bp, heart_rate, etc.), consciousness scales	2017-2018

Key Features

• Contains vital signs data not available in Patientliggaren:
  • Blood pressure (systolic_bp, diastolic_bp)
  • Heart rate (heart_rate)
  • Respiratory rate (respiratory_rate)
  • Oxygen saturation (oxygen_saturation)
  • Temperature (temperature)
  • Consciousness level (consciousness_rls, consciousness_gcs)
• Contains death_within_7days and death_within_30days indicators
• Includes discharge_diagnosis (ICD-10 codes)

Limitations and Data Quality Issues

• Limited linkage: Only 76.7% of SVAR records could be linked to a patient_id, substantially limiting integration with other datasets
• Triage level reliability: Triage data is generally unreliable, with 95%+ of walk-ins marked as Green and completely unreliable data from Helsingborg and Ystad hospitals
• Vital parameters quality:
  • Substantial missing data
  • Values may not represent the first measurement taken
  • Unclear temporal relation to the ED visit
• Chief complaint accuracy: Known to be inaccurate in many cases
• Referral information: The referral_elsewhere field has substantial missing data

Data Integration

Linking Approaches

The datasets can be linked through common identifiers:

• patient_id: Links all records for a specific patient across all SEM datasets
• contact_id: Links records related to a specific healthcare contact/encounter (available in Patientliggaren)

Integration Caveats

When integrating these datasets:

1. SVAR linkage limitation: Only 76.7% of SVAR records have a valid patient_id/contact_id for linking to other datasets, meaning any analysis incorporating SVAR data will exclude approximately 23.3% of emergency visits
2. Temporal alignment: Ensure proper alignment of timestamps when combining datasets
3. Variable consistency: Some similar concepts are represented differently between systems