ORIGINAL STUDY

Clinical Applicability of the International Classification of Disease and Related Health Problems (ICD-9) Glaucoma Staging Codes to Predict Disease Severity in Patients With Open-angle Glaucoma

Anjali S. Parekh, MD,* Ali Tafreshi, BS,* Syril K. Dorairaj, MD,w and Robert N. Weinreb, MD*

to provide more specific detail with regard to disease Purpose: The aim of the study was to determine how the Interna- diagnosis and severity in a particular patient. The purpose tional Classification of Disease and Related Health Problems of this characterization would be to allow third party (ICD-9) Glaucoma Staging Codes (GSC) relate to visual field payers and insurers to better understand the needs of global indices and average retinal nerve fiber layer (RNFL) thick- a given patient, profile the complexity of a physician’s ness in patients with primary open-angle glaucoma (POAG). practice, and appropriately allocate resources.2 Another Methods: Over a 6-week period, charts of consecutive patients were suggested benefit of this system is the potential to do prospectively reviewed. Included patients had optic nerve head meaningful research based on a more defined and specific damage consistent with glaucoma as well as reliable functional and Medicare database.2 After refinements required by the imaging tests. Patients were divided into early-stage, moderate- government’s ICD-9 Coordination and Maintenance stage, and severe-stage glaucoma using the ICD-9 GSC guidelines. Committee, the GSC were approved for implementation on Relationships between mean deviation (MD), pattern SD (PSD), 3 average RNFL thickness, and the staging system were evaluated. October 1, 2012. Many staging systems have been proposed during the Results: A total of 616 patient charts were evaluated and 135 past 40 years.4 However, patients with glaucoma present patients met the inclusion criteria. Of the 135 patients, both eyes with varying severity along a continuum, and a widely (270 eyes) were evaluated and the worse eye (by MD) was included accepted system of staging has not been adopted.5 Under- in the analysis. Using the ICD-9 GSC guidelines as a basis for standing how this new system relates to clinically diagnosed defining severity, MD was significantly lower in the severe-stage compared with the moderate-stage group (P < 0.01) and in the disease severity is of significant importance, as resources, moderate-stage when compared with the early-stage group reimbursements, and research will likely be influenced by (P = 0.020). PSD was significantly higher in the severe-stage this system. compared with the moderate-stage group (P < 0.01) and in the This study aimed to determine how the ICD-9 GSC moderate-stage compared with the early-stage group (P < 0.01). relate to visual field global indices and average retinal nerve Average RNFL thickness was not significantly lower in the severe- fiber layer (RNFL) thickness, widely used measures of stage compared with the moderate-stage group (P = 0.05); how- functional and structural testing in clinical practice. ever, it was significantly lower in the severe-stage and moderate- stage groups compared with the early-stage group (P < 0.01). Conclusions: ICD-9 GSC showed a significant relationship between METHODS worsening visual field indices and RNFL loss with increasing dis- Subjects ease severity in patients with POAG. This was an observational cross-sectional study that Key Words: glaucoma staging codes, open-angle glaucoma, visual included patients from a glaucoma specialist’s practice at field the Shiley Eye Center (University of California, San Diego). Participants were evaluated with a comprehensive clinical (J Glaucoma 2014;23:e18–e22) examination along with imaging and functional tests. The University of California San Diego Human Subjects Committee approved all the protocols, and methods he overall burden of glaucoma is likely to intensify in described attended to the tenets of the Declaration of Tthe future because of the increasing aging population Helsinki. and numbers of patients at elevated risk for the disease.1 To Over a 6-week period, the charts of consecutive prepare for this possibility, a new glaucoma staging system patients were prospectively reviewed. Each subject’s chart was developed by an American Glaucoma Society work- underwent evaluation by 2 reviewers and the following group. The Glaucoma Staging Codes (GSC) were designed information was compiled and recorded: patient age, race, review of medical history, best-corrected visual acuity, slit- lamp biomicroscopy, intraocular pressure (IOP) using Received for publication August 1, 2013; accepted September 25, 2013. From the *Hamilton Glaucoma Center, University of California, La Goldmann applanation tonometry, gonioscopy, dilated Jolla, CA; and wMayo Clinic, Jacksonville, FL. fundoscopy, stereoscopic optic disc photography, optic Disclosure: The authors declare no conflicts of interest. nerve head optical coherence tomography (OCT), and Reprints: Anjali S. Parekh, MD, Hamilton Glaucoma Center, Uni- standard automated perimetry (SAP). To be included, all versity of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093 (e-mail: [email protected]). patients had glaucomatous optic neuropathy diagnosed by Copyright r 2013 by Lippincott Williams & Wilkins the presence of neuroretinal rim thinning, excavation, or DOI: 10.1097/IJG.0000000000000033 RNFL defects in addition to reliable functional and e18 | www.glaucomajournal.com J Glaucoma  Volume 23, Number 1, January 2014 J Glaucoma  Volume 23, Number 1, January 2014 Clinical Applicability of the ICD-9 GSC

