How ACIP Turns Vaccine Evidence Into a Recommendation

When a new vaccine reaches your clinic, the recommendation that comes with it is not a summary of a single trial. It is the output of a two-step public process run by the Advisory Committee on Immunization Practices (ACIP): first the committee rates how certain the science is, then it decides, in the open, whether and for whom the vaccine should be recommended. Understanding those two steps explains a lot about why some recommendations are firm defaults and others are left deliberately open.

Key points#

• ACIP separates two questions: how good is the evidence, and given that evidence, what should we recommend?
• GRADE answers the first by rating certainty about each outcome as high, moderate, low, or very low.
• The Evidence to Recommendations (EtR) framework answers the second by scoring seven domains, from the size of the public health problem to equity.
• A smaller work group builds the analysis; the full committee then votes in public, and the CDC Director reviews the result.
• The phrasing of the final recommendation, especially "shared clinical decision-making," signals how much room is left for an individual choice.

The problem the process is built to solve#

A promising trial and a good policy are not the same thing. Data can look convincing yet leave real uncertainty, and even solid data does not decide, on its own, whether a vaccine is worth recommending for a whole population. Cost, feasibility, public acceptance, and fairness all sit outside the trial. ACIP's answer is to handle the science and the policy question with two different tools so that neither one stands in for the other. A strong dataset does not get to declare itself a strong recommendation, and thin data does not get dressed up as certainty.

The committee formally adopted GRADE for the first job in 2010. For the second, it adopted a modified Evidence to Recommendations framework by a unanimous vote at its February 2018 meeting, as the committee's published framework update in the MMWR describes.

Step one: rating the evidence with GRADE#

GRADE stands for Grading of Recommendations, Assessment, Development and Evaluation. It rates certainty, not the number of studies. A work group starts by naming the outcomes that actually matter to patients, such as symptomatic illness, hospitalization, or serious adverse effects. For each of those outcomes it looks at the whole body of evidence and assigns one of four certainty levels: high, moderate, low, or very low. In ACIP's earlier tables these were numbered as evidence types 1 through 4.

Randomized trials usually enter at high certainty and can be marked down for problems such as risk of bias, inconsistent results across studies, indirect comparisons, imprecision, or hints of publication bias. Observational studies generally enter lower and can rise if the signal is strong and consistent.

The useful honesty here is in what a low rating means. "Low certainty" about a harm does not say the harm is real or that it is imaginary. It says the studies in hand cannot settle the question yet. That humility carries straight into the next step.

Step two: weighing seven domains with EtR#

Rated evidence still has to be turned into a decision. The EtR framework does this by making the work group answer a fixed list of questions, each a separate domain. Six domains came with the 2018 framework, and a seventh, equity, was added in 2020.

• Public health problem. Is this disease actually a priority in the group under consideration?
• Benefits and harms. How large are the good effects, how large are the bad ones, and which way does the balance tip?
• Values. How much do the affected people care about these outcomes, and how much does that vary from person to person?
• Acceptability. Will clinicians and the public accept the recommendation?
• Feasibility. Can it realistically be carried out in everyday practice?
• Resource use. Is this an efficient use of money, judged with cost-effectiveness analysis?
• Equity. Would the recommendation widen, narrow, or leave unchanged the health gaps between groups?

Every domain gets an explicit written judgment, backed by the graded evidence plus extra inputs such as economic models and provider surveys. Writing them down is the whole point. Later, anyone can see which domains carried the decision and which rested on assumption.

From a small work group to a public vote#

Most of this analysis happens inside a work group, a smaller body that gathers the studies and assembles the EtR assessment. The work group does not set policy. It brings its evidence tables and domain judgments to the full committee at a public meeting, where members debate them, take public comment, and vote. A majority of voting members carries a recommendation. When a key piece of information is missing, the committee can hold off rather than force a decision.

Even a yes vote is not the last word in federal guidance. Under the committee's described process, a recommendation that ACIP approves goes to the Director of the CDC for review and, if adopted, is published in the Morbidity and Mortality Weekly Report. At that point it becomes official CDC guidance.

Reading the wording, and why it matters at the bedside#

The framework also decides how a recommendation is phrased, and the phrasing is not decoration. A routine or risk-based recommendation sets the default to vaccinate: absent a contraindication, the patient in that group should be offered the vaccine. A recommendation against does the reverse.

A third category, shared clinical decision-making, sets no default either way. The CDC reserves it for situations where some individuals may benefit but vaccinating the entire group is unlikely to produce a clear population-level gain. Here the choice is meant to happen in conversation between a person and their clinician, guided by that person's own risk factors, values, and preferences rather than a blanket rule. For a family medicine or internal medicine clinician, spotting this label is a cue to slow down and actually have the conversation, not to reach for a default that was never set.

Why the visible machinery is the payoff#

The strength of this system is that it can be inspected. The GRADE tables and EtR judgments behind any given vaccine recommendation are public, so a reader can trace exactly which domains drove it. That does not remove judgment from the process. Benefit, harm, values, and equity cannot be collapsed into one number, and thoughtful experts can weigh them differently. What the framework does is make the reasoning legible, so that both agreement and disagreement land on specific evidence rather than on the general authority of a committee.