Risk Factors for Post–Open-Globe Injury Endophthalmitis

Abstract

Purpose:

The objective of our project is to use the National Inpatient Sample Database to identify risk factors for endophthalmitis in cases of open-globe injury (OGI).

Methods:

This is a cross-sectional observational study of 48 627 cases of OGI from the National Inpatient Sample Database. We performed regression analysis using IBM SPSS Statistics 23. Codes from the International Classification of Disease, Ninth Revision were used to identify ocular findings and conditions. Variables with P values less than .05 on univariate analysis were included in the multivariable regression model; Bonferroni correction was applied to these results.

Results:

Of the 48 627 cases of OGI, 37 440 (77.0%) occurred in the adult group (21 years and older). Overall, 1018 (2.1%) cases developed posttraumatic endophthalmitis, with 74.5% cases in the adult group. Endophthalmitis developed in 293 (4.5%) eyes with an intraocular foreign body (IOFB). Results of binary logistic regression showed the clinical findings of traumatic cataracts, hypopyon, vitreous inflammation, corneal ulcers, or IOFBs were associated with an increased risk of developing endophthalmitis after OGI. Conversely, orbital fractures, rupture-type injuries, and intraocular tissue prolapse were associated with a decreased likelihood of being diagnosed with endophthalmitis.

Conclusions:

Endophthalmitis developed in 2.4% of pediatric OGIs and 2.0% of adult OGIs. Traumatic cataract, hypopyon, vitreous inflammation, corneal ulcer, keratitis, retinal detachment, IOFB, and diabetes increased the risk of post-open-globe endophthalmitis.

Introduction

Open-globe injuries (OGIs) are among the most common causes of monocular blindness, with an estimated incidence rate of 3.5 per 100 000 people worldwide and 2 to 3.8 per 100 000 people in the United States. ^{1 -4} One of the most feared complications of OGIs is endophthalmitis, which can have devastating consequences on anatomic and functional outcomes. The incidence of posttraumatic endophthalmitis (PTE) is low and noted in 1.0% to 11.9% of cases. ^5,6 The likelihood of developing PTE depends on a myriad of factors. Researchers have found that the presence of an intraocular foreign body (IOFB), lens rupture, delayed primary globe repair, age older than 50 years, female sex, large wound size, location of the wound, ocular tissue prolapse, placement of an intraocular lens, contaminated wound, isolation of a virulent organism, rural locale, retained lens particles, absence of antibiotic use, bow and arrow injuries, household injuries, and injuries involving a wooden object may increase the likelihood of developing infection. ^{6 -11} Some studies have shown that iris prolapse, hyphema, high-speed IOFBs, and prompt primary repair are associated with a decreased likelihood of developing endophthalmitis. ^6,8

The objective of this paper is to use a large national data set to evaluate risk factors for an uncommon disease, PTE. Various demographic, associated systemic, and ocular variables were analyzed to determine their impact on the likelihood of developing endophthalmitis after an OGI.

Methods

The inpatient data from the 2002 to 2014 National Inpatient Sample (NIS) Database was used to identify OGIs. The Healthcare Cost and Utilization Project, under the sponsorship of the Agency for Healthcare Quality and Research, created the NIS Database with the aim of sampling and analyzing inpatient hospital admission data. From the NIS Database, we extrapolated various demographic variables, systemic comorbidities, and ocular findings. Because the data were deidentified by the Healthcare Cost and Utilization Project, no informed consent was needed or obtained.

The NIS Database provides data pertaining to a patient’s hospital stay from admission to discharge. The database provides information on demographics, medical diagnoses (acute conditions and chronic diseases), procedures, and logistics of 1 specific hospital admission; it does not include information from outpatient medical records. Furthermore, the NIS Database does not follow patients longitudinally and cannot provide details of the patient’s health care after the hospital stay. The NIS Database uses the International Classification of Disease, Ninth Revision (ICD-9) for diagnosis codes in the timeline 2002 to 2014.

OGIs were included in this study if the admitting primary diagnosis was acute OGI or if the secondary diagnosis of OGI was coded concurrently with globe repair performed during that specific hospitalization. All cases had a secondary diagnosis of endophthalmitis. Cases of OGI as a secondary diagnosis without documentation of globe repair were excluded from the diagnosis because these patients could have represented readmissions with ocular sequelae.

