Leaving against medical advice as a signal of unmet care needs in adult sickle cell disease hospitalizations

A retrospective analysis of New York City hospitalizations reveals that the high rate of sickle cell disease patients leaving against medical advice, which is strongly associated with subsequent readmissions, serves as a critical indicator of unmet care needs and systemic barriers rather than solely patient responsibility.

The Gist

The Big Picture: A Broken Elevator in a Skyscraper

Imagine the healthcare system as a massive skyscraper. For most people, the elevator works fine: you get in, you go up to get treated, and you leave when you're ready.

But for adults with Sickle Cell Disease (SCD), the elevator is often broken. They have a chronic, painful blood condition that causes sudden, severe "crises" (like a car engine overheating). When they go to the hospital, they are often in excruciating pain.

This study looked at what happens when these patients get frustrated with the "elevator" and decide to leave before the doctors say they are safe. In medical terms, this is called Leaving Against Medical Advice (LAMA).

The researchers asked a simple but profound question: Are these patients just being difficult, or is the hospital failing them?

The Main Findings: It's Not Just "Bad Behavior"

The study analyzed thousands of hospital visits in New York City over two years. Here is what they found, using our elevator analogy:

1. The "Exit Rate" is Shockingly High

• The Stat: About 14% of Sickle cell patients left the hospital early. Compare that to the general population, where only 4% leave early.
• The Analogy: If you walked into a restaurant and saw 1 out of every 7 customers storming out of the kitchen before their food was served, you wouldn't just blame the customers. You'd wonder if the kitchen is on fire, the food is cold, or the staff is rude.

2. The "Broken Elevator" Varies by Building

• The Stat: The rate of people leaving varied wildly from hospital to hospital. Some hospitals had a 5% exit rate, while others had a 30% exit rate.
• The Analogy: Imagine a city with 45 different skyscrapers. In some buildings, the elevators work perfectly. In others, the doors keep slamming shut, or the buttons are broken. The study found that even after adjusting for how sick the patients were, some hospitals still had way more people leaving than others. This suggests the problem isn't the patients; it's the hospital's service.

3. The "Heavy Hitters" (Repeat Visitors)

• The Stat: A small group of patients (about 7% of the people) accounted for 40% of all the hospital visits. These are the people who are in the hospital constantly.
• The Analogy: Think of these patients as the "frequent flyers" of the hospital. They know the system better than anyone. If they are the ones leaving, it's a huge red flag that the system isn't working for the people who need it most.

4. The "Boomerang Effect"

• The Stat: Over 50% of the patients who left early came back within 30 days. Only 38% of those who stayed came back.
• The Analogy: It's like jumping off a moving train because you're angry at the conductor, only to realize you're stranded on the tracks and have to run back to the station immediately. Leaving didn't solve their problem; it just made it worse. They left because their pain wasn't managed, and they came back because the pain was still there.

Why Do They Leave? (The "Why" Behind the "What")

The paper argues that we need to stop blaming the patient. When a patient with Sickle Cell leaves early, it's often a distress signal.

• The Pain Gap: Studies show these patients often wait longer for pain medication than people with broken bones or kidney stones.
• The Bias Gap: Some doctors might think the patient is "drug-seeking" rather than actually in pain. This is like a mechanic telling a driver, "You're just trying to get a free tune-up," when the car is actually on fire.
• The System Gap: The hospital might not have the right protocols or staff trained to handle this specific disease.

The Takeaway: Fixing the Elevator

The authors suggest that hospitals shouldn't just count how many people leave and say, "Well, those patients are difficult." Instead, they should treat "Leaving Against Medical Advice" as a warning light on the dashboard.

What should hospitals do?

• Listen: Ask the patients why they left. Was the pain ignored? Was the wait too long?
• Train: Teach doctors and nurses about Sickle Cell so they don't judge the patients.
• Plan: Create a "pain plan" before the patient even gets to the hospital, so they don't have to fight for relief every time.
• Support: Bring in social workers to help with the stress and logistics of the disease.

The Bottom Line

This paper is a wake-up call. It tells us that when patients with Sickle Cell Disease leave the hospital early, it's not a failure of the patient's character. It is a failure of the healthcare system to meet their needs.

If we want to stop the "boomerang effect" (patients leaving and coming back sicker), we need to fix the elevator so they don't feel the need to jump out in the first place.

Technical Summary

1. Problem Statement

Sickle Cell Disease (SCD) is a life-threatening inherited blood disorder affecting approximately 100,000 people in the US, predominantly of African ancestry. Patients with SCD face significant systemic barriers in healthcare, including provider bias, undertreated pain, and structural racism. Consequently, SCD hospitalizations are associated with exceptionally high rates of adverse outcomes:

• Readmissions: Over 30% of adult SCD hospitalizations result in a 30-day readmission (nearly triple the national average).
• Leaving Against Medical Advice (LAMA): SCD patients are roughly four times more likely to leave the hospital against medical advice compared to the general population.

Current hospital performance metrics often exclude or overlook LAMA discharges, implicitly treating them as patient failures rather than indicators of systemic care gaps. This study aims to challenge that assumption by analyzing LAMA patterns as a signal of unmet care needs.

