Unexpected Term NICU Admissions: a Marker of Obstetrical Care Quality?

Published In

American Journal of Obstetrics and Gynecology

Document Type


Publication Date



We want to comment on the terminology used to describe the study design of the recent paper, Unexpected term NICU admissions: a marker of obstetrical care quality, which analyzed data from the Consortium for Safe Labor.1 The study abstract identifies the study as a retrospective cross-sectional study. The defining feature of cross-sectional studies is that they do not have a temporal direction (ie, retrospective, prospective, or ambidirectional). This is because they involve data collection at just 1 point in time about things that may exist in the past or present, and study eligibility is not based on whether a study outcome has already occurred.2

The term, retrospective cross-sectional study, is thus an oxymoron. The word retrospective lacks a consensus definition in epidemiology,3,4 but it has been used at various times to mean the following: (1) de novo collection of exposure data after outcome occurrence; (2) the case-control study design in general; or (3) the feature of historical cohort studies that distinguishes them from other cohort studies: the temporal ordering of follow-up of exposed and unexposed individuals before initiation of research (ie, in the past).2,4 In the current situation, we suspect that the term retrospective was intended to mean a study in which preexisting data were used to address a question not specifically envisioned at the time the data were collected.

The distinction between prospective and retrospective was clearer in the era when information on exposures and confounders was usually collected by direct interview of participants. However, it has become blurred in the present era of widely available electronic records of clinical interactions, insurance claims, vital records, environmental monitoring, etc, particularly when those records can be linked at the individual level. Therefore, in the current era, these terms “do not readily convey a clear message about the study” (page 96).2

Because of the substantial challenges in accurately applying study design categories to modern data resources, when study design distinctions must be made, we advocate dividing only between experimental and observational studies. Even this distinction is not simple and is vulnerable to classification error: for instance, when randomized controlled trials are analyzed for secondary purposes to study only exposures that were not randomized, they are effectively observational studies. However, focusing on this distinction might provide the most crucial information while also enabling accurate designation of study designs and, in turn, facilitate the identification of potentially important biases.


© 2019 Elsevier Inc. All rights reserved.



Persistent Identifier