Measurement of quality of stroke care with national electronic health records: a prospective cohort study during and after the COVID-19 pandemic

This prospective cohort study demonstrates that linking national electronic health records in England significantly improves stroke case ascertainment and reveals a rising stroke incidence alongside suboptimal antihypertensive prescribing and socioeconomic disparities in "home-time," thereby validating the utility of integrated data for monitoring and enhancing stroke care quality beyond the pandemic era.

The Gist

Imagine the healthcare system as a massive, bustling city. For years, the city's "Stroke Department" has been trying to keep a perfect ledger of everyone who has had a stroke, how well they were treated, and how they recovered. But until now, they've been trying to do this with only a few scattered notebooks. Some notebooks were kept in the hospitals, some in the local clinics, and some in the morgue. Often, these notebooks didn't talk to each other, meaning the city was missing pieces of the puzzle.

This study is like a team of detectives who finally decided to glue all those notebooks together into one giant, digital master book. They did this to see the real picture of stroke care in England, especially during the chaotic years of the COVID-19 pandemic.

Here is what they found, broken down into simple stories:

1. The "Missing Pieces" Puzzle

The Analogy: Imagine trying to count how many people entered a stadium. If you only count the people walking through the main gate (the hospital), you miss the people sneaking in through the side doors (primary care) or the people who didn't make it past the turnstiles (those who died very quickly).

The Finding: The researchers found that if they only looked at hospital records, they were missing a huge chunk of the story.

• 10.8% of non-fatal strokes were only recorded in a doctor's office (primary care), never making it to the hospital.
• 19.4% of fatal strokes were only found in death records.
• The Lesson: To get a true count of how many people are having strokes, you have to look at every source. When they combined all the data, they saw that the number of strokes was actually rising, not falling, even during the pandemic.

2. The "Medicine Cabinet" Check

The Analogy: After a stroke, doctors prescribe a "safety kit" of medicines to prevent a second one. Think of it like a car's safety features: you need brakes (blood pressure meds), airbags (blood thinners), and a good engine tune-up (cholesterol meds).

The Finding: The team checked the "receipts" (dispensed medication records) to see if people actually got these safety kits.

• The Good News: Most people got their "airbags" (blood thinners) and "engine tune-ups" (cholesterol meds).
• The Bad News: A lot of people were missing their "brakes." Only about 44% of people got blood pressure medication, even though high blood pressure is a major cause of strokes.
• The Twist: This wasn't just a pandemic problem. It was a long-standing issue. Older people and those with other health problems were the least likely to get these "brakes," even though they needed them the most.

3. The "Home-Time" Score

The Analogy: Traditionally, doctors measured recovery by asking, "Did you survive?" This study introduced a new, more human metric called "Home-Time." Imagine a clock that starts when you leave the hospital. Every day you spend at home, the clock ticks up. Every day you spend back in the hospital, the clock stops. The goal is to maximize the "Home-Time."

The Finding:

• On average, people spent about 166 days at home in the first six months after a stroke.
• The Inequality Gap: The clock ticked slower for people who were older, lived in poorer areas, or had more severe strokes.
• The Pandemic Effect: In 2020 (the height of the pandemic), people spent less time at home. But by 2023, as things returned to normal, people started spending more time at home again, suggesting the system is slowly healing.

4. The "Pandemic Distortion"

The Analogy: Think of the pandemic as a giant fog that rolled over the city. During the fog, people were scared to go to the hospital.

• The Result: The number of recorded non-fatal strokes dropped (because people stayed home), but the number of deaths went up. This suggests that many people were having strokes but weren't getting the care they needed because they were afraid to seek help.

The Big Takeaway

This study is like upgrading the city's traffic cameras from grainy black-and-white to high-definition color. By linking all the data sources, the researchers showed us:

1. We are missing people: We need to look beyond hospitals to find everyone having a stroke.
2. We are missing medicines: We are failing to give enough people the blood pressure meds they desperately need to prevent a second stroke.
3. We need a new ruler: "Home-time" is a better way to measure if our healthcare system is actually helping people get back to their lives.

In short: The system is getting better at seeing the whole picture, but it still has work to do to ensure that every stroke survivor gets the full "safety kit" and spends as much time as possible in their own living room, not a hospital bed.

Technical Summary

1. Problem Statement

National stroke audits, such as the Sentinel Stroke National Audit Programme (SSNAP) in the UK, are resource-intensive and primarily capture data from hospitalized patients. Consequently, they may miss:

• Patients with milder strokes managed in primary care or emergency departments without admission.
• Patients who die very quickly (pre-hospital or in ED).
• Data on secondary prevention medications dispensed in the community.
• Longitudinal functional outcomes (e.g., time spent at home) beyond the acute hospital stay.

The study aims to evaluate whether linked national electronic health records (EHRs) can improve case ascertainment, measure quality metrics not routinely captured (incidence, medication dispensing, disability), and assess the impact of the COVID-19 pandemic on stroke care quality in England.

2. Methodology

Study Design & Data Sources

• Design: Prospective cohort study.
• Population: 425,675 adults (≥18 years) with a first-ever stroke in England between January 1, 2020, and December 31, 2023.
• Data Linkage: The study utilized the NHS England Secure Data Environment (SDE) to link four distinct national datasets:
  1. SSNAP: Sentinel Stroke National Audit Programme (hospital admissions).
  2. HES-APC: Hospital Episode Statistics (Admitted Patient Care).
  3. GDPPR: General Practice Extraction Service Data for Pandemic Planning and Research (primary care records).
  4. NHSBSA: NHS Business Services Authority (medication dispensing records).
  5. ONS: Office for National Statistics (death registrations).
• Linkage: Performed via the Master Person Service using pseudo-identifiers.

