The sanatorium dossiers are the heaviest in the archive's institution shelf, and the weight is not paranormal. It is the arithmetic of a disease that killed at scale before antibiotics — and of the buildings raised to manage the dying. Between roughly 1900 and 1950, America built hundreds of tuberculosis hospitals on hillsides and in pine woods, places engineered around the only treatments that existed: rest, altitude, fresh air, and distance from the living. Streptomycin, discovered in 1943, ended the era in a decade. The buildings outlasted their purpose, and the ghosts moved into the vacancy.
The architecture of managed dying
Waverly Hills is the file everyone knows, and the file's first dispute is arithmetic. Tour literature puts the death toll at 63,000; the surviving patient death index — an ongoing transcription maintained by Kentucky Historic Institutions — supports a figure in the thousands, dreadful enough without the inflation. What is documented: Louisville's wetlands posted one of the country's worst tuberculosis death rates; the hospital opened in 1910, replaced by the five-story Tudor-Gothic pile in 1926; treatments ran to artificial pneumothorax — the deliberate collapse of a lung — and a five-hundred-foot enclosed tunnel moved the dead out of sight of the living. What is only reported: the nurse in Room 502, in either of two competing tellings, neither confirmed by any record that has surfaced.
The design was the therapy. At Seaside in Waterford — the 1934 children's hospital on the Sound — Cass Gilbert wrapped the wards in heliotropic terraces reaching toward the water, because sunlight was medicine. Undercliff opened in 1910 on a Meriden mountain for tubercular children; it was renamed and repurposed until demolition took the building in 2013 and left only the ghost-children reports. Nopeming — an Ojibwe phrase rendered as "out in the woods" — was Minnesota's first county sanatorium in 1912, quarantine as site plan: fifty beds grown to three hundred, a nurses' training school, then a nursing home until the 2000s, when the ghost hunters moved in.
Self-sufficient worlds, left standing
Glenn Dale is the quarantine file: when tuberculosis filled Washington's hospitals, the District bought two hundred acres of Maryland countryside, close enough for the railway and far enough for the air, and took the indigent first. Twenty-one Colonial Revival buildings went empty in 1982 and the abandonment made the legend — the file notes drily that boiler-plant smoke rising years later was mistaken for activity. Indiana State ran as its own town — two hundred acres, dairy, power plant, steam tunnels linking eleven buildings — and its sudden shutdown left beds, equipment and patients' suitcases in place, which is why the photographs look the way they do.
Two files deserve separate shelving. DeJarnette Sanitarium in Staunton is listed by internet consensus as Virginia's most haunted building, and its dossier is unusual in one respect: the documented record is worse than the legend. Dr. Joseph DeJarnette lobbied for Virginia's 1924 sterilization law, championed the cause that produced Buck v. Bell, and publicly admired the Nazi program; the state sterilized roughly eight thousand people before 1979 and apologized in 2002. The ghost hunters in the halls are, whatever else is true, in the right building. And St. Albans in Radford has already lived three lives — an 1892 Lutheran boys' school, a hospital from 1916, a paranormal venue — with insulin-coma treatments and 6,509 patients in 1945 on the documented side.
What the file cannot say
The record is blunt about the ratio: in these buildings, the verified history is almost always heavier than the haunting attached to it. The epidemic is documented; the nurse in 502 is not. The archive keeps both because both are real kinds of evidence — one of what happened, one of what the building does to people now.
Several of these sites are fenced, condemned or dangerous, and no access line in a dossier overrides a posted sign. The tuberculosis dead do not need a trespasser; they need the reader to notice that the legend's worst numbers are usually smaller than the file's true ones.