Patient Safety and Quality

Accurate secondary diagnosis codes improve risk adjustment of inpatient mortality rates

If the completeness and precision of coding secondary diagnoses could be improved, risk adjustment for quality reporting would be considerably more accurate, concludes a new study. Accurate hospital mortality rates depend on proper adjustment of patient risk factors for dying in the hospital, including severity of illness on admission.

Anne Elixhauser, Ph.D., of the Agency for Healthcare Research and Quality (AHRQ) and colleagues wanted to see if more complete coding of secondary diagnoses, which are currently underreported because of ICD-9-CM coding rules, would improve the risk adjustment of inpatient mortality rates. They also looked to see if adding laboratory values at the time of admission to claims data would improve predictions of inpatient mortality. They analyzed claims data and abstracted clinical data for 5 medical conditions and 3 surgical procedures from 188 hospitals in Pennsylvania.

They found substantial potential benefits by combining present-on-admission (POA) coding (which describes if diagnoses were present on admission to the hospital or if they originated during the stay) with expanded use of a few currently available secondary diagnosis codes. Current rules preclude the coding of signs and symptoms that affect the risk of inpatient mortality once a final diagnosis is established, even though ICD-9-CM codes are available. For example, once a diagnosis of stroke is established, symptoms such as coma can no longer be recorded in claims data.

The study also found that adding numerical laboratory results to claims data substantially improved predictions of inpatient mortality. They recommend three steps to improve risk adjustment of inpatient mortality rates. First, claims data should be enhanced by accurately reporting POA for secondary diagnoses for all patients. Second, there should be more consistent and complete use of existing ICD-9-CM secondary diagnosis codes demonstrated to be important predictors of hospital outcomes, such as coma, severe malnutrition, and abnormal vital signs. Finally, traditional claims data should be merged with about two dozen laboratory values that are usually obtained at admission and that are available electronically from most U.S. hospitals. These steps would eliminate costly abstraction of medical records, which contributed little to improving the accuracy of risk-adjusted inpatient mortality rates in this study.

More details are in "Modifying ICD-9-CM coding of secondary diagnoses to improve risk-adjustment of inpatient mortality rates," by Michael Pine, M.D., M.B.A., Harmon S. Jordan, Sc.D., Dr. Elixhauser, and others, in the January/February 2009 Medical Decision Making 29, pp. 59-81.

Reprints (AHRQ Publication No. 09-R052) are available from the AHRQ Publications Clearinghouse.

Return to Contents
Proceed to Next Article