Influence of Socio-Demographic Characteristics on Eye Care Expenditure: Data from the Medical Expenditure Panel Survey 2007

Abstract

Objective:

To evaluate the association between sociodemographic factors and eye care expenditure and to assess the burden of ocular expenditure compared to total health care expenditure.

Methods:

A retrospective analysis of ocular expenditure in participants of the 2007 Medical Expenditure Panel Survey. Data from 20,620 unique participants aged ≥18 years were evaluated for eye care expenditure by demographic characteristics.

Results:

A total of 22% of the studied population had eye care expenditures in 2007. Demographic factors significantly associated with higher probability of having eye care expenditures included older age (65+ years 35%, 45–64 years 23%, <45 years 17%), female sex (female 26%, male 19%), higher educational attainment (greater than high school education 25%, less than high school education 17%), having insurance (private 24%, uninsured 13%), and visual impairment (mild 31%, none 22%). Older age, female sex, higher educational attainment, having insurance, and presence of visual impairment were also significantly associated with higher mean eye care expenditure. In those with eye care expenditure, the mean ratio between eye care and total medical expenditure was 24%, with uninsured patients spending 42% of their medical care expenditure on eye care.

Conclusions:

Demographic factors are associated with both the probability of having ocular expenditure and the amount of expenditure. Of all factors examined, insurance status has the most potential for modification. Policy makers should consider these numbers when devising the terms by which eye care coverage will be provided under the Patient Protection and Affordable Care Act.

INTRODUCTION

Previous studies have shown that sociodemographic factors play a role in health care expenditure and use.^1–6 Shi evaluated the role of race and ethnicity on perceptions of access and objective health care use.¹ They found that while blacks and Hispanics were less likely to report difficulties in accessing care, objective measures of use suggested that minorities had less access to care than whites. Sex-specific differences in medical expenditure have been described for various conditions including obesity,⁷ rheumatoid arthritis,⁸ and hypertension,⁹ with women generally incurring higher costs compared to men. Only a few studies have specifically examined associations between sociodemographic factors and eye care expenditure.^10,11 Lam found that women, those with public-only insurance and those with less than a high school education had higher mean annual glaucoma medication expenditure than their counterparts.¹⁰

There is a knowledge gap, therefore, with regards to how eye care expenditure varies across socioeconomic subgroups. Specific questions to be studied include how socioeconomic factors are associated with the probability of having eye care expenditure along with the overall mean eye care expenditure for specific subgroups. Since the Medical Expenditure Panel Survey (MEPS) has comprehensive information on eye care-specific expenditure, the aim of this study was to close this knowledge gap by evaluating the influence of such factors on ocular as compared to medical expenditure and to understand the particular burden of ocular expenditure vis-à-vis total health care expenditure

This study had a secondary goal, which was to provide baseline levels of eye and medical care expenses in the year prior to the implementation of major health care reform legislation. The Patient Protection and Affordable Care Act (PPACA) was passed in 2010 and has the potential to improve both total healthcare and ocular healthcare access and use in the coming years. The data presented in this study can be used to document eye care expenditure within important population subgroups of sex, race, and educational level, which can be used to monitor changes in ocular healthcare expenditure as health care reform is implemented in the coming years. Such monitoring is important as we seek both to control health care costs while improving access, especially in historically underserved populations.

METHODS

Sample

The MEPS is an annual survey of families and individuals, their medical providers, and employers across the United States. Designed to be representative of the US population, MEPS provides the most complete source of data on health care use and cost, as well as health insurance coverage.¹² The overall response rate was 56.9% in 2007. Sampling weights were applied in order to ensure that the resulting sample was nationally representative of US households, and included adjustment for oversampling of race/ethnic groups and survey non-response.

