Hospitalization for Nontraumatic Disorders of the Eye and Ocular Adnexa: Analysis of the Florida Agency for Health Care Administration DataSet

Objective  To study the demographic features and patterns of hospital admissionin Florida for nontraumatic disorders of the eye and ocular adnexa.

Methods  The public data set from the Florida Agency for Health Care Administrationfor 2001 was used to identify persons hospitalized for 24 hours or longerfor nontraumatic disorders of the eye and ocular adnexa by using International Classification of Diseases, Ninth Revision, Clinical Modification codes.

Results  In 2001, there were 2137 hospital admissions for nontraumatic disordersof the eye and ocular adnexa, most of which were for infections or neuro-ophthalmologicdisorders. The median length of stay was 3.0 days (mean ± SD, 3.4 ±3.8 days). On average, 1 patient was admitted per month to 180 Florida hospitals.Eighty-three patients (3.9% of eye admissions) were hospitalized for 10 daysor longer and accounted for 18.1% of total hospital-patient days. Prolongedhospital stay was positively associated with hospital transfer (P < .001) and facial cellulitis (P = .04).A trend for positive association with Medicaid coverage was also observed(P = .07).

Conclusions  Nontraumatic eye care composes a small proportion of all inpatient care(< 0.1%) in Florida. Few of these patients require prolonged hospitalizationbut use a large proportion of inpatient care on the basis of the percentageof gross charges. An opportunity exists to improve hospital efficiency andimprove eye care by targeting the patients at highest risk for prolonged hospitalstay.

Modern hospitals are important collective investments of communitiesand the place most people think of going when they are seriously ill or injured.Hospitals are typically the caregivers of last resort for persons withoutmedical insurance and the hub of medical education. During the past decade,hospitals have come under intense financial and legislative pressure to reducewaste, cut costs, and improve efficiency. The effects of these forces on hospitalsare complex and often difficult to selectively measure. Data from the Agencyfor Healthcare Research and Quality, for instance, have shown declining hospitalstays but rising hospital charges throughout the United States from 1993 through2000.¹

Despite the central role hospitals have in our health care system, thereis little population-based information on inpatient eye care. Previous studiesof hospital eye care have dealt with ocular injury.^2-5 Tobetter understand the nature, distribution, and volume of inpatient eye carenot related to trauma, we studied the hospital discharge data set of the FloridaAgency for Health Care Administration (AHCA), Tallahassee, Fla, for 2001.

The public-use hospital discharge data set from AHCA for Florida for2001 was accessed for clinical and demographic information for patients dischargedwith primary diagnoses for eye and ocular disease by using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). This comprehensive set of codesincluded benign and malignant neoplastic diagnoses of the eye, conjunctiva,eyelids, lacrimal gland, ocular adnexa, and orbit (Table 1).

The 2001 AHCA data set includes discharge summaries from all nonfederalFlorida hospitals except state tuberculosis hospitals and state mental healthhospitals. After data are entered into the system, they are subjected to formattingand logic checks. The primary hospital submitting patient information mustthen certify the data are correct and also verify the accuracy of a summaryreport before it is released by the AHCA.

We retrieved patient information (age, race, and sex), primary admissiondiagnosis, principle procedure code, principal payer, day of week of admission,total gross charges for hospitalization, hospital, length of stay, days beforeprocedure, source of admission (physician referral, clinic referral, healthmaintenance organization referral, hospital transfer, emergency room, courtor law enforcement, nursing home, other health care facility transfer, andother), and discharge status (1 of 9 levels). Admission type was classifiedby the admitting physician as emergency, urgent, or elective; when applicable,admission type was classified by the coding technician as newborn or other.Management systems for medical insurance, such as health maintenance organizationor preferred provider organization, were reported according to the primaryinsurance carrier (eg, Medicare, Medicaid, commercial, government, workers'compensation, CHAMPUS, etc).

