Risk-Set Matching Cuts Bias in Estimating Effects of Hospital-Acquired Infections

Studies that do not consider time of infection may overestimate the effect of hospital-acquired infections on patient outcomes, according to the results of a new study published in the American Journal of Epidemiology.

“The estimates of the impact attributable to [hospital-acquired infections] on length of stay and costs were dramatically larger using conventional matching compared to risk set matching,” study author David Watson, PhD, from the Children’s Minnesota Research Institute, Children’s Hospitals and Clinics of Minnesota, Minneapolis, and colleagues wrote.

Although hospital-acquired infections pose significant health care and cost burdens, the investigators noted that methods used for estimating the effects of these infections can be subject to selection bias.

In an interview with Contagion®, Dr. Watson explained that many studies do not account for the point in time during the hospitalization at which hospital-acquired infections occur, and so current research may overestimate the effect of hospital-acquired infections on costs and length of stay.

“The problem with these methods is that they often are comparing infected patients to uninfected patients who were already discharged by the time the infection occurred, thus they are not comparable patients,” he said.

“As a solution, we proposed the risk-set matching method, which is just a fancy way of saying we made sure the uninfected comparison group was still in the hospital at the corresponding time of the hospital-acquired infection.”

In the newest study, Dr. Watson and colleagues compared the results of using conventional matching versus risk-set matching to assess the impact of hospital-acquired infections.

They analyzed 237,625 pediatric inpatient encounters from 160 hospitals, with 374 (0.16%) of the encounters experiencing a laboratory-confirmed hospital-acquired infection.

According to the investigators, conventional matching estimated the attributable length of hospital stay due to hospital-acquired infection as 31 additional days on average, while risk-set matching estimated it as 12 additional days.

Similarly, the conventional method estimated the hospital-acquired infection-associated costs to the hospital at an additional $66,400, while risk-set matching estimated costs at $31,900.

Dr. Watson believes the results of this study will be important for providing better and less biased estimates of the effect of hospital-acquired bloodstream infections on health care costs, length of hospital stay, and patient mortality.

“I think previous studies have not incorporated time of infection mainly because most databases, especially large claims databases, do not have information on the time of infection,” he told Contagion®.

“However, in studying rare events like hospital-acquired infections, we often need to use these large databases in order to have a reasonable number of infected patients. Thus, I'm interested in developing statistical methods that somehow account for time of infection when time of infection is unknown.”

Dr. Watson hopes that risk-set matching eventually becomes more commonly used in epidemiologic research. “The method is appropriate for hospital-acquired infections, as well as for other preventable hospital-acquired conditions like pressure ulcers or air embolisms,” he concluded.