Skip to main content
THE TREATMENT REGISTRY
Patient journey

Outcome-data methodology

Aggregate complication, revision and mortality rates published on each /outcomes/[procedure] page (and rolled up onto the matching procedure detail page) are produced by a transparent, auditable pipeline. This page documents the rules.

Sources

  • PubMedvia NIH E-utilities (esearch, esummary, efetch). For each procedure with a registered query, the top peer-reviewed systematic-review results are fetched. Their abstracts are scanned for rate language; their PMIDs are written to the registry's source file (sources.json) under the slugpubmed-<PMID>.
  • Operator-published quality pages from individual clinic websites (Quality / Outcomes / Patient Safety pages). Where structured rates can be extracted with sample size and reporting period, they are written to clinic.outcomeData[]. Operator-self-reported figures are presented separately from the systematic-review aggregates because methodology and case selection vary.

Extraction rules

A percentage value is recorded as a rate only when it appears within an 80-character window after one of the registered rate-kind keywords:

  • Complication rate — keywords: complication, complications, adverse event(s), morbidity rate, hospital-acquired infection.
  • Revision rate — keywords: revision (surgery / procedure / operation), reoperation, reintervention, secondary surgery.
  • Mortality rate — keywords: 30-day / perioperative / in-hospital / operative / overall mortality, death rate.

Values above 60% are rejected as implausible for these outcome classes (most commonly false positives from incidence figures or subgroup statistics).

The cross-citation agreement gate

A procedure-level rate is published only when at least two independent peer-reviewed citations both report a value, and the spread between the highest and lowest reported value is within a factor of about 2.5×. Wider disagreement is treated as a signal that the sub-populations differ (operative technique, case mix, length of follow-up) and no single aggregate is written.

When the gate passes, the median of the reported values is written as the headline rate, and the minimum and maximum are written as the range. Every contributing PMID is cited.

Caveats

  • Operator-self-reported figures should not be directly compared with the systematic-review aggregates above. Reporting standards, case selection, and what constitutes a complication vary by operator.
  • Abstract-text extraction will miss rates that the authors summarised numerically in the full text but not in the abstract. Editorial review is welcome — submit corrections at /corrections.
  • Country-band aggregates and individual case-rate variation are not the same as patient-level risk. Speak with a clinician about your specific risk profile.