For the purpose of this study, diagnostically related ICD-9-CM codes were combined to simplify analysis and minimize potentialcoding variations due to semantic differences in making clinical diagnoses.For example, endophthalmitis included 3 ICD-9-CM codes:36000 endophthalmitis, 36001 acute endophthalmitis, and 36002 panophthalmitis.

Annual rates of admission were calculated by using 2000 US census dataextrapolated to the study year and were expressed as number of hospital admissionsper 100 000 population.

The SAS System release 8.01 for Windows (SAS Institute Inc, Cary, NC)was used to analyze the data. Frequencies were reported as percentages. Prolongedlength of stay was defined as hospitalization for 10 days or more. Frequencydistributions were examined before and after stratification according to lengthof stay. Crude and adjusted odds ratios were calculated by using logisticregression.⁶ The adjusted odds ratios werederived from 1 full model that contained the following variables: age, malesex, black race (as compared with white), Medicaid enrollees (as comparedwith self-pay, commercial insurance, or other), emergency or urgent admission(as compared with elective admission), and source of admission (hospital transfer,as compared with other sources). We did not detect collinearity among theseindependent variables.

After deleting records that had missing values for the dependent variableand/or independent variables, records of patients who were not black or white,and records of patients who were discharged to another hospital, 1698 recordswere available for logistic regression. The records of patients who were dischargedto another hospital (n = 45) were deleted to minimize the probability of includingmultiple records for a single patient in the multivariate analysis. The χ² test was used to compare patients deleted from theregression analysis because of an incomplete data field, according to frequencyof primary admission diagnosis, with those who remained in the analysis. Anodds ratio greater than 1.00 indicated that the variable increased the oddsof prolonged length of stay, while an odds ratio less than 1.00 indicatedthat the variable protected against it. Ninety-five percent confidence intervalswere calculated in the traditional manner. Results were considered significant(P ≤ .05) if the confidence interval excludedthe null value of 1.00.

The potential role of secondary diagnoses in hospital length of staywas assessed by examining the frequency distribution of these diagnoses accordingto length of stay. Differences between the 2 groups were tested for statisticalsignificance by using the Fisher exact test.

From January 1, 2001, through December 31, 2001, 268 Florida hospitalshad 2 343 138 admissions with a hospital stay of 24 hours or more.Of these admissions, 2858 were for primary disorders of the eye and ocularadnexa. Seven hundred twenty-one admissions (0.03%) for injuries to the eyeand ocular adnexa were excluded and are the data set of a separate study.The remaining 2137 (0.09%) admissions to 180 hospitals for nontraumatic disordersof the eye and ocular adnexa composed the data set of this study.

More than 95% of patients (n = 2034) listed a primary residence ZIPcode in Florida; 29 (1.4%) resided outside the country. The median numberof patients admitted per hospital was 6 (mean ± SD, 11.9 ± 17.2).

The nontrauma admissions involved 1011 male patients (47.3%) and 1126female patients (52.7%) and were listed with 204 different ICD-9-CM codes. One thousand three admissions (46.9%) were classifiedas emergency, 624 as urgent (29.2%), and 510 as elective (23.9%). Nearly 48%of the admissions were through hospital emergency departments, and 41% weredirect admissions by staff physicians. There were 56 hospital transfers (2.6%).The mean ± SD age at admission was 46.5 ± 29.2 years. One thousandfour hundred two patients were white (65.6%), 336 were black (15.7%), and332 were Hispanic (including both black and white Hispanic; 15.5%) (Table 2). The most patients were admittedon Tuesday (17.6%) and the least on Sunday (9.1%).

The median length of hospital stay was 3.0 days (mean ± SD, 3.4± 3.8 days; range, 1-92 days). Five hundred fifty-seven patients (26.1%)were discharged after 1 day, 495 (23.2%) after 2 days, and 381 (17.8%) after3 days. Eight-three patients (3.9%) remained hospitalized between 10 and 92days. The total number of days spent hospitalized for eye care was 7256 (ie,hospital-patient days). Eighty-two percent of all eye patients were dischargedhome, 3.0% to a skilled nursing home, and 4.3% to home care with supervisionof a health care organization. Because there were no hospital deaths, theterms "discharge" and "admission" are used interchangeably.

