On the use of medical records for characterizing brain lesions.(Report)

In studies of neurogenic communication disorders, medical records commonly are used to characterize brain lesions, to make decisions regarding participant inclusion, and to draw general conclusions regarding brain-behavior relationships. The validity of using existing medical records has not been directly evaluated. The purpose of this article is to draw attention to the potential problems with relying exclusively on medical records to characterize neurological lesions. Examples from a study of language in adults with right hemisphere brain damage are used to highlight discrepancies between imaging reports taken from existing medical records and structural images obtained at the time of the study. The discussion of factors that may contribute to discrepancies between the imaging reports includes the scanning method and protocol used, interrater reliability for reading neuroradiologic images, the effect of time, and neurological changes associated with normal aging.

**********

The search for the nature of brain-behavior relationships has been ongoing for nearly two centuries but is still an active search. The description of these relationships can be as broad as comparing right versus left hemisphere functions or as specific as isolating a region of a particular gyrus responsible for specific language functions. The advent of widely used imaging technology has both enhanced our ability to pinpoint such relationships and spawned new questions and problems.

Studies of neurogenic communication disorders frequently rely on reports from participants' medical records to identify site and/or size of lesion. Decisions regarding participant inclusion and group classification are made on the basis of these reports, along with conclusions regarding brain-behavior relationships. In typical studies of neurogenic language disorders, the screening process for potential participants involves a review of medical records. Participants included in a study are those whose medical records provide evidence of a lesion in the hemisphere, lobe, or structure of interest. Many times this information is from a computed tomography (CT) or magnetic resonance imaging (MRI) scan that was obtained within a few hours or days postonset of stroke. In some cases the radiology imaging report is not available, and information is obtained from a neurologist's summary of the CT/MRI scan(s). Some studies include participants who exhibit physical or behavioral signs of unilateral stroke (e.g., hemiparesis) even in the presence of a negative brain scan. Participants may be classified as having anterior versus posterior lesions (in relation to the central sulcus) based on the initial CT/MRI reports. In combination with behavioral data obtained from experimental language tasks or diagnostic tools, conclusions about localization of function may be made from those classifications.

While use of medical records is common in the screening process, the validity of this practice has not been directly addressed. This article draws attention to the potential problems with relying exclusively on medical records to characterize neurological lesions and make assumptions about brain-behavior relationships. Examples are provided from a study of individuals with right hemisphere stroke.

Data for the current discussion were obtained from 12 participants who were recruited for a study of language disorders associated with right hemisphere brain damage. They were included in the study after a review of medical records for each person indicated lesion(s) restricted to the right hemisphere (reports were based on CT and/or structural MRI scans). All participants were at least 1 year postonset of stroke. As part of the research study, a structural MRI scan was completed. Comparisons between the imaging reports from medical records and the recent MRI (see Table 1) indicate more, or more extensive, lesions in each of the 12 participants. Based on the criteria for the original language study (lesions due to stroke restricted to the right hemisphere), 6 of these 12 individuals (Participants 2, 3, 4, 9, 10, & 11) were subsequently excluded. Had the MRIs not been completed, language data from all 12 participants would have been analyzed as a group, thus potentially distorting the conclusions regarding right hemisphere language function.

There are various factors that may contribute to the discrepancies between the imaging reports, including the scanning method and protocol used, interrater reliability for reading CT/MRI scans, the effect of time, and neurological changes associated with normal aging. Each of these factors will be discussed. The focus of the ensuing discussion will be on neurological lesions due to stroke.