Fisher exact text and Χ² analysis were used for descriptive statistical analysis. The demographic, systemic, and ocular variables listed in Table 1 were used in our regression model. Table 2 shows the results of the Χ² analysis; the proportions (expressed as percentages) were calculated using 1018 as the total number of OGIs that developed endophthalmitis and 47 609 cases that did not develop endophthalmitis. Univariate and multivariable binary logistic regression analyses (Table 3) were performed using IBM SPSS 23. A P value of less than .05 was significant on univariate analysis. To reduce the likelihood of type 1 error, we applied Bonferroni correction, with a P value less than .004 considered statistically significant.

Table 1.

Tested Variables and International Classification of Disease, Ninth Revision (ICD-9) Codes.

Variable	ICD-9 codes
Retinal detachment	36100	36100	36101	36102	36103	36104	36105	36106	36107
	36181	36181	36189	3619	3612
Serous and hemorrhagic choroidal detachment	36361	36362	36363	36370	36371	36372
Iridocyclitis	36400	36401	36402	36403	36404	36410	36411	36421	36422
	36423	36424	3643
Hypopyon	36405
Corneal ulcer	37000	37001	37002	37003	37004	37005	37006	37007
Keratitis	37020	37021	37022	37023	37024	3708	3709
Orbital fracture	8026	8027
Hypotony	36030	36031	36032	36033	36034
Traumatic cataract	3662	36620	36621	36622	36623
Vitreous inflammation	37929	36012	36004
Open-globe injury	8710	8711	8712	8714	8715	8716	8717	8719
Penetrating injury	8710	8711	8714	8715	8716	8717
Rupture injury	8712
Intraocular foreign body	8715	8716
Intraocular tissue prolapse	8711	37926
Endophthalmitis	36000	36001	36002	36003	36013	36019	09842

Table 2.

Results of Χ² Test.

	No. (%) of selected cohort with OGI
Variable	No endophthalmitis (n = 47 609)	Endophthalmitis (n = 1018)	P
Sex			.44
Women	12 994 (27.6)	269 (26.4)
Men	34 107 (72.4)	749 (73.6)
Age, y			.03
0-20	10 766 (22.7)	260 (25.5)
21+	36 682 (77.3)	758 (74.5)
Race			.001
White	19 748 (52.2)	427 (54.1)
Black	7549 (20.0)	120 (15.2)
Hispanic	7487 (19.8)	175 (22.2)
Asian/Pacific Islander	876 (2.3)	28 (3.5)
Native American	391 (1.0)	*
Other	1749 (4.6)	34 (4.3)
Injury details
Penetrating OGI without IOFB	33 237 (69.8)	534 (52.4)	< .001
Intraocular foreign body	6258 (13.1)	293 (28.8)	< .001
Rupture	6145 (12.9)	73 (7.2)	< .001
Ocular variables
Corneal ulcer	219 (0.5)	84 (8.3)	< .001
Hemorrhagic and serous choroidal detachment	796 (1.7)	19 (1.9)	.63
Hypopyon	34 (0.1)	114 (11.2)	< .001
Hypotony	199 (0.4)	*	.02
Intraocular tissue prolapse	20 985 (44.1)	208 (20.4)	< .001
Keratitis	48 (0.1)	14 (1.4)	< .001
Orbital fracture	2943 (8.3)	20 (2.0)	< .001
Retinal detachment	1654 (3.5)	89 (8.7)	< .001
Traumatic cataract	3551 (7.5)	149 (14.6)	< .001
Vitreous inflammation	29 (0.1)	29 (2.8)	< .001
Systemic variables
Diabetes without chronic complications	3702 (8.0)	97 (9.8)	.03
Diabetes with chronic complications	521 (1.1)	24 (2.4)	< .001

Abbreviations: IOFB, intraocular foreign body; OGI, open-globe injury.

*Denotes cells with a small number (1-10) of patients.

Table 3.

