Scoring Methodology

Overview

ForThePatient.org assigns every Medicare-certified healthcare facility with sufficient data a composite quality score on a 1.0–10.0 scale. Scores are built from 4–5 weighted components specific to each facility type, derived entirely from government-mandated reporting data published by the Centers for Medicare & Medicaid Services (CMS).

Our methodology is designed around three principles: every score must be reproducible from public data, every component weight must be defensible, and every formula must be published. If you can identify a flaw in our methodology, please file a dispute — we will investigate within 30 days.

Data Sources

All scoring data comes from CMS's Provider Data Catalog, which publishes 24 core datasets across 7 facility types. These datasets are updated quarterly and are based on government-mandated reporting — facilities cannot choose what to report or how to frame it.

As of engine v2.x (see the notice above), two additional compulsory sources are inputs: hospital survey-enforcement records (CMS Form 2567 statements of deficiencies, via CMS's QCOR system) and Medicare payment-penalty files (HRRP and HAC program results).

We use no self-reported data, no survey responses beyond CMS's own HCAHPS/HHCAHPS surveys (which are government-administered), and no data from industry groups, hospital associations, or commercial rating systems.

Additional context data — such as teaching status, ER availability, NICU, trauma center designation, and Case Mix Index — comes from the CMS Provider of Services (POS) file and the IPPS Impact File. These fields provide filtering and context but do not affect composite quality scores.

Scoring Framework

Percentile Ranking

The foundation of our scoring is per-measure national percentile ranking. For each quality measure, we compute a national distribution across all facilities that report that measure. Each facility's value is then converted to a percentile (0–100) against that distribution.

This approach is critical because different measures have vastly different scales and distributions. A readmission rate of 15% and a patient satisfaction score of 72% cannot be meaningfully compared without normalization. Percentile ranking transforms every measure to a common 0–100 scale that represents the facility's standing among its peers.

Multiple measures within a component are averaged (equal weight within a component) to produce a component percentile, which is then scaled to 1.0–10.0.

Exponential Recency Decay

Where CMS provides multiple reporting periods for the same measure, we apply exponential recency decay to weight recent data more heavily. The decay function reduces the weight of older data exponentially, so that a facility's score reflects its current quality trajectory rather than a flat historical average.

The specific half-life depends on the measure update frequency. For quarterly measures, the half-life is approximately 6 months. For annual measures, the half-life is approximately 18 months.

Score Classification

Update — engine v2.4 (see the notice at the top of this page): labels are now assigned per facility type from fixed, published thresholds, not from the single global bands below. The bands shown here reflect v1.9 and no longer match live labels; the full per-type threshold table ships with the v2.4 rewrite.

Composite scores are classified into five quality tiers, plus a sixth classification for facilities lacking sufficient data:

Hospitals

5,421 facilities · 4,597 scored

Hospital scoring uses 5 components drawn from CMS Hospital Compare datasets, plus a CMS star blend and HRRP enforcement overlay.

Component	Weight	Source
Mortality 30-day mortality rates across major conditions. Uses general_info group comparison counts (CMS no longer publishes individual MORT_30 measures).	35%	complications_deaths.csv, general_info.csv
Safety & Infections Healthcare-associated infections (CLABSI, CAUTI, SSI, MRSA, C. diff) using standardized infection ratios (SIR).	25%	healthcare_infections.csv
Readmissions 30-day readmission rates for AMI, heart failure, pneumonia, COPD, hip/knee replacement, and CABG.	20%	readmissions_deaths.csv
Patient Experience HCAHPS survey results: communication, responsiveness, environment, discharge information, overall rating.	15%	patient_experience.csv
Timely & Effective Care Process measures including ED wait times, sepsis treatment, blood clot prevention, and immunization rates.	5%	timely_effective_care.csv

CMS Star Blend

Where CMS publishes an overall hospital quality star rating (1–5 stars), we incorporate it as a 30% blend with our independently calculated composite. This acknowledges CMS's risk adjustment work while maintaining our independent methodology as the dominant signal. The star rating is converted to our 1.0–10.0 scale: 1 star = 2.0, 5 stars = 10.0.

