Eye injuries after anesthesia, although infrequent, may result in visual impairment. Previous studies have not defined the risk factors associated with these injuries. To study the cause of these injuries and to determine incidence data, the authors reviewed the records from a 4.5-y period of patients who sustained eye injuries after anesthesia and nonocular surgery.
The records were surveyed of 60,965 patients who underwent anesthesia for nonocular surgery at our institution from January 1988 to July 1992. Eye injuries were identified by examining several sources of information, including quality improvement data, postoperative visits or telephone calls, and examination of medical records of postoperative patients who required an ophthalmology consultation. Records were reviewed to determine the types and causes of eye injuries. Logistic regression, with univariate and multivariate analysis, was used to identify the importance of possible risk factors.
Of 34 patients who sustained eye injuries (0.056%), the most common was corneal abrasion (n = 21). Other injuries were conjunctivitis, blurred vision, red eye, chemical injury, direct trauma, and blindness. Independent factors associated with a higher relative risk of eye injury were long surgical procedures (odds ratio, 1.16 per hour of anesthetic care; CI, 1.1 to 1.3), lateral positioning during surgery (odds ratio, 4.7; CI, 2 to 11), operation on the head or neck (odds ratio, 4.4; CI, 2.2 to 9.0), general anesthesia (odds ratio, 3.0; CI, 2.2 to 38), and surgery on a Monday (odds ratio, 2.7; CI, 1.4 to 5.3). In only 21% of cases was a specific cause of injury identified.
For reasons not understood, eye injuries are more common in patients who are older or who undergo lengthier procedures. The cause of injury was usually not determined by the authors' methods. Until the mechanisms producing perioperative eye injuries are better understood, it is difficult to identify strategies to prevent this type of injury.
Key words: Anesthesia, complications: blindness; corneal abrasion; Eye: trauma.
Eye injuries after anesthesia for nonocular surgery are relatively uncommon, but when they do occur, there may be discomfort or pain and sometimes loss of vision. The most commonly reported complication in the literature is corneal abrasion. Its cause during anesthesia has been related to direct trauma to the cornea from face masks, surgical drapes, or other foreign objects that accidentally contact the eye. Another possible, although unproved cause may be decreased basal tear production secondary to general anesthesia. The reported incidence of corneal abrasion varies depending on the methods used to detect it; one small prospective study in patients undergoing general surgical procedures (using fluorescein staining of the corneal epithelium) reported an incidence of 44%, whereas a larger prospective study in a selected group of patients undergoing neurosurgical procedures (assessed by clinical symptoms) reported a 0.17% incidence of corneal abrasions. [1,4]
Other types of reported eye injuries have included injures from toxic chemicals allowed to contact the eyes during induction of anesthesia and various degrees of visual loss, possibly related to pressure on the eye or to optic nerve or retinal ischemia. Although constituting only 3% of cases in the ASA Closed Claims Study, eye injuries were associated with significantly greater monetary settlements compared with claims for nonocular injuries. In that series, corneal abrasion was also the most common eye injury. Examination of factors associated with eye injuries, in a large series of patients having surgery, has not been previously reported. In a relatively small survey of patients having day surgery at two centers, the incidence of an "eye problem" was 0.6% and 0.1% in 524 and 886 patients, respectively. Furthermore, the clinical impression that eye injuries are more likely in patients undergoing surgical procedures involving the face, head, or neck, or in patients positioned other than supine (e.g., prone) has not been confirmed. To examine these issues, we studied the incidence of eye injuries after anesthesia for nonocular surgery in our academic anesthesia practice.
Materials and Methods
Data collection for this study was approved by the Institutional Review Board. We evaluated the incidence of eye injuries in all patients older than 6 y who received anesthetic care for nonocular surgery at the University of Chicago Medical Center from 1 January 1988 through 30 June 1992. In our practice, demographic data on every patient receiving anesthetic care are routinely recorded for quality improvement, research, and billing purposes on a standard form on the day anesthetic services are rendered. Forms (Figure 1) are completed by residents, certified nurse anesthetists, or faculty personally involved in the anesthetic care; in all cases, faculty are required to examine forms of patients under their care for completeness and accuracy before submission. Information from these forms constitutes the basic data set contained within the department's Continuous Quality Improvement Database (CQID), which is maintained using the Fox Pro (Microsoft, Redmond, WA) program on a Novell-based personal computer server system. Demographic data for the study population appear in Table 1.