Case Definition & Phenotyping

• Index Stroke: Defined as the earliest record across all sources. Records within 30 days were considered the same event.
• Stroke Subtyping: Prioritized in the order: SSNAP > HES-APC > GDPPR > ONS Deaths.
  • Ischaemic: ICD-10 I63 or SNOMED-CT equivalents.
  • Haemorrhagic: ICD-10 I61.
  • Unknown: ICD-10 I64 or mixed records.
• Fatal vs. Non-fatal: Fatal defined as death from any cause within 30 days of the index stroke.

Outcome Measures

1. Stroke Incidence: Age- and sex-standardized annual rates per 100,000.
2. Secondary Prevention: One-year cumulative incidence of dispensed medications (antiplatelets, anticoagulants, antihypertensives, lipid-lowering agents) stratified by stroke type.
3. Home-time: A proxy for functional disability, defined as the number of days alive and not in hospital within 180 days post-stroke.

Statistical Analysis

• Ascertainment: UpSet plots used to visualize overlap between data sources.
• Incidence: Direct standardization to the European Standard Population (ESP).
• Medication Dispensing: Competing risk analysis (death as a competing event) and multivariable Cox proportional hazards models to estimate adjusted hazard ratios.
• Home-time: Multivariable linear regression to estimate adjusted mean days at home, controlling for age, sex, deprivation, comorbidities (Charlson index), and stroke severity (NIHSS).

3. Key Contributions

• Comprehensive Case Ascertainment: Demonstrated that relying on a single source (e.g., hospital data) significantly underestimates stroke burden. Combining all sources increased case capture, identifying 10.8% of non-fatal ischaemic strokes exclusively in primary care and 19.4% of fatal strokes exclusively in death records.
• Novel Quality Metrics: Successfully operationalized "home-time" as a person-centered outcome derived from routine data, revealing disparities linked to socioeconomic status and clinical severity.
• Medication Adherence Proxy: Utilized dispensing data (NHSBSA) rather than prescribing data, providing a more accurate proxy for actual medication adherence and secondary prevention coverage.
• Pandemic Impact Analysis: Quantified how healthcare-seeking behavior shifts during lockdowns affected recording quality (decreased non-fatal recordings, increased fatal recordings).

4. Key Results

Incidence & Ascertainment

• Standardized Incidence: Rose from 227.6 (95% CI 226.1–229.0) per 100,000 in 2020 to 244.8 (95% CI 243.4–246.3) in 2023.
• Source Overlap: SSNAP captured 51.8% of cases; GDPPR captured 71.5%. The highest incidence rates were found in primary care data, followed by hospital data.
• Pandemic Effect: During early 2020 lockdowns, non-fatal stroke recordings dropped while stroke-related deaths rose, suggesting a shift in healthcare-seeking behavior and recording sensitivity.

Secondary Prevention (Medication Dispensing)

• Ischaemic Stroke (1-year dispensing):
  • Antiplatelet/Anticoagulant: 89.1%
  • Lipid-lowering: 80.5%
  • Antihypertensives: 44.5% (Significantly suboptimal)
• Haemorrhagic Stroke:
  • Antihypertensives: 46.6%
  • Lipid-lowering: 41.1%
  • Anticoagulants: 13.5% (Expectedly low due to bleeding risk, but highlights the need for careful management).
• Disparities: Medication dispensing declined with increasing age and comorbidity burden. There was little variation by ethnicity or region.
• Trends: No significant change in dispensing rates between 2020 and 2023, indicating that under-treatment of hypertension is a systemic, long-term issue rather than a pandemic-specific disruption.

Home-time (Functional Outcome)

• Mean Home-time: 166.6 days (out of 180) overall.
• Determinants of Reduced Home-time:
  • Age: 90+ years old had ~27.6 fewer days at home compared to 15-59 age group.
  • Severity: NIHSS >22 resulted in ~21.7 fewer days at home.
  • Deprivation: Most deprived quintile had 1.6 fewer days than the least deprived.
  • Time Trend: Home-time improved post-pandemic, rising from 164.4 days (2020) to 169.3 days (2023).

5. Significance & Implications

• Public Health Burden: The rising stroke incidence highlights a growing public health challenge that persists despite pandemic disruptions.
• Clinical Optimization: The suboptimal dispensing of antihypertensives (only ~45% coverage) represents a critical, addressable gap in secondary prevention, particularly for older and comorbid patients who stand to gain the most.
• Audit Modernization: The study validates the use of linked EHRs to create a "learning health system." It suggests that future national audits should integrate primary care and dispensing data to capture the full spectrum of stroke patients, including those not admitted to hospitals.
• Policy Recommendations:
  • Implementation of the "Sudlow review" recommendations to allow routine national linkage of primary care data.
  • Adoption of "home-time" as a standard metric in stroke audits to better reflect patient-centered outcomes and socioeconomic disparities.
  • Investigation into the systemic causes of low antihypertensive prescribing in stroke survivors.

Limitations Noted:

• Potential misclassification of stroke subtypes (lower PPV in death records).
• Inability to distinguish between clinical contraindications and failure to prescribe (dispensing data does not capture intent).
• Exclusion of private sector prescriptions and recurrent strokes.
• Data truncation issues in SSNAP uploads during the study period.

In conclusion, this study demonstrates that linked national EHRs provide a robust, scalable framework for monitoring stroke care quality, revealing significant gaps in secondary prevention and functional outcomes that routine hospital audits miss.