Booster seats are supposed to keep children too old for child seats safe by positioning them so the seat belts fit.

Now, comprehensive tests on these widely used products are raising new concerns about boosters.

According to a study by the Insurance Institute for Highway Safety and the University of Michigan Transportation Research Institute, a number of booster seats don't do enough to protect children in a crash and could even contribute to internal injuries.

In a recently released report, the Insurance Institute said 13 of the 41 booster seats it tested performed so poorly that it can't recommend them. The research group, which is funded by the insurance industry, said it considers 10 models the "best bets," and five more models are good bets. The rest are fair.

Another key finding: Children's proportions and vehicle designs vary so much that even a good booster can be misused. Parents should make sure that a booster fits their child and their car before buying the seat.

Boosters are designed for children -- typically ranging from about 4-9-years-old -- who are too big for toddler restraints but still too small to use adult safety belts alone. Most toddler seats are for kids who weigh up to 40 pounds. A booster seat is supposed to raise the child so the vehicle's belts rest properly across the pelvis and chest. A poor fit can result in the lap belt resting against the abdomen and causing internal injuries in a crash. Shoulder belts can cause injuries if they are too high and stretch across a passenger's neck.

The best bets, according to the study, include the Fisher-Price Safe Voyage, the Britax Parkway, the LaRoche Bros. Teddy Bear, the backless Graco TurboBooster, the backless Combi Kobuk and the Recaro Young Style. Some high-back seats convert to backless designs, so the group tested the configurations separately. The Safeguard Go made the "best bets" list when used in its backless mode, and the highback versions of the Volvo booster and Britax Monarch made the list. Adrian Lund, president of the insurance group, says he'd expect the seats in this group to improve seat-belt fit for children in nearly any vehicle.

The main problem with the boosters that the group doesn't recommend is that they place the lap belt on the child's vulnerable abdomen. The list includes the Cosco Dorel Summit, Traveler and Alpha Omega, the Cosco Highback Booster, the Graco CarGo Zephyr, the Evenflo Generations and Evenflo Chase Comfort Touch, and the Compass B505 and B510. Also, the Dorel/Safety 1st Prospect and Intera, as well as the Evenflo Big Kid Confidence, were in the worst list when used in highback versions and the Safety Angel Ride Ryte was when used backless.

Dorel Juvenile Group said its seats "meet or exceed all federal regulations." Graco also said its products meet "all relevant regulations."

Still, safety experts say the risk of injury to a child is less likely with a booster -- even a less-effective model.

At the same time, a good rating doesn't guarantee a proper fit. Different back-seat angles, for instance, can change the way seat belts fit children.

The Insurance Institute's test seats ranged in price from about $20 to $200, and the prices didn't necessarily correlate with the seats' performance. The tests also showed that both high-back and backless designs can perform well.

More states have required boosters in recent years. Safety experts generally set 4-feet, 9-inches, as the height at which most children can start using a seat belt without a booster.

Still, the use of booster seats has been on the rise for years. Among children up to 7-years-old, 80 percent were using some kind of child restraint in 2007, compared with 51 percent in 1999, according to a long-term study by the Children's Hospital of Philadelphia and the Association of International Automobile Manufacturers. Booster-seat use among 4 and 5-year-olds increased to 88 percent from 30 percent in the same period.