WASHINGTON — As it races to revamp Medicaid by allowing work requirements for the first time, the Trump administration is failing to enforce federal rules directing states to assess the impact of the change on low-income patients who rely on the half-century-old safety-net program, a Times analysis shows.

None of the eight states that the administration has cleared to implement a Medicaid work requirement has in place a plan to track whether Medicaid enrollees find jobs or improve their health, two goals often touted by administration health officials.

And nine of the 17 states that have sought federal permission to implement Medicaid work mandates have been allowed by the Trump administration to proceed with their applications despite failing to calculate the number of people who could lose coverage, according to a review of state and federal Medicaid records.

Federal regulations issued under the Obama administration direct states seeking permission to experiment with new Medicaid policies to, in most cases, estimate effects on coverage before the initiative starts, and then independently evaluate the impact of the programs after they begin to ensure they are achieving their goals.

Nevertheless, Arkansas, which the Trump administration celebrated in June as the first state to implement a Medicaid work requirement, still has no approved research plan, even though the state has already removed more than 18,000 people from Medicaid coverage for failing to comply with its mandate.

“For any new medical treatment, we require rigorous evaluation to assure it is safe and effective,” said Dr. Benjamin Sommers, an internist at Harvard University who has extensively studied the effect of Medicaid policies nationwide. “We should be equally vigilant that these changes in policy are working as intended, as they could have far-reaching effects on patients’ health.”

Critics say the administration and the states appear to be systematically ignoring or weakening the requirement for independent analysis, perhaps because they fear the results.

“There is a lot of hiding the ball here,” said Joan Alker, executive director of the Georgetown University Center for Children and Families, a research organization that is tracking the administration’s efforts to revamp Medicaid rules.

“We know that health insurance coverage is very popular,” Alker added. “So taking it away, as some of these plans do, is not something the administration wants to talk about.”

Some states seeking permission to implement Medicaid work requirements have projected that Medicaid enrollment will decline, as has occurred in Arkansas.

Alabama, for example, acknowledges in its application that about 16,000 people will probably lose standard Medicaid coverage as a result of its mandate, state filings indicate.

However, state officials seeking permission to implement work mandates in Arizona, Arkansas, Kansas, Michigan, New Hampshire, Oklahoma, South Dakota, Tennessee and Wisconsin did not project in any detail how their experiments would affect Medicaid enrollees’ coverage, according to a review of hundreds of pages of state documents filed with the federal Center for Medicare and Medicaid Services, or CMS.

Several of those states, including Arizona, Oklahoma and Wisconsin, make no mention of the enrollment impact of the work requirement.

In New Hampshire, Medicaid officials claimed that enrollment “will not change materially.”

South Dakota officials similarly asserted: “Coverage losses will be small,” though the state did not offer evidence in its waiver application to support this claim.

Tennessee’s work requirement application — which was filed Dec. 28, months after substantial coverage losses in Arkansas were already being reported — acknowledged that some people could lose Medicaid as a result of its mandate.

Tennessee officials did not make any estimates, however, concluding: “It is not possible to reliably project the magnitude of this decrease in enrollment at this time.”

The lack-of-coverage projections come despite 2012 federal regulations specifying that requests to implement Medicaid experiments “will not be considered complete” unless states include “an estimate of the expected increase or decrease in annual enrollment.”

Nevertheless, the Trump administration deemed all the state Medicaid applications complete, CMS documents show.

The same federal regulations also mandate that states seeking permission to implement Medicaid experiments, also known as demonstrations, develop plans for independent evaluations that involve an “empirical investigation of the impact of key programmatic features of the demonstration.”

Meeting the research requirements can be difficult for states, acknowledged Sara Rosenbaum, an authority on Medicaid law at George Washington University’s Milken Institute School of Public Health.

But the rules are there for a reason, she explained. “States are drawing down hundreds of millions of dollars in taxpayer money. If they want to experiment, we want to know whether what they are doing is working, what we are gaining, what we might be losing and what the effects are on patients.”

The Trump administration’s senior Medicaid official, Seema Verma — who as CMS administrator has cheered Medicaid work requirements — declined to be interviewed.

In a written response to questions, however, a CMS spokesperson said the agency does not believe states must do enrollment calculations. “Transparency regulations do not require that states provide precise numerical estimates of coverage impacts, which are difficult to predict in many types of demonstrations.”

The spokesperson said CMS is working to develop evaluation strategies for states that are implementing work requirements, which CMS also calls “community engagement,” or CE. “We will soon be releasing guidance for all CE states to support robust evaluation,” the spokesperson said.

In public speeches, Verma has repeatedly said she wants to reduce administrative burdens on states to make it easier for them to experiment with Medicaid.

“State leaders are closer to the people and the problems they face every day,” Verma said at a recent conference convened by the conservative American Legislative Exchange Council, or ALEC. “For too long, states have looked to Washington with a ‘Mother, may I?’ approach, and Washington has placed unworkable restrictions on states.”

CMS late last year also scaled back requirements on states that want to loosen health insurance rules put in place by the 2010 Affordable Care Act, often called Obamacare.

And last spring, the Trump administration canceled a major study of a Medicaid experiment in Indiana that requires low-income patients there to pay more for their medical care.

The Healthy Indiana Plan, which Verma helped develop as a consultant before joining the Trump administration, has drawn increasing scrutiny amid evidence that enrollees are not making required payments and consequently are losing health protections.

The administration’s retreat from independent analysis has alarmed researchers as well as many physicians, hospitals and patient advocates, who warn that the Trump administration’s rush to reshape safety net programs without adequate review risks harming low-income people who rely on Medicaid.

This kind of analysis has frequently complicated the Trump administration’s efforts to reshape healthcare policy, however.

When White House and congressional Republicans tried to roll back the healthcare law in 2017, for example, studies by the nonpartisan Congressional Budget Office indicated that tens of millions of Americans would probably lose health coverage. That research helped sink the GOP repeal campaign.

Similarly, when a federal judge last year blocked Kentucky’s Medicaid work requirement plan and noted it was inconsistent with the program’s purpose of providing health protections to low-income Americans, the judge cited an analysis by the state that estimated as many as 95,000 people could lose coverage.

Arkansas’ early experience with a Medicaid work requirement suggests such mandates will likely cause a significant number of people to lose Medicaid, in part because they fail to adequately report that they are working or seeking work, as required. That makes independent analysis of the new initiatives so crucial, experts say.

In November, the independent, nonpartisan Medicaid and CHIP Payment and Access Commission called for a pause in Arkansas’ Medicaid work requirement after the state reported nearly 9,000 people were cut from Medicaid in the first month of the mandate amid widespread confusion about how enrollees were supposed to report their work activities. The state website where enrollees are supposed to report, for example, goes offline every day between 9 p.m. and 7 a.m.

“We are highly concerned about the disenrollment,” commission Chairwoman Penny Thompson wrote in a letter to Health and Human Services Secretary Alex Azar, citing the “absence of sufficient measures and data to interpret early results and guide adjustments.” Three months later, Azar still has not responded to the commission’s letter.

Nor has the Trump administration slowed its approval of Medicaid work experiments. Eight states, including Kentucky, which got a second CMS approval, have now been cleared.