ICD-9 Staging Definitions Each patient’s visual field was evaluated, and the ICD- 9 GSC3,6 were used to determine disease severity. Two graders evaluated each visual field and staged both eyes of each patient; however, as defined by the system, the worse eye was used to classify each patient’s level of disease and was used for analysis. If a patient had 2 eyes that were equally damaged, the worse of the 2 eyes, based on involvement of fixation, or poorer global indices was marked as the eye for analysis. The worse eye was the only eye included in the analyses. Disease severity was graded as follows3,6:

Mild or Early-stage Glaucoma It is defined as optic nerve abnormalities consistent with glaucoma but no visual field abnormalities on any white-on-white visual field test, or abnormalities present only on short–wavelength-doubling perimetry (Fig. 1).

Moderate-stage Glaucoma It is defined as optic nerve abnormalities consistent with glaucoma and glaucomatous visual field abnormalities in 1 hemifield, not within 5 degrees of fixation (Fig. 2).

Severe-stage Glaucoma It is defined as optic nerve abnormalities consistent with glaucoma and glaucomatous visual field abnormalities in both hemifields, and/or loss within 5 degrees of fixation in at least 1 hemifield (Fig. 3). FIGURE 1. Mild-stage or early-stage glaucoma: no visual field abnormalities on any white-on-white visual field test or abnor- Statistical Analysis 3 malities present only on short–wavelength-doubling perimetry. Continuous variables were compared between the groups using 1-way analysis of variance (ANOVA) with imaging tests as described below. Patients must have undergo visual field testing and OCT imaging within 1 year of the clinical examination. Subjects with coexisting cornea, retinal disease, uveitis, or nonglaucomatous optic disc neuropathy were excluded from the study. Patients underwent SAP with 24-2 Swedish Interactive Threshold Algorithm (Carl Zeiss Meditec Inc., Dublin, CA). Only reliable tests with r33% fixation losses and false negatives and <15% false positives were included. For each test, visual field index, mean deviation (MD), and pattern SD (PSD) were recorded. Visual fields were reviewed to identify the presence of artifacts such as lid and rim artifacts, fatigue effects, inattention, or inappropriate fixation. In addition, they were reviewed for the presence of nonglaucomatous abnormalities such as homonymous hemianopia. RNFL imaging was carried out using Cirrus SD-OCT (software version 4.5; Carl Zeiss Meditec Inc.). Cirrus uses a superluminescent diode scan with a center wavelength of 840 nm and an acquisition rate of 27,000 A-scans per second at an axial resolution of 5 m. The protocol used for RNFL thickness evaluation, the optic disc cube, is based on a 3-dimensional scan of a 6Â6mm2 area centered on the optic disc, where information from a 1024 (depth)Â200Â200-point parallelepiped is collected. Then, a 3.46-mm-diameter circular scan is automatically placed around the optic disc, and the information about para- papillary RNFL thickness is obtained. To be included, all images were reviewed for noncentered scans and had to have signal strength >6, the absence of movement artifacts, FIGURE 2. Moderate-stage glaucoma: visual field abnormalities and good centering around the optic disc. in one hemifield, and not within 5 degrees of fixation.3 r 2013 Lippincott Williams & Wilkins www.glaucomajournal.com | e19 Parekh et al J Glaucoma  Volume 23, Number 1, January 2014