The MEPS 2007 total health care expenditure variable “totexp07” was used to calculate medical expenditure. This variable included expenditure information related to office visits, hospital outpatient and inpatient visits, emergency room visits, dental care, home health care, vision aids, other medical equipment and services, and prescribed medicines. To obtain eye care-related expenditure, a comprehensive list of conditions related to eye care were first selected using Clinical Classification Software (CCS) codes 086 through 091. The CCS was developed by the Agency for Healthcare Research and Quality based on the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM). The ICD-9-CM’s multitude of codes are collapsed into a smaller number of clinically meaningful categories that are more useful for presenting descriptive statistics than are individual ICD-9-CM codes.¹³ The selected conditions were then used to identify any medical event related to eye care. Type of medical event included inpatient and outpatient care, office visits, emergency room visits, prescription drugs, eye glasses and contact lenses. Expenditures related to all eye care medical events were summed for each individual participant to obtain total eye care expenditure. For the study period, 20,620 unique participants aged 18 years or older were identified and included in the analysis.

Sociodemographic Data

Self-reported demographic and insurance information of qualified participants was obtained from the MEPS Full-Year Consolidated Data Files. Sociodemographic data collected included sex, age (18–44 years, 45–64 years, 65 years and greater), race/ethnicity (white non-Hispanic [NH], black NH, other NH, and Hispanic), self-reported health insurance status (private, public only, and uninsured), education level (less than high school education, high school graduate, greater than high school education), and poverty level (poor, near poor, low-, mid-, and high-income). To determine poverty level, family income, measured as a percentage, was calculated by dividing total family income by the applicable poverty line (based on family size and composition). The resulting percentages were grouped into three categories: poor/near poor (<125%), low income (125% to <200%), and middle/high income (200% and above).

Self-reported visual impairment (VI) data was also collected. To determine VI, participants were asked a set of questions. Their answers were summarized into a five category VI status variable: 1 – No difficulty seeing; 2 – Some difficulty seeing, can read newsprint; 3 – Some difficulty seeing, cannot read newsprint, can recognize familiar people; 4 - Some difficulty seeing, cannot read newsprint, cannot recognize familiar people but not blind; 5 – Blind. Participants falling into category 2 and 3 were designated as “mild VI”. Participants falling into category 4 and 5 were designated as “severe VI”. Finally, the presence of diabetes was determined by a positive response to the question “Have you ever been told by a doctor or other health professional that you had diabetes or sugar diabetes?” The main outcome measure was eye care expenditure by sociodemographic characteristics.

Statistical Analyses

All statistical analyses were performed using SAS 9.2 (SAS Institute Inc., North Carolina), SUDAAN 10 (RTI International, North Carolina), and Stata 11 (StataCorp LP, College Station, Texas) statistical packages. Multivariable zero-inflated Poisson regression was performed to study sociodemographic characteristic variables that predicted health care expenditure. The zero-inflated Poisson was used due to a large number of zero values and a highly skewed distribution of expenditure variables. The first part of the regression applied a logistic regression analysis to evaluate which factors predicted having any ocular or other medical expenditure. Next, Poisson regression analysis was performed to evaluate which factors were associated with higher incurred costs in those with expenditure.¹⁴

For the logistic regression results, we present the predicted probability and 95% confidence interval (CI) of having any expenditure for various sociodemographic characteristics. The predicted mean and 95% CI conditional on the probability of having expenditure is given for the Poisson portion of the model. In those with eye care expenditure, the ratio of eye care to total expenditure was calculated by dividing the overall conditional mean for eye care by the overall conditional mean for total expenditure. The University of Miami Institutional Review Board reviewed and approved the study, which was conducted in accordance with the principles of the Declaration of Helsinki.

RESULTS

A total of 20,620 unique participants aged 18 and older were identified using the MEPS 2007 database. Of these, 4306 (22%, weighted percent) reported eye care expenditures during the year (Table 1). Differences in sociodemographic characteristics were seen between the groups, with female sex, older age, greater than high school education, private insurance, and high income being more common in the group with reported eye care expenditure (all differences significant to p<0.05). Within the same population, 16,924 (85%, weighted percent) reported medical expenditures during the same year (Table 1).

TABLE 1.

Demographic characteristics of participants with and without eye care expenditure, 2007 Medical Expenditure Panel Survey Data.

