Major US healthcare database has missing pieces

A major source of American health information data contains a handful of glaring flaws related to health risks, say Johns Hopkins researchers in a study published online Wednesday in the journal PLOS ONE.

The Nationwide Inpatient Sample (NIS) critically underreports several important health measures, say Susan Hutfless, M.S., Ph.D., and Elie Al Kazzi, M.D., M.P.H., of the Johns Hopkins University School of Medicine. Alcohol and tobacco use, as well as data on patients' weight and body mass, are dramatically underreported in the NIS. Missing information in the database leads to inaccurate health quality assessments and could have a devastating economic impact on hospitals that see the sickest patients.

The Centers for Medicare and Medicaid Services (CMS) uses NIS data to determine hospitals’ risk for readmission or surgical complications, both of which can lead to lower hospital reimbursements. Obesity status and alcohol abuse are used to calculate the risk, but are often not recorded in the database.

During the course of hospital admission, patients are typically weighed and asked questions about alcohol and tobacco use. That information, however, often does not travel from a patient’s history to his or her hospital bill. Since the NIS is based on billing data, only the information in the patient’s bill is recorded.

“Hospitals that are better at using billing codes to record the alcohol, tobacco and obesity status of their patients will be classified as having higher-risk patients,” “Their reimbursement will be lower than hospitals that don’t include the information in their billing records."

The missing information has implications beyond hospital reimbursement.

The National Institutes of Health uses the NIS to spot trends in the burdens that certain diseases put on the U.S. health system. “If there are problems with coding obesity and tobacco, which are the two greatest contributors to mortality worldwide,” Hutfless says, “it is reasonable to assume that there may be errors in the way we perceive the burdens of certain diseases on the U.S. health care system.”

The NIS consists of the billing data collected from 20 percent of all patients discharged from randomly selected academic and specialty hospitals — a total of about 6.8 million hospital stays from more than 1,000 hospitals across the country. The study’s authors compared NIS information with the Behavioral Risk Factor Surveillance System (BRFSS), a federally sponsored telephone-administered survey where more than 500,000 American adults answer questions about their health, including their height and weight, as well as their tobacco and alcohol habits.

Comparing the NIS with the BRFSS, the study's authors found significant differences in statistics related to these areas. Both nationally and state to state, the NIS statistics in those areas were far lower than those generated by direct data collection.

“Risk factors in the areas we studied are inaccurate,” says Al Kazzi. “If this discrepancy isn't fixed, it will have a significant effect on provider reimbursement.”