FIGURE 3. Severe-stage glaucoma: glaucomatous visual field abnormalities in both hemifields, and/or loss within 5 degrees of fixation in at least one hemifield.3

Tukey honestly significant difference (HSD) test for post There was no significant difference between the groups in hoc analysis. In the event, significant differences between any of the other characteristics (Table 1). the groups were found in ANOVA; analysis of covariance MD was significantly lower in the severe-stage glau- (ANCOVA) was applied to adjust for age, central corneal coma group compared with the moderate-stage group thickness (CCT), and race. Linear regression analysis was (P < 0.01) and the moderate-stage compared with the used to evaluate the relationship between RNFL thickness early-stage group (P = 0.020; ANOVA with Tukey HSD and SAP indices MD and PSD across groups; R2 and the test). PSD was significantly higher in the severe-stage corresponding P-values were calculated for each parameter. glaucoma group compared with the moderate-stage glau- Statistical analyses were made using JMP software version coma group (P < 0.01) and in the moderate-stage group 10.0.0 (SAS Institute, Cary, NC). A P-value of <0.05 was compared with the early-stage group (P < 0.01). Although considered statistically significant. the average RNFL thickness values was not significantly lower in the severe-stage group compared with the moderate-stage group (P = 0.05), there was a significant RESULTS difference in the severe-stage and moderate-stage groups A total of 616 charts were evaluated and 135 patients compared with the early-stage group (P < 0.01). met the inclusion criteria. Of the excluded patients, 351 The difference in the MD value remained significant patients had glaucomatous disease other than primary when using analyses of covariance (ANCOVA) and open-angle glaucoma (POAG), 73 patients were excluded adjusting for age, CCT, race, and IOP (F = 15.85, because of unreliable functional or imaging tests, and 57 P < 0.01). The PSD value also remained significantly dif- patients had coexisting cornea, retinal disease, uveitis, or ferent when adjusting for these characteristics (F = 18.97, nonglaucomatous optic disc neuropathy. Of the 135 P < 0.01). Using ANCOVA to adjust for age, CCT, race, patients, both eyes (270 eyes) were evaluated and the worse and IOP, RNFL thickness values were still significantly eye was the only eye included in the analysis. Eyes were different between groups (F = 12.30, P < 0.01). categorized as early-stage glaucoma (n = 20), moderate- stage glaucoma (n = 24), and severe-stage glaucoma (n = 91) using the ICD-9 GSC.3,6 Table 1 presents the DISCUSSION demographic and clinical characteristics of the study When the ICD-9 GSC were compared with visual field groups. The mean age of the 3 POAG groups was indices, there was a statistically significant relationship to 62.4 ± 14.2 years (range, 38 to 82 y) for the early-stage worsening MD and PSD with increasing severity of disease. glaucoma, 72.7 ± 11.1 years (range, 45 to 89 y) for the When evaluated against other measurements of glaucoma- moderate-stage glaucoma, and 71.2 ± 9.0 years (range, 34 tous damage such as average RNFL loss, a statistically to 87 y) for the severe-stage glaucoma. The early-stage significant relationship was not found between the severe- glaucoma group was significantly younger than the other 2 stage compared with the moderate-stage group; however, groups (P < 0.01), and the early-stage glaucoma group had there was a statistical difference between the severe-stage significantly higher IOP values than the other 2 groups. and moderate-stage compared with the early-stage group. e20 | www.glaucomajournal.com r 2013 Lippincott Williams & Wilkins J Glaucoma  Volume 23, Number 1, January 2014 Clinical Applicability of the ICD-9 GSC