	Participants without eye care expenditure		Participants with eye care expenditure		Participants without medical expenditure		Participants with medical expenditure
Characteristic	n a	Weighted %	n a	Weighted %	n a	Weighted %	n a	Weighted %
Total	16,314	77.6	4306	22.4	3696	15.2	16,924	84.8
Sex
Male	7874	51.1	1647	39.1	2334	67.3	7187	45.1
Female	8440	48.9	2659	60.9	1362	32.7	9737	54.9
Age group (years)
18–44	8661	52.9	1484	36.3	2733	73.1	7412	44.9
45–64	5427	33.5	1631	37.2	846	23.5	6212	36.3
65 and over	2226	13.6	1191	26.5	117	3.5	3300	18.8
Race
White non-Hispanic	8434	66.5	2773	75.8	1181	47.5	10,026	72.3
Black non-Hispanic	2711	12.1	585	8.9	671	15.5	2625	10.6
Other non-Hispanic	1206	6.8	298	5.9	309	8.6	1195	6.2
Hispanic	3963	14.6	650	9.4	1535	28.3	3078	10.8
Education
Less than high school	4350	19.5	776	12.9	1352	28.5	3774	16.1
High school	5079	31.4	1272	29.1	1260	36.5	5091	29.9
Greater than high school	6728	49.1	2233	58.1	1037	34.9	7924	54.0
Insurance type
Private insurance	9674	67.5	3029	76.4	1484	47.8	11,219	73.4
Public insurance only	3163	15.0	935	17.2	421	8.9	3677	16.7
Uninsured	3477	17.5	342	6.4	1791	43.3	2028	10.0
Poverty
Poor/near poor	3598	15.4	735	12.4	1038	20.8	3295	13.7
Low income	2687	13.4	557	10.9	806	18.5	2438	11.8
Mid/high income	10,029	71.2	3014	76.8	1852	60.8	11,191	74.5
Visual impairment (VI)
No VI	15,312	94.3	3824	89.6	3570	96.5	15,566	92.7
Mild VI	864	5.3	419	9.6	110	3.4	1173	6.8
Severe VI	55	0.3	43	0.9	8	0.2	90	0.5

^an = unweighted number

All differences significant to p<50.05

In multivariable analysis, sociodemographic factors significantly associated with higher probability of having eye care expenditure included older age, female sex, having insurance, and higher educational attainment (Table 2). Older age, female sex, and having insurance were also significantly associated with higher probability of having medical expenditure (Table 2).

TABLE 2.

Two-stage multivariable regression analysis of eye care and total medical expenditure by demographic factors, 2007 Medical Expenditure Panel Survey data.

