New research identifies significant variation in health outcomes across hospitals

BOSTON, December 14, 2016A pioneering study of over 22 million hospital admissions found significant variation in health outcomes across the United States. Patients in low-performing hospitals (bottom 10%) are three times more likely to die and 13 times more likely to experience complications than those in high-performing hospitals (top 10%).

These are the key findings of The Boston Consulting Group (BCG) and research partners Ariadne Labs, Harvard T.H. Chan School of Public Health, Alerion Institute, Johns Hopkins School of Medicine, University of Michigan Medical School, and the University of Rochester Department of Public Health. The research was published today in the peer-reviewed scientific journal PLOS ONE.

The study is the most comprehensive analysis of health outcomes variation in the United States to date, covering 22 million hospital admissions across states where over half of the US population resides.   It analyzes 24 specific health outcomes including many widespread illnesses such as cardiovascular disease, diabetes, and pneumonia.

“Quite simply, this study found that where you live can determine if you live,” said Barry Rosenberg, MD, a Chicago-based BCG partner in the firm’s Health Care practice and lead author of the study. “Americans do not fully appreciate the alarming extent of outcomes variation that exists among US hospitals. If you call 911, do you want your loved one’s heart attack treated at a hospital with a 4% death rate or a 16% death rate? The closest hospital may not always be the best hospital.”

Even after extensive risk adjustment, the research found that large variations in hospital performance persisted across geographies.   These differences in hospital outcomes could not be fully explained by differences in patient demographics, health, or a variety of macro health system factors. For example, even after risk adjustment:

  • The probability of dying in the hospital after an acute event, such as a heart attack or stroke, is more than twice as high at low-performing hospitals vs. high-performing hospitals.
  • Patients were nearly 20 times more likely to experience central line infections at low performing hospitals vs. high performing hospitals, and over three times as likely to contract postoperative sepsis.
  • Preventative health systems also showed significant variability in performance, with over five times the rate of admissions for short-term complications from poorly controlled diabetes in poor performing counties vs. high-performing counties.

Challenging conventional wisdom, the study identified regions with poor-performing hospitals that serve high-income, largely white populations and many regions with high performing hospitals that serve low-income, minority populations. The observed variation in hospital outcomes exists across the US and can be found both between states and within states.

“It’s been known that hospitals vary in quality, but it hadn’t been recognized exactly how large the variation is between high performing and low performing hospitals,” said author Atul Gawande, MD, MPH, executive director of Ariadne Labs, professor at Harvard T.H. Chan School of Public Health, and surgeon at Brigham and Women’s Hospital in Boston. “This clearly indicates that greater outcomes transparency and greater focus on performance improvement are necessary.  We have a tremendous opportunity to improve outcomes for patients.”

For the past decade, researchers and policy makers have focused attention on geographic variation in health care cost and utilization, but limited attention on outcomes. “We need to expand the focus to include measuring and improving outcomes that matter to patients, such as death rates and complication rates. Only if hospitals measure and share their risk-adjusted outcomes will we have an efficient health care market with a high rate of innovation and competition to improve quality of health care,” said Stefan Larsson, MD, PhD, a senior partner and global leader of BCG’s Payer and Provider practice.

BCG CEO Rich Lesser, another co-author of the study, commented: “As the US prepares for its next generation of health care reform, these findings reinforce that it is not just about how much to spend on health care, it’s also about how to create an environment that accelerates the sharing of best practices across the broader health care system.”

The present study uses all-payer data from 2011 from the Agency for Healthcare Research and the Quality Healthcare Cost and Utilization Project/ State Inpatient Databases (HCUP/SID).   Due to changes in the HCUP/SID database design in 2012 that removed most hospital identifiers needed for geographic mapping, it is not possible to update the analysis with more recent data.

A copy of the article, “Quantifying geographic variation in health care outcomes in the United States before and after risk-adjustment,” can be obtained from the PLOS ONE website at

BCG’s interactive tool visually displaying all 24 risk-adjusted outcomes is available at