| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
Disability Statistics Center, University of California, San Francisco.
Address correspondence to Mitchell P. LaPlante, Disability Statistics Center, Institute for Health & Aging, 3333 California St., Rm. 340, San Francisco, CA 94118. E-mail: laplant{at}itsa.ucsf.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. A two-part multivariate regression model of the probability of PAS use and hours of help received was developed to control for need level, living arrangements, and other characteristics that may differ between persons with met and unmet needs and to determine the shortfall in hours associated with unmet need.
Results. Individuals with unmet need for personal assistance with two or more of the five basic ADLs have a shortfall of 16.6 hours of help per week compared with those whose needs are met. The relative shortfall is twice as great for persons who live alone as for those who live with others. People who live alone and have unmet needs fare worse than people with unmet needs who live with others, and both groups are more likely than those whose needs are met to experience adverse consequences, including discomfort, weight loss, dehydration, falls, burns, and dissatisfaction with the help received.
Discussion. Overall, just 6.6% of needed hours are unmet among the 3.3 million people needing help in two or more ADLs. We estimate the annual cost of eliminating unmet need among persons with incomes under 300% of the Supplemental Security Income level between $1.2 and $2.7 billion for those living alone and from $2.2 to $7.1 billion for those living with others.
COMMUNITY-RESIDING adults of all ages with needs for personal assistance services (PAS) in daily activities are at risk of all or some of their needs being unmet, which can reduce their quality of life, compromise safety, and increase risks of a number of adverse consequences (Allen & Mor, 1997
; Desai, Lentzner, & Weeks, 2001; Lima & Allen, 2001). These include injuries from falls, burns, weight loss, dehydration, discomfort, and other problems that can further worsen health and disability and increase risks of institutionalization and death. Minimizing unmet need is the primary goal underlying long-term care policy.
PAS refers to human help provided to individuals in specific activities that are generally obligatory for bodily maintenance and for living in the community, comprising the activities of daily living (ADLs; bathing, dressing, transferring from a bed or chair, toileting, and eating) and the instrumental activities of daily living (IADLs; such as taking medications and shopping for groceries). PAS include all help, whether hands-on, standby, or supervisory, whether paid or unpaid. Unmet need occurs when assistance from others is needed but not provided or inadequate.
There are many causes of unmet need. Over 85% of all of the hours of assistance people receive with ADLs and IADLs are provided by family and friends (LaPlante, Harrington, & Kang, 2002). Informal helpers, who may have to balance other responsibilities, including work and child care, are often limited in the amount of help they can provide. People with low informal resources, such as those who live alone, are especially likely to have unmet needs and depend largely on formal assistance, if affordable and available. Public PAS remain biased toward institutional rather than community living (Harrington et al., 2000), and people who live in states that do not provide enough home- and community-based services may be at greater risk of unmet need (LeBlanc, Tonner, & Harrington, 2000, 2001; Muramatsu & Campbell, 2002).
To address the magnitude of the problem of unmet need for PAS, it is useful to know what the national prevalence of unmet need is, what populations are most at risk, and how much of their needs are lacking. Surveys of individuals are convenient for measuring unmet need because obtaining professional judgments on a large scale is prohibitively expensive (Williams, Lyons, & Rowland, 1997). The prevalence of survey-discovered unmet need lies between one fifth and one third of elderly people with PAS needs (Desai et al., 2001; Kennedy, 2001; Lima & Allen, 2001; Manton, 1989; U.S. General Accounting Office, 1988; Williams et al., 1997).
The need for PAS spans a continuum, measured often by the number and type of ADLs a person requires human help with. The ADLs define a hierarchy of needs (Katz & Akpom, 1976), with dependence in eating defining the most severe need. The IADLs define a lesser degree of need for PAS. An important measurement issue concerns who determines need. Bradshaw (1972) outlined several distinctions in operationally defining need: normative need, which is assessed by experts; felt need, the level of help the individual feels he or she needs; expressed need, the amount of help received or demanded by the individual; and comparative need, disparities in help received. Conflict arises between normative and felt need, as when a professional feels that institutional placement is warranted but the individual may not want it or when the individual wants help that a professional will not give. Differences are known to exist between professional and self-assessments of need. In one study, self-rated ADL disability was greater than that observed by professionals (Kelly-Hayes, Jette, Wolf, D'Agostino, & Odel, 1992). Professionals assess need differently than individuals, but neither is necessarily better than the other (Verbrugge & Sevak, 2002).