HRRP Enforcement

Update — engine v2.3 (see the notice above): HRRP no longer produces an enforcement deduction of the kind described here. It is now part of a separate, half-weight payment signal that never flags or caps a score.

Hospitals participating in the Hospital Readmissions Reduction Program (HRRP) with an excess readmission ratio greater than 1.0 receive an enforcement deduction proportional to the excess. This captures Medicare payment penalties that the quality components alone may not fully reflect.

Nursing Homes

14,713 facilities · 14,713 scored

Nursing home scoring has the most complex enforcement overlay due to the breadth of regulatory actions CMS takes against underperforming skilled nursing facilities.

Component	Weight	Source
Health Inspections Deficiency counts and severity from state survey inspections. Higher deficiency counts and more severe findings reduce the score.	40%	health_inspections.csv
Quality Measures Long-stay and short-stay quality measures: falls, pressure ulcers, UTI rates, physical restraint use, antipsychotic medication use.	25%	quality_measures.csv
Staffing Adjusted nursing hours per resident day. Includes RN, LPN, and CNA staffing ratios, adjusted for case mix.	15%	staffing.csv
Penalties & Enforcement Civil monetary penalties (fines) and payment denials imposed by CMS.	15%	penalties.csv
Complaint Surveys Deficiencies found during complaint-triggered surveys, indicating issues reported by residents, families, or staff.	5%	health_inspections.csv (complaint type)

Enforcement Overlay

Nursing home enforcement uses a proportional deduction model rather than a flat deduction. This prevents floor clustering — when facilities with already-low composites receive flat penalties that push them all to the minimum score, making it impossible to distinguish degrees of poor quality.

Enforcement actions include: civil monetary penalties (fines) from individual penalty rows only (no double-counting), payment denials, abuse citations, and Special Focus Facility (SFF) designations. Deficiency severity feeds the inspection component, not the enforcement overlay — this prevents double-counting between quality measurement and regulatory action.

Dialysis Centers

7,557 facilities · 7,557 scored

Component	Weight	Source
Clinical Outcomes Standardized mortality ratio (SMR), standardized hospitalization ratio (SHR), and standardized readmission ratio (SRR). These use confirmed standardized ratios, not raw rates.	30%	facility.csv (DFC)
Adequacy & Safety Dialysis adequacy (Kt/V), vascular access measures (fistula rate, catheter rate), hypercalcemia, and phosphorus management.	25%	facility.csv (DFC)
Patient Survival Long-term patient survival rates and transfusion rates. Percentile-ranked against national distributions.	25%	facility.csv (DFC)
Infection Control Bloodstream infection rates (BSI) and healthcare-associated infections specific to dialysis facilities.	15%	facility.csv (DFC)
Patient Experience ICH CAHPS survey results: communication, care quality, providing information.	5%	facility.csv (DFC)

CMS Star Blend

Where CMS publishes a Dialysis Facility Compare star rating, it is incorporated as a 10% blend. Dialysis star ratings receive a lower blend weight than hospitals because the DFC star methodology has fewer risk-adjustment dimensions.

Home Health Agencies

12,251 facilities · 8,885 scored

Component	Weight	Source
Quality of Care Timely initiation of care, drug education, fall risk assessment, depression screening, and flu/pneumonia vaccination rates.	35%	quality_measures.csv (HHC)
Patient Outcomes Improvement in ambulation, bed transferring, bathing, pain management, and breathing. Also includes acute care hospitalization rate and ER use rate.	30%	quality_measures.csv (HHC)
Patient Experience HHCAHPS survey results: professional care, communication, overall rating, willingness to recommend.	15%	patient_experience.csv (HHC)
Process Measures How consistently the agency follows clinical best practices in assessment and treatment planning.	15%	quality_measures.csv (HHC)
Potentially Avoidable Events Rates of unplanned hospitalizations and ER visits that could have been prevented with better home-based care.	5%	quality_measures.csv (HHC)

All 5 home health components use actual CMS measures — no synthetic or imputed values. Home health agencies that report on fewer than the minimum number of measures are classified as Unrated.

