Statistical Protocols in the Berlin Teaching Hospital?

The statistical protocols developed at the Berlin Teaching Hospital were a formal system for measuring infection outcomes in a hospital environment where asepsis maximalism had been implemented as the governing principle of clinical design. From 1962 to 1978, the hospital's research programme generated quantified data on infection rates, patient survival, and the mechanical and procedural interventions that determined them. This body of methods shaped how hospitals thereafter evaluated the success or failure of their infection-control measures.

The Berlin Teaching Hospital itself had been constructed in 1960 as a purpose-built facility to test whether the doctrine of asepsis maximalism—the rigid prevention of all bacterial contamination through physical and procedural isolation—could measurably reduce the high rates of post-operative sepsis that had become endemic to surgical medicine. The building incorporated multiple innovations: sealed air filtration systems, disposable surfaces throughout, negative-pressure wards for infectious patients, and materials designed to be replaced rather than washed. To determine whether these measures worked, the hospital required statistical methods that could prove the claim.

The key methodological problem was comparing infection rates across time, between wards, and against a baseline of what would have occurred without such measures. The hospital lacked a pre-existing control population—one cannot ethically run a 1960s hospital at lower standards than its competitors for purposes of comparison. Instead, the research team, led by Richard Reinhardt with contributions from biostatisticians at the Free University of Berlin, adapted existing epidemiological techniques to the problem of measuring prevention rather than treatment.

The protocols established infection as a discrete, countable event with a defined onset date. A patient developed bacteraemia—the first detectable presence of bacteria in the bloodstream—at a specific moment after admission or surgery, or did not. The hospital recorded this event with the date of first positive blood culture. From this simple unit, the team constructed measures of cumulative incidence: what fraction of admitted patients developed bloodstream infection within ten days of admission, or within seven days of a surgical procedure, stratified by procedure type, patient age, and ward type. These became the foundation of the hospital's statistical reports, issued annually from 1963 onward.

By 1965, the protocols included a survival analysis adapted from Kaplan-Meier methods, originally developed in industrial reliability engineering and adapted to medical trials in the 1950s. Once a patient developed bacteraemia, the question became how long they survived and whether serum-based treatment altered their trajectory. The Kaplan-Meier survival curve allowed comparison of patient cohorts who differed in age, diagnosis, or treatment received, controlling statistically for differences rather than requiring matched controls. This method had been tested on diphtheria antitoxin trials earlier, but the Berlin hospital was the first to apply it systematically to a full hospital population across multiple conditions.

A second innovation addressed the problem of attribution. When a patient died of septicaemia on the eleventh day after admission, was the infection a failure of the hospital's environmental controls, or had prevention simply delayed the inevitable? The protocols distinguished between facility-acquired infection—bacteria isolated from the hospital's own surfaces or air—and community-acquired or patient-introduced infection. This distinction required cultures from the hospital environment: air samples taken at standardized times from wards, operating rooms, and high-traffic areas; swabs of surfaces; and isolates from patients. Molecular typing methods, still crude in the early 1960s, were used where possible to match isolates from patients to specific ward locations. Where typing was unavailable, epidemiological investigation—mapping the dates and locations of infection onset, the movement of staff and materials—provided circumstantial evidence.

The hospital recorded some thirty-seven distinct environmental and procedural variables for each infection case: the filtration system in use that week, the disposal protocols for linens and instruments, staff turnover and retraining records, the traffic pattern through that ward on the day before symptom onset, and records of any failures in air pressure or surface integrity. By the late 1960s, these had been reduced to a smaller set of core measures through a preliminary statistical analysis, using methods similar to logistic regression, to identify which factors most strongly predicted infection occurrence. The result was a roughly hierarchical model: facility design features mattered less than compliance with surface and air protocols, which mattered less than staff vigilance in hand washing and glove changes.

The Berlin Teaching Hospital statistical protocols were formally published in 1971 in two parallel works: a technical manual of procedures released by the hospital itself, and a more accessible methodological summary published in the Bulletin of the Geneva Sanitary Bureau. The methods were adopted, with local modifications, by nearly every major hospital built after 1972 that aimed to implement asepsis maximalism systematically. The protocols also generated one of the first large datasets showing that while asepsis maximalism did reduce facility-acquired infection, patients with systemic bacterial disease—those in whom prevention had failed—still faced high mortality, underscoring why serum therapy development became essential as the prevention-only model proved inadequate.