Certainly, if need is a continuum, then unmet need is also a continuum. Some people may lack help entirely, whereas others may need just a little more help. A professional may judge that a person needs a lot of help when he or she receives only a little. Few studies exist comparing self-reports and professional assessments of unmet need. One study suggests that self-ratings of unmet need may be conservative compared with professional ratings of unmet need (Morrow-Howell, Proctor, & Rozario, 2001).
To advance the understanding of unmet need beyond mere prevalence, it is useful to know how much help is lacking and what the cost might be to address it. Quantifying unmet need for PAS in terms of the shortfall in hours of help is an important step in that direction. This analysis builds on a prior national study estimating hours of PAS received among adults of all ages (LaPlante et al., 2002) to examine the relationship of hours with unmet need. We define unmet need as the gap between felt need and expressed need. We use a large nationally representative survey to estimate the shortfall in hours associated with unmet need by comparing the hours of help received by people who report that their needs are met with those whose needs are unmet, after controlling for population differences in need levels and demographics (comparative need). We then estimate the cost of eliminating the unmet-need shortfall in hours and demonstrate that unmet need is associated with a variety of adverse consequences.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
, 1998b). The 202,560 respondents of all ages (93% Phase I response rate) were screened for disability, using an extensive set of criteria including functional limitation, specific disabling diagnoses, perception of disability, and use of disability-related services. A sample of 32,788 individuals was reinterviewed in Phase II (89% Phase II response rate), also known as the Disability Followback Survey. Data collection began in September 1994 and concluded in April 1997, with reinterviews taking place between 7 and 26 months after the Phase I interview (normally distributed about the mean of 14 months). Phase I excluded individuals in nursing homes, prisons, and large residential facilities. About 3.1% of the eligible adult sample died before follow-up, 3.4% refused, 3.0% could not be located, and 1.4% were not interviewed for unknown reasons. A small number of respondents moved into institutions and, although interviewed, were excluded from this analysis (n = 335). The sample for this analysis is restricted to adults with met or unmet need for help in any IADL/ADL (n = 9,646).
Measures
Data from Phase II of the NHIS-D are used to estimate met and unmet need for assistance with ADLs/IADLs and to estimate hours of paid and unpaid help received. Individuals were asked if they were helped in any of 5 ADL and 10 IADL activities, including hands-on help, supervision, or stand-by help, and whether they needed help or more help (activities a person does not do for reasons other than health are excluded). Individuals who lacked help or needed more help in one or more ADL/IADL activities are defined as having unmet need. For each of up to four helpers, the respondent was asked (a) how many days the helper helped during the last 2 weeks and (b) how many hours per day the helper provided help. The two measures were multiplied for each helper, summed over helpers, and divided by 2, to yield the total hours received per week.
The survey contains a number of demographic variables that are used in this analysis. Age and living arrangement were measured in Phase II; gender, race, and individual income were measured in the initial interview. A measure of the presence of cognitive symptoms (Phase I) was constructed by combining affirmative responses on having "a lot of trouble concentrating long enough to complete everyday tasks," being "frequently confused, disoriented, or forgetful," or having "Alzheimer's disease or another type of senile disorder." A second variable indicating cognitive impairment was whether the respondent had a proxy interview in Phase II because of poor memory, senility, confusion, Alzheimer disease, or other mental condition.
Several additional questions were asked about experiencing discomfort and other problems when individuals could not do activities as frequently as they needed to (Phase II). Additional questions were asked about falls, bedsores, contractures, dissatisfaction with the person's primary helper, and problems of being left alone. In some cases, respondents were asked if the problem was caused by a lack of help, thus attributing causality to unmet need. This battery yielded 53 measures that we used for this study. We expected that people with unmet need would be more likely to experience these problems and be more likely to be dissatisfied.
