Validation of case correctness and time intervals agreement in the Swedish registry of cardiopulmonary resuscitation using emergency medical dispatch data, 2015-2024

This retrospective validation study of the Swedish Registry of Cardiopulmonary Resuscitation (2015–2024) reveals that while unit response times align closely with emergency dispatch data, the registry's total response times are consistently shorter than dispatch records, indicating a systematic underestimation likely due to discrepancies in EMS case number accuracy and timing definitions.

The Gist

Imagine Sweden has a giant, national "Hall of Fame" for heart attacks that happen outside the hospital. It's called the SRCR. Every time a paramedic team rushes to save someone whose heart has stopped, they are supposed to write a report and add it to this Hall of Fame. This report is crucial because it helps doctors and officials figure out how to save more lives in the future.

But here's the problem: Is the Hall of Fame actually accurate?

This study is like a massive "spot check" or a detective investigation. The researchers wanted to see if the reports in the Hall of Fame (the SRCR) matched up perfectly with the Dispatch Center's own digital logs (the EMDC). Think of the Dispatch Center as the "911 control room" that sends the paramedics out. They have a stopwatch that starts the moment they get the call and stops the moment the paramedics arrive.

Here is what the investigation found, broken down into simple stories:

1. The "Name Tag" Problem (Case Numbers)

Imagine you go to a party, and the host asks you to write your name on a guest list.

• The Goal: The host (the Registry) wants your exact name so they can find you later.
• The Reality: Sometimes people forget to write their name at all. Sometimes they write "John Smith" instead of "John A. Smith." Sometimes they write "J. Smith."
• The Finding: In Sweden, about 98% of the time, the paramedics wrote some kind of ID number that helped link the two lists. However, only about 40% of the time did they write the perfect ID number.
  • The Analogy: It's like having a library where 90% of the books have a label, but only 40% have the correct barcode. If you try to find a specific book using a computer, you might miss the ones with the wrong labels. This makes it hard to combine data from different sources later on.

2. The "Stopwatch" Game (Time Intervals)

The researchers compared two stopwatches:

• Stopwatch A (The Dispatch Center): Starts when the 911 operator answers the call. Stops when the ambulance arrives. This is the "Gold Standard" because it's automatic and precise.
• Stopwatch B (The Registry Report): This is what the paramedics write down later, often from memory or a clipboard.

The researchers looked at two different ways to measure time:

A. The "Drive Time" (Unit Response Time)

• Definition: How long it takes the ambulance to drive from the station to the patient.
• The Result: Perfect match! The two stopwatches agreed almost exactly (within a fraction of a second).
• The Metaphor: This is like two runners starting a race at the same time and finishing at the same time. The paramedics are very good at knowing when they actually arrived at the scene.

B. The "Total Wait Time" (Total Response Time)

• Definition: How long it takes from the moment the 911 call is made until the ambulance arrives.
• The Result: Big mismatch! The Registry report said the ambulance arrived 80 seconds faster than the Dispatch Center's clock said.
• The Metaphor: Imagine you order a pizza. The restaurant's computer says it took 40 minutes to deliver. But when you look at your own watch, you know it took 50 minutes.
  • Why? The researchers suspect the paramedics are "rounding up" or "rounding down" the numbers. If they arrive at 10:04, they might write "10:05" or "10:00" to make the math easier. Also, sometimes the paramedic who writes the report isn't the first one to arrive; they might be the second unit, but they write down the time the first unit arrived, or vice versa.
  • The Danger: If a report says "We got there in 5 minutes" but it actually took 6.5 minutes, it makes the system look better than it really is. This is dangerous because it hides the real delays that need fixing.

3. The "Simpson's Paradox" Twist

The paper mentions a tricky statistical thing called "Simpson's Paradox."

• The Analogy: Imagine you look at the average speed of cars in a city. If you mix all the data together, it looks like everyone is driving at a steady 60 mph. But if you look at specific neighborhoods, you see that in the city center, everyone is stuck at 20 mph, and on the highway, everyone is doing 100 mph. The "average" hides the truth.
• The Finding: When the researchers looked at all the data mixed together, the times looked similar. But when they broke it down by year or region, they saw huge differences. Some areas were consistently faster in the reports than in reality.

The Bottom Line

This study is a "health check" for Sweden's heart attack data.

• Good News: The system is mostly working. Most cases are recorded, and the time it takes to drive to the scene is recorded accurately.
• Bad News: The total time from "Call 911" to "Arrival" is often recorded as being faster than it actually is. This is likely because humans are bad at writing down exact seconds, and they tend to round numbers to make them look neat.

What needs to happen?
The authors suggest that instead of asking paramedics to write down times on paper or type them in later, we should use GPS technology. Just like your phone knows exactly when you walk into a building, ambulance GPS could automatically stamp the time when the vehicle enters the "heart attack zone." This would stop the guessing, the rounding, and the "fast-forwarding" of time, giving us a true picture of how fast we are really saving lives.

Technical Summary

1. Problem Statement

The Swedish Registry of Cardiopulmonary Resuscitation (SRCR) is the national quality registry for out-of-hospital cardiac arrests (OHCA) in Sweden. While it is a critical tool for surveillance and quality improvement, its validity relies on accurate case ascertainment and precise recording of "Utstein" time variables (e.g., response times).

