National Health and Nutrition Examination Survey

The Retinal Imaging subsection of the Ophthalmology Component tests for the presence of diabetic retinopathy, age-related macular degeneration, and other retinal conditions. Two forty-five degree non-mydriatic digital retinal images of each eye were obtained using an ophthalmic digital imaging system (1).

In 2012, digital images of participants with a cup-to-disc ratio (CDR) greater than or equal to 0.6 were re-read by ophthalmologists from Johns Hopkins University with attention to features relevant to glaucoma. 
 

All examined participants aged 40 years and older who did not meet any of the exclusion criteria. Persons could be excluded from the retinal imaging component due to blindness (i.e., unable to see light with both eyes open), eye infections, or eye patches on both eyes (1).

Retinal images of all examined participants aged 40 years and older who had a vertical CDR 0.6 or greater were re-graded in 2012

The retinal imaging exam was performed using an ophthalmic digital imaging system (Retinal Photography) to assess the presence of retinal diseases.

Forty-five degree non-mydriatic digital images of the retina were captured from survey participants aged 40 years and older using the Canon Non-Mydriatic Retinal Camera CR6-45NM. Technicians conducting the exams were trained in the use of the digital imaging system. The room was darkened while these images were captured to allow the pupils to dilate naturally. No pharmacologic dilation was used (1).

The exam was conducted with the participant seated in a chair in front of the Canon CR6-45NM, with a Canon 10D camera back (6.3 megapixels per image), with the flash default (DA) set to 2. Please refer to the NHANES Ophthalmology Procedures Manual in the NHANES website for camera settings. Participants were asked to focus on a target for proper field positioning while two digital images per eye were obtained. The first image was centered on the macula (field 2) and the second on the optic nerve (field 1). The first eye to be imaged was determined by an automated system. The digital images were copied to DVDs and shipped to the University of Wisconsin for grading using standardized methods used for other populations.

Field 1 and 2 images provided photographic documentation of the optic disc, macula, and substantial portions of the temporal arcades.

During external alignment of the pupil, its size was estimated, once for each eye, prior to taking the first image for that eye. Each digital image was initially evaluated as it appeared on the back of the digital camera. Technicians were trained to determine when an image should be repeated, such as an image being too dark or the person blinked. After taking the required four images, repeat images were captured for those that were obviously ungradable. The maximum number of images allowed was 12 (i.e., 3 per field) (1).

Digital images were evaluated by graders at the University of Wisconsin. If an eye was judged to have a treatable pathologic condition that posed an imminent threat to vision, the University of Wisconsin notified NCHS within three working days of receipt of the images. Further information can be found in the NHANES Ophthalmology Procedures Manual available on the NHANES website.

In 2012, retinal images that were determined to have a CDR 0.6 or greater were re-read by ophthalmologists at Johns Hopkins University to look for additional indicators suggestive of the presence of glaucoma. Specifically, the optic disc images were evaluated. A tablet-based review system was used to grade the images.

As a quality control check, 200 images were randomly selected from eyes having a vertical CDR less than 0.6. All optic disc images assessed as 0.6 or greater were selected by the photo reading center and re-graded. 
 

A variety of quality control procedures were used to ensure the accurate and reliable collection of these data. Calibrations of the Canon forty-five degree non-mydriatic camera were performed prior to conducting examinations at each survey location. Prior to data collection, a multistage training was conducted during which the technician was required to image a minimum of 25 volunteers (50 eyes). Six images captured by each technician were sent to graders for evaluation. At regular intervals, NCHS and contractor staff observed digital imaging examinations to evaluate whether appropriate procedures and techniques were being followed. Midyear and annual retraining sessions were conducted with health technicians to review data collection procedures, provide feedback, review grader evaluation comments, and reinforce proper protocols and technique.

Digital images of the retina were sent to graders under contract to the University of Wisconsin, Madison for reading. The grading protocol for assessing retinopathy, macular degeneration, and other retinal lesions was successfully performed in NHANES III. The grading team included nine experienced graders (a preliminary grading coordinator, two preliminary graders, and six detail graders).

The digital images were evaluated in semi-quantitative fashion by independent graders using EyeQ Lite, an image processing software for storage, retrieval, and manipulation of digital images. The graders viewed each retinal image with a high resolution monitor, using the EyeQLite image processing software and database, and referenced the written protocol and the digital photographic standards to evaluate retinal abnormalities.

Three distinct steps comprised the grading process: early review for pathology, preliminary grading, and detail grading. A preliminary grading coordinator reviewed images for significant pathology and notified NCHS of pathology within 3 days of receiving the images. All images were further graded by at least two graders (a preliminary grader and a detail grader). Upon systematic grading, if the first two graders did not agree on pathology, a third grader graded the eye. If two of the three graders disagreed, the image was evaluated by an adjudicator to make a final decision. Refer to the NHANES Digital Grading Protocol on the NHANES website for more detail (2).

The reading center provided a photo quality report generated from the graders’ evaluations of all photographs which was used to evaluate re-training of technicians.

For details on the QA/QC process for this component, please refer to the NHANES Ophthalmology Procedures Manual on the NHANES website.

Inspection, calibration, and maintenance of the equipment and supplies were performed on a regular basis. The NHANES Ophthalmology Procedures Manual details the equipment quality control procedures (1).