The median gross charge for hospital admission, not including professionalfees, was $8665 (mean ± SD, $11 307 ± $12 345) (Table 2). The median gross charge for patientshospitalized 9 days or fewer was $8357 (mean ± SD, $10 158 ±$7540) and for 10 days or more was $26 418 (mean ± SD, $39 744± $41 165) (Table 2).The most frequent payer sources were Medicare (35.1%), commercial insurance(22.8%), and Medicaid (15.3%) (Table 2).The total number of uninsured patients was estimated by combining the 3 payercategories: charity, other, and self-pay. With this definition, 190 patients(8.9%) lacked insurance coverage at the time of discharge. The mean chargefor hospitalized eye care in Florida was $3325 per day, which is comparablewith other types of inpatient care on the basis of national averages (Table 3).

The 2 most common reasons for nontrauma admission (ie, principal diagnoses)were acute inflammation (cellulitis) of the orbit (19.9%) and diplopia and/orcranial nerve palsies (11.1%) (Table 4).The 10 most frequent diagnostic categories accounted for approximately 75%of all admissions (Table 4). Overall,these hospitalizations were not procedure intense; no single procedure wasperformed in more than 50 patients (ie, in fewer than 2.3% of all hospitalizedeye patients). The 7 most frequent hospital procedures were as follows: removalof surgical implant in 49 patients, destruction of retinochoroidal lesionin 47 patients, scleral buckling procedure in 47 patients, mechanical vitrectomyin 40 patients, orbital surgery in 39 patients, artery biopsy in 29 patients,and cerebral angiography in 29 patients.

The 83 patients (3.9%) who were hospitalized for 10 days or longer accountedfor 1310 hospital-patient days, or 18.1% of the total days of inpatient care.Their gross cost for hospitalization was $3 298 752, or 13.7% ofthe total eye care–related charge of $24 160 922. Prolongedhospitalization was significantly associated with hospital transfer (adjustedodds ratio, 7.38; 95% CI, 3.18-17.13) (Table 5). There was a positive trend for prolonged stay with Medicaidcoverage (adjusted odds ratio, 1.93; 95% CI, 0.95-3.95). No associations werefound with age, race, or type of admission (Table 5). There were only minor differences in the distributionof diagnoses for patients with prolonged stay, except for a 3-fold increasein keratitis. The most frequent diagnoses for prolonged stay were acute orbitalinflammation (21 patients [25.3%]) and keratitis and/or ulcer (15 patients[18.1%]).

There were 624 unique secondary admission diagnoses. The 3 most commonwere systemic hypertension (7.6%), atrial fibrillation (3.3%), and chroniclung disease (2.4%). When the proportion of the 12 most common secondary diagnoseswere examined according to length of stay, only 2 showed a statistically significantdifference between groups: patients with systemic hypertension were more likelyto be discharged within 9 days (P = .004), whilethose with facial cellulitis were more likely to have a prolonged hospitalization(P = .04).

The Florida AHCA database revealed that there were 2137 hospital admissionsfor nontraumatic disorders of the eye and ocular adnexa in 2001, or fewerthan 0.1% of all hospital admissions for the year. The number of patientsadmitted for ocular trauma was even lower (721 admissions), or approximatelyone third the annual total for nontraumatic admissions. These data confirmwhat casual observation has suggested for some time—that ophthalmologypatients compose a small fraction of inpatient health care. Most eye admissionswere not discretionary; they were usually for serious vision-threatening disorders(eg, keratitis), potentially life-threatening disorders (eg, orbital cellulitis),or neuro-ophthalmological emergencies (Table 4).

