Changes in Veterans’ Use of Outpatient Care From 1992 to 2000

Abstract

Objectives. During the mid-1990s, the Veterans Health Administration (VHA) reorganized and placed greater emphasis on high-quality primary care. To determine whether the reorganization was associated with changes in patterns of out-patient VHA use, we sought to evaluate changes in characteristics of veterans who use VHA outpatient services between 1992 and 2000.

Methods. We merged 2 waves of the National Survey of Veterans to determine changes in patterns of outpatient care use. We evaluated the extent to which veterans who received outpatient care received that care from the VHA.

Results. The odds ratio for VHA-only outpatient care relative to non-VHA–only care in 2000 relative to 1992 was 1.75 (95% confidence interval [CI]=1.51, 2.04), and the odds ratio for dual relative to non-VHA-only care was 1.22 (95% CI=1.08, 1.37). Veterans who were older, had low incomes, and had no additional health insurance coverage were most likely to increase their use of VHA outpatient care.

Conclusions. Our results suggest that the VHA is increasingly serving veterans who have trouble accessing the private health care system.

In the mid-1990s, the administrative structure and clinical priorities of the Veterans Health Administration (VHA) were reorganized to ensure predictable and consistent provision of cost-effective, high-quality care. Key elements of the reorganization included a universal primary care system for all veterans, eligibility reform, and a performance management program that assessed, among other areas, delivery of preventive care.1 Evaluations of the VHA system since this reorganization indicate marked increases in provision of outpatient preventive and disease-specific care, as well as disease-specific decreases in urgent care visits, VHA bed days, and mortality.2^,3 In addition, the care provided by the VHA appears to be at least comparable to, if not better than, care provided outside of the VHA2^,4 However, these evaluations have focused solely on veterans accessing VHA care.

During this same period, shifts in the health care landscape also were occurring outside the VHA. The number of Americans lacking health insurance coverage rose, and employers and insurance companies shifted more of the costs of providing care to individual consumers.5^,6 In addition, costs of medical care, drug costs in particular, continued to rise faster than costs in other sectors of the economy.7^,8 While all health care consumers were affected, specific groups were potentially more vulnerable to these changes, including those without medical insurance, low-income populations, the elderly, and those with heavy disease burdens.9

One option available to veterans that is not available to other populations who have difficulty accessing or paying for care is the VHA. Part of the VHA mandate is to care for veterans who would have difficulty obtaining care from other health service providers.10 One way the VHA accomplishes this mandate is by providing inexpensive or free care (including prescription medications) to veterans who have difficulty affording such services elsewhere. Given the changes in the VHA and the broad trends in US health care markets, little is known about shifts in how veterans access the VHA system.

In this study, we examined how the characteristics of veterans who use outpatient VHA services have changed since 1992. We hypothesized that, given both the reorganization of the VHA and the changing health care environment outside the VHA, veterans who might have difficulty obtaining needed care elsewhere used VHA services more frequently. Our specific aims were to (1) describe trends in both VHA and non-VHA out-patient health care use among veterans from 1992 to 2000, (2) identify characteristics associated with veterans’ use of VHA versus non-VHA outpatient health care services, and (3) measure changes in these characteristics between 1992 and 2000.

METHODS

Study Population

The data for these analyses were derived from 2 different rounds of the National Survey of Veterans, the first conducted in 1993 (in which questions focused on the year 1992) and the second conducted in 2001 (in which questions focused on 2000). The National Survey of Veterans is a nationwide telephone survey designed to assist the Veterans Administration (VA) in making resource allocation decisions by evaluating the demographic profiles, health needs, and patterns of use of VA benefits among US veterans. The 2 surveys examined here included veterans who both used and did not use VHA care. Detailed descriptions of both the 1993 and 2001 surveys are available elsewhere.10^–13

Although the 2 survey waves were not identical, similar methodologies were used. In short, the 1993 survey comprised a probability sample of 11645 noninstitutionalized veterans residing in the United States and Puerto Rico; veterans who had service-connected disabilities or who had used VHA medical facilities in 1992 were oversampled. The 2001 survey comprised a similar probability sample of 20048 noninstitutionalized veterans, with oversampling of particular VHA priority groups, women, Hispanics, and African Americans. The larger sample size in 2001 was necessary for greater precision in estimating the sizes of the 7 health care priority groups assessed.

