Postoperative vision loss and visual disturbances in a Canadian provincial administrative data repository.

Abstract

Purpose: Population-based, health administrative database analyses of postoperative vision loss (POVL) have primarily examined International Classification of Diseases (ICD) diagnosis codes for ischemic optic neuropathy (ION) and retinal artery occlusion (RAO) after cardiac and spine surgery. Other nonspecific diagnosis codes describing significant vision loss have been mostly overlooked. We sought to investigate their contribution to the broader epidemiology of POVL after a variety of surgical procedures.

Methods: Using administrative data from Manitoba, Canada, we identified hospital admissions where patients underwent spine, cardiac, and other common inpatient and ambulatory surgeries between 1987 and 2017. To avoid misclassifying pre-existing vision loss as POVL, we excluded admissions where patients had ICD diagnosis codes suggestive of pre-existing vision loss in at least 5 years of preoperative data. Postoperative vision loss outcomes included specific diagnosis codes, such as RAO and ION, and nonspecific diagnosis codes for blindness, visual disturbances, and visual field defects.

Results: We excluded 158,730 admissions for suspected pre-existing vision loss, including 345 admissions with POVL diagnosis codes. We included 170 POVL cases in 596,241 admissions. Nonspecific diagnosis codes accounted for 140 (82%) cases and were associated with risk factors previously reported for ION, RAO, and cortical blindness. Overall, 24 (14%) cases were RAOs, and privacy restrictions precluded analysis of the remaining 6 (4%) cases. The incidence of RAO and nonspecific diagnosis codes was highest after cardiac and spine surgery. Still, other types of surgery accounted for 87 (62%) of the nonspecific diagnosis code cases and 11 (46%) of the RAO cases.

Conclusion: In surgical patients without a history of vision loss, new diagnosis codes for blindness, visual field defects, and visual disturbances other than ION and RAO may represent an important but overlooked aspect of POVL epidemiology.