Information regarding ocular complications was derived from several sources used by our department to evaluate all anesthetic-related complications. All complications of any type possibly related to anesthesia that occurred in the perioperative period were recorded in the CQID. The information was transmitted from the various sources to either of two faculty members (J.P.E. or B.D.S.), who then entered the data into the CQID. No attempt was made during the study period to prospectively inquire in particular about eye injuries; rather eye injury was merely one of a large collection of complications that were routinely recorded. The Department of Anesthesia and Critical Care has a mandatory reporting procedure: Anesthetic-related complications are reported by the anesthetic care team, postanesthetic care nurses, and by a nurse assigned to postoperative follow-up.
Approximately 95% of inpatients were seen on postoperative follow-up within 3 days of surgery by a full-time nurse employed by the Department of Anesthesia and Critical Care. This nurse examined patients' charts for the presence of anesthetic-related complications. If possible, patients were asked to describe any symptoms that were present. The nurse did not have access to the CQID database information before seeing the patients; rather a list was made available of patients requiring postoperative follow-up.
Another source was word-of-mouth reporting of anesthetic-related complications to five to seven full-time anesthesia faculty and the departmental chairperson, all of whom were members of the Quality Improvement Committee. There was no "critical incident" reporting system; rather, word-of-mouth recording occurred at the discretion of the anesthetic faculty member. A half-time employee telephoned outpatients within 1 week of surgery to inquire about anesthetic complications; however, only about 50% of patients were available for follow-up. The caller inquired about any complaints that were related to the perioperative period and did not use a specific questionnaire. A fourth method was to evaluate the medical records of all patients who required inpatient ophthalmology consultations within 3 months of their date of surgery. (It was deemed unlikely that perioperative eye injuries would first become apparent beyond that period of time). It is possible that some patients may have been lost to follow-up, particularly because we did not have access to records of patients that may have been seen by ophthalmologists outside of our institution.
Anesthetic and hospital records of patients with suspected anesthetic-related eye injuries were reviewed by two attending anesthesiologists (S.R. and S.B.). All of the hospital and anesthetic records of the 34 patients who sustained ocular injuries were available. From the CQID, patient name and date of surgery, anesthesia faculty, resident or certified nurse anesthetist, type of anesthetic, ASA physical status, operative location and position, inpatient or outpatient status, surgical scheduling status, patient age, sex, race, duration of anesthesia, and anesthetic complications were already known. For patients who received general anesthesia, use of endotracheal intubation was also listed. Among the 34 patients whose charts were examined, no discrepancies were apparent between the patient records and the CQID. When a new ocular injury appeared in the postoperative period, an attempt was made to determine the cause of injury based on notations in the hospital records. We did not try to interview the anesthesia team members involved in the care of these patients because many were no longer at the institution, or we believed that recall of the generally distant events would be poor. Although in some instances a specific mechanism of injury was easily identifiable (e.g., one case in which an intravenous pole accidentally fell into a patient's eye), in other cases we could only speculate as to the cause of injury.
After identifying the patients who sustained eye injuries, their demographic information was separated from the CQID, leaving two distinct groups of patients: Nonocular injured, and ocular injured (Table 1). Summary data for age and duration of anesthetic care are reported as mean +/- standard deviation. We used SAS Release 6.1 (SAS Institute, Cary, NC) to examine the influence of the various demographic variables on the occurrence of eye injuries. Odds ratios and their 95% confidence intervals (CI) were estimated for these speculated risk factors. Because the outcome was rare, these represent estimates for the corresponding relative risks. The factors examined included ASA physical status, age, hours of anesthetic care, sex, race, hospital stay (inpatient vs. outpatient), surgical scheduling (elective compared with urgent or emergent case), surgical position (supine, sitting, jackknife, lateral, or prone), anatomic site of the operative procedure (head or neck vs. other), day of week of surgery, endotracheal versus nonendotracheal intubation, and type of anesthetic (general, epidural, sedation, or other).