In 1993, the survey was administered to 5529 veterans selected via random-digit dialing (RDD) and 6116 veterans randomly selected from VA files. In 2001, the survey was administered to 12956 veterans selected through RDD and to 7092 veterans randomly selected from VA files. The VA files were composites of several different computerized files, including VHA enrollment files and the Veterans Benefits Administration Compensation and Pension File. Veterans in these files have not necessarily used the VHA system; however, their inclusion was necessary to ensure adequate representation. To be eligible for either survey, veterans were required (1) to have served their full active duty obligation in the military, (2) to not be on active duty at the time of the survey, and (3) to not have received a dishonorable discharge. Veterans residing in nursing homes had to have been residing there for less than 6 months and had to have a principal residence elsewhere.

In the 1993 RDD sample, 68.3% of the veterans identified via preliminary screening questions were eligible and completed the interview; 9.5% were determined to be ineligible after more detailed screening; 14.4% were eligible but did not take part; and 7.8% had died, could not be interviewed, or, after being located, could not be contacted. In the 2001 RDD sample, 64.9% of the veterans identified through preliminary screening questions were eligible and completed the interview; 11.5% were determined to be ineligible after more detailed screening; 3.1% were eligible but did not participate; and 20.5% had died, could not be interviewed, or, after being located, could not be contacted.

In 1993, 33.9% of the veterans identified via VA lists had not used any services in 1992 and were not included in the VA list sample. Of the remaining veterans selected from VA lists, 53.0% completed the survey; 7.3% refused to participate; and the remainder had died, had been institutionalized, or could not be located. List identification improved in 2001; only 8.8% of the sample was ineligible, and, of those who were eligible, 59.2% completed the survey and 3.6% refused to participate.

Outcome and Covariates

The primary measure assessed was self-reported receipt of outpatient care. In 1993, veterans were considered to have received outpatient care if they replied yes to the following question: “During 1992, did you go for any outpatient visits or receive any kind of medical care on an outpatient basis?” In 2001, they were considered to have received such care if they answered yes to “In the last 12 months, did you get outpatient care for yourself? For example, doctor visits, urgent care, routine exams, medical tests, or shots?” The 2001 survey, but not the 1993 survey, asked about receipt of other types of care (e.g., emergency, psychiatric); we focus only on outpatient care for which comparable questions existed in both surveys.

Categories of outpatient care use were as follows: no use, non-VHA care only, VHA care only, and dual care. Veterans were considered to be nonusers if they reported having received no outpatient care in the past year. They were considered to have used non-VHA–only care only if they indicated that they had received all of their outpatient care in the past year from non-VHA providers. Veterans were considered to be users of VHA-only care if they indicated that, in the past year, they had received outpatient care only from the VHA. Finally, veterans were considered to be dual care users if they indicated that they had received outpatient care from both VHA and non-VHA providers in the past year. The outcome was defined as described earlier, rather than as a count of use frequency, because we were most interested in where people access care (as opposed to the extent to which they access care).

Independent predictor variables assessed included age, gender, marital status (married vs other), self-reported race/ethnicity (White, African American, Hispanic, other), education level (high school or less, more than high school), pretax total family income (adjusted for inflation to 2000 dollars among 1992 survey respondents), number of chronic conditions reported, and self-reported difficulty with an activity of daily living (including bathing or showering, getting dressed, getting in and out of chairs or bed, walking across a room, climbing stairs, eating, using the toilet or getting to the toilet, and controlling bladder or bowel).14^–18 In addition, we determined whether veterans had a service-connected disability, a known predictor of use of the VHA system.19 Service-related disabilities, defined as injuries or diseases incurred or aggravated during active military service, can affect veterans’ retirement pay, disability severance pay, separation incentive payments, and amount of VHA compensation paid.

