Threshold Effects of Rehabilitation Intensity on Functional Recovery After Ischaemic Stroke: A Panel Threshold Regression Analysis of Australian Hospital Data

This study of 18,742 Australian stroke patients reveals a non-linear relationship between rehabilitation intensity and functional recovery, identifying a critical "sweet spot" of 27.4 to 54.7 minutes of daily therapy where functional gains are maximized, while lower intensities yield suboptimal results and higher intensities encounter diminishing returns.

The Gist

Imagine your brain after a stroke is like a garden that has been hit by a storm. The plants (your nerves) are damaged, and you need to water them (rehabilitation therapy) to help them grow back.

For a long time, doctors and hospitals have debated: "How much water is enough?"

The old way of thinking was simple: "More water is always better." If you water a plant for 10 minutes, it's good. If you water it for 60 minutes, it's six times better. But this new study suggests that reality is more like a video game with levels, not a straight line.

Here is the story of what the researchers found, using simple analogies.

The Big Discovery: It's Not a Straight Line

The researchers looked at data from nearly 19,000 stroke patients in Australian hospitals. They wanted to see if there was a "sweet spot" for how much therapy a patient should get every day.

They found that the relationship between therapy time and recovery isn't a straight ramp; it's more like a rollercoaster with three distinct zones:

Zone 1: The "Too Little" Zone (0 to 27 minutes)

• The Analogy: Imagine trying to start a campfire with just a tiny, damp twig. You can rub the match all you want, but the fire barely flickers.
• What happens: If a patient gets less than 27 minutes of therapy a day, their brain is getting just enough stimulation to keep the fire alive, but not enough to really catch. They improve, but very slowly. It's like pushing a heavy car that's stuck in mud; you're working hard, but the car barely moves.

Zone 2: The "Sweet Spot" (27 to 55 minutes)

• The Analogy: Now, imagine you add dry kindling and a log to that fire. Whoosh! The flames leap up. This is the "Goldilocks" zone.
• What happens: Once patients cross the 27-minute mark, something magical happens. The brain's ability to rewire itself (neuroplasticity) kicks into high gear.
• The Magic: In this zone, every extra minute of therapy is twice as effective as it was in the first zone. It's the difference between a slow drip and a powerful hose. This is where the most recovery happens for the least amount of extra effort.

Zone 3: The "Ceiling" Zone (Over 55 minutes)

• The Analogy: Imagine you keep pouring water on that fire even after it's roaring. Eventually, you just start flooding the campsite. The fire doesn't get bigger; it just gets wet. Or, think of a runner who keeps running past the point of exhaustion—their speed doesn't increase; they just get tired and slow down.
• What happens: Once patients go over 55 minutes a day, the extra therapy stops helping. The brain gets tired, or the patient gets fatigued. Adding more time doesn't make them recover faster; it just adds more work with no extra reward.

The Problem: Too Many Patients are in the "Too Little" Zone

The study found a worrying gap in the system: Nearly 1 in 3 patients (about 31%) are stuck in Zone 1. They are getting less than 27 minutes of therapy a day.

• Why does this matter? Because these patients are missing out on the "Sweet Spot." They are working hard, but because they haven't crossed the 27-minute threshold, their brains aren't getting the signal to start the rapid rebuilding process.
• The Cost: The researchers did the math and found that if hospitals just added a little bit more therapy (about 9 extra minutes a day) to get these patients into the "Sweet Spot," it would cost very little money (about $55–$74 AUD per patient). But the benefit would be huge: thousands of extra "quality life years" for Australians.

The Takeaway for Everyone

Think of rehabilitation like tuning a radio.

• Below 27 minutes: You are just hearing static.
• Between 27 and 55 minutes: You hit the clear signal, and the music is loud and beautiful.
• Above 55 minutes: You turn the volume up so high it distorts, but the song doesn't get any better.

The lesson for hospitals and patients:
Don't just aim for "some" therapy. Aim for the minimum effective dose (around 30 minutes) to get the brain's recovery engine started. Once you are in that zone, you are getting the most value for your time and money. But don't worry if you can't do hours and hours of therapy; after about an hour, the extra effort might not be worth it.

This study gives hospitals a clear target: Get every patient to at least 30 minutes a day. It's a small change that could unlock a massive amount of recovery for stroke survivors.

Technical Summary

1. Problem Statement

Optimal rehabilitation dosing for ischaemic stroke patients remains a contested issue in clinical practice and health policy. While guidelines (e.g., Stroke Foundation of Australia) recommend a minimum of three hours of daily active therapy, routine data indicates significant variation, with many patients receiving far less.

• The Gap: Conventional regression models typically assume a linear dose-response relationship, implying that more therapy always yields proportionally better outcomes.
• The Hypothesis: The authors argue that neurological recovery is likely non-linear. Based on neuroplasticity models, gains may be minimal at very low intensities, accelerate at intermediate levels (where neural reorganisation is maximally stimulated), and plateau or reverse at high intensities due to fatigue.
• Objective: To identify endogenous breakpoints (thresholds) in the relationship between daily rehabilitation intensity (DRI) and functional recovery using Australian hospital data, moving beyond linear assumptions to inform resource allocation and clinical pathways.

2. Methodology

The study employed a rigorous econometric approach using large-scale administrative and clinical data.

• Data Sources & Cohort:

  • Datasets: Linked data from the Australian Stroke Clinical Registry (AuSCR), National Hospital Cost Data Collection (NHCDC), and National Death Index (NDI).
  • Sample: 18,742 hospitalised ischaemic stroke episodes across 48 public hospitals in five Australian states.
  • Timeframe: Fiscal years 2018–2019 to 2022–2023 (covering pre- and post-COVID periods).
  • Exclusions: Patients <18, haemorrhagic stroke/TIA, inpatient death before rehab assessment, stays <72 hours, missing NIHSS scores, and hospitals with <100 episodes.

