Intravenous Fluids—A Test Case for Learning Health Systems

“Explanatory trials evaluating the efficacy and safety of adding a new drug or device to current clinical care (A vs A-plus) are traditionally powered to detect the smallest difference in clinical outcomes that would justify incorporating the new treatment into care, considering its added risks, costs, and burdens. For example, the median minimal clinically important difference in mortality targeted by acute care trials has been approximately 8%, implying that smaller differences in mortality might not justify incorporation of the new treatments into care. These types of trials ask, “is this new treatment better than current care by enough to offset the added cost and risk?” In contrast, when attempting to select between two treatments (A vs B) that are already extensively used in current care, widely available, and similar in cost, the concept of a minimal clinically important difference is less relevant. If availability and cost are similar, then the clinical question is simply, “which treatment is more likely to produce a better patient outcome?” The choice between balanced crystalloids and saline exemplifies such a comparative effectiveness question, and the unique challenges faced in determining which of two common treatments is better when essentially any difference in patient outcomes would be sufficient to drive practice.

Balanced crystalloids (like lactated Ringer) and saline are the two commonly available options for intravenous isotonic crystalloid infusion. Both have been used for over a century, are widely available, and cost approximately $1 per liter to manufacture. More than 30 million patients in the United States receive these fluids each year. Because treatment with these fluids is so ubiquitous, differences in outcomes that might seem small at the level of the individual equate to massive differences in patient outcomes at the level of the health care system and society. For example, a true 1% absolute difference in mortality between balanced crystalloids and saline would equate to at least 200 000 preventable deaths annually (more deaths than are caused by any disease in the United States except heart disease or cancer). A true 0.1% absolute difference would equate to at least 20 000 preventable deaths annually (similar to all deaths from influenza in the United States each year). If tens or hundreds of thousands of deaths could be prevented every year at minimal added cost by simply using whichever of the two fluids results in even slightly better outcomes, what could be the argument against using the slightly better fluid? [..]

In just 16 months with only institutional funding, Bledsoe and colleagues used tools within the electronic health record to (1) enroll nearly 150 000 eligible patients, (2) separate the two study groups with regard fluid receipt (balanced crystalloids vs saline), and (3) collect, analyze, and report clinical outcomes. Had the health system applied these exact same methods, but replaced the before-and-after design with a parallel-group, cluster-randomized, or cluster-crossover design, the results would have provided rigorous evidence on the effect of fluid choice on outcomes at a scale unprecedented for any comparative effectiveness question in acute care.

Going forward, health systems should collaborate with researchers, clinicians, patients, communities, ethicists, and regulators to develop the ethical frameworks, regulations, and research methods required to (1) embed trials into clinical care to identify small differences between common treatments that translate into large improvements in patient outcomes and (2) implement the treatments found to be better into care at scale. Although transforming our current health systems into learning health systems will be challenging, for the millions of patients who receive common but unstudied treatments in clinical care each year, the costs of not doing so are simply too great.”

The editorial refers to an article published in the same issue. The research team developed a decision support tool to ask providers ordering normal saline to consider ordering lactated Ringer solution instead. The alert was suppressed if a patient’s sodium was less than 125 mmol/dL, serum potassium was greater than 6 mmol/dL, or the active problem list suggested the patient had or was at risk for elevated intracranial pressure.

Full editorial, MW Semler, JAMA Network Open, 2022.5.3