Results of Univariate and Multivariable Regression With Bonferroni Correction.^a

	Univariate		Multivariable
Variable	OR (95% CI)	P	OR (95% CI)	P
Sex
Women	1.00	(Reference)	–	–
Men	1.06 (0.92-1.22)	.42	–	–
Age, y
0-20	1.00	(Reference)	–	–
21+	0.85 (0.74-0.99)	.03	–	–
Race
White	1.00	(Reference)	1.00	(Reference)
Black	0.72 (0.59-0.87)	< .001	0.75 (0.61-0.93)	.01
Hispanic	1.05 (0.89-1.25)	.54	–	–
Asian/Pacific Islander	1.49 (1.02-2.17)	.04	1.20 (0.78-1.85)	.40
Native American	0.59 (0.24-1.41)	.24	–	–
Other	0.90 (0.64-1.27)	.55	–	–
Injury details
Penetrating injury without IOFB	1.00	(Reference)	–	–
IOFB	2.53 (2.20-2.91)	< .001	1.89 (1.60-2.24)	< .001
Rupture	0.64 (0.50-0.82)	< .001	0.49 (0.37-0.64)	< .001
Ocular variables
Corneal ulcer	19.42 (14.98-25.18)	< .001	20.27 (15.16-27.10)	< .001
Hemorrhagic and serous choroidal detachment	1.17 (0.75-1.84)	.48	–	–
Hypopyon	175.92 (119.28-259.47)	< .001	175.72 (115.74-266.77)	< .001
Hypotony	2.23 (1.16-4.3)	.02	2.25 (0.95-5.32)	.07
Intraocular tissue prolapse	0.33 (0.28-0.38)	< .001	0.35 (0.29-0.42)	< .001
Keratitis	14.29 (7.90-25.85)	< .001	6.57 (3.30-13.10)	< .001
Orbital fracture	0.22 (0.14-0.35)	< .001	0.37 (0.23-0.57)	< .001
Retinal detachment	2.67 (2.14-3.33)	< .001	2.22 (1.74-2.83)	< .001
Traumatic cataract	2.13 (1.79-2.54)	< .001	1.51 (1.23-1.85)	< .001
Vitreous inflammation	47.84 (28.45-80.44)	< .001	22.41 (10.80-46.50)	< .001
Systemic variables
Diabetes without chronic complications	1.26 (1.02-1.56)	.03	1.87 (1.50-2.35)	< .001
Diabetes with chronic complications	2.21 (1.47-3.34)	< .001	3.55 (2.25-5.61)	< .001

Abbreviations: IOFB, intraocular foreign body; OGI, open-globe injury; OR, odds ratio.

^aVariables with “–” in the Multivariable column were not included in the multiple regression model because they were not significant (P < .05) on univariate analysis. The P values found through the multivariable regression model were adjusted using the Bonferroni correction method. In bold are the variables that were significant on univariate analysis, multivariable analysis, and after Bonferroni correction. The Bonferroni adjusted alpha cutoff value is .004.

Results

We identified 48 627 cases of OGI in the NIS Database after applying discharge weights. Table 2 shows a breakdown of our data by demographic variables, systemic comorbidities, and ocular findings and presents the findings of our Χ² analysis. Of these cases, 11 026 (22.7%) cases occurred in the pediatric group (aged 20 years and younger), and 37 440 (77.0%) cases occurred in the adult group (aged 21 and older). Men composed 71.7% of cases of OGI documented in the NIS Database. Endophthalmitis developed in 1018 (2.1%) cases; 25.5% of cases of PTE occurred in the pediatric group and 74.5% occurred in the adult group. Of the cases of OGI in the pediatric group, 260 (2.4%) developed endophthalmitis; this figure was 758 (2.0%) in the adult group. Furthermore, endophthalmitis developed in 293 (4.5%) eyes with an IOFB. Diabetes with and without chronic complications, retinal detachment (RD), hypopyon, traumatic cataracts, corneal ulcers, keratitis, IOFBs, hypotony, and vitreous inflammation were more common in patients who developed PTE (P < .05). Conversely, orbital fractures, rupture-type injuries, penetrating injuries without IOFBs, and intraocular tissue prolapse were less common in patients who developed PTE (P < .05).