Analyses

In all of the analyses described here, sampling probability weights were used to ensure that parameter estimates would be nationally representative and to adjust for the different sampling schemes used in the 2 surveys. In the case of both surveys, weights for veterans identified from VHA files were determined from selection probabilities adjusted by responses to screening questions and lack of response to the main interview. The survey administrators calculated weights for veterans identified by RDD in 1993 by computing veterans’ household cluster weights, raking the household weights to Current Population Survey household totals, and adjusting for interview nonresponse.11 (Raking is a statistical procedure designed to improve the reliability of survey estimates and correct for bias because of missed households without telephones and with unlisted numbers.) Survey administrators calculated these weights for 2001 by computing inverse probabilities of selection, adjusted for household-level and individual-level nonresponse and the possibility of multiple telephone lines, and then raking the weights to the Census 2000 Supplementary Survey.12^,13 In the case of both years, we combined weights to produce composite weights for use with the combined list and RDD samples and transformed them to adjust for the original sample size.

Our analysis involved several components. First, we compared the demographic, social, and clinical characteristics of the veteran populations in the 2 study years. Second, we calculated outpatient health care use by year and trends in care from 1992 to 2000 after adjusting for potential confounders. In the model focusing on outpatient care versus no outpatient care, we used logistic regression; in the model focusing on amount of VHA outpatient care—among those accessing such care—we used multinomial logit regression. The categories used in the multinomial logit regression were non-VHA only, VHA only, and dual use.

Third, in an attempt to understand the characteristics most responsible for the trends observed, we reran the multinomial logit model, initially stratified by year and then including both years together with an interaction term between year and all of the other independent predictor variables. This last model enabled us to determine whether there were statistical differences between the odds ratios (ORs) for the covariates by year. For ease of interpretation, we present odds ratios from the stratified models and significance values for differences between years determined from the interaction terms in the nonstratified model. Individuals with missing data were eliminated from the multivariate analyses; thus, the final models were run on 90% of the sample eligible for inclusion. SAS (version 8.0; SAS Institute Inc, Cary, NC) was used in conducting all analyses.

RESULTS

Table 1 ▶ presents the characteristics of the respondents from both surveys. Statistically significant differences reflect the aging veteran population, including greater percentages of veterans who (1) are older (mean age: 52.95 years in 1992 vs 58.18 in 2000), (2) have chronic conditions (mean number of conditions: 1.99 in 1992 vs 2.52 in 2000), and (3) have a disability associated with an activity of daily living (16.8% in 1992 vs 25.6% in 2000). Table 2 ▶ illustrates how overall use of outpatient care changed from 1992 to 2000. From 1992 to 2000, there was a decrease in the percentage of veterans reporting no use of any outpatient care (from 48.2% to 25.9%). Overall, veterans exhibited increased rates of use of non-VHA outpatient care (from 43.9% to 58.3%), VHA-only care only (from 3.0% to 6.8%), and dual care (from 4.9% to 9.0%).

TABLE 1—

Characteristics of Sample: National Survey of Veterans, 1993 and 2001

	1992 (n = 11 645)	2000 (n = 20 048)	χ2Pa
Age, y, mean (SD)	52.95 (0.14)	58.18 (0.10)	<.0001
Male, %	95.6	94.1	<.0001
Race/ethnicity, %			<.0001
White	85.3	84.4
African American	8.2	8.8
Hispanic	3.9	2.6
Other	2.6	4.2
Married, %	77.4	73.3	<.0001
Education beyond high school, %	53.1	58.8	<.0001
Income, $, %			<.0001
0–25 000	32.1	28.3
25 001–50 000	35.7	32.8
≥ 50 001	32.2	38.9
No health insurance coverage, %	9.5	10.6	<.0134
No. of chronic conditions, mean (SD)	1.99 (1.99)	2.52 (2.26)	<.0001
Activity of daily living disability, %	16.8	25.6	<.0001
Service, connected disability, %	9.9	13.8	<.0001

Note. Values were weighted to reflect probability of sampling. Numbers of missing values were as follows: age, 3 missing from 1992 and 87 missing from 2000; marital status, 21 missing from 1992; education, 63 missing from 1992 and 47 missing from 2000; income, 645 missing from 1992 and 2181 missing from 2000; and insurance status, 239 missing from 1992.

^aP values for weighted comparisons: χ² test for trend for ordinal variables, χ² test for nominal and binary variables, and t test for continuous variables.