	Eye expenditure				Total medical expenditure				Ratio
Characteristic	Probability of having expenditure (SE)	95% CI	Mean expenditure (SE)*	95% CI	Probability of having expenditure (SE)	95% CI	Mean expenditure (SE)*	95% CI	Eye/total (%)†
Age (years)
Less than 45	0.17 (0.01)ab	0.16–0.19	56.53 (4.26)ab	48.18–64.88	0.80 (0.00)ab	0.80–0.81	294.32 (81.56)ab	2134.46–2454.17	32
45–64	0.23 (0.01)c	0.22–0.24	117.27 (6.68)c	104.18–130.36	0.88 (0.00)c	0.88–0.89	5532.89 (251.68)c	5039.61–6026.17	22
65 and over	0.35 (0.01)	0.33–0.37	324.43 (23.72)	277.94–370.91	0.95 (0.01)	0.94–0.96	7675.09 (272.77)	7140.47–8209.71	16
Sex
Male	0.19 (0.01)a	0.18–0.20	103.04 (5.96)a	91.36–114.72	0.80 (0.00)a	0.80–0.81	3895.37 (154.59)a	3592.38–4198.36	29
Female	0.26 (0.01)	0.25–0.27	139.74 (6.32)	127.36–152.12	0.89 (0.00)	0.89–0.90	4930.83 (131.62)	4672.85–5188.80	21
Race
White non-Hispanic	0.23 (0.01)a	0.22–0.24	121.81 (5.33)	111.36–132.25	0.88 (0.00)ab	0.87–0.88	4583.01 (120.24)ab	4347.35–4818.67	22
Black non-Hispanic	0.20 (0.01)	0.18–0.21	124.24 (16.96)	91.00–157.49	0.81 (0.01)	0.80–0.83	4772.71 (354.49)c	4077.93–5467.50	28
Hispanic	0.22 (0.01)	0.20–0.24	133.07 (12.98)	107.63–158.52	0.79 (0.01)	0.78–0.81	3508.40 (196.47)	3123.32–3893.48	33
Education
Less than high school	0.17 (0.01)ab	0.16–0.19	89.46 (8.20)ab	73.38–105.54	0.82 (0.01)b	0.80–0.83	4463.13 (262.54)	3948.57–4977.70	24
High school	0.21 (0.01)c	0.20–0.22	112.85 (7.83)c	97.51–128.19	0.82 (0.01)c	0.81–0.83	4189.62 (180.42)	3836.00–4543.24	24
Greater than high school	0.25 (0.01)	0.24–0.26	141.33 (7.28)	127.06–155.59	0.88 (0.00)	0.87–0.89	4621.51 (152.18)	4323.23–4919.78	24
Insurance
Uninsured	0.13 (0.01)ab	0.11–0.15	62.41 (8.17)ab	46.39–78.42	0.69 (0.01)ab	0.67–0.71	1821.84 (150.63)ab	1526.61–2117.07	42
Public only	0.21 (0.01)c	0.19–0.23	104.45 (8.87)c	87.06–121.83	0.90 (0.01)	0.88–0.91	5650.83 (284.88)c	5092.48–6209.18	19
Private	0.24 (0.01)	0.23–0.25	137.21 (7.02)	123.45–150.98	0.89 (0.00)	0.88–0.89	4543.95 (137.85)	4273.76–4814.14	24
Poverty
Poor/near poor	0.22 (0.01)	0.20–0.24	121.26 (15.73)	90.43–152.08	0.84 (0.01)	0.83–0.86	5290.31 (370.58)b	4563.98–6016.63	24
Low income	0.21 (0.01)c	0.19–0.22	111.12 (10.23)	91.07–131.18	0.83 (0.01)c	0.81–0.84	4935.45 (272.64)c	4401.08–5469.81	24
Mid/high income	0.23 (0.01)	0.22–0.24	125.00 (5.34)	114.53–135.46	0.85 (0.00)	0.85–0.86	4204.48 (117.13)	3974.91–4434.04	24
Visual impairment (VI)
No VI	0.22 (0.00)ab	0.21–0.23	113.08 (4.48)ab	104.30–121.87	0.85 (0.00)a	0.84–0.85	4352.26 (107.64)b	4141.30–4563.22	25
Mild VI	0.31 (0.02)	0.28–0.33	214.67 (21.84)	171.85–257.48	0.89 (0.01)	0.87–0.91	5605.90 (327.37)c	4964.27–6247.53	21
Severe VI	0.35 (0.05)	0.25–0.44	331.49 (108.15)	119.52–543.46	0.87 (0.04)	0.79–0.95	4002.58 (645.16)	2738.10–5267.06	19

^*Conditional mean expenditure for those with expenditure

^†Ratio of eye to total medical expenses was calculated using only the population that had eye care expenditure

^aMeans/probabilities in first group are significantly different to second group (p<0.05)

^bMeans/probabilities in first group are significantly different to third group (p<0.05)

^cMeans/probabilities in second group are significantly different to third group (p<0.05)

CI, confidence interval; SE, standard error

In individuals with eye care or medical expenditure in 2007, several sociodemographic factors were significantly associated with the degree of expenditure (multivariable Poisson regression analysis, Table 2). A strong association was found between a person’s age and the amount of money spent on eye care. Individuals over the age of 65 years spent a mean of $324 per year, whereas individuals younger than 45 years spent an average $57 per year. Sex also affected eye care expenditure, with women spending a mean of $140 compared to $103 in men. Other factors associated with higher mean eye care expenditure included increased educational attainment (greater than high school versus less than high school) and having insurance (public and private insurance holders versus the uninsured). While income was not significantly associated with eye care expenditure, there was an inverse relationship between income and total medical care expenditure with those in the poor category spending more than mid- and high-income individuals (poor/near poor $5290, 95% CI $4564–6017, versus mid/high $4204, 95% CI $3975–4434).