2. Methodology

Study Design & Data Source:

• Type: Retrospective analysis of inpatient discharge records.
• Setting: New York City hospitals (one of the largest SCD populations in the US).
• Period: January 1, 2022 – November 30, 2023 (with December 2023 data used for readmission tracking).
• Data Source: SPARCS (Statewide Planning and Research Cooperative System), an all-payer administrative database.
• Population: Adults (≥18 years) discharged with a principal diagnosis of SCD (ICD-10-CM codes D57.x, excluding trait D57.3).
• Exclusions: Hospitals with <20 SCD hospitalizations; records with missing admission/discharge dates or patient IDs.

Outcomes:

• Primary Outcome: Hospital-level LAMA rates.
• Secondary Outcome: 30-day all-cause readmissions, stratified by LAMA status.

Statistical Analysis:

• LAMA Modeling: Used Bayesian hierarchical logistic regression with a hospital-level random intercept to stabilize estimates for facilities with low event counts.
  • Covariates: Sex, age group, ED admission, severity of illness (APR-DRG), and Elixhauser comorbidities (excluding HIV).
  • Metric: Adjusted LAMA rates calculated as (Predicted Cases / Expected Cases) × 100.
• Readmission Modeling: Hierarchical logistic regression to determine the association between LAMA and 30-day readmission.
  • Models: Model 1 adjusted for clinical characteristics; Model 2 added demographic variables (race/ethnicity, insurance, neighborhood fixed effects).
• Sensitivity Analyses:
  1. Added a crossed patient-level random intercept to account for correlation among patients with multiple hospitalizations.
  2. Excluded patients with 10 or more SCD hospitalizations to test the influence of "super-utilizers."

3. Key Results

Descriptive Statistics:

• Volume: 7,939 SCD index hospitalizations vs. 1,358,810 all-cause hospitalizations.
• LAMA Rates:
  • SCD: 14% (1,121 discharges).
  • All-cause: 4% (54,755 discharges).
• Patient Concentration: A small subset of patients drives the volume. Approximately 7% of unique SCD patients had 10+ hospitalizations, accounting for 40% of total SCD hospital volume. In contrast, <1% of all-cause patients had 10+ admissions, contributing only 3% of volume.

Hospital-Level Variation:

• Adjusted hospital-level SCD LAMA rates ranged from <5% to 30% (Interquartile Range: 10–20%).
• Crude rates exceeded 30% in several high-volume facilities.
• SCD LAMA rates were consistently higher than all-cause LAMA rates across most facilities, even after adjustment.
• Sensitivity Findings: Accounting for multi-visit patient correlation modestly compressed the variation (IQR 11–19% vs. 10–20%), but significant hospital-level variation persisted. Excluding patients with 10+ visits reduced the overall SCD LAMA rate from 14% to 11%, yet variation remained.

Readmission Outcomes:

• 30-Day Readmission Rates:
  • SCD LAMA discharges: 54% (607/1,121).
  • SCD Non-LAMA discharges: 38% (2,577/6,575).
  • All-cause LAMA: 26%; All-cause Non-LAMA: 12%.
• Risk Association: LAMA was significantly associated with higher odds of 30-day readmission:
  • SCD: Adjusted Odds Ratio (aOR) = 1.63 (95% CI: 1.39–1.91).
  • All-cause: aOR = 2.10 (95% CI: 1.96–2.26).

4. Key Contributions

1. Reframing LAMA: The study provides empirical evidence that LAMA in SCD is not solely a patient behavioral issue but a robust signal of unmet care needs and systemic gaps.
2. Quantifying Variation: It demonstrates that significant variation in SCD LAMA rates exists across hospitals even after adjusting for patient case-mix, suggesting that facility-level factors (e.g., pain management protocols, provider bias) play a critical role.
3. The "Super-Utilizer" Phenomenon: The analysis highlights that a small fraction of patients (those with 10+ admissions) drive a disproportionate share of SCD volume and LAMA events, necessitating targeted care coordination strategies.
4. Readmission Link: It establishes a strong causal link between LAMA and subsequent readmissions, reinforcing the clinical and economic cost of early departures.

5. Significance and Implications

• Policy & Quality Metrics: The findings argue for the inclusion of LAMA as a quality metric in SCD care. Currently, excluding LAMA from performance measures hides systemic failures.
• Clinical Interventions: The authors propose actionable steps for hospitals, including:
  • Root cause analysis for every SCD LAMA event.
  • Standardizing clinical pathways for vaso-occlusive episodes (VOE).
  • Implementing individualized pain plans visible in Electronic Health Records (EHRs).
  • Investing in social work and care coordination, particularly for multi-visit patients.
• Addressing Structural Racism: By linking LAMA to unmet needs and provider bias, the study underscores the need to address structural racism and stigma in healthcare delivery to improve outcomes for the SCD population.

Limitations:

• Reliance on administrative data (potential under-reporting of LAMA if coded as routine discharge; lack of granular clinical data like pain management timing).
• Study confined to NYC (urban academic centers), limiting generalizability to other markets.
• Indirect standardization may not fully account for unmeasured patient complexity.

Conclusion:
Leaving against medical advice in adult SCD hospitalizations is a critical indicator of healthcare system failure rather than patient non-compliance. Monitoring and addressing the drivers of LAMA is essential to reducing readmissions and improving the quality of life for patients with SCD.