Hospitals in Florida, on average, admitted 1 patient for nontraumaticeye care per month, which is a relatively low rate, considering the technicalexperience many of these admissions demand. Low rates of hospital admissionmight raise quality of care issues because of a possible association betweenincreased patient volume and better clinical outcome, which have been documentedacross a range of specialties.^7,8 Thevolume-outcome relationship has not been specifically studied for urgent orinpatient eye care.^7,8

Another finding in this survey that deserves comment was the skeweddistribution of length of hospital stay, with fewer than 4% of all hospitalizedpatients accounting for more than 18% of all hospital-patient days. The averagecost of individual hospitalization for this minority of patients was nearly4 times the overall average. Because of the high cost of inpatient care, insurancecarriers no longer allow hospitalization without justification of medicalnecessity. Prolonged length of stay is probably a reliable surrogate measureof disease severity. A better understanding of the reasons for prolonged lengthof stay could lead to measures that reduce the risk for extended hospitalcare. Our ability to more thoroughly study this high-use group, however, waslimited with a public access data set. The 2 variables significantly associatedwith prolonged hospital stay were hospital transfer and facial cellulitis,while the association with Medicaid status approached statistical significance(P = .07). Facial cellulitis appears to be a logicalsign of more severe inflammation in patients admitted for orbital and eyelidcellulitis.

The association between Medicaid coverage and increased hospital usesuggested by these data is plausible on the basis of the results of otherstudies.^9-12 Aninverse association of socioeconomic status with disease severity and clinicaloutcome has been reported for a variety of disorders, but the causes of theserelationships are not well understood.¹⁰ Resultsof some studies suggest that the increase in disease severity found with Medicaidenrollees is because of poor access to outpatient care, which in urban areasis often limited to emergency departments.^11,12 Medicaidpolicies that restrict availability to outpatient medications may furtherincrease the risk of disease progression and hospitalization.¹³

The interpretations of the AHCA database are subject to several qualifications.The total number of recorded admissions could include persons who were dischargedfrom one hospital and readmitted shortly thereafter to another. There wasno reliable means to identify this type of discharge-readmission transfer.Transfers do not alter the total number of individual admissions during thestudy year, but they artificially inflate the frequency of some admissiondiagnoses. Because this data set was a public document, it was not possibleto study case-specific information that could have clarified certain findings.For instance, without a medical record review, there was no means of determiningwhether a prolonged hospital stay was due to medical necessity or the patient'slack of social support. Given current financial realities, however, it seemsunlikely that many prolonged hospitalizations would not be medically justified.The total amount of inpatient eye care is greater than can be estimated bymeans of simply totaling primary discharge diagnoses because this method doesnot take into account admissions to veterans hospitals. Without review ofthe medical record, the interpretation of secondary admission codes is fraughtwith hazard because there is no method to estimate the clinical relevance(ie, need) for hospitalization, if any, on the basis of a secondary diagnosis.

Logistic regression analysis for prolonged hospitalization excludespatients whose data fields are not complete, which is why data for 439 patientswas not included. These 439 patients were similar to the patients includedin the analysis with respect to 7 of the 10 most common admission diagnoses.Of the 3 admission diagnoses that were overrepresented in the deleted group,only 1 (retinopathy of prematurity) is related to prolonged length of stay.The potential effect of deleting these patients' data should be small giventhat retinopathy of prematurity represents only 3% of diagnoses.

Nontraumatic disorders of the eye and ocular adnexa represent a smallproportion of all inpatient care in Florida. Few of these patients requireprolonged hospitalization but use considerable resources. A better understandingof the risk factors associated with prolonged hospitalization might lead tomore effective hospital use and perhaps reduced ocular morbidity.

Corresponding author: Curtis E. Margo, MD, MPH, Watson Clinic, Ophthalmology,1600 Lakeland Hills Blvd, Lakeland, FL 33805 (e-mail: cmargo@watsonclinic.com).

Submitted for publication January 30, 2003; final revision receivedAugust 24, 2003; accepted September 10, 2003.