The presence of mild VI, as compared to no VI, was associated with the presence and degree of ocular expenditure, with an approximate doubling of mean eye care expenditure in those with mild VI (Table 2). Patients with VI were not only more likely to spend more on eye care but were also more likely to use and spend more on medical care.

Adding presence of a diabetes diagnosis to the model, patients with a self-reported diagnosis of diabetes were more likely to have eye care expenditure than those without diabetes (28% versus 22%, respectively, p<0.001). In patients who incurred eye care expenditures, a diagnosis of diabetes was associated with an approximately 2-fold increase in mean expenditure compared to those without diabetes when adjusting for all other variables ($199.17, 95% CI $162.41–235.92 versus $111.95, 95% CI $103.09–120.81, respectively, p<0.001).

In order to evaluate the burden of ocular expenditure in comparison to overall health expenditure, a ratio was calculated for each patient which divided expenditure related to ocular care by total health care expenditure. In the 4306 patients with eye care expenditures, the mean ratio of ocular expenditure to total health expenditure was 24% (standard error of mean 1%). Looking at ratios within the different sociodemographic subgroups, patients without insurance had the highest mean ratio and spent almost half of their medical expenditure on eye care (Table 2).

DISCUSSION

Because of its population-based design, MEPS provides the best estimates for eye care costs in the US population. Using this nationally representative dataset, our study is the first to document overall eye care expenditure in overall sociodemographic subgroups. We found that age, sex, education, and insurance status were associated with the presence of ocular expenditure. While the etiologies behind these discrepancies are not clear, it is important to recognize the role of sociodemographic factors when considering the myriad determinants of health. Of all factors examined, insurance status has the most potential for modification, especially considering that this group spent the most money (proportionally) on eye care.

VI was associated not only with the presence and degree of ocular expenditure, but mild VI was also associated with the presence and degree of overall medical expenditure. This finding is in line with that of Frick et al. who found that those with mild VI had an additional $1043 in total expenditure compared to those without VI.⁵ However, unlike Frick who found an additional $2221 in health expenditure in those with severe VI, we did not find that severe VI significantly impacted expenses. There are several reasons for this discrepancy including a different patient population in the two studies (MEPS 1996–2002 versus 2007), different definitions for mild and severe VI based on patient self-report, and differing study methodologies (simple 2 part models compared to zero-inflated negative binomial models). To address whether VI categorization impacted our findings, we re-ran our analysis using the same VI categories as Frick et al. and again, did not find an increased mean medical expenditure in the severe VI group compared to the mild VI group (data not shown). Interestingly, Wong et al., using a prospective cohort methodology, found that severity of VI (mild, moderate or severe) was not associated with annual total personal out-of-pocket costs for eye care.¹¹

As with all retrospective reports, our study has limitations which need to be considered when interpreting the study results. While we can postulate on the underlying reasons behind disparities in eye care use, a limitation of the study is its inability to directly look at such causes. Furthermore, MEPS data relies in part on individual self-report which may not be entirely accurate. For example, for the VI questions, the answers selected depend on whether the individual interpreted the question to mean VI with or without spectacles. Despite these limitations, the strength of the MEPS dataset is that it looks specifically at the patterns of health expenditure and therefore provides the best estimates for eye care and total health care costs.

In summary, we found that older individuals, women, private insurance holders, and those with higher educational attainment were more likely to have eye care expenditure. Mild VI influenced both the presence and degree of eye care and total medical care expenditure. In patients with eye care expenditure, a significant proportion of total medical expenditure was spent on eye care (~25%), with the highest proportional spending in those without insurance. This has implications for policy makers as it suggests that eye care costs are an important aspect to consider when making decisions on what services to cover in the PPACA. Furthermore, detailed eye care expenditure estimates for key population subgroups provide an important baseline measure as we move forward as a nation with implementation of health care reform. Future monitoring of these expenditures in the MEPS and other population-based surveys will enable us to determine if eye care access and use disparities are being reduced.