TABLE 2—

Changes in Use of Outpatient Care, 1992–2000: National Survey of Veterans, 1993 and 2001

	1992		2000
	Unweighted No. (%)	Weighted No. (%)	Unweighted No. (%)	Weighted No. (%)
No outpatient care use	3631 (31.2)	5613.9 (48.2)	4489 (22.4)	5 196.3 (25.9)
Non-VHA–only care	4215 (36.2)	5110.8 (43.9)	9988 (49.8)	11 682.0 (58.3)
VHA-only care	2116 (18.2)	344.7 (3.0)	2483 (12.4)	1 372.4 (6.8)
Dual care	1683 (14.5)	575.7 (4.9)	3088 (15.4)	1 796.9 (9.0)

Note. VHA = Veterans Health Administration.

It is clear from Table 2 ▶ that outpatient care use increased over time, but to provide a better understanding of which types of care increased relative to others, Table 3 ▶ presents adjusted odds of receipt of different types of outpatient care in 2000 versus 1992, stratified by age. Confirming the result shown in Table 2 ▶, the odds ratio for receipt of any outpatient care in 2000 relative to 1992 was overwhelmingly significant: 2.45 (95% confidence interval [CI] = 2.31, 2.59). Among those who used any outpatient care, use of both VHA-only care (OR = 1.75; 95% CI = 1.51, 2.04) and dual care (OR = 1.22; 95% CI = 1.08, 1.37) increased relative to use of non-VHA–only care. However, when these adjusted odds were stratified by age group, a different pattern emerged. Relative to non-VHA–only outpatient care, VHA-only care significantly increased among veterans younger than 65 years as well as those 65 years or older, but dual use significantly increased only among those 65 years or older.

TABLE 3—

Adjusted Odds Ratios for Use of Outpatient Care, 2000 vs 1992: National Survey of Veterans, 1993 and 2001

	Any Care vs None, OR (95% CI)	VHA Only vs Non-VHA Only, OR (95% CI)	Dual vs Non-VHA Only, OR (95% CI)
All veterans	2.45 (2.31, 2.59)	1.75 (1.51, 2.04)	1.22 (1.08, 1.37)
Veterans younger than 65 years	2.38 (2.22, 2.54)	1.54 (1.27, 1.87)	0.79 (0.68, 0.91)
Veterans 65 years or older	2.75 (2.46, 3.08)	2.26 (1.77, 2.88)	2.15 (1.75, 2.63)

Note. OR = odds ratio; CI = confidence interval; VHA = Veterans Health Administration. Adjusted ORs were based on 28 470 veterans (89.8%) with no missing data. ORs were adjusted for age, gender, marital status, race, education completed, income level, lack of health insurance coverage, number of chronic conditions, presence of an activity of daily living disability, and presence of a service-related disability. Values were weighted to reflect probability of sampling.

We next examined the characteristics of veterans who used VHA-only and dual out-patient care in comparison with those who used only non-VHA care (Table 4 ▶). Both VHA-only care and dual users were more likely to have no additional health insurance coverage in 2000 than in 1992. For example, relative to a non-VHA–only user, the odds of a VHA-only user having no additional health insurance coverage were 4.90 (95% CI= 3.37, 7.13) in 1992 and 8.06 (95% CI=6.66, 9.75) in 2000. There were no other significant differences among VHA-only care users between 1992 and 2000; however, there were a number of other differences among dual users.

TABLE 4—

Odds of Any VHA Outpatient Care vs Non-VHA Outpatient Care Only in 1992 and 2000: National Survey of Veterans, 1993 and 2001