TABLE 1. One-Way Analysis of Variance (ANOVA) Mean and 95% Confidence Intervals (CI) for Demographic and Clinical Characteristics of the Study Groups as Defined by AGS Guidelines POAG Parameters Early Stage (N = 20) Moderate Stage (N = 24) Severe Stage (N = 91) P Age (y) 62.4 (55.8-69.1) 72.7 (68.0-77.4) 71.2 (69.3-73.0) 0.0015 CCT (mm) 553.2 (536.4-570.0) 539.2 (525.5-552.9) 543.6 (536.0-551.3) 0.423 Race 0.335 White 20 17 63 Hispanic 3 0 8 Black 0 2 5 Other 1 5 15 IOP (mm Hg) 18.4 (16.3-20.5) 15.5 (13.4-17.5) 14.4 (13.1-15.7) 0.020 MD À1.53 (À2.25-0.81) À3.72 (À4.82-2.61) À9.03 (À10.46-7.60) < 0.001 PSD 1.84 (1.52-2.15) 4.41 (3.29-5.53) 6.99 (6.17-7.82) < 0.001

AGS indicates American Glaucoma Society; CCT, central corneal thickness; IOP, intraocular pressure; MD, mean deviation; POAG, primary open-angle glaucoma; PSD, pattern SD.

In this specific population of patients with POAG, the when attempting to assess progression of disease by this ICD-9 GSC appears to successfully differentiate patients system. Subtle changes in disease would not be identifiable, based on visual field global indices and RNFL thickness. and in order for a patient to advance from early to Although SAP testing can be variable given the subjective advanced disease, the severity of the glaucoma would have nature of the exam, RNFL thickness has been shown to to significantly progress. have low measurement variability and good reproduci- The results of this study do have some limitations. It bility.7 The inability to show a statistically significant dif- did not attempt to assess the reproducibility of this grading ference between average RNFL thickness in the moderate system between various clinicians. Such an analysis is and severe groups may have been related to the fact that in something that should be carried out in the future. In severe cases of glaucoma, a clear floor effect for structural addition, a sample size sufficient for data analysis was used; measurements has been described.8 After a certain level of however, it may not be sufficient for extrapolation of results visual field damage, RNFL thickness will change little to the entire US population. As only patients from an despite deterioration on visual field, and it has been pro- academic tertiary care practice were included, it is likely posed that functional tests may better follow glaucomatous that the recruitment introduced a selection bias. Larger loss in advanced cases.8 evaluations in more diverse practice settings would be Automated visual field testing has been the gold required to broaden the validity of this staging system. standard for the past 10 years and it provides the single Another limitation of this study is that patients with sig- most useful indicator of disease severity, given the wide nificant ocular comorbidities were excluded. In a practical range of visual field parameters.4 However, perimetry in clinical setting, patients with visual field loss from other itself is a subjective psychophysical testing method, thus causes would alter disease categorization. Had these any classification system that is based on this type of data patients been included, the results of the association can never be completely accurate and reproducible.5,9 between visual field global indices and severity of disease as Although the ICD-9 GSC includes the need for optic nerve defined by this coding system may have been altered. head damage, other clinical factors that may be used to Moreover, this study analyzed RNFL thickness using OCT. assess severity of disease such as optic nerve head topo- There are several other modalities available in clinical graphic parameters, nerve fiber layer analysis, or ganglion practice to characterize RNFL thickness. cell loss are not currently included in the coding system. Studies have shown that the direct costs of glaucoma Including objective testing in the classification of disease management typically increase with worsening disease severity may be a beneficial way to offset patient-related severity.10 One may postulate that as disease severity wor- error that often occurs during perimetry. sens, increased resource consumption will be prompted by a There are several advantages to having a staging sys- physician’s desire to slow disease progression.11 However, tem such as the one presented. As suggested, physician’s given the number of clinical scenarios that exist in glau- practices could be better characterized and patients coma, this is not always the case. For example, if a patient requiring more resources may be more appropriately is labeled as having mild disease, one could say that perhaps identified. In the future, a classification system could be they only should have annual or biannual examinations as a implemented as a screening system to direct patients with cost-effective strategy. However, in glaucoma, one may moderate or severe disease to glaucoma specialists. More- need multiple visual fields and IOP recordings to decide the over, in institutions where care is stratified, patients level of disease and the treatment required to prevent pro- with moderate or severe disease could be seen in a more gression. To determine this information, several visits timely manner or more frequently than those with mild would be necessary. Alternatively, if a patient is labeled as disease. Although this classification system was not having severe disease but has been clinically stable for an designed as a research tool for precise staging of the disease, extensive period of time, indefinite resource consumption it is possible that it could be used for this purpose. How- might be unnecessary. As such, resource allocation and ever, the coding categories are very broad (eg, advanced reimbursement based on disease severity must be consid- and very advanced glaucoma are not differentiated in ered in relative terms that allow for medical judgment in the severe-stage group), and this could be an issue assessing the need for treatment. r 2013 Lippincott Williams & Wilkins www.glaucomajournal.com | e21 Parekh et al J Glaucoma  Volume 23, Number 1, January 2014