Table 3 shows the results of our univariate and multivariable analyses. After applying Bonferroni correction to reduce the likelihood of type 1 error, the clinical findings of corneal ulcer, keratitis, hypopyon, RD, traumatic cataract, vitreous inflammation, IOFBs, and diabetes with chronic complications had a positive association with PTE. Conversely, the presence of orbital fractures, rupture-type injuries, and intraocular tissue prolapse were associated with a decreased risk of PTE. The corresponding odds ratios (ORs) for the variables are presented in Table 3.

Conclusions

In our study, men accounted for 71.7% of OGIs. We found that the prevalence of endophthalmitis was 1.7% in patients without IOFBs and 4.5% in patients with IOFBs. The pediatric population accounted for 22.7% of cases with OGI, and 2.4% of these eyes developed endophthalmitis; similar figures have been reported in the literature. ^{3,6 -17} The literature cites many variables that may increase the risk of infection; these include retained IOFB, lens rupture, delayed primary globe repair, age older than 50 years, female sex, large wound size, location of the wound, ocular tissue prolapse, placement of an intraocular lens, contaminated wound, rural locale, retained lens particles, absence of antibiotic use, isolation of virulent pathogens, pure corneal injuries, needle-related trauma, shorter lacerations, bow and arrow injuries, household injuries, and injuries involving a wooden object. ^{5 -11}

Presence of a traumatic cataract after an OGI had a 50% higher chance of being associated with endophthalmitis in our study. It has been previously documented in the literature that lens disruption is a predictor of developing PTE. ^8,11 Researchers have postulated that disruption of the lens and its fibers facilitates pathogen entry into the vitreous cavity, leading to inflammation. Other theories suggest that damage to the lens may disrupt normal aqueous outflow, thereby impairing clearance of microorganisms. ^8,18,19 Additionally, fragmentation and/or retention of lens particles can by itself cause noninfectious severe intraocular inflammation resembling an infection. ^{20 -22} Thus, presence of traumatic cataract may increase the likelihood of being diagnosed with either sterile or infectious endophthalmitis.

Studies have found that patients with diabetes are at an increased risk of developing endophthalmitis postoperatively. ^{23 -27} In our study, the risk of PTE in patients with diabetes increased by 2- to 4-fold. The mechanism by which diabetes increases the risk of endophthalmitis may involve a debilitated immune system secondary to altered antigen response and impaired phagocytic capabilities. ^{28 -30} With increased severity of diabetes, defined as presence of diabetes-related chronic complications, the risk of PTE doubled (OR, 1.89-3.60) when compared with diabetic patients with diabetes without chronic complications.

It is not surprising that the presence of hypopyon and vitreous inflammation were strong risk factors for developing endophthalmitis. These variables, being signs of infection, have a strong association with endophthalmitis. ^6,29,31,32

Our analysis found that the presence of an IOFB was a significant risk factor for developing PTE (OR = 1.89). Foreign bodies may be contaminated with bacteria, and their entry into the globe directly inoculates the vitreous. ⁶ The rates of endophthalmitis development found in our analysis (1.7% in cases without IOFB vs 4.5% in cases with IOFB) are in concordance with those presented in previous studies. Studies have found that the incidence of PTE without IOFB is between 3.1% and 30%; however, this figure may be as high as 60% in cases that involve a foreign body. ^{5,6,33 -38}

Researchers have further stratified IOFBs by location and have found that posterior foreign bodies pose a greater risk for developing endophthalmitis than do anterior foreign bodies. ^6,37 A limitation of the NIS Database is that it does not specify the intraocular location site of the IOFB; for this reason, we could not differentiate between anterior and posterior foreign bodies. Furthermore, some studies have found that the material of the IOFB may have an impact on the risk of developing endophthalmitis. ⁶ Duan et al found that cases with organic-material IOFBs (eg, twig injury) had the highest risk of developing endophthalmitis, whereas glass and plastic IOFBs had the lowest risk. ³⁸ Researchers also theorize that high-velocity foreign bodies (ordnance-related and fireworks accidents) may be associated with a decreased risk of developing infection because the high-speed projectiles can heat up through increased air resistance and can become self-sterilized before impact. ⁶ In their study of 79 US military soldiers who experienced combat-related OGI, Colyer and colleagues found that no eyes developed endophthalmitis even with delayed IOFB removal. ³⁵ Although most clinicians advocate prompt removal of the IOFB to decrease the risk of developing infection, ^{6,36 -38} delayed IOFB removal may not increase the risk of PTE. ³⁵