Hospice

6,970 facilities · 6,321 scored

Component	Weight	Source
Quality of Care Pain assessment and management, dyspnea screening, treatment preferences documented, beliefs and values addressed.	30%	quality_measures.csv (hospice)
Patient & Caregiver Experience CAHPS Hospice Survey: communication, timeliness, emotional and spiritual support, symptom management, overall rating.	25%	patient_experience.csv (hospice)
Care Patterns Hospice visits in last days of life, nurse visit frequency, proportion of care delivered in the home vs. inpatient.	25%	quality_measures.csv (hospice)
Appropriateness Median length of stay, very short stays (under 7 days), very long stays, live discharge rate.	15%	general_info.csv (hospice)
Staffing & Services Breadth of services provided: social work, chaplain, bereavement, volunteer hours per patient.	5%	general_info.csv (hospice)

Inpatient Rehabilitation Facilities (IRF)

1,221 facilities · 1,149 scored

Component	Weight	Source
Functional Outcomes Change in self-care and mobility scores from admission to discharge, as measured by the IRF-PAI assessment. Uses continuous numeric values (not text performance categories).	40%	quality_measures.csv (IRF)
Safety & Infections Infection rates (CAUTI, CLABSI), falls with major injury, and pressure ulcer incidence in the rehabilitation setting.	25%	quality_measures.csv (IRF)
Community Discharge Rate of patients discharged back to the community (home or home with services) vs. to another institution.	20%	quality_measures.csv (IRF)
Process & Timeliness Medicare spending per beneficiary, potentially preventable readmissions within 30 days, and efficiency measures.	15%	quality_measures.csv (IRF)

Long-Term Care Hospitals (LTCH)

319 facilities · 316 scored

Component	Weight	Source
Functional Outcomes Change in mobility and self-care from admission to discharge, using LTCH CARE Data Set assessment values.	30%	quality_measures.csv (LTCH)
Safety & Infections CLABSI, CAUTI, MRSA, C. diff rates in the long-term acute care setting. Standardized infection ratios where available.	30%	quality_measures.csv (LTCH)
Discharge Outcomes Discharge to community rate, and potentially preventable readmission rate within 30 days of LTCH discharge.	25%	quality_measures.csv (LTCH)
Ventilator & Pressure Injuries Ventilator liberation rates and new or worsened pressure ulcer/injury rates, critical indicators for LTCH populations.	15%	quality_measures.csv (LTCH)

Enforcement Overlay

Update — engine v2.2/v2.3 (see the notice at the top of this page): hospital enforcement is now two-tier. Genuine, active survey findings can flag a facility and cap its score; HRRP/HAC payment penalties are a separate, half-weight payment signal that never flags. Boilerplate citations are excluded, and severity comes from active findings only. The description below reflects v1.9 and remains accurate for nursing homes.

Enforcement is an independent layer applied on top of composite quality scores. Quality data feeds components; enforcement captures regulatory actions. No measure feeds both a component and the enforcement overlay — this prevents double-counting.

Enforcement actions currently applied:

• Hospitals: Hospital Readmissions Reduction Program (HRRP) — excess readmission ratio > 1.0 triggers a proportional deduction derived from the readmissions component data.
• Nursing homes: Civil monetary penalties (fines), payment denials, abuse citations, and Special Focus Facility (SFF) designations. Uses proportional deduction formula to prevent floor clustering.
• Dialysis, home health, hospice, IRF, LTCH: Enforcement data is displayed when available but does not currently affect composite scores. As CMS enforcement data for these types matures, enforcement overlays will be added with appropriate methodology updates.

