Given the complexity of medical decision making and the myriad questions that arise during the care of individuals, an expectation that every causal question be addressed with a randomized clinical trial (RCT) is not realistic. Nevertheless, large administrative databases linked with electronic health records, coupled with new statistical methods for extracting causal information from raw data, can complement clinical trial evidence, enabling a “learning health care system.” Yet despite continued advances in epidemiological and statistical methods and the advent of “big data,”1 there is concern that inferences from observational data can lead to poor health care decisions by misrepresenting association for causation.
Scatter plots of results from empirical comparisons of propensity score analyses (y-axis) and corresponding randomized clinical trial (RCT) results (x-axis). Markers denote comparisons between observational and randomized study estimates for the same research question (similar populations, interventions, and outcomes); statistically significant differences (P < .05) are shown in orange. The dotted lines indicate lack of effect in RCTs (vertical lines) and observational studies (horizontal lines). Values lower than 1 indicate that the new treatment evaluated in the trial was more effective than the more established treatment; observational study results are expressed in the same way as the corresponding trial results. Markers in the top-right and bottom-left quadrants in each panel indicate agreement between randomized and observational results with respect to the direction of effects. Markers in the top-left and bottom-right quadrants indicate discordant direction of effects between designs. Black dashed diagonal lines indicate the line of identity (perfect agreement) between RCT and observational study results; gray dashed lines demarcate observational study relative risks that are between 0.67 and 1.5 times those produced by the corresponding RCT results. The term “relative risk” is used to denote risk, odds, or hazard ratio estimates, as reported in the 3 empirical analyses contributing data to this figure.
Some tools below are only available to our subscribers or users with an online account.
Download citation file:
Web of Science® Times Cited: 2
Customize your page view by dragging & repositioning the boxes below.
More Listings atJAMACareerCenter.com >
Users' Guides to the Medical Literature
Fixed-Effects vs Random-Effects Models: An Analogy
All results at
Enter your username and email address. We'll send you a link to reset your password.
Enter your username and email address. We'll send instructions on how to reset your password to the email address we have on record.
Athens and Shibboleth are access management services that provide single sign-on to protected resources. They replace the multiple user names and passwords necessary to access subscription-based content with a single user name and password that can be entered once per session. It operates independently of a user's location or IP address. If your institution uses Athens or Shibboleth authentication, please contact your site administrator to receive your user name and password.