In conclusion, in patients with POAG, the system 5. Brusini P, Johnson CA. Staging functional damage in investigated in this study showed a significant relationship glaucoma: review of different classification methods. Surv between worsening visual field indices and RNFL loss with Ophthalmol. 2007;52:156–179. increasing disease severity. Further studies looking at a 6. American Glaucoma Society. Glaucoma Staging Codes Teaching Module. 2012. larger population with other disease entities is needed to 7. Leung CK, Cheung CY, Weinreb RN, et al. Retinal nerve fiber know if this system is applicable on a more global scale. layer imaging with spectral-domain optical coherence tomog- raphy: a variability and diagnostic performance study. Oph- REFERENCES thalmology. 2009;116:1257–1263. 1. Fremont AM, Lee PP, Mangione CM, et al. Patterns of care 8. Leite MTZL, Weinreb RN, Rao HL, et al. Structure-function for open-angle glaucoma in managed care. Arch Ophthalmol. relationships using the Cirrus spectral domain optical coher- 2003;121:777–783. ence tomograph and standard automated perimetry. J Glau- 2. Reynolds A, Mattox C. Taking glaucoma codes to the next coma. 2012;21:49–54. level. Review of Ophthalmology 2011. Available at: http:// 9. Mills RP, Budenz DL, Lee PP, et al. Categorizing the stage of www.revophth.com/content/d/glaucoma_management/c/31327/. glaucoma from pre-diagnosis to end-stage disease. Am J Accessed July 12, 2011. Ophthalmol. 2006;141:24–30. 3. Ronald L, Fellman M, Cynthia G, et al. American Academy 10. Lee PP, Walt JG, Doyle JJ, et al. A multicenter, retrospective Of Ophthalmology. Savvy Coder: coding & reimbursement pilot study of resource use and costs associated with severity of know the new glaucoma staging codes. EyeNet Magazine; disease in glaucoma. Arch Ophthalmol. 2006;124:12–19. October 1, 2011. 11. Rein DB, Wittenborn JS, Lee PP, et al. The cost-effectiveness 4. Ng M, Sample PA, Pascual JP, et al. Comparison of visual of routine office-based identification and subsequent medical field severity classification systems for glaucoma. J Glaucoma. treatment of primary open-angle glaucoma in the United 2012;21:551–561. States. Ophthalmology. 2009;116:823–832.

e22 | www.glaucomajournal.com r 2013 Lippincott Williams & Wilkins