Methodology

This page explains how OpenChart Health processes and displays CMS hospital and nursing home quality data. All data shown on this site is sourced from the Centers for Medicare & Medicaid Services (CMS) Provider Data Catalog and related CMS publications. The site is a data aggregator: it republishes federal data with statistical uncertainty made visible. It does not produce ratings or rankings of its own.

Two-Tier Measure Descriptions

Each measure is described twice. A plain-language gloss (8th-grade reading level) is presented first for readability — this is our consumer-friendly interpretation, not a CMS definition. Where available, the verbatim CMS measure definition follows in a labeled block: "CMS defines this measure as: ...". The CMS definition is the legally authoritative description; the plain-language gloss is supplementary. The two are presented together so any drift between them is visible to readers.

Why No Directional Color Coding

Measure values render in neutral gray, not red or green. Color coding a value as "good" or "bad" relative to an average is an editorial judgment, and averages themselves carry uncertainty. Visual position relative to the national average — shown on the benchmark bar and the distribution histogram — communicates the comparison without imposing a third-party verdict. Color is reserved for two threshold-based signals: tail-risk states defined by external standards (immediate jeopardy citations, Special Focus Facility status, abuse findings) and repeat deficiencies cited across multiple inspection cycles. These are CMS's own categorical determinations, not statistical comparisons.

Tail Risk in the Primary View

Mortality, infections, complications, and other adverse-event measures appear in the primary profile view alongside other measures. They are not buried in a sub-panel. Adverse events are low-frequency and high-severity by nature; presenting them at the same level as higher-frequency process measures keeps the asymmetry visible to readers. Each tail-risk measure is flagged with a category badge.

Three Reporting States: Reported, Suppressed, Not Reported

A measure can be in three distinct states, each displayed prominently rather than hidden. Reported values include a numeric value, sample size, period label, and any applicable footnote codes.Suppressed values are ones CMS has withheld — usually because the sample is too small to publish without privacy or reliability concerns. The reason is shown alongside the suppression notice. Not reported values are ones the provider did not submit to CMS at all. This is a distinct state from suppression, and it is surfaced at full visual weight: a missing report is itself information about the provider.

Interval Estimates

Every measure on this site displays an interval estimate alongside the point value. Interval estimates reflect statistical uncertainty — the range within which the true value plausibly falls given the available data. A narrower interval means more precision; a wider interval means more uncertainty.

Where CMS publishes interval bounds for a measure, we display those values directly. For measures where CMS publishes a rate and sample size but not interval bounds, we calculate a 95% Bayesian credible interval using a Beta-Binomial model. This includes patient experience survey measures (HCAHPS), where CMS adjusts for patient mix but does not publish interval bounds — the sampling uncertainty from finite survey counts is real and the adjusted percentage is treated as a binomial proportion. The prior is informed by the CMS-published state average rate for the measure when available, or the CMS-published national average rate as a fallback. When neither CMS-published average is available, an uninformative prior is used. This approach provides meaningful uncertainty estimates while appropriately shrinking extreme rates from very small samples toward the population average.

Patient Experience Survey Intervals

HCAHPS patient experience scores are adjusted by CMS for patient mix (age, education, language, etc.) to allow fair comparison across hospitals. CMS does not publish interval bounds for these measures, but the sampling uncertainty from the number of completed surveys is real. A hospital where 50 patients completed the survey has much more uncertainty than one where 5,000 did.

We calculate credible intervals for the primary response on each survey question (the "Always," "Definitely Yes," or "9-10" percentage) by treating the adjusted percentage as a binomial proportion with the number of completed surveys as the denominator. These intervals are labeled as "calculated using a Bayesian Beta-Binomial model" to distinguish them from CMS-published intervals.

Why Some Intervals Are Very Wide

For HGLM-adjusted measures (mortality, readmissions, complications), CMS-published intervals are often wider than users expect. This is because the CMS hierarchical model accounts for uncertainty between hospitals, not just within a single hospital's data. A hospital with 1,800 patients may still have a wide interval because the model is conservative about how much the data should shift the estimate away from the national average.

On the distribution histogram, this means the blue shaded region (showing the plausible range for a hospital's result) may cover a large portion of the national distribution. This is not a display error — it reflects the genuine statistical uncertainty in these estimates. When the blue zone spans most of the distribution, it means the data does not provide strong evidence that this hospital differs from average.

Small Sample Sizes

Measures based on fewer than 30 cases carry an amber warning. With small samples, the observed rate is highly uncertain and may not reflect the hospital's typical performance. The distribution histogram visualizes this — a wide blue zone on a small-sample measure shows that the true rate could fall almost anywhere in the national distribution.

Standardized Infection Ratios (SIR)

Healthcare-associated infection measures use the CDC's Standardized Infection Ratio (SIR). A ratio of 1.0 means the hospital had exactly as many infections as expected given its patient volume and case mix. Below 1.0 means fewer infections than expected; above 1.0 means more.