Likelihood ratio chi-square analysis was used to determine the significance of differences in proportions of patients with or without perioperative eye injuries. The analysis consisted of three main components. First, univariate analysis was performed. Second, regardless of the results of the univariate analysis, logistic regression was performed using all factors (except ASA physical status, which was virtually identical among the groups) in a multivariate analysis to study the independent effects of risk factors when other factors were controlled for ("full model"). Aside from removal of hospital stay (inpatient vs. outpatient) and surgical scheduling (elective vs. urgent/emergent; see footnote to Table 1), no stepwise model building was used. Third, risk factors that were statistically significant in the full multivariate analysis were then included in a smaller logistic regression model, again without stepwise model building ("reduced model"). Finally, the model obtained with forward stepwise analysis and the same set of candidate predictor variables using the likelihood chi-squared criterion was the same as the reduced model. For age and duration of anesthesia, t testing was used to compare results between groups. In all instances, P < 0.05 was considered statistically significant.
During the 4.5-y period we reviewed, 60,965 anesthetics were administered for nonocular surgery (Table 1), with an overall incidence of eye injury of 0.056% (34 patients). Patients with eye injuries, compared with those without such injuries, were older (51.2 +/- 16.6 y vs. 44.1 +/- 20.7 y; P = 0.017 by t test), whereas the duration of their anesthetics was greater (4.7 +/- 3.0 vs. 2.9 +/- 2.6 h; P = 0.001 by t test). The odds ratio for eye injury was greater for older patients (1.17; CI, 1.0 to 1.4), but multivariate analysis showed that age was not an independent risk factor. Duration of anesthesia, however, was an independent risk factor for eye injury; per hour of anesthetic time, risk was increased for patients undergoing lengthier procedures (odds ratio, 1.16; CI, 1.1 to 1.3). Among eye-injured patients, 50% were men and 50% were women, compared with 38% and 62%, respectively, for the noninjured (NS). Not associated with a greater risk of injury were ASA physical status, sex, and race.
Inpatients were more likely to sustain eye injury than were outpatients (odds ratio, 10; CI, 1.7 to 59). Insufficient numbers of patients prevented assessment of the independent influence of this risk factor. Compared with other days of the week, risk of eye injury was increased in patients whose surgery was performed on a Monday (odds ratio, 2.7; CI, 1.4 to 5.3); this finding represented an independent risk factor. Furthermore, risk of eye injury was greater in patients operated on electively, compared with those having urgent and emergent procedures (odds ratio, 12.6; CI, 2 to 77). Numbers of patients were insufficient, however, to assess the independent influence of this risk factor.
Use of general anesthesia was an independent risk factor associated with eye injury. General anesthesia had been used in 33 of 34 patients (97%) who sustained an eye injury, whereas 63.6% of uninjured patients received general anesthesia (odds ratio, 9.2; CI, 2.2 to 38). Risk of eye injury was increased when endotracheal intubation was used, compared with mask ventilation (odds ratio, 3; CI, 1.4 to 6.5) in the univariate analysis, but this finding reflects only the association of endotracheal intubation with general anesthesia. Among patients who received general anesthesia, endotracheal intubation conferred no added risk (P = 0.88).
The risk of eye injury was greater in patients whose surgery involved the head or neck, and such surgery constituted an independent risk factor. In this group of patients, the odds of sustaining eye injuries were 4.4 times as great (CI, 2.2 to 9.0) compared with those whose surgery involved other regions. For all types of surgical procedures, the incidence of eye injuries was greater in patients operated on in the lateral position (odds ratio, 4.7; CI, 2 to 11), which was also an independent risk factor. In six patients operated on in the lateral position, who then subsequently were found to have corneal abrasions, it was difficult to discern the relationship between the position of the affected eye (up or down) and the occurrence of injury. Specifically, injury occurred in the "down" eye in two cases, in the "up" eye in three cases, and in both eyes in another.
The most common injury was corneal abrasion (21 patients). Others included conjunctivitis or red eye (7 patients), blurred vision persisting longer than 1 day (3 patients), chemical injury, permanent vision loss, and eyelid hematoma (1 patient each). With one exception, the diagnosis of corneal abrasion was made within 1 to 3 h after the conclusion of anesthesia. All patients with corneal abrasion were seen by a consultant ophthalmologist within 24 hours after onset of symptoms, and treatment was started soon after. None sustained permanent injury.
