Mortality outcomes after removal from end-of-life registers: A prospective cohort-controlled study

This prospective cohort study demonstrates that GP-led removal of patients from end-of-life registers via a systematic digital review process is safe, identifies a distinct group with significantly better survival outcomes, and supports the integration of such digital systems into routine practice and national guidance.

The Gist

Imagine a hospital or a doctor's office keeps a special "Watch List" for patients who are very sick and might not have much time left. This list helps doctors and nurses make sure these patients get extra care, support, and attention when they need it most. It's like having a safety net for people in their final chapter of life.

But here's the tricky part: What happens if a patient on that list suddenly starts feeling better? Or what if, after a closer look, the doctor realizes the patient actually has a good chance of living for several more years?

For a long time, no one really knew what to do with those patients. Should they stay on the list forever? If they are taken off, is it safe? Could they get hurt because they lose that special attention?

This study is like a detective story that answers those questions.

The Story of the "Digital Detective"

The researchers in Wolverhampton, UK, set up a clever system. They used a digital "robot" assistant (think of it like a super-smart librarian) to scan through the medical records of everyone on the "Watch List."

This robot didn't make the final decision. Instead, it acted like a traffic light, flashing a signal to the doctors: "Hey, look at this patient. They don't seem to have the usual warning signs of being in their final year. Let's take a closer look."

The doctors then sat down and reviewed these patients personally. They asked themselves: "Is this person still in the 'last year of life' category, or have they turned a corner?"

The Big Discovery

Out of 422 people on the list, the doctors decided to take 33 people off the "Watch List." That's about 8 out of every 100.

The big fear was: "If we take them off the list, will they die sooner because they aren't getting that extra care?"

The answer was a big NO.

In fact, the study found that:

1. They lived just as long as healthy people: The 33 people who were taken off the list had survival rates almost identical to a group of people who were never on the list to begin with.
2. They lived much longer than those who stayed: The people who stayed on the list had a much higher chance of passing away within a year compared to the ones who were removed.

The "Safety Net" Metaphor

Think of the "Watch List" like a life jacket for a swimmer in rough water.

• The Problem: Sometimes, the water calms down, and the swimmer is actually a strong swimmer who doesn't need the jacket anymore. But if they keep wearing it, it's heavy and uncomfortable.
• The Old Way: Just leave the jacket on forever because you're scared to take it off.
• The New Way (This Study): The doctor checks the water. If it's calm, they say, "Okay, you can take the jacket off." But here's the genius part: They don't just throw the jacket away. They put it in a "backpack" (a digital surveillance list) and check on the swimmer every six months. If the water gets rough again, they can put the jacket back on immediately.

Why This Matters

This study proves that lists should be alive and changing, not static.

• It's not about giving up: Taking someone off the list doesn't mean the doctor is giving up on them. It means the doctor is saying, "You are doing better than we thought, and you don't need the 'end-of-life' label right now."
• It saves confusion: It stops patients from being labeled as "dying" when they are actually stable.
• It's safe: Using a digital system to double-check the list ensures that no one is accidentally forgotten. If they are taken off, they are still watched closely, just in case.

The Bottom Line

This research shows that we can be brave enough to update our lists. By using technology to help doctors review who is on the "end-of-life" list, we can ensure that:

1. People who need the most help get it.
2. People who are recovering aren't stuck with a label that doesn't fit.
3. Everyone is kept safe with a digital "safety net" that watches over them even after they leave the list.

It's a reminder that in healthcare, flexibility and regular check-ins are just as important as the initial care.

Technical Summary

1. Problem Statement

End-of-life (EOL) registers in primary care are designed to identify patients in their last year of life to facilitate proactive, coordinated care. While national guidance emphasizes the importance of early identification, there is a critical gap in understanding the review and removal process.

• The Issue: Patients on EOL registers may stabilize or have their prognosis revised, yet there is little evidence regarding the safety or outcomes of removing patients from these registers.
• The Risk: Without structured review, patients may remain on registers unnecessarily, potentially leading to misaligned care goals, increased clinical workload, and "labeling" effects. Conversely, premature removal could lead to missed care opportunities.
• The Knowledge Gap: It is unknown how frequently GPs remove patients, the clinical characteristics of those removed, and whether removal correlates with adverse mortality outcomes.

2. Methodology

Study Design:

• Type: Prospective, cohort-controlled observational study.
• Setting: Eight general practices in Wolverhampton, UK (a multi-ethnic, socioeconomically deprived area).
• Population: All adults (≥18 years) on the EOL register (EOL_R) within the participating practices ($n=422$).
• Follow-up: 15 months.

