Interpretation: this is a large, broad sample rather than the full review population. The facility-level cap prevents a handful of heavily reviewed locations from overwhelming the discovery step, while year stratification preserves coverage across the 2019–2026 window.
Interpretation: the review population is highly polarised: 89.2% of reviews are either one or five stars. That pattern is common in voluntary review data - people are most likely to post after an especially good or especially poor experience. The rising annual count reflects both review activity and the available source data; it should not be read as growth in hospital demand. The 2026 count is incomplete.
Interpretation: California, Texas and Florida account for 39.2% of the sampled reviews, so the national result is not geographically even. State rating differences are descriptive, not a league table: facility mix, review behaviour and source coverage vary substantially by state. The map is best used to understand coverage before making comparisons.
Interpretation: the four largest facility labels - hospital, emergency room, medical clinic and medical center - contribute roughly 71% of reviews. The taxonomy is nevertheless broader than emergency care alone: it also covers scheduled access, billing, maternity and pediatric care, behavioural health, rehabilitation, privacy and accessibility.
Interpretation: the final taxonomy is the main analytical product of this run. It defines a manually reviewed vocabulary of 29 topics that can be applied consistently to every review in a subsequent labelling step.
It identifies a coherent vocabulary for hospital experience analysis: the major dimensions people discuss and how fragmented source wording can be consolidated into stable topics.
To say how common each topic is, compare facilities, or chart topic sentiment over time, each review must be assigned one or more final topics. Discovery-stage topic counts cannot substitute for review-level prevalence.
The run produced a concise, operational taxonomy spanning access, staff interactions, administration, facilities, clinical care, coordination, patient rights and specialty care. The taxonomy defines what can be measured in the subsequent review-labelling step; it does not by itself show which topics are most prevalent.
Next analytical step: label reviews with the 29 final topics, then calculate topic prevalence, sentiment or rating by facility, state and year. That is the point at which the taxonomy becomes a measurement system and supports reliable time-series or geographic comparisons.