	VHA Only vs Non–VHA Only			Dual Care vs Non–VHA Only
	1992, OR (95% CI)	2000, OR (95% CI)	Pa	1992, OR (95% CI)	2000, OR (95% CI)	Pa
Age	1.01(1.00, 1.03)	1.01 (1.00, 1.02)	.4670	0.99 (0.98, 0.99)	1.01 (1.01, 1.02)	<.0001
Male	2.21 (1.10, 4.44)	1.62 (1.17, 2.24)	.4144	0.89 (0.61, 1.31)	0.78 (0.62, 1.00)	.5631
Race/ethnicity
White	1.00	1.00		1.00	1.00
African American	2.65 (1.83, 3.82)	1.89 (1.53, 2.33)	.1095	1.70 (1.24, 2.34)	1.26 (1.03, 1.55)	.1145
Hispanic	1.54 (0.78, 3.01)	0.87 (0.55, 1.37)	.1616	1.51 (0.92, 2.48)	1.20 (0.82, 1.76)	.4624
Other	1.99 (0.93, 4.27)	1.99 (1.47, 2.69)	.9953	1.11 (0.60, 2.05)	1.28 (0.96, 1.72)	.6740
Married	0.67 (0.50, 0.90)	0.85 (0.73, 1.00)	.1375	0.86 (0.67, 1.09)	1.18 (1.02, 1.36)	.0219
Education above high school	0.90 (0.68, 1.20)	0.80 (0.70, 0.93)	.4629	1.37 (1.11, 1.68)	1.00 (0.88, 1.13)	.0105
Income, $
0–25 000	14.65 (7.61, 28.20)	7.63 (6.05, 9.62)	.0605	1.43 (1.09, 1.89)	3.16 (2.64, 3.78)	<.0001
25 001–50 000	3.02 (1.52, 6.02)	2.58 (2.05, 3.25)	.6630	0.95 (0.74, 1.21)	2.05 (1.74, 2.40)	<.0001
≥ 50 001	1.00	1.00		1.00	1.00
No health insurance coverage	4.90 (3.37, 7.13)	8.06 (6.66, 9.75)	.0108	1.11 (0.74, 1.67)	2.26 (1.82, 2.81)	.0021
No. of chronic conditions	1.13 (1.06, 1.20)	1.13 (1.09, 1.16)	.9728	1.16 (1.11, 1.22)	1.22 (1.19, 1.25)	.0858
Activity of daily living disability	1.47 (1.08, 2.00)	1.15 (0.98, 1.35)	.1150	1.29 (1.02, 1.64)	1.26 (1.10, 1.44)	.8375
Service-connected disability	6.82 (5.09, 9.14)	5.06 (4.30, 5.95)	.0739	4.80 (3.85, 5.99)	4.48 (3.90, 5.15)	.5955

Note. OR = odds ratio; CI = confidence interval; VHA = Veterans Health Administration. Values were weighted to reflect probability of sampling.

^aP value for interaction term between year and variable of interest. Values less than .05 indicate that odds ratios by year are significantly different from each other.

In 2000, increases were evident in the relative numbers of dual users who were older, married, and at lower income levels, while there was a decrease in the relative number with more than a high school education. For example, relative to a non-VHA–only user, the odds of a dual user having an income of $25 000 or less versus an income above $50 000 were 1.43 (95% CI = 1.09, 1.89) in 1992 and 3.16 (95% CI = 2.64, 3.78) in 2000. Finally, VHA-only care and dual out-patient users were more likely than those who used only non-VHA care to be African American, to have a greater number of chronic conditions, and to have a service-related disability; these patterns did not change from 1992 to 2000.

DISCUSSION

The analyses described here indicate that there were complex shifts in patterns of out-patient care use among US veterans between 1992 and 2000. Veterans began using out-patient services more frequently, independent of multiple predictors of care including age and number of comorbidities. Among veterans who reported use of outpatient care services, there were shifts over time in the groups that received their care in the VHA. In particular, from 1992 to 2000, VHA use among veterans receiving outpatient care services was increasingly reported by those most likely to have difficulty affording care elsewhere, including veterans who were older, did not have additional medical insurance, and were in lower income brackets. Much of the increased use of VHA outpatient services by these vulnerable veteran populations was driven by dual users of VHA and non-VHA care rather than by users of VHA care alone.

These findings reconfirm the importance of the VHA as a “safety net” provider of care.10 As out-of-pocket costs for prescription medications and other care increased during the 1990s, more veterans not only turned to the VHA for care but also were disproportionately represented by those who might be least able to pay such costs. In 1992, among dual users, younger veterans were more likely to use the VHA; however, by 2000 this trend had reversed. Contrary to the widespread perception that veterans leave VHA care once they become eligible for Medicare, our results indicate that as the veteran population is aging, it is increasingly accessing the VHA system.