Context-Only Fields

Several data fields appear on facility detail pages but do not affect composite quality scores. These fields provide filtering and interpretive context:

• Teaching status — from CMS POS file. Teaching hospitals may have different outcome profiles due to case complexity, but we do not adjust scores for teaching status.
• Emergency Room availability — from POS file.
• NICU availability — from POS file.
• Trauma center designation — from POS file.
• Cardiac catheterization lab — from POS file.
• Case Mix Index (CMI) — from IPPS Impact File. Higher CMI indicates more complex cases, which correlates with but does not cause lower quality.
• Bed count — from POS file.
• Ownership type — from POS file (government, nonprofit, proprietary).

The decision to keep these fields context-only is deliberate: teaching hospitals may have higher mortality because they take harder cases, but that is a reason for interpretation, not an adjustment that would mask real quality differences.

Unrated Facilities

Facilities with insufficient data for a composite score receive a score of None (not 0.0) and are classified as "Unrated." This prevents false "worst" rankings for facilities that simply lack data.

Roughly one in ten facilities is currently Unrated. They appear on the map as gray markers. See our medical disclaimer for more information on what Unrated means for patients.

Validation

Our scoring engine ships with an automated validation suite (currently validator v2.5) covering score-range validity, component-weight verification, per-type label coverage and distribution guards, enforcement logic, geographic distribution, and edge cases.

Two permanent fixture facilities — one confirming case that must carry the enforcement flag, and one counter-example that must never carry it on boilerplate, expired findings, or a routine payment penalty — are re-checked on every enforcement or classification change.

Validation runs after every scoring-engine update and before every database seed. Full validation documentation, including current guard results, will ship with the v2.4 rewrite of this page.

Versioning & Change Log

Our scoring engine is versioned. The engine now in production is v2.4. The detailed documentation on this page covers v1.9; the v2.0–v2.4 changes are summarized below and will be fully documented in the forthcoming rewrite (see the notice at the top of this page). Every change to component weights, scoring formulas, enforcement logic, or data sources increments the version number and is documented here.

• v2.4 (June 2026) — Score classification moved to per-type, criterion-referenced thresholds, calibrated once against the June 2026 national distributions and then frozen. Threshold changes relabel only — they never silently re-score.
• v2.3 (June 2026) — Two-tier hospital enforcement: the warning flag restricted to active survey enforcement; HRRP/HAC payment penalties demoted to a separately displayed, half-weight payment signal (no flag, no cap). The hospital flag rate fell from 45.5% to 7.4%.
• v2.2 (June 2026) — Enforcement accuracy audit: boilerplate deficiency citations (26.9% of source rows) excluded at ingest and at scoring; enforcement severity derived from active findings only; plain-language summaries added for every surfaced finding.
• v2.0–v2.1 (May–June 2026) — Hospital survey-enforcement intake (CMS Form 2567 statements of deficiencies via QCOR) with exponential recency decay, per-severity lingering windows, and resolution-date imputation.
• v1.9 (April 2026) — Added POS file integration for context fields. IPPS Case Mix Index. No scoring formula changes.
• v1.8 (March 2026) — Fixed nursing home enforcement floor clustering (proportional deduction). Rebuilt all 5 home health components from actual CMS measures.
• v1.7 (March 2026) — Fixed dialysis score compression. Switched to percentile ranking on actual numeric rates.
• v1.6 (March 2026) — Corrected IRF and LTCH measure code dictionaries against actual CMS data. Added per-measure national distributions.
• v1.5 (March 2026) — Fixed hospital mortality scoring to use general_info group comparison counts (CMS removed individual MORT_30 measures).

All previous versions of the scoring engine are retained for reproducibility. If you need access to a prior version for research purposes, contact hello@forthepatient.org.