Nutrition

III. CONTRIBUTION OF THE TITLE III NUTRITION PROGRAM TO PARTICIPANTS’ 24-HOUR

DIETARY INTAKE AND SOCIAL CONTACTS

One of the goals of the Elderly Nutrition Program (ENP) is to improve the dietary intakes of program participants by directly providing them with nutritionally balanced meals. The Older Americans Act (OAA), as amended, specifies that nutrition projects are to provide at least one hot or other appropriate meal per day to participants, five or more days a week. [ In rural areas, if approved by the State Unit on Aging (SUA), the number of meals per week may be fewer than five, if this frequency is not feasible or for other reasons.] Furthermore, program meals must comply with the Dietary Guidelines for Americans, set forth by the U.S. Department of Agriculture (USDA) and U.S. Department of Health and Human Services (DHHS), and meet standards for food energy and selected nutrients based on the Recommended Dietary Allowances (RDAs) of the National Research Council (NRC). For congregate meals, another program goal is to attract isolated elderly people to the meal sites in order to facilitate social interaction and delivery of other nutrition and supportive services that they need.

This chapter presents evaluation findings on the contribution of the Title III meals program to participants’ daily intake of nutrients and opportunities for socialization. It has two sections. The first examines the contribution of the program to participants’ 24-hour dietary intakes and number of monthly social contacts, by presenting evidence on the fraction of participants’ daily dietary intake and monthly social contacts derived from program sources. In this section, we also compare participants’ dietary intake from program meals with the RDAs and other dietary recommendations. The second section examines the contribution of the program by estimating differences between the daily dietary intakes and number of monthly social contacts for program participants and eligible nonparticipants, controlling statistically for other participant characteristics related to these outcomes.

The evaluation found that Title III meal program participants’ average dietary intake from the program meal generally meets or exceeds the OAA requirement of one-third of the RDAs for most nutrients. Participants’ average intake from program meals of total fat and saturated fat as a proportion of total calories is slightly higher than the recommended levels. Intake of carbohydrates as a percentage of total food energy is below recommendations. Overall, more than 40 percent of the average participants’ total daily intake on a day that they attend or receive meals from the program is derived from program meals. Consistent with previous research, the evaluation found that program participants’ dietary intakes relative to the RDAs exceed those of eligible nonparticipants for the days on which participants receive a program meal. The dietary improvements were generally stronger for congregate than home-delivered meal program participants. Of the 18 nutrients studied, congregate participants’ average daily intakes of all nutrients except iron, folate, and Vitamin B₁₂ are higher than those of congregate-eligible nonparticipants, at a statistically significant level. Home-delivered meal program participants’ average daily intakes of Vitamin A, Vitamin D, riboflavin, calcium, phosphorous, potassium, zinc, and magnesium are higher than those of home-delivered eligible nonparticipants, at a statistically significant level. The evaluation also found that program participants average more social contacts per month than eligible nonparticipants, using a broad definition of contacts that includes assistance from public or private home-and community-based long-term care service providers. Although methodological limitations make it impossible to attribute these differences definitively to causal effects of the program, as discussed more fully in the text, the differences are probably at least partially attributable to the ENP.

The remainder of the chapter describes these findings in greater detail.

A. DIETARY INTAKE AND SOCIALIZATION FROM PROGRAM SOURCES

1. Participants’ Dietary Intake from Program Meals
The in-person survey asked Title III meal program participants to recall all the foods they ate during the previous 24 hours. Participants were also asked the source of their meals. From these responses, weidentified all foods and beverages that came from the Title III program. For congregate participants, we included foods they consumed at the meal site, as well as any program foods taken home and eaten during the recall period.

Intake of Food Energy and Nutrients from Program Meals.
The OAA requires nutrition providers to serve meals that comply with the RDAs. Program meals must achieve a minimum of one-third of the RDA, if one meal is provided to participants per day; a minimum of two-thirds of the RDA, if two meals are provided per day; or 100 percent of the RDA, if three meals are provided per day. Thus, over a period of time, program meals for each eating occasion (breakfast, lunch, and dinner) are to average a minimum of one-third of the RDAs for specified nutrients. [ Chapter IV presents findings on the nutrient content of program meals as served or offered . It shows that the average program meal meets the explicit program target of providing at least one-third of the relevant RDAs. In this section, we address the issue of whether participants ' intake per program meal meets or exceeds one-third of the RDAs. Even though program meals as offered meet one-third of the RDAs, on average, participants ' average intake of nutrients from program meals may be less because they might not eat all of what is served to them.]