Less than 3% of responses are missing for the ADL need for help items and <6% for the IADL items. Cognitive symptoms were missing in <1% of records. As item nonresponse is generally low on these measures, we assume, when it occurs, that the individual does not have the characteristic in question. Reason for proxy interview was not ascertained for about 20% of those with proxy interviews; in these cases, we assume that a proxy response was needed for reasons other than cognitive ones.
The data set includes imputation for individual income based on a detailed hot deck imputation procedure carried out by the University of Michigan Survey Research Center (U.S. NCHS, 1998a, 1998b). In the sample with a need for help in IADLs/ADLs, 33.3% had some income components imputed, the most frequent being earnings and interest income.
The hours data used for the dependent variable included data we imputed previously for 19.5% of persons who said they were helped in the last 2 weeks but had missing data for the number of days or hours per day. We used multiple imputation, an iterative approach generally considered the most appropriate method when dependent variable data are missing (Crawford, Tennstedt, & McKinlay, 1995; Rubin, 1987). Indeed, missing hours were not random: Those with missing data were actually somewhat less disabled. The imputation involved estimating three separate regression models for hours received from paid helpers, a primary informal helper, and secondary informal helpers. Independent variables included demographics, living arrangement, income, Medicare and Medicaid coverage, supply of helpers, and IADL and ADL activities the helpers helped with. The imputation introduces a random component to the regression parameters, based on their estimated variances, and the prediction is replicated 10 times to yield a distribution of values, which were then averaged and summed over the three models to yield imputed total hours (LaPlante et al., 2002).
The 53 adverse consequence items had low nonresponse, generally under 8%, with the exception being three questions regarding needing help eating (about 16%) and several satisfaction questions (8–15%).
Analysis Methods
We employed statistical methods to estimate the shortfall in hours associated with unmet need. The hours data are nonnormal in two ways: There are some cases with 0 hours and some with a high number of hours. Using multivariate regression methods, we estimated a two-part model of PAS utilization (Duan, Manning, Morris, & Newhouse, 1983), which takes into account both types of skewness: The first step predicts whether or not any hours of help are received by weighted logistic regression, and the second step predicts the hours of help among recipients by weighted least-squares (WLS) regression, using survey sampling weights in both. In the latter, we use the logarithm of hours as the dependent variable to eliminate right skewness. Each person's level of need is measured by 15 dummy variables for the specific IADLs/ADLs they need help with. The log-transformed dependent variable in the WLS regression makes the contributions of the activities multiplicative rather than additive. That is, the logarithm of the total hours received does not equal the sum of the logarithms of the hours received for each activity. We include a count of multiple activities to correct for this effect, an integer equal to the number of activities minus 2, if help is needed in two or more activities, or 0 otherwise.
Four dummy variables measure cognitive difficulties: having cognitive symptoms, being unable to manage money, being unable to manage medications, and having a proxy interview because of a cognitive condition. Living arrangements are measured using two dummy variables: living alone and living with a spouse, the comparison group being those who live with others. To identify individuals with low incomes based on Supplemental Security Income (SSI) eligibility criteria, two additional dummy variables are included classifying personal income at or below 100% of SSI federal benefit rate ($446 per month in 1994 and $458 per month in l995 if living alone; $669 per month in l994 and $687 per month in l995 if living with a spouse) and from 101% to 300% of the SSI federal benefit rate level, the comparison group being those with incomes >300% of the federal benefit rate. Although individuals on Medicaid may retain $2,000 in tangible assets, the survey questions were not detailed enough to make further distinctions on assets. The unadjusted data suggest that unmet need has a larger impact among people with two or more ADLs (as discussed herein; Table 1), so we wished to test that. We further wished to distinguish unmet-need effects by whether people live alone or with others. Four dummy variables measure the effects of unmet need for four population groups of interest: people living alone or living with others and having fewer than two or two or more ADL needs. These groups are compared with those who have no unmet need. We developed the models first introducing unmet need and living arrangements, then level of need, and finally, age, race, gender, and income. We deleted variables not significant in either model. In the WLS model, we ascertained that the model residuals were normally distributed.
