Methodology & Data Sources

Data Source

Hospital quality data comes from the Centers for Medicare & Medicaid Services (CMS) Hospital Compare program, the most comprehensive publicly available source of hospital quality information in the United States. CMS collects and publishes standardized quality measures for Medicare-certified hospitals, including overall star ratings (1 to 5 stars), patient experience survey results (HCAHPS), healthcare-associated infection rates from CDC NHSN, unplanned readmission rates, mortality rates for specific conditions, timely and effective care measures, and Medicare spending per beneficiary. The full dataset is published via the CMS Provider Data Catalog. This is the same dataset used by Medicare.gov to help beneficiaries compare hospitals.

Medical device adverse event reports come from the FDA's MAUDE (Manufacturer and User Facility Device Experience) database via the openFDA API, covering reports submitted by manufacturers, healthcare facilities, and patients about medical devices. These reports are raw submissions, not confirmed safety findings, and PlainHospital presents them exactly as documented in that caveat on every device page.

Data Vintage and Update Frequency

CMS updates Hospital Compare data periodically, typically on a quarterly basis. Quality measures reflect data collection periods that vary by measure, patient experience data is based on surveys collected over a 12-month period, while infection and mortality rates may reflect data from multiple years to ensure statistical reliability. PlainHospital refreshes its database when CMS publishes updated Hospital Compare files. FDA MAUDE device reports are a periodic snapshot we take from the openFDA API, not a live mirror of it: openFDA refreshes its own feed weekly, while our copy is re-fetched on our ETL cycle and can lag it. Treat the device pages as accurate as of our snapshot rather than as of today, and check the FDA's own MAUDE search for anything filed since.

Cross-stage populations (do not mix)

Several numbers on the same page come from different official populations. PlainHospital never divides or averages them against each other as if they shared a denominator.

• CMS overall star average vs hospital count: Statewide and national average star ratings are computed only over hospitals with a published overall rating. The catalog hospital_count includes unrated facilities; whenever an average sits next to a count, PlainHospital discloses "N of M rated" so the average is never read as covering the full catalog.
• CMS measure windows: On a hospital page, infection SIRs, HCAHPS experience scores, mortality/readmission measures, spending, and visits each carry their own CMS start_date–end_date. Dataset temporalCoverage on the hospital page is the min–max of those measure windows for that facility, not a single portal refresh stamp applied as if every measure shared one year.
• MAUDE event type vs patient outcome: Headline "Death / Injury / Malfunction" counts are FDA event_type (one code per report). The Patient Outcomes chart is FDA patient.sequence_number_outcome (a report can list multiple outcomes). An event-type Death count is not the same population as an outcome labeled Death, and percentages in each chart use that chart's own total.
• MAUDE yearly trends: Year bars are counts of reports by date_received year for that device or manufacturer. They sum to the entity's total events (modulo sparse/unparseable dates); they are not a second independent universe overlaid on CMS hospital counts.
• CMS vs MAUDE: Homepage hospital counts and star-rating shares use only CMS Hospital Compare. Device-safety surfaces use only MAUDE. Coverage percentages never mix the two sources.

Coverage

PlainHospital covers Medicare-certified hospitals nationwide, including general acute care hospitals, critical access hospitals in rural areas, and specialty hospitals. Coverage is limited to facilities that participate in Medicare, private hospitals that do not accept Medicare are excluded. Some hospitals may lack overall star ratings if they have insufficient data across the required quality measure groups for the CMS star rating methodology. Separately, PlainHospital covers medical devices with FDA MAUDE adverse event reports and VA medical facilities with published Access to Care wait-time data; these are distinct FDA and VA datasets, not part of the CMS Hospital Compare coverage figure above.

How CMS Star Ratings Work

CMS assigns overall hospital star ratings from 1 to 5 stars using a composite of quality measures across five groups:

• Mortality (22% weight): Death rates for heart attack, heart failure, pneumonia, COPD, stroke, and coronary artery bypass graft
• Safety of care (22%): Rates of serious complications and patient safety events
• Readmissions (22%): 30-day readmission rates for multiple conditions
• Patient experience (22%): HCAHPS survey scores on communication, responsiveness, cleanliness, and overall rating
• Timely & effective care (12%): Process measures for emergency department and outpatient care

CMS uses hierarchical clustering to assign star ratings based on the distribution of composite scores across all qualifying hospitals. PlainHospital displays star ratings exactly as assigned by CMS without modification.

Key Measures Explained

• HCAHPS (Patient Experience): Hospital Consumer Assessment of Healthcare Providers and Systems, a standardized survey of recently discharged patients
• HAI (Healthcare-Associated Infections): Standardized Infection Ratios (SIRs) compared to national baselines. SIR < 1.0 means fewer infections than predicted
• MSPB (Medicare Spending Per Beneficiary): Medicare spending for each hospital compared to the national median

Accuracy Commitment

PlainHospital reproduces CMS Hospital Compare data exactly as published. Star ratings, quality measure scores, infection ratios, and patient experience percentages are presented without editorial modification or proprietary reweighting. Hospital names, addresses, and ownership types are displayed as reported to CMS. When CMS updates its star rating methodology or publishes revised quality data, PlainHospital incorporates the changes in its next database refresh.

Corpus placement (#N of M)

Rated hospital detail pages carry a corpus-relative placement derived from CMS Hospital Compare overall star ratings, never a hand-picked list:

• Hospitals - ranked by overall CMS star rating (star_rating DESC, name ASC); highest = #1. Matches the top-rated hospital listing. Hospitals with a NULL star rating are excluded from the denominator.
• States - ranked by average CMS star among rated hospitals in the state (highest = #1).
• Devices - ranked by FDA MAUDE total adverse-event count (total_events DESC); most events = #1.

Ties break on name/slug ASC so rebuilds stay stable. These placements are informational only and are not medical advice.

Limitations

• Star ratings are backward-looking, they reflect past performance data collected over multi-month to multi-year periods, not a guarantee of current care quality.
• Not all hospitals have sufficient data for all quality measure groups. Missing measures can affect whether a hospital receives a star rating and may influence the final rating calculation.
• Specialty hospitals (oncology centers, orthopedic facilities, cardiac hospitals) may be scored differently than general acute care hospitals, and direct comparison across hospital types should be made with caution.
• CMS star ratings use a complex methodology involving hierarchical clustering and latent variable modeling. Hospitals near rating boundaries may shift between star levels with small changes in underlying measures.
• This data does not constitute medical advice. Always consult with your healthcare provider about care decisions. PlainHospital is not affiliated with CMS or any government agency.

Contact

Questions about our methodology? Contact us.