Our data also found a significant positive association of corneal ulcers and keratitis with PTE. The NIS Database does not provide information regarding when the corneal ulcers were diagnosed. It is possible that corneal ulcers were the initiating pathology that led to OGI and endophthalmitis. ³⁹ Eight percent of our endophthalmitis post-OGI cases had a corneal ulcer. In the setting of corneal ulcer and OGI, the present study noted an 18-fold greater chance of the eye having endophthalmitis than not. Studies have documented progression of infectious keratitis to endophthalmitis in 0.3% to 6.1% of cases. ^40,41 Endophthalmitis secondary to keratitis or corneal ulcers is associated with particularly poor outcomes and a high rate of enucleation. ^{42 -44}

Similarly, the presence of an RD in our study was also associated with PTE; the reason for this may be that RD implies that the severity of the trauma was great enough to involve the posterior segment. Previous research has demonstrated that cases of concurrent endophthalmitis and RD are associated with a poor visual prognosis. ⁴⁵

Orbital fractures in the setting of OGI had a negative association with PTE. Furthermore, patients with rupture-type injuries had a decreased risk of PTE as compared with patients with penetrating injuries. For orbital fractures and rupture-type injuries, this decreased risk may be due to the type and etiology of the injury and subsequent location of the globe rupture. Previous research has found that trauma caused by a blunt force is a more common cause of orbital fracture than trauma caused by a sharp object. ^{46 -48} Similarly, blunt trauma causes globe ruptures and is more likely to involve the posterior aspect of the globe, zones 2 and 3. ^47,49 The posterior region of the globe is sequestered from the surrounding environment; for this reason, posterior ruptures may not have as high a risk of contamination as anterior scleral or corneal open wounds and thus may be at a reduced risk of developing endophthalmitis. ⁵⁰

Prolapse of intraocular tissue was associated with a 60% decreased risk of PTE. The prolapsed tissue through the wound may seal the opening and preclude pathogens from entering the eye. ^18,50,51 In a study of 558 OGIs (PTE prevalence of 0.9%) by Andreoli et al, endophthalmitis developed in 0.1% of patients with uveal prolapse, and 0.7% of patients without uveal prolapse; however, this trend was not statistically significant. ⁵¹ Soheilian and colleagues, on the other hand, found that prolapse of ocular tissue increased the risk of developing PTE, ⁵² suggesting this promotes exposure to infectious agents, thereby increasing the risk of infection. ⁵³ Further research needs to be performed to evaluate the degree to which intraocular tissue prolapse affects the risk of developing PTE.

Given the structure of the NIS Database and its reliance on ICD-9 codes, several limitations to this study exist. The ocular and systemic diagnoses evaluated in our regression model were established through ICD-9 codes, and their reliability depends on the accuracy of the individual coding the data. Because the NIS Database is not specific to ophthalmology, we were not able to analyze the data with respect to variables not encoded by ICD-9; for this reason, visual acuity, order of established diagnoses, and afferent pupillary defects, among other variables, could not be analyzed. Additionally, laterality of a diagnosis cannot be established, which can introduce some bias in the data. The NIS Database does not provide information regarding the sequence of diagnoses; however, we can be reasonably sure that certain variables (diabetes, intraocular tissue prolapse, orbital fracture, IOFB, penetrating injury, rupture injury, etc) occurred before the diagnosis of PTE was made.

In summary, our analysis of the NIS Database revealed multiple comorbidities associated with PTE. The presence of posttraumatic cataracts, hypopyon, RD, vitreous inflammation, corneal ulcers, and IOFBs was associated with increased risk of infection. Conversely, orbital fractures, rupture-type injuries, and intraocular tissue prolapse were protective against infection. Patients with diabetes, especially those with diabetes-related chronic complications, were at a significantly high risk, approximately 4-fold, of developing PTE. This fact needs to be further elucidated in future studies.