|
The predicted probability of receiving any help is calculated using the exponentiated model parameters from the logistic model. To retransform the predictions of the WLS regression (predicted logarithm of hours) into a linear scale (predicted hours), we exponentiate the model prediction and multiply by a retransformation or "smearing" factor (Duan, 1983; Duan et al., 1983), obtained by averaging the exponentiated residuals. Heteroscedasticity across populations requires separate smearing factors (Manning, 1998). We observe smaller mean squared residuals for people living alone (versus living with others) or having two or more ADLs (versus less than two) and calculate separate smearing factors for the four populations thus defined. We then multiply the predicted probabilities of use by the predicted hours of use. Finally, we estimate the predicted hours of PAS for those with unmet need, along with their expected hours if their unmet need were to be eliminated, by setting the relevant unmet need dummy variable to zero. The difference in expected hours represents an estimate of the shortfall in hours, the hours that would be required to fulfill the person's needs.
The weekly shortfall is multiplied by 52 weeks to obtain an annual estimate (which is allowable because the survey is conducted continuously). PAS costs are assumed to be $10 per hour on average, based on average state Medicaid reimbursement data for PAS collected from a survey of state officials in l998–1999 (LeBlanc et al., 2000, 2001). Finally, we use a bootstrap procedure (resampling with replacement) to calculate 95% confidence intervals around the estimated cost.
We hypothesized that the percentage of people experiencing adverse consequences would be greater on every measure for those with unmet versus met need, which we tested using a one-tailed t test. Because the estimates are based on a sample of the population, they are subject to sampling error. All estimates are calculated using sampling weights, and sampling errors have been calculated directly using STATA survey data commands (Stata Corp., 2003), which take into account the geographic clustering of the sample design by Taylor series approximation.
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
1 of the 15 IADLs and ADLs, 21.4% of whom have unmet need (Table 1). The percentage with unmet need is lowest for those who need help only in IADLs (18.3%) and generally increases with higher numbers of ADLs, from 23.6% with one ADL need to 34.6% at four ADLs, but declines to 25.9% at five ADLs. Most people with unmet need get some help but do not get all the help they need. We find that only 545,000 adults of the 3.2 million with unmet need receive no help at all, a population with a low level of disability, mostly needing help in IADLs only, and in particular the single activity of heavy housework (not shown). For most people with unmet need, the issue is not whether they get help, but how much help they get.
Relationship of Unmet Need to Hours of Help
We expect that people with unmet need would have fewer hours of help than those whose needs are met. We find only a small difference in the mean hours of help received between those with unmet versus met need among people who need help with just IADLs or only one ADL (<2 hours per week) but large and highly significant differences for those needing help in two or more ADLs. This group amounts to 949,000 people—almost a million—with perceived unmet need, who average 16.1 fewer hours per week than those whose needs are met (see Table 1). Persons with unmet needs have fewer hours for each number of ADLs above one that they need help with, and the differences in mean hours are statistically significant for those needing help in two, three, or five ADLs. If we combine those with two or more ADLs, there is a large significant difference in mean hours for those with unmet needs, but not for those with fewer than two ADLs.
These differences in mean hours received, however, might be due to differences between the met-need and unmet-need populations with respect to level of need and other individual characteristics that influence the amount of help received. We use multivariate methods to control for such differences to estimate the true shortfall among people with unmet need for PAS.
Estimating the Unmet-Need Shortfall in Hours
The sample means and standard deviations are shown in Table 2 for variables tested in the regressions, contrasting those with and without unmet need. Adults with unmet need are less likely to get help but receive about the same average hours when they do get help. They are more likely than those with no unmet need to be non-White, to be female, and to live alone. They are less likely to live with a spouse, have a greater number of ADLs and IADLs, are more likely to have cognitive symptoms but less likely to have a proxy interview for cognitive or other reasons, and are slightly more likely to have incomes below 100% of the SSI level.
