Designing for Reliability
Processes, Systems, and Tools

Reliability science
in healthcare

Reliability science is the systematic study of how systems can be designed to perform their intended function consistently and without failure. Applied to healthcare, it asks: given that humans are fallible and systems are complex, how do we design processes that produce the right outcome with a consistently high degree of probability?

At 80% reliability, a process fails once in every ten opportunities. At 99% reliability, once in every hundred. At 99.9%, once in every thousand. The ambition of high reliability is to move healthcare processes from the 80–95% range — where most currently operate — toward the upper end of this spectrum for the most critical care processes.

Moving along this spectrum requires layering multiple interventions — each adding an additional defense — so that the failure of one does not immediately produce harm. This is the practical application of the Swiss cheese model: not eliminating holes, but ensuring that when one hole appears, others do not align with it.

Standardization
protecting judgment by reducing unnecessary variation

Standardization is one of the most powerful and most misunderstood tools in reliable process design. It is frequently resisted in healthcare with the argument that clinical care cannot be standardized — that each patient is different and clinical judgment must prevail.

This misunderstands what standardization is for. Standardization does not eliminate clinical judgment. It protects it. By removing unnecessary variation from routine processes — how a central line is inserted, how a medication is prepared and labeled, how a surgical site is marked — standardization frees cognitive capacity for the genuinely complex decisions that require individual clinical judgment.

The WHO Surgical Safety Checklist — a standardized 19-item checklist completed before every surgical procedure — reduced surgical mortality by more than 40% in its original study across eight hospitals in eight countries. Not because surgeons did not know the items on the checklist, but because the structured process made it dramatically less likely that any item would be forgotten under pressure.

The hierarchy of
reliability interventions

Not all reliability interventions are equally effective. Human factors science describes a hierarchy of interventions, ranked from least to most reliable in terms of their ability to prevent errors independent of human performance.

At the bottom sit education and training — the most commonly used interventions in healthcare and the least reliable, because they depend entirely on individuals remembering and applying what they were taught. Moving up the hierarchy are warnings and reminders — alerts, checklists, and signage that prompt the right behavior. More reliable than training alone, but still dependent on human attention.

Higher still are forcing functions — system designs that make the wrong action physically impossible or automatically prevented. A medication dispensing system that will not release a drug without pharmacist verification. An anesthetic machine that will not allow a hypoxic gas mixture. These interventions work regardless of how tired or pressured the individual is, because they remove the human from the error pathway.

FMEA — Failure Mode and Effects Analysis — applies this thinking prospectively. Before a new process is implemented, teams ask: in what ways could this process fail? For each failure mode, what is the likelihood, detectability, and potential harm? Those with the highest combined risk score become priority for redesign — before any patient is harmed.

Key concepts
from this lesson