How dangerous is a day in hospital? A model of adverse events and length of stay for medical inpatients

Med Care. 2011 Dec;49(12):1068-75. doi: 10.1097/MLR.0b013e31822efb09.


Background: Despite extensive research into adverse events, there is no quantitative estimate for the risk of experiencing adverse events per day spent in hospital. This is important information for hospital managers, because they may consider discharging patients earlier to alternative care providers if this is associated with lower risk, but other costs and benefits are similar.

Methods: We model adverse events as a function of patient risk factors, hospital fixed effects, and length of stay. Potential endogeneity of length of stay is addressed with instrumental variable methods, using days and months of discharge as instruments. We use administrative hospital episode data for 206,489 medical inpatients in all public hospitals in the state of Victoria, Australia, for the year 2005/2006.

Results: A hospital stay carries a 5.5% risk of an adverse drug reaction, 17.6% risk of infection, and 3.1% risk of ulcer for an average episode, and each additional night in hospital increases the risk by 0.5% for adverse drug reactions, 1.6% for infections, and 0.5% for ulcers. Length of stay is endogenous in models of adverse events, and risks would be underestimated if length of stay was treated as exogenous.

Conclusions: The results of our research contribute to assessing the benefits and costs of hospital stays-and their alternatives-in a quantitative manner. Instead of discharging patients early to alternative care, it would be more desirable to address underlying causes of adverse events. However, this may prove costly, difficult, or impossible, at least in the short run. In such situations, our research supports hospital managers in making informed treatment and discharge decisions.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Cross Infection / epidemiology
  • Drug-Related Side Effects and Adverse Reactions / epidemiology
  • Health Services Research / statistics & numerical data
  • Hospital Administration / statistics & numerical data*
  • Humans
  • Length of Stay / statistics & numerical data*
  • Patient Discharge / statistics & numerical data
  • Patient Safety / statistics & numerical data*
  • Pressure Ulcer / epidemiology
  • Quality of Health Care / statistics & numerical data
  • Risk Factors