|
|
In Table 4, the shortfall in hours estimated by the regression models is shown for people with two or more ADLs who have unmet need, by living arrangement and income level. As explained in Analysis Methods, the estimated shortfall is the model prediction of the difference between needed and received hours. Overall, people needing help in two or more ADLs and reporting unmet need receive 56.2 hours of help and are estimated to need an additional 16.6 hours of help per week. Those living alone receive 24.1 hours of help and need an additional 18.7 hours, a relative shortfall of 43.7% of their total needed hours. Those who live with others receive substantially more help (66.1 hours) but need an additional 16.0 hours; their relative shortfall is a more modest 19.5%. People with fewer than two ADLs with unmet need are omitted owing to their relatively small shortfall (4 hours) and lesser policy import.
|
Negative Consequences of Unmet Need and Dissatisfaction
Even people whose needs are usually met can experience adverse consequences owing to lack of help, if a helper is not available at the right time and place, or if the quality of help is inadequate. But we would generally expect people with perceived unmet needs to have a higher frequency of adverse experiences. For all comparisons where a significant difference is observed, people with unmet needs have a higher rate of adverse consequences. Of individuals who need help with two or more ADLs, those with unmet needs have significantly greater probability of adverse consequences than those whose needs are met on 29 of 34 measures tested (Table 5). These measures include discomfort, distress, mobility restriction, doing activities oneself that the person needs help with, and more serious concerns such as going hungry, running out of food, getting burned, unintentional weight loss, and dehydration. Many of these measures specifically attribute lack of help as the cause of the events, such as going hungry because no help was available to eat.
|
|
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
), the problem of unmet need for PAS is magnified among people who live alone, 45% of whom have unmet need, compared with 26% of those who live with others. This study demonstrates that an association exists between perceived unmet need and reduced hours of help, independent of level of disability, race, age, and income level. People who live alone and have unmet need (almost a quarter-million people) are estimated to lack 18.7 hours of help per week (95% confidence interval = 12.3–25.1) or 44% of the hours they need. Those with unmet needs who live with others lack 16.0 hours per week (95% confidence interval = 7.7–24.4) or 20% of the hours they need. Unmet need is not a measure of individuals' insatiable demands for more help. The data confirm that unmet need is associated with higher rates of adverse consequences on 48 of 53 measures tested, including discomfort, going hungry, losing weight, dehydration, falls, injuries due to falls, burns, and dissatisfaction, at much higher rates than those whose needs are met, particularly for people who live alone. These are serious problems that compromise the safety, comfort, and hygiene of individuals with unmet needs, reducing their ability to live independently and increasing their risk of institutionalization and possibly death. Many of these events are specifically attributed by the respondent as resulting from lack of help. Inferring causality on other measures lacking such attributions, like contractures and bedsores occurring in the last 3 months, is hazardous because of the cross-sectional design of the survey. The occurrence of such events may have precipitated a need for more than the available level of help. Nevertheless, the weight of the evidence clearly suggests that unmet need is associated with and attributed to much greater adversity.
Limitations to these findings include the fact that need and unmet need are self-reported or proxy reported, whereas professional assessment could yield different results. Hours for those whose needs are met are the standard against which the unmet-need shortfall is defined, and those hours are likely to be influenced by whether they are self-reported, proxy reported, or assessed by a professional. Item nonresponse necessitated imputation and proxy reporting raised questions of potential bias in reporting levels of need, unmet need, and hours. However, we determined that results were robust even when imputed hours and proxy reports were separately excluded in the analysis (results not shown). Nevertheless, professional assessment could yield a different estimate of the shortfall.
Working-age adults account for half the prevalence of met and unmet need among those needing help in two or more ADLs. We analyzed the working-age population separately and found the results robust, although the relative shortfall in needed hours is somewhat smaller than for all ages combined (18% versus 23%, results not shown). However, unmet need among people living alone is more an issue for the elderly, because two thirds of those living alone and needing more help are age 65 or older. It is notable that both age and being non-White have positive relationships to hours received after controlling for level of need, which may represent sociocultural differences in informal helping.