Overall, a specific cause for the eye injuries could only be discerned in 21% of cases. In most corneal abrasions (79%) or conjunctivitis (86%), a specific cause could not be determined. The most common findings were accidental loosening or removal of tape covering the eyes, leading to exposure keratopathy (2 patients), iodine prep solution dripping into the eye (1 patient), and trauma caused when an intravenous pole fell into an anesthetized patient's eye (1 patient). Eyes of all patients (with one exception) had been taped closed, and in only one patient was it documented that the tape had been dislodged. A chemical burn developed in a portion of the sclera in one patient when cidex solution spilled onto the eye from an incompletely rinsed reusable anesthetic mask. Permanent unilateral blindness developed in another patient from ischemic optic neuropathy after lumbar spinal fusion. Deliberate hypotension and hemodilution were used in this patient.
We found an overall incidence of eye injury in nonocular surgery of 0.056%. Our study is the first to describe the overall incidence of eye injury in a large, heterogeneous patient population. Compared with smaller studies of selected patient populations, we believe these data are useful for estimating incidence of low-risk events such as eye injuries and trying to evaluate risk factors. Most of the data examined in this study were collected prospectively and then analyzed retrospectively. All data had been entered into a previously established quality improvement database. Four methods were used to collect the information that was entered into the database: (1) self-reporting by faculty, residents, certified nurse anesthetists, and postanesthesia care nurses into the database, and review of postoperative patients by a nurse employed by our department, (2) "word-of-mouth" reporting by faculty to the chairperson and/or CQI committee; (3) telephone follow-up of outpatients, and (4) examination of ophthalmology consultations within 3 months of the date of a patient's surgical procedure. We do not know the precise "capture" rate of any of these methods, and our study did not try to identify eye injuries during the prospective data collection. Of these methods, the postoperative nurse follow-up was probably the most accurate, because more than 95% of the patients were seen. It is possible that we missed some cases of eye injury. However, we suspect the number of cases missed was small and not likely to have significantly altered our findings. Most of the eye injuries described resulted in significant pain and discomfort for the patients; as a result, it is not likely that many were left unreported to the physicians or nurses caring for these patients.
The biggest source of error was probably in outpatients, who may have been lost to follow-up. Yet we contacted 50% of these patients; because outpatients constituted 23% of the database (14,127 patients), about 7,000 patients were not contacted. Assuming the incidence of eye injury in these patients to be the same as in the overall population we studied (0.056%), it is possible that three or four patients could have been overlooked. It is improbable that these "misses" would have altered our findings significantly, although the point estimates for the effects of inpatient status and of general anesthesia would likely decrease.
Consistent with other reports, corneal abrasion was the most common injury in this series. Presence of corneal abrasion is suggested by obvious clinical symptoms. The denuded or injured corneal epithelium provokes a characteristic response consisting of tearing, miosis, photophobia, and foreign body sensation in the eye. The diagnosis may be confirmed by staining the cornea with fluorescein; injured areas stain yellow under blue cobalt light. In a previous large prospective study of 4,652 neurosurgical patients, the overall incidence of corneal abrasion was 0.17%. These authors found a greater incidence of corneal abrasion in patients operated on while in the prone position for lumbar laminectomy compared with other neurosurgical procedures. Either taping the eyes closed or applying ointment did not affect the incidence of corneal abrasion. Two other considerably smaller prospective studies of 76 and 127 adult patients found no corneal abrasions and no advantage in applying ointment rather than taping the eyes closed. [11,12]Other smaller series have reported a considerably greater incidence. In one study, using fluorescein staining of the corneal epithelium in patients whose eyes were left open during general anesthesia, 44% of the patients showed evidence of corneal injury, but only 3% complained of clinical symptoms. 
Corneal abrasion was the single most common injury in the database of eye injuries of the ASA Closed Claims Study, in which a cause for corneal abrasion could only be determined for 20% of the claims, despite the availability of patient records, depositions, and other materials. This is similar to the percentage of patients in whom we were able to determine a cause of injury.