Images with a CDR 0.6 or greater were re-graded in 2012 to look for indicators suggestive of glaucoma. As a quality control measure, 200 optic disc images were randomly selected from eyes with a vertical CDR of less than 0.6 (considered “normal” for glaucoma for this re-grading process) and evaluated for signs suggesting the presence of glaucoma. The validation study was considered successful.

Automated data collection procedures for the survey were used to collect the digital imaging data. Digital images were captured by a non-mydriatic camera. The captured images were then transferred to the survey database system.

Data on indicators of glaucoma (quality, vertical CDR, glaucoma, disc heme, size, excavation, notch, tilted) were provided to NCHS by the ophthalmologists with Johns Hopkins University involved in the 2012 study of re-grading images with a CDR 0.6 or greater.

During the editing process, all data were examined for logical inconsistencies and technician or equipment errors. Edits of the data were performed when errors were detected.

Users are encouraged to note carefully the “English Instructions” describing the various variables and suggested analysis in the codebook documentation.

Note that eye conditions are provided for the worse eye, right eye, and left eye. The following can be determined from the variable names: OP - Ophthalmology component; third letter - data was gathered through examination (X), it is an administrative variable (A), or a derived variable (D). Worse eye is defined by “U”, right eye by “D”, and left eye by “S”. The remaining letters refer to the corresponding data gathered or eye condition.

OPASCCT2 (Retinal Imaging Exam Status Comment) is a quality control variable that allowed NCHS to monitor the component and the reasons technicians provided for why an exam was not done or incomplete. The data were not collected for analytic reasons but for quality control purposes. This variable is only provided in the data release file to allow analysts to have some information on missing data and possible reasons for non-response.

References to the Early Treatment Diabetic Retinopathy Study (ETDRS) standards are made in the codebook, thus familiarization with the standards is helpful.

Retinal Imaging data can be combined with the Ophthalmology Frequency Doubling Technology data to assess for the presence of glaucoma.  Visual acuity data can be used in analysis where noted in the English Instructions.

Please refer to the Analytic Guidelines on the NHANES website to obtain further details on the use of sample weights and other analytic issues (3).

The NHANES Digital Grading Protocol which can be found on the NHANES website provides definitions and a detailed description of the grading process (2).

The following procedures were used in the 2012 re-grading of the optic disc images:

• The images were inserted into a tablet-based review system (Truthmarker for iOS) and graded in the following areas:
  • Quality: Excellent, Good, Fair, Poor, Ungradable
  • Vertical CDR: 0.0 to 1.0 in increments of 0.1, Ungradable
  • Glaucoma: No, Possible, Probable, Definite, Ungradable
  • Disc Heme: No, Maybe, Yes, Ungradable
  • Size: Small, Average, Large, Ungradable
  • Excavation: No, Maybe, Yes, Ungradable
  • Notch: None, Inferior, Superior, Both, Ungradable
  • Tilted: No, Yes, Ungradable

• Three glaucoma specialists graded each image and then the results were adjudicated if necessary.
  • Quality:
    • If at least 2 of 3 graders agreed on quality, and the third grader was within one level on the scale, that grade was assigned to the image.
    • If there was no agreement between at least 2 graders or if the third grader was off by two or more levels, the image was reviewed in the presence of all graders to achieve consensus.
    • Images judged to be ungradable using this process were not assessed regarding the criteria below.
  • Vertical CDR:
    • If at least 2 of 3 graders were within 0.1 on CDR, the median value was assigned to the image.
    • If there was disagreement of at least 0.2 between any two graders, the image was re-reviewed in the presence of all graders and consensus was reached.
  • Glaucoma:
    • If at least 2 of 3 graders provided the same grade, and the third grader was within 1 level, that grade was assigned to the image.
    • If there was no agreement between at least 2 graders or if the third grader was off by two or more levels, the image was reviewed in the presence of all graders to achieve consensus.
  • Disc Heme:
    • If at least 2 of 3 graders provided the same grade, that grade was assigned to the image.
    • If there was no agreement between at least 2 graders, the image was reviewed in the presence of all graders to achieve consensus.
  • Size:
    • If at least 2 of 3 graders provided the same grade, that grade was assigned to the image.
    • If there was no agreement between at least 2 graders, the image was reviewed in the presence of all graders to achieve consensus.
  • Excavation:
    • No consensus grade was assigned. The grades of all three graders are were included then an algorithm was applied to determine the final value:
      • If two of the three grades were the same, that grade was assigned to the image
      • If the grades were maybe, no, ungradable, the grade assigned to that image was "ungradable"
      • If the grades were maybe, yes, ungradable, the grade assigned to that image was "ungradable"
      • If the grades were maybe, no, yes, the grade assigned to that image was "yes"
  • Notch:
    • No consensus grade was assigned. The grades of all three graders are were included then an algorithm was applied to determine the final value:
      • If two of the three grades were the same, that grade was assigned to the image
      • If the grades were superior, inferior, and both, the grade assigned to that image was "both"
      • If the grades were three different grades and one of those grades was “none”, the grade assigned to that image was "none"
  • Tilted:
    • No consensus grade was assigned. The grades of all three graders are were included then an algorithm was applied to determine the final value:
      • If two of the three grades are the same, that grade was assigned to that image
      • If the grades were three different grades (no, yes, ungradable), the grade assigned to that image was "ungradable"