The "expected" rate is based on a national baseline period. As infection prevention has improved nationally, the average hospital now performs better than the baseline — which is why the national average SIR for many infection types is below 1.0. This does not mean the baseline is wrong; it means hospitals have collectively improved since it was set.

CMS Direction Indicators

Each measure displays which direction CMS designates as associated with better outcomes — either "lower is better" (e.g., death rates, infection ratios) or "higher is better" (e.g., patient experience scores, compliance rates). This directional guidance comes from CMS technical documentation, not from this site's editorial judgment.

Distribution Histograms

Each measure card includes a histogram showing how all hospitals nationally are distributed on that measure. The histogram uses 25 bins computed from all reporting hospitals for the same measure and period.

• Blue shading marks the bins that fall within this hospital's interval estimate — the plausible range for the true value. A wider blue zone means more uncertainty.
• Blue dashed line shows this hospital's observed value.
• Orange dashed line shows the national average.

Comparing Two Providers

The compare page renders two providers side by side on the same measures. For each measure, both providers' values appear on a single horizontal axis with their credible intervals as whisker-and-band marks and the national average as a reference line. When the two intervals overlap, a note states that the difference between the providers may not be meaningful given the available data; when they do not overlap, the note states that the difference appears meaningful. This is a conservative factual statement, not a verdict. Trend lines for both providers are overlaid on the same chart so trajectory differences are visible without scanning between separate charts.

State averages are intentionally omitted from the side-by-side interval plot. Two providers from different states would have different state-level reference values for the same measure, and showing only one (or both) would clutter the visualization without clarifying the comparison. State averages remain available on the per-provider benchmark bar.

Sort and Filter, Not Ranking

The Explore view lets users sort hospitals or nursing homes by any CMS-reported measure and filter by state, subtype, or name. This is factual ordering of CMS data — the same operation as sorting a spreadsheet column. The default sort when a measure is selected is alphabetical by provider name, not by value: the tool does not editorialize which end of the value distribution matters. When a user sorts by value, the column renders with full context (sample size, period label, footnote codes, suppression state) so the sorted view is not stripped of the qualifications that accompany every value elsewhere on the site.

Nursing Home Specifics

Five-Star Ratings. CMS publishes a Five-Star overall rating with three domain sub-ratings (health inspection, staffing, quality measures). Star ratings are ordinal categorical outputs, not continuous rates — credible intervals do not apply to them. Instead of placing uncertainty bands on the stars, the site shows the constituent measures that feed each domain so the inputs are visible.

Scope and Severity Codes. Inspection deficiencies are coded A through L. Codes A–C indicate no harm with minimal risk; D–F indicate no harm with potential for more than minimal harm; G–I indicate actual harm short of immediate jeopardy; J–L indicate immediate jeopardy to resident health or safety. J–L citations and repeat deficiencies in the same category across multiple inspection cycles are the only states where color encoding is used (orange).

Ownership Data. CMS publishes nursing home ownership records identifying named individuals, organizations, and management companies. The site displays this data as structural information without editorial characterization. Association with a facility does not establish a causal relationship between ownership structure and quality outcomes; ownership and quality are republished from separate CMS datasets and presented in the same view so the juxtaposition is visible. Cross-facility patterns within an ownership group are presented factually (e.g., "N of M facilities cited for Category X in their most recent inspection cycle") without narrative interpretation.

Special Focus Facility (SFF) Status. SFF and SFF Candidate designations are CMS determinations for facilities under intensive oversight due to inspection history. These appear at the top of a facility profile with full visual prominence — they are gate conditions, not metadata. The abuse icon, when CMS has flagged a substantiated finding, is treated the same way.

Population Context Limitations

CMS risk-adjusts most outcome measures for clinical factors (age, comorbidities, diagnosis) but does not fully adjust for patient socioeconomic characteristics. Hospitals serving higher proportions of low-income or dual-eligible patients may show higher rates on readmission and mortality measures partly due to patient population factors, not care quality alone. The site does not currently include hospital-level socioeconomic population data that would aid direct interpretation. The SES disclosure block accompanies any measure group containing SES-sensitive measures so the limitation is visible wherever it applies.

Data Sources and Refresh Cadence

All data is sourced from the CMS Provider Data Catalog and related CMS publications. Individual measure sources, including dataset name and reporting period, are listed on each measure card under the Source toggle.

Hospital quality data is typically refreshed quarterly by CMS, with annual refreshes for some measures. Nursing home Provider Information and inspection records refresh monthly. SNF QRP and SNF VBP refresh on federal fiscal year cycles. The site re-ingests CMS data after each publication; the date a provider's record was last refreshed is shown in the page footer.

Measure methodology occasionally changes between reporting periods. CMS signals these via footnote code 29. Trend lines that cross a methodology boundary are flagged so cross-period comparisons do not silently mix incompatible metrics.