Can data-driven benchmarks be used to set the goals of healthy people 2010

UMMS Affiliation

Department of Quantitative Health Sciences

Publication Date


Document Type



Benchmarking; *Data Interpretation, Statistical; *Evidence-Based Medicine; Female; Forecasting; *Health Planning Guidelines; Health Priorities; Humans; Mammography; Mass Screening; Middle Aged; Reproducibility of Results; United States


Bioinformatics | Biostatistics | Epidemiology | Health Services Research


OBJECTIVES: Expert panels determined the public health goals of Healthy People 2000 subjectively. The present study examined whether data-driven benchmarks provide a better alternative.

METHODS: We developed the "pared-mean" method to define from data the best achievable health care practices. We calculated the pared-mean benchmark for screening mammography from the 1994 National Health Interview Survey, using the metropolitan statistical area as the "provider" unit. Beginning with the best-performing provider and adding providers in descending sequence, we established the minimum provider subset that included at least 10% of all women surveyed on this question. The pared-mean benchmark is then the proportion of women in this subset who received mammography.

RESULTS: The pared-mean benchmark for screening mammography was 71%, compared with the Healthy People 2000 goal of 60%.

CONCLUSIONS: For Healthy People 2010, benchmarks derived from data reflecting the best available care provide viable alternatives to consensus-derived targets. We are currently pursuing additional refinements to the data-driven pared-mean benchmark approach.


Am J Public Health. 1999 Jan;89(1):61-5.

Journal/Book/Conference Title

American journal of public health

PubMed ID


Related Resources

Link to Article in PubMed