hospital_variance

How far apart providers actually are on one published quality measure, read off a public measure corpus rather than declared by whoever needed the number. measure_reliability requires between_provider_variance and ships no default, because the defaulted input is the one that decides the answer. This tool reads that input instead, for one named measure, out of a corpus this world holds. measure_id is required and there is no default. A variance is in the units of one measure and is not a variance in any other, so serving whichever measure happened to be first would be this world choosing which question was asked. The refusal for a missing measure_id names every measure this cut carries. The cut is one state's facilities against the measures CMS publishes for them, and it carries no readmission measure at any horizon under any spelling. Both the measure identifier and the measure name were searched for every spelling a readmission goes under, on the same run that returned every measure name in the table, and the search returned nothing. A caller who asks for a readmission measure is told the corpus has none, by name, rather than handed the nearest mortality measure: substituting one for the other would choose their conclusion by a different route than the default they were refused. The estimator ADDS where the obvious formula subtracts, and that is not a slip. These are risk-standardised rates, posterior means already shrunk toward the national rate, so the spread of the published scores understates the spread of the true rates and subtracting a sampling term would subtract the shrinkage twice. By the law of total variance the true spread is the observed spread of the scores plus the mean posterior variance behind them, which is an addition, never below the observed spread, and never negative. The posterior variance is read off each provider's own published interval as a half-width over the normal quantile for that interval, and that reading is the one step here that is not arithmetic, so it is stated on every answer. The median asymmetry of the intervals is served beside it, per measure, so a reader can see what the half-width reading costs on a lopsided interval. Scale is the trap and it does not look like an error. A variance is in the square of its measure's units, those units differ by measure, and they are not written in the table: they are CMS's measure specification, which this world has not read and does not claim to. So the answer states the scale as data, carrying the measure's own observed minimum, median, maximum and mean beside the figure, and serves both candidate conversions, one for a percentage and one for a rate per thousand, choosing neither. A variance in one measure is not a variance in any other. No roster, no ordering, no verdict. The source names hospitals and this answer never does: the cut behind it projects a few numbers per provider and no identifier at all, ordered by those numbers rather than by anything about a facility. An argument key asking for an order, for a threshold, or for the source's own better-or-worse column is refused by name rather than dropped, because a caller who sent one believes this world has an opinion about it and a silent drop leaves them believing it.

Use this form for new arguments. Follow concrete links for known values. Fetch-only readers may be unable to submit new URLs.

Exact input contract

{
  "additionalProperties": false,
  "properties": {
    "measure_id": {
      "description": "CMS's identifier for the measure, matched exactly. No case folding and no near miss: a measure id is CMS's own token, and guessing at a caller's spelling would be this world deciding which measure they meant.",
      "enum": [
        "COMP_HIP_KNEE",
        "Hybrid_HWM",
        "MORT_30_AMI",
        "MORT_30_CABG",
        "MORT_30_COPD",
        "MORT_30_HF",
        "MORT_30_PN",
        "MORT_30_STK",
        "PSI_03",
        "PSI_04",
        "PSI_06",
        "PSI_08",
        "PSI_09",
        "PSI_10",
        "PSI_11",
        "PSI_12",
        "PSI_13",
        "PSI_14",
        "PSI_15",
        "PSI_90"
      ],
      "type": "string"
    }
  },
  "required": [
    "measure_id"
  ],
  "type": "object"
}

Ready links