We found several factors that may be involved in producing perioperative eye injuries. Patients with eye injuries tended to have undergone longer operative procedures. The incidence of eye injury was greater with increasing age, but this is due to the fact that older patients tend to undergo more surgery requiring general anesthesia, more operations on the head and neck, and more surgical procedures performed in the lateral position. It is also possible that the decreased tear production resulting from general anesthesia is an additional risk factor ; in this respect, it is important that corneal abrasion only occurred in one patient not subjected to general anesthesia. The rarity of this complication despite the many patients undergoing general anesthesia indicates that other, currently unknown factors in addition to whatever effect is produced by general anesthesia per se may be responsible.
Position of the patient may play an important role, because the incidence of eye injury was greater in positions other than supine, but only lateral positioning was an independent risk factor. Similarly, operative location was related to injury; head and neck surgery was associated with a greater incidence of injury. Although the second finding would be expected (perhaps due to the possibility of direct trauma by the surgeon or surgical assistants), it is somewhat less clear why lateral positioning is associated with a greater risk of injury. One possibility is trauma due to rubbing of the eye against a cushion used to support the head. However, the eyes of all patients (with one exception) had been taped closed, and in only one patient was it documented that the tape had been dislodged. The finding of no consistent relationship between position of the eye ("up" or "down") and the presence of injury suggests that the mechanisms responsible for the association between position and injury are not known.
We identified a cause for corneal abrasion in only two patients, who sustained exposure keratopathy, a preventable complication. In cases when no specific cause was found, it is possible that direct, undetected trauma was also involved. Others have noted the possibility of various objects contacting the eye accidentally (e.g., the anesthesiologist's identification card, bracelet, watch, stethoscope, and so forth). [2,13,14]Our finding of an increased risk of injury in patients undergoing endotracheal intubation, compared with mask ventilation, may support this speculation. However, endotracheal intubation was not an independent risk factor, because we could not distinguish its effect from that of general anesthesia itself. Perhaps the proximity of the anesthesiologist to the patient during procedures for which general anesthesia is used as well as that of any objects attached to the physician are far more likely to result in unintended contact with the patient's eye. Because the cornea is the principal barrier to infection in the eye, prompt recognition and treatment (antibiotics and patching of the eye) are essential to prevent further injury. Nearly all of the corneal abrasions in this series were diagnosed in the immediate postoperative period, and treatment begun almost immediately thereafter, which may explain the lack of permanent sequelae in these patients.
Several cases of conjunctivitis or red eye were identified. Two of the cases of red eye were due to a reaction to eye ointment. We did not have sufficient numbers of patients to accurately assess the risk factors associated with conjunctivitis.
A curious and unexpected finding of the study was the greater risk of eye injury in patients operated on on Monday. We have no clear explanation for this finding, which, to our knowledge, has never been documented for any other type of anesthetic-related injury. It would be interesting to examine our database for the presence of other, perhaps more common injuries (e.g., dental damage), to determine whether some consistent pattern associating anesthetic-related injuries and day of the week could be discerned.
Blindness occurred in only one patient in our series, and fortunately reports of this outcome have been rare. Blindness involves damage to the posterior segment of the eye (retina or optic nerve) or may be related to damage to the visual cortex. Mechanisms possibly responsible for these injuries include compression, hypotension, hypoxia, and anemia; however, the literature describing these complications is mostly anecdotal. [6,15,16]Data describing the influence of anesthesia or perioperative changes in oxygen delivery on the retina and optic nerve are only available in animals. [17-19]
We found that eye injuries after anesthesia, although uncommon, are often of indeterminate cause. Corneal abrasion was the most common injury. Eye injury occurs more frequently in association with head and neck surgery and is more likely to occur when the patient is positioned laterally, undergoes general lengthy anesthesia, and has surgery on Monday. Future studies of the cause of these injuries, if performed, preferably would be done in an entirely prospective manner. Because the incidence is so low, such a study would require far more patients than we studied. Until the mechanisms producing perioperative eye injuries are better understood, it is difficult to identify preventive strategies that are likely to decrease the risk of this type of injury.