Send to

Choose Destination
Am J Kidney Dis. 2001 Feb;37(2):267-75.

Limitations of the facility-specific standardized mortality ratio for profiling health care quality in dialysis.

Author information

Home Dialysis Therapies and Health Care Information Services, Dialysis Services, Fresenius Medical Care, North America, Lexington, MA, USA.


Health care quality is assessed by profiling measures of care and/or health outcomes. However, such tools to measure outcome as standardized mortality ratio (SMR) are often used without thorough validation of their strengths and limitations. Our study compared the dialysis facility-specific SMR and SMR-based rating using different statistical methods and followed them over time. All Fresenius Medical Care, North America dialysis facilities (n = 377) that contributed patient data from 1993 to 1995 (>103,500 patient-years) were included. Four distinct statistical methods (US Renal Data System [USRDS], Poisson, logistic, and Cox regression) were used to compute facility-specific SMRs and rank and classify facilities. The analysis compared the SMR and SMR-based rating of dialysis facilities between SMR method and over time. Different methods produced statistically significant differences in SMR distribution (P < 0.05). The USRDS method produced SMR values that decreased over time (P < 0.001). Based on 90% confidence intervals to determine outliers, the SMR-dependent ranking of dialysis facilities varied by method (P < 0.001). SMR-based ranking was stable over time except for the USRDS method (P < 0.001). Contingency table analysis showed up to a 33% total misclassification rate between SMR methods when ranking facilities. The facility-specific SMR and SMR-based ranking are both sensitive to statistical technique. Because the SMR yields different results in a year and over time and because there is no demonstrable gold standard, conclusions based on any one technique are unstable and unreliable. Regulatory monitoring, actions, and/or performance awards should be avoided based on this measure. However, a facility-specific SMR estimated in any valid way may be useful as an epidemiological research tool.

[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Elsevier Science
Loading ...
Support Center