Proprietary Research · Healthcare Services Intelligence

Healthcare is
Stabilizing

We read 1.2 million sentences from 33,000 SEC filings using 29 ML classifiers. The data shows exactly when healthcare broke down — and when it recovered.

March 2026 · 1.2M sentences analyzed · Biotech Alpha Analytics

The Scale of the Problem

How do you figure out if healthcare is recovering? You can't just read one company's filings. You need to read all of them — and you need a way to cut through the noise.

1.2M
Sentences analyzed
(2018–2026)
33K
SEC filings
(10-K and 10-Q)
82
Healthcare companies
(SIC 8000–8093)

We built 29 ML classifiers that read every sentence and answer specific questions: Is this company talking about improving utilization? Pricing pressure? Revenue growth? Going concern risk?

But raw classifier scores are noisy. Quarter-to-quarter, they bounce around. A 10-K always scores higher than a 10-Q. You can't see trends with the naked eye.

The solution: Compare each quarter to its own recent history using a rolling z-score. This normalizes for seasonality and noise, and reveals when something statistically unusual is happening. Above +2σ = breakout. Below -2σ = breakdown. These are rare events (<5% of quarters).

29 Classifiers Tell the Story

We combine 18 positive classifiers (growth, strength, recovery) and 11 negative classifiers (pressure, decline, risk) into a single sector health score. Then we measure it against recent history using a rolling z-score.

18
Positive classifiers
(growth, strength)
11
Negative classifiers
(pressure, decline)
±2σ
Breakout threshold
(top/bottom 2.5%)
Sector Z-Score: Statistical Breakdowns and Breakouts
Z-score measures how far the current quarter deviates from the rolling 8-quarter average. Gray band = normal range. Outside ±2σ = statistically significant signal.
How to read this: The gray band shows where 95% of quarters fall (normal operations). When the line drops below -2σ, something truly bad is happening. When it rises above +2σ, the sector is outperforming its own history. COVID and Delta created two clear breakdowns. Recovery came in three waves: 2022 Q2, 2024 Q2, and 2025 Q2.

The Five Signals Since 2020

Q3 2020: COVID Peak
-2.4σ — The worst quarter in our dataset. Utilization collapsed, labor costs spiked, elective procedures halted. 672 filings all telling the same story.
Q3 2021: Delta Wave
-2.1σ — Just as recovery was starting, Delta hit. Another statistically significant breakdown, though less severe than the initial COVID shock.
Q2 2022: First Recovery
+2.6σ — The first true breakout. Labor markets stabilizing, volumes returning, management language shifting from crisis to recovery.
Q2 2024: Second Wave
+2.2σ — Continued momentum. Pricing power holding, utilization improving, reimbursement secure. The sector was genuinely healing.
Latest Signal (Q2 2025): +2.3σ — The third consecutive recovery breakout. This isn't noise; this is a statistically significant trend. The sector has entered a new phase.

How We Built This

This analysis required processing 1.2 million sentences from 33,000 SEC filings. Each sentence was evaluated by 29 ML classifiers, producing billions of probability scores. We then aggregated by quarter and applied rolling z-score normalization to reveal statistically significant events.

33,000 SEC filings
→
1.2M sentences
→
29 ML classifiers
→
Rolling Z-Score
→
Breakout signals
Z-Score = (Current − Rolling 8Q Average) ÷ Std Dev
Compares each quarter to its recent history. Above ±2σ = statistically rare event (<5% of quarters).

The 29 Classifiers

We combine 18 positive classifiers and 11 negative classifiers from healthcare-specific and general-purpose models:

Positive (18)
utilization_improving, pricing_power, reimbursement_improving, procedure_volume_growing, membership_growing, occupancy_increasing, same_facility_growth, service_line_expanding, star_rating_improving, premium_pricing_power, reimbursement_secured, volume_inflecting, mlr_stable, revenue_growing, above_expectations, positive_outcome, indication_expansion, market_expansion
Negative (11)
utilization_declining, pricing_pressure, reimbursement_pressure, revenue_declining, below_expectations, cost_cutting, going_concern_warning, financing_desperation, supply_chain_issue, regulatory_setback, negative_outcome

Key Takeaways

  • 1 COVID created two distinct breakdowns. Q3 2020 (-2.4σ) and Q3 2021 (-2.1σ) were statistically rare events — the worst quarters in our dataset.
  • 2 Recovery came in waves. Three breakout quarters (Q2 2022, Q2 2024, Q2 2025) show a sustained recovery pattern, not a one-time bounce.
  • 3 The sector has stabilized. After five years of volatility, filings are returning to the normal range. The crisis language is fading.
  • 4 This is a leading indicator. SEC filings reflect management's operational reality before it shows up in earnings. The z-score turned positive before the stocks did.