One potential explanation for this finding is the VHA’s drug benefit. Medicare recipients without some form of supplemental insurance had to pay for their medications during the period under study, while the VHA provides medications at no cost to the poorest veterans and at a low price to all other veterans.19 In addition, access to these reduced cost medications became easier during the study period, making it even more attractive for veterans to seek VHA care. In October 1996, Congress passed the Veterans’ Health Care Eligibility Reform Act (Pub L No. 104-262). This legislation paved the way for the creation of a medical benefit package that included, among other benefits, coverage for prescription drugs under the VHA national formulary system. Our results support this potential explanation in that the relative pattern of growth among dual users primarily involved those 65 years or older.

Our results demonstrating that the VHA has increased the amount of care it provides to low-income and uninsured veterans are consistent with a recent report indicating increases in the number of veterans without additional health insurance who use the VHA.20 As mentioned, the VHA underwent a large reorganization in the mid-1990s. In a historical account of the reorganization, Kizer and Demakis noted that, in the early 1990s, a number of evaluations performed by both internal and external investigators identified serious operational and managerial problems within the VHA.1 The VHA was criticized for being too specialized and hospital focused, and this organizational structure was thought to lead to uncoordinated, episodic care. As a means of addressing this situation, the VHA developed a plan to transform the VHA through the use of population health and managed care principles.1 As a consequence, an increased emphasis was placed on out-patient care and delivery of preventive services.21^,22 Evaluations of preventive care delivery programs within the VHA during the 1990s indicate that, among other positive signs, rates of preventive care service delivery improved.23^,24

In addition, there is evidence that the VHA is at least as effective as non-VHA facilities in terms of delivering preventive health care services.2^,4^,24^–27 In fact, improved quality of VHA care relative to non-VHA care may be yet another force driving patients to use the VHA. McGlynn and colleagues showed that, overall, adults in the United States do not receive recommended amounts of preventive, acute, and chronic care, while Asch and colleagues showed that individuals who obtain care from the VHA (even dual care users) are much more likely to receive these recommended services than non-VHA users (including nonveterans).4^,28 Our work extends this picture of the VHA and confirms that the organization has significantly improved in its mandate to provide universal care for all veterans regardless of ability to pay. This is reflected in the relative growth of low-income dual users. Changes in both health care finances and VHA administration appear to be influencing the trends observed in this study; however, we were unable to identify the independent effects of these forces.

This study involved limitations. First, when pulling together 2 different surveys, one must always consider whether design differences are responsible for the results produced. The sampling methodology was not identical in the 2 surveys assessed here, but the data sources used to identify potential participants were identical, and the sampling algorithms were similar. In addition, completion rates were similar from year to year. Weighting procedures were similar in the 2 survey years and were anchored to census data, and all analyses incorporated weights. Second, although the wording of questions was similar from year to year, there were slight variations, and questions may have been interpreted differently in different years. However, we can find no reason why older veterans, poorer veterans, veterans without health insurance, or sicker veterans would systematically interpret these questions differently from other veterans.

Third, veterans were asked to recall outpatient use for an entire year, and there is some indication that recall was less than perfect. In the case of the 1992 survey list sample, responses were corroborated with administrative files. Agreement with VHA records was good for questions regarding age and service-related disability status; however, about 25% of the list sample respondents provided answers in disagreement with VHA records regarding the medical care they had received during 1992.9 However, this high misclassification rate may have been because of poorly managed VHA records as opposed to poor recall, given that by 2000 the ineligibility rate resulting from list misclassification had dropped from 33.9% to 8.8%. Finally, it is likely that a greater proportion of low-income than high-income veterans did not have telephones and may have been underrepresented in our sample, limiting the generalizability of our results to certain of the most disadvantaged groups of veterans (e.g., homeless veterans).

In conclusion, an important mandate of the VHA is to serve vulnerable veterans. Since the reengineering of the organization in the mid-1990s, it seems to have made significant strides toward fulfilling this mandate by increasing the amount of care it provides to older, low-income veterans, as well as those without additional health insurance coverage. Although there remain opportunities for improvement, the VHA continues to play a vital role in the US health care safety net.