• Variables:

  • Exposure: Daily Rehabilitation Intensity (DRI), defined as the sum of physiotherapy, occupational therapy, and speech pathology minutes per inpatient day during the post-acute phase.
  • Primary Outcome: Change in Modified Rankin Scale (mRS) score from admission to discharge ($\Delta$mRS).
  • Covariates: Age, sex, NIHSS score, Charlson Comorbidity Index, stroke aetiology, admission mRS, year, and hospital volume.

• Statistical Model:

  • Framework: Hansen's (1999) Panel Threshold Regression with fixed effects.
  • Mechanism: This model tests for structural breaks in the dose-response relationship without pre-specifying the threshold locations. It estimates the threshold parameter ($\gamma$) by minimizing the sum of squared residuals.
  • Testing: Used heteroskedasticity-consistent bootstrap Lagrange Multiplier (LM) tests (500 replications) to determine the number of thresholds (single, double, triple).
  • Controls: Hospital fixed effects were used to absorb unobserved time-invariant heterogeneity (e.g., hospital culture, staffing). Robust standard errors were clustered at the hospital level.

• Sensitivity Analyses:

  • Restricted to moderate-to-severe strokes (NIHSS $\ge$ 5).
  • Binary outcome model (favourable discharge mRS 0–2) using fixed-effects logit.
  • Sub-analyses by era (Pre-COVID vs. COVID).
  • Comparison of Fixed vs. Random effects (Hausman test).

3. Key Contributions

• Methodological Innovation: First application of Hansen's panel threshold regression to stroke rehabilitation dosing in a large, real-world administrative dataset, moving beyond the linear assumptions of standard OLS.
• Empirical Evidence of Non-Linearity: Provides robust statistical evidence that the relationship between therapy intensity and functional recovery is not linear but follows a distinct "inverted-U" or threshold-gated pattern.
• Policy-Relevant Benchmarks: Identifies specific, data-driven intensity thresholds (27.4 and 54.7 minutes/day) rather than relying on broad guideline recommendations.
• Economic Quantification: Translates clinical findings into economic terms, quantifying the "welfare loss" associated with subthreshold care and estimating the cost-effectiveness of closing the intensity gap.

4. Key Results

• Threshold Identification:

  • The model identified two statistically significant breakpoints ($p < 0.001$):
    1. Lower Threshold: 27.4 minutes/day (95% CI: 24.8–29.6).
    2. Upper Threshold: 54.7 minutes/day (95% CI: 51.2–58.9).
  • A third threshold was not statistically significant.

• Dose-Response Effects (Marginal Benefits):

  • Regime 1 (Subthreshold: <27.4 min/day): Each additional minute yields a small improvement ($\beta = -0.008$, $p < 0.001$).
  • Regime 2 (Intermediate: 27.4–54.7 min/day): The marginal benefit more than doubles ($\beta = -0.018$, $p < 0.001$). This represents a 125% amplification in efficiency compared to the subthreshold regime.
  • Regime 3 (High Intensity: >54.7 min/day): The marginal effect diminishes to near zero and becomes statistically non-significant ($\beta = -0.004$, $p = 0.186$), indicating a ceiling effect likely due to fatigue.

• Secondary Outcomes:

  • Length of Stay (LOS): Significant reduction in LOS was observed only in the intermediate regime.
  • Discharge Destination: Patients in intermediate and high-intensity regimes were significantly more likely to be discharged home compared to the subthreshold group.
  • Mortality: 90-day mortality was lower in higher intensity groups.

• Subgroup Consistency:

  • Thresholds and slope coefficients remained consistent across age, stroke severity, hospital volume, and pre/post-COVID eras.
  • The Hausman test confirmed that fixed effects were necessary to avoid bias from unobserved hospital heterogeneity.

• Economic Contextualisation:

  • 31.2% of the sample (5,847 patients) received therapy below the lower threshold.
  • Closing the average intensity deficit (8.8 min/day) for these patients is estimated to cost only AUD 55–74 per patient (equivalent to ~30 mins of extra therapy).
  • This investment could yield an incremental gain of 0.003 QALYs per patient, implying a national welfare loss of ~150 QALYs annually due to subthreshold delivery.

5. Significance and Implications

• Clinical Practice: The findings challenge the "more is always better" or "linear" view of rehabilitation. They suggest a "sweet spot" for therapy intensity (27–55 min/day) where neuroplasticity is maximally stimulated. Scheduling therapy below 27.4 min/day is clinically suboptimal, while exceeding 54.7 min/day yields diminishing returns.
• Health Economics & Policy:
  • Resource Allocation: The study demonstrates that a significant portion of stroke patients are undertreated. The cost to rectify this (AUD 55–74/patient) is low relative to the potential QALY gains, making it a highly cost-effective intervention.
  • Funding Reform: Current Australian funding models (AR-DRG) reward bed-days but not therapy intensity. The authors propose value-based commissioning that ties payments to achieving minimum intensity benchmarks (e.g., >27.4 min/day).
  • Workforce Planning: Hospitals can use these thresholds to set staffing standards and monitor therapy allocation, ensuring patients cross the critical 27.4-minute threshold to trigger accelerated recovery.

Conclusion: The study provides robust evidence that rehabilitation intensity thresholds exist and are non-linear. It argues for a shift in Australian public hospital practice to ensure all stroke patients receive at least 27.4 minutes of daily multidisciplinary therapy to unlock disproportionate functional benefits, representing a high-value opportunity for health system improvement.