One-third of nursing homes residents are admitted to the hospital at least once each year, and half of those admissions could be avoided. Preventing them could protect hundreds of thousands of older adults from potential harm and save Medicare billions of dollars.

The problem is neither new nor surprising. But it is tough to fix. Last week, the federal Centers for Medicare and Medicaid Services (CMS) announced a second stage of what it calls an Initiative to Reduce Avoidable Hospitalizations among Nursing Facility Residents.

This effort began three years ago, when CMS chose seven organizations to run model programs aimed at reducing hospitalizations. Some are university-based health systems, one is a provider network, others are community-based non-profits. Together, they partnered with 146 nursing facilities from Nebraska to New York City.

Did the experiment work? The jury is still out, according to an evaluation by the consulting firm RTI International. While several efforts show promise, according to RTI, it is too early to know what works and what does not.

Why is reducing hospitalizations of nursing home residents so hard? In part, it is because it is not one problem. It is many.

Nursing homes send their patients “out” to hospitals for lots of reasons. Here are just a few: Their staffs are not trained to address common medical problems, they don’t have the tools to communicate changes in a resident’s health status to physicians, fear of litigation, the adult children of residents demand it, and government payment models encourage it.

For all those reasons, hospitalization is often the default option when a nursing home resident suffers a decline in health status. If she spikes a fever, the facility will call 911. Same if she has breathing problems. Or if she falls.

To some degree, the market—with a nudge from government– is already addressing some of these issues. Medicare now penalizes hospitals that have excessive readmissions. And because many come from nursing facilities, hospitals have begun responding with both sticks and carrots.

The stick: Some hospitals began closely tracking discharges, admissions, and readmissions. And they stopped discharging to facilities that were most likely to send patients back.

The carrots: Some hospitals are working closely with cooperating nursing homes and physicians to reduce those 911 calls. They are improving communications, including combining health IT systems. They are training nursing home staff to identify potential patient problems before they blossom into a crisis and teaching staff how better to respond if they do. Some hospitals are even embedding nurses or even physicians in high volume nursing homes.

In many states, better physician orders for end-of-life care, sometimes called MOLSTs or POLSTs, are helping to clarify the desires of patients and their families about hospitalization. And increasingly, residents of nursing homes are signing Do Not Hospitalize orders to avoid the trauma of an end-of-life trip to an emergency department or intensive care unit.

All of this may help reduce readmissions. But Medicare and Medicaid still provide powerful incentives for nursing homes to send residents to the hospital. For instance, bed-hold rules mean facilities can be paid to keep an empty bed while a resident is hospitalized.

And it is almost always more lucrative for a nursing facility to take a Medicare post-acute patient (at reimbursement of about $600 per day) than a Medicaid long-stay resident (at the Medicaid rate of about $150 per day). The key difference: the facility can get the higher Medicare rate only if the patient has first been hospitalized for at least three days. The rate is good for 100 days after a discharge. And every hospitalization can start the clock ticking again.

There is yet another complication: this one largely of the government’s own making. In response to Medicare’s crackdown on what it considers unnecessary hospital admissions, hospitals have been increasingly keeping patients under “observation” status rather than admitting them (Full disclosure: I am an unpaid board member of a community hospital).

This change saves Medicare a lot on hospital costs (on average, a hospital is paid about one-third less for an observation patient than for an admission). And Medicare may save even more on post-acute care. If a patient has not been admitted for those three days, Medicare does not have to pay for a post-discharge nursing home stay—though, oddly, it may do so anyway.

The growth of observation stays confuses the nursing home readmission discussion: If a patient goes from a nursing home to a hospital as an observation patient, there was no hospital admission. Thus, it can claim it has reduced admissions even though that resident was, for all but payment purposes, hospitalized.

It is long past time for the government, hospitals, nursing homes, doctors, and patient advocates to get together and reduce those needless hospitalizations. Lives are at risk.