Access to paid help is critical for people who live alone because they are much more likely to receive paid help than are people who live with others (62% versus 24%; p <.001, not shown). Of people who live alone, those whose needs are met are more likely than those whose needs are unmet to get paid help (70.2% versus 50.3%; p <.001, not shown). More than half of that paid help is reported to be paid by Medicare or Medicaid. Among people living with others, the difference in getting paid help by unmet need is much smaller (25.7% versus 19.4%; p <.001, not shown). Thus, expanding access to paid PAS appears crucial in reducing unmet need among those who live alone. A limitation of the data is that family helpers were not asked if they were paid, which may accentuate differences between paid and unpaid help by living arrangement.
If the estimated shortfall in hours were to be provided through public funds, the cost of eliminating unmet need among people who live alone with low incomes ranges from $1.2 to $2.7 billion, a relatively small amount. However, the cost for people who live with others is more than twice as large, from $2.2 to $7.1 billion.
People who live with others get twice as many hours of help per person, independent of need and unmet need. Is the lower level of help provided to people who live alone a problem? Actually, people who live alone and whose needs are met are no more likely to experience adverse consequences than people who live with others and have their needs met, which suggests that the difference in hours is not a problem. It seems more likely that the difference in hours represents a different accounting of help received. After all, among people who live alone, helpers come in to the home usually with a set of goals to be achieved in a limited time. It is a different case when the person lives with someone who provides help continuously throughout the day. A different accounting does not imply that the unmet-need shortfall in hours for people who live with others is not real. They do have higher risk of adverse consequences when their needs are unmet, although not as extreme as for those who live alone.
There are some problems to be dealt with in reducing unmet need for PAS. The potential for greater publicly financed help to displace unpaid help has been acknowledged (Kemper, 1992; Moscovice, Davidson, & McCaffrey, 1988). Some research suggests that paid services complement, rather than substitute for, unpaid care (Liu, Manton, & Aragon, 2000). The potential substitution effect is much less an issue for people who live alone, only one third of whom are helped by children or relatives, than it is among people who live with others and have unmet needs, of whom about 90% are helped by relatives, usually a spouse.
Greater public provision of help to low-income individuals is required to reduce unmet need among those who live alone, but it is not the only remedy for those who live with others. Other strategies may be considered to increase informal hours for those living with others by defraying the personal costs that helpers and their families incur, such as providing tax credits for helpers, and by other financial incentives to foster more informal help. However, these strategies do not focus on unmet need and would provide resources to those whose needs are already being met (or their caregivers) and who have lower risk of adverse consequences than those with unmet need.
Finally, human help is not always the best solution even for those who feel they need it. Trying to eliminate unmet need only by fulfilling people's wishes for more human help, without considering alternative ways to accomplish activities through use of assistive technology (Verbrugge & Sevak, 2002) and environmental modification (Lawton, 1982), could result in less empowerment, greater learned dependency, and greater societal expense.
Legal precedents such as the Supreme Court's Olmstead decision (U.S. Government Accounting Office, 2001) and legislative initiatives such as the Medicaid Community Attendant Services and Supports Act (MiCASSA; Harkin & Spector, 2001) would enable individuals to live in the least restrictive setting they desire. However, such efforts to reduce society's reliance on institutions cannot work if individuals' needs are not well met in the community.
Previous research, by looking only at the prevalence of unmet need for PAS, creates a false impression that unmet need is a large and costly problem to resolve. In fact, only 6.6% of all needed hours of help are unmet among adults needing help in two or more ADLs. The reduction, if not the elimination, of unmet need for PAS is a financially achievable goal for the nation and one that long-term care policy should focus on.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received for publication August 28, 2002. Accepted for publication July 21, 2003.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  J. C. Lima, S. M. Allen, F. Goldscheider, and O. Intrator Spousal Caregiving in Late Midlife Versus Older Ages: Implications of Work and Family Obligations J. Gerontol. B. Psychol. Sci. Soc. Sci., July 1, 2008; 63(4): S229 - S238. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Meng, B. Friedman, A. W. Dick, B. R. Wamsley, G. M. Eggert, and D. Mukamel Effect of a Voucher Benefit on the Demand for Paid Personal Assistance Gerontologist, April 1, 2006; 46(2): 183 - 192. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Allen, L. Resnik, and J. Roy Promoting Independence for Wheelchair Users: The Role of Home Accommodations Gerontologist, February 1, 2006; 46(1): 115 - 123. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