• The Gap: Previous estimates of SRCR coverage (often cited as ~80%) were derived from local audits rather than national validation against an external reference.
• The Risk: Without validation against a gold standard, systematic errors in case numbers and time intervals may go undetected. This could lead to selection bias in research, inaccurate temporal trend analyses, and flawed regional benchmarking.
• The Objective: To validate the correctness of EMS case numbers and the agreement of time intervals (unit response time and total response time) in the SRCR by linking them to Emergency Medical Dispatch Center (EMDC) data, which serves as an operational, real-time reference.

2. Methodology

Study Design & Data Sources:

• Type: Retrospective validation study covering January 1, 2015, to December 31, 2024.
• Primary Data: 57,708 OHCA records from the SRCR (56,969 after deduplication).
• Reference Data: EMDC-indexed suspected OHCA events from SOS Alarm (17 regions) and Sjukvårdens Larmcentral (SvLc, 4 regions).
• Ethics: Approved by the Swedish Ethical Review Authority.

Linkage Strategy:
The authors employed a hierarchical matching algorithm to link SRCR records to EMDC records:

1. Exact Match: Direct matching of EMS case numbers.
2. Partial Match: Matching based on common formatting variations (e.g., missing prefixes/suffixes, different separators).
3. Manual Review: For unlinked records, manual matching was performed using event date, municipality, region, and timestamps to identify the most plausible EMDC counterpart.

• Note: Case number formats vary significantly between SOS Alarm (e.g., 19-9876543) and SvLc (e.g., 20260101-100-26), requiring robust parsing logic.

Time Interval Definitions:

• Unit Response Time: Time from EMS dispatch to arrival at the scene.
• Total Response Time: Time from call answer/start to arrival at the scene.
• Metric: Differences were calculated as $(SRCR - EMDC)$ in seconds.

Statistical Analysis:

• Agreement: Quantified as the percentage of fully correct and partially correct EMS case numbers.
• Time Differences: Modeled using Bayesian quantile regression with an asymmetric Laplace likelihood (implemented in brms).
• Output: Posterior median differences with 95% Credible Intervals (CrI), analyzed by year and healthcare region.

3. Key Contributions

• First National Validation: This is the first comprehensive study to validate SRCR case ascertainment and time intervals against a national EMDC dataset.
• Quantification of Data Quality: It provides concrete metrics on the "completeness" and "correctness" of the registry's primary identifier (EMS case number).
• Identification of Systemic Bias: The study isolates specific sources of error, such as manual rounding in the registry versus automated system generation in dispatch centers.
• Methodological Framework: Demonstrates a hierarchical linkage approach that can be applied to other national registries lacking automated external data integration.

4. Key Results

Case Number Correctness:

• Missing Data: 1.8% (1,004) of SRCR records lacked an EMS case number entirely.
• Partial vs. Full Correctness:
  • The proportion of records with partially correct case numbers (missing dispatch codes but linkable) was ~90% until 2020, declining to ~85% in 2022–2024.
  • The proportion of fully correct case numbers increased from 32% (2015–2019) to >40% (2020–2024).
• Regional Variation: Stockholm showed the highest quality (98% fully correct), likely due to automated workflows where the registry report is a mandatory step in finalizing patient records. Two regions had 0% fully corrected numbers.

Time Interval Agreement:

• Unit Response Time (Dispatch to Arrival): High agreement between sources.
  • Median difference: -0.3 seconds (95% CrI: -3.9 to 4.0).
  • Interpretation: The time from dispatch to arrival is recorded consistently in both systems.
• Total Response Time (Call Answer to Arrival): Significant discrepancy.
  • SRCR times were consistently shorter than EMDC times.
  • Median difference: 80.9 seconds (95% CrI: -84.7 to -77.0).
  • Regional Extremes: Central Sweden showed the largest discrepancy (-132 seconds), while Stockholm showed the least (-32 seconds).
  • Temporal Trend: The discrepancy was largest in 2015 (>2 minutes) and improved slightly by 2024 (-67 seconds), but remained significant.

Analysis of Discrepancies:
The authors attribute the underestimation of total response time in SRCR to:

1. Rounding: SRCR timestamps are often entered manually as HH:MM, leading to rounding down.
2. Workflow Differences: EMDC times are system-generated; SRCR times may be retrospective or based on the reporting unit's arrival rather than the first unit's arrival.
3. Simpson's Paradox: Aggregated data masked regional variations, which were only visible when stratified by region/year.

5. Significance and Implications

• Research Validity: Researchers using SRCR data for epidemiological studies can rely on case numbers (with caveats on missing/partial data) and unit response times. However, total response times should be treated with caution as they are systematically underestimated by approximately 1.3 minutes.
• Registry Governance: The study highlights the need for:
  • Standardizing EMS case number formats across all 21 Swedish regions.
  • Integrating EMDC data directly into the SRCR to automate time-stamping and reduce manual entry errors.
• Future Improvements: The authors propose using GPS-based geofencing to automatically capture arrival and departure times, eliminating rounding and manual entry biases.
• Policy: The findings suggest that the "80% coverage" estimate may be an overestimation if based on unlinked data, and that the "gold standard" response interval (Call Answer to Arrival) requires automated linkage to be accurately measured in national registries.

Conclusion:
While the SRCR is a robust tool for OHCA surveillance, this validation reveals systematic underestimation of total response times and variable case number quality. Integrating EMDC data and automating timestamp collection are critical next steps to ensure the registry's data supports accurate clinical and policy decisions.