Intervention & Digital Infrastructure (PRADA System):

• The study utilized the PRADA (Proactive Risk-Based and Data-Driven Assessment of Patients at the End of Life) digital pathway.
• BOT (Robotic Process Assessment): An automated algorithm scanned integrated health data (primary, secondary, and community) for five specific EOL process markers:
  1. The "Surprise Question" (GP would not be surprised if the patient died within a year).
  2. Electronic Palliative Care Coordination System (EPaCCS) documentation.
  3. ReSPECT (Recommended Summary Plan for Emergency Care and Treatment) documentation.
  4. Advanced EOL care plans.
  5. Coded contact with Specialist Palliative Care teams.
• Workflow: GPs reviewed all 422 registered patients using the PRADA dashboard. They made a binary decision: Retain or Remove from the register based on clinical judgment.
• Safety Mechanism: Patients removed from the register were not discarded; they were placed on a "digital EOL_R tracker" for mandatory surveillance review at 6 months (or sooner if clinically indicated).

Cohorts & Controls:
The study analyzed four distinct groups:

1. Removed Cohort: Patients removed from EOL_R after GP review ($n=33$).
2. Retained Cohort: Patients remaining on EOL_R ($n=389$).
3. Propensity-Matched Control: A 1:1 matched cohort of patients not on the EOL_R, matched for age, sex, ethnicity, Index of Multiple Deprivation (IMD), and multimorbidity ($\ge$3 conditions) ($n=354$).
4. Residual General Population: The remaining unselected population ($n=38,303$).

Statistical Analysis:

• Mortality was the primary outcome, sourced from hospital statistics and NHS Strategic Tracing Service.
• Comparisons used Chi-square tests for proportions, ANOVA for means, and Cox regression for survival analysis.

3. Key Results

Prevalence of Removal:

• GPs removed 33 patients (7.8%) from the EOL register following systematic review.
• All removed patients were "BOT-negative" (lacking any of the five automated EOL process markers), suggesting they were registered based on clinical intuition prior to the digital audit.

Mortality Outcomes:

• Removed vs. Retained: The mortality rate in the removed group was significantly lower than in the retained group (9.1% vs. 39.6%, $p<0.001$).
• Removed vs. Matched Control: The mortality rate in the removed group was statistically indistinguishable from the tightly propensity-matched control cohort (9.1% vs. 7.1%, $p>0.05$).
• Negative Predictive Value (NPV): The decision to remove a patient had an NPV for mortality of 90.9% (30 out of 33 removed patients were alive at the end of the study).
• Sub-analysis of BOT-Negative Patients: Among patients on the register who lacked EOL process markers (BOT-negative), those who were retained had a significantly higher mortality rate (35.6%) than those who were removed (9.1%), indicating that GP judgment successfully discriminated between high-risk and lower-risk patients within the "grey" zone.

Demographics:

• Removed patients were younger and had fewer urgent care events (A/E attendances/Non-Elective Admissions) compared to the retained group.
• Notably, the removed group had a higher prevalence of nursing home residence compared to the matched control, yet still demonstrated survival rates equivalent to the control group.

4. Key Contributions

1. Validation of Dynamic Registration: The study provides empirical evidence that EOL registration should be a dynamic process. It confirms that a significant minority of patients (approx. 8%) stabilize or no longer meet the criteria for EOL care, and removing them is safe.
2. Safety of De-escalation: It demonstrates that GP-led removal, supported by digital governance, does not lead to adverse mortality outcomes. The removed cohort's survival was equivalent to a general population cohort with similar comorbidities.
3. Integration of Clinical Judgment: The study highlights the value of combining automated data signals (BOT) with human clinical judgment. While the BOT identified markers, the GP's decision to remove patients successfully identified a low-risk subgroup that the automated system alone might not have distinguished from high-risk patients.
4. Governance Framework: It proposes a novel "fail-safe" model where removal does not mean abandonment. The "digital tracker" ensures continued monitoring, addressing the fear that removal leads to loss of care.

5. Significance and Implications

• Clinical Practice: Supports the implementation of routine, structured reviews of EOL registers. It suggests that GPs can confidently de-escalate care for stabilizing patients without fear of missing deaths, provided a governance framework (like the tracker) is in place.
• Health Informatics: Validates the utility of "robotic process" tools not just for case-finding (inclusion), but for quality assurance and review (exclusion). It shows how digital systems can operationalize the "Surprise Question" and other markers into a workflow that prompts necessary human review.
• Policy & Guidance: Calls for national guidance to explicitly include protocols for removal from EOL registers, not just inclusion. It argues that current incentive frameworks (like the Quality and Outcomes Framework) should reward the accuracy of the register (removing those who no longer qualify) rather than just the volume of registrations.
• Resource Efficiency: By identifying patients who no longer require intensive EOL planning, healthcare systems can reduce unnecessary clinical workload and focus resources on patients with a genuine, imminent need for palliative care.

Conclusion:
The study concludes that systematic, digitally supported GP review of EOL registers is feasible, safe, and clinically valuable. It identifies a distinct patient subgroup with a substantially better prognosis than those retained, supporting a shift toward a dynamic, evidence-based approach to end-of-life care registration.