Patrick Taffé1, Patricia Halfon2, Matthieu Halfon3. 1. Center for Primary Care and Public Health, University of Lausanne, Lausanne, Switzerland. Electronic address: patrick.taffe@unisante.ch. 2. Center for Primary Care and Public Health, University of Lausanne, Lausanne, Switzerland. 3. Division of Nephrology and Hypertension, Lausanne University Hospital, Lausanne, Switzerland; Transplantation Center, Lausanne University Hospital, Lausanne, Switzerland.
Abstract
BACKGROUND AND OBJECTIVES: The Bland and Altman's limits of agreement (LoA) method is the most commonly used statistical method to assess the bias and precision of a new measuring device (it has been cited over 40,000 times as of March 2019). What is less known is that the LoA method can be dramatically misleading. METHODS: A new statistical methodology, which circumvents these deficiencies, has recently been published and made available in the R and Stata statistical packages. We aimed at introducing and illustrating with a small data set on blood pressure (BP) measurements, taken by two different oscillometric devices, the use of this new methodology to a clinical audience. RESULTS: For diastolic BP, the LoA method was particularly misleading, as it identified differential and proportional biases of opposite signs compared with the new methodology. Regarding systolic BP, the LoA method strongly overestimated both the differential and proportional biases, for both devices. CONCLUSION: The LoA method may be dramatically misleading and does not allow one to estimate the precision of each measurement method. We recommend the use of the newly developed statistical methodology instead.
BACKGROUND AND OBJECTIVES: The Bland and Altman's limits of agreement (LoA) method is the most commonly used statistical method to assess the bias and precision of a new measuring device (it has been cited over 40,000 times as of March 2019). What is less known is that the LoA method can be dramatically misleading. METHODS: A new statistical methodology, which circumvents these deficiencies, has recently been published and made available in the R and Stata statistical packages. We aimed at introducing and illustrating with a small data set on blood pressure (BP) measurements, taken by two different oscillometric devices, the use of this new methodology to a clinical audience. RESULTS: For diastolic BP, the LoA method was particularly misleading, as it identified differential and proportional biases of opposite signs compared with the new methodology. Regarding systolic BP, the LoA method strongly overestimated both the differential and proportional biases, for both devices. CONCLUSION: The LoA method may be dramatically misleading and does not allow one to estimate the precision of each measurement method. We recommend the use of the newly developed statistical methodology instead.