On the basis of an analysis of a single day’s 24-hour dietary recall, Title III congregate participants’ average intake of nutrients per program meal exceeds the one-third RDA requirement for all nutrients studied, often by substantial amounts (Table III.1). [ The majority of meal program participants receive just one program meal daily (88 percent of congregate participants and 87 percent of home-delivered participants). For participants who received more than one program meal during the recall period, we standardized their intake from program meals to a per-meal per-day basis, so it could be meaningfully compared to one-third of the RDA. For example, if a participant received two program meals daily, intakes from these two meals for each nutrient were summed and divided by two (the number of meals) to derive a measure of intakes on a per-meal basis.] For example, congregate participants’ average intake of protein from program meals is 58.4 percent of the daily RDA, or nearly twice the per-meal standard of 33.3 percent. The typical congregate participant’s intakes of the critical nutrients calcium, folate, and magnesium exceed one-third of the RDAs. Overall, for 10 of the 18 nutrients examined, two-thirds or more of Title III congregate participants’ intakes per program meal meet or exceed one-third of the RDAs.

TABLE III.1
PARTICIPANTS’ DAILY NUTRIENT INTAKE PER PROGRAM MEAL

(As a Percentage of RDAs)

	Title III Congregate Meal Participants				Title III Home-Delivered Meal Participants
Nutrient	Mean	Median	Percentage Exceeding One-Third of the RDA		Mean	Median	Percentage Exceeding One-Third of the RDA
Food Energy	33.1	32.0	46		26.6	27.1	31
Protein	58.4	55.8	86		50.6	51.4	73
Vitamin A	69.6	43.0	62		66.6	41.3	59
Vitamin C	61.0	46.8	66		48.5	33.7	51
Vitamin D	46.9	47.4	60		47.0	50.0	61
Vitamin E	38.0	35.1	53		28.3	26.4	35
Thiamin	50.6	47.0	76		44.6	43.0	65
Riboflavin	54.4	52.9	78		48.8	48.6	71
Niacin	56.8	53.8	83		45.8	44.8	69
Vitamin B6	37.5	35.0	54		32.8	32.5	46
Folate	46.1	41.0	66		41.1	35.3	53
Vitamin B12	91.3	70.0	80		74.2	64.5	73
Calcium	40.5	40.5	56		39.1	42.6	58
Iron	45.2	43.5	72		36.9	36.6	58
Phosphorous	59.1	57.5	85		52.7	55.2	75
Potassium	57.5	56.7	86		47.8	48.7	72
Magnesium	35.2	33.6	51		29.4	29.4	41
Zinc	33.2	30.8	43		28.2	27.7	35
Unweighted Sample Size	1,039	1,039	1,039		815	815	815

Source: Elderly Nutrition Program Evaluation, Participant survey, weighted tabulations.

Notes: Includes 89 participants who received but did not consume a program meal during the 24-hour recall period (23 congregate participants who attended the meal site and usually eat a program meal but did not that day because of medical tests or other reasons; 66 home-delivered meal program participants who received a program meal but chose not to eat the meal during the recall period, saving it for another time). Tabulations are weighted to be representative of a cross-section of participants receiving Title III meals on a given day. Persons who received more than one program meal during the recall period had their intakes summed and divided by the number of program meals received, so their intake could be compared to the one-third RDA standard.

RDA = Recommended Dietary Allowance.

For zinc and food energy, however, fewer than one-half of congregate participants attain one-third of the RDAs.

Title III home-delivered meal program participants’ average intakes per program meal meet or exceed the one-third RDA requirement for all nutrients except food energy, Vitamin E, Vitamin B₆, magnesium, and zinc (Table III.1). The typical home-delivered meal program participant consumes 50.6 percent of the RDA for protein, 41.1 percent of the RDA for folate, and 39.1 percent of the RDA for calcium from a program meal. For significant percentages of home-delivered participants, however, intakes from the program meal do not attain the one-third RDA requirement for the nutrients examined. Fewer than half of home-delivered participants achieve one-third of the RDA for food energy, Vitamin E, Vitamin B₆, magnesium, and zinc.

It is important to note that eight percent of Title III home-delivered participants (or 66 elderly people) did not report consuming any food items from a program meal during the recall period. These individuals received zeros in the calculation of the mean (median) intake from program meals for each nutrient. Since most Title III home-delivered participants receive five meals per week, most of these individuals probably received a program meal during the recall period, but (1) they chose not to eat it, either because they planned to eat it some other day, or because they did not like what was served; (2) they failed to identify food items consumed as coming from the program; or (3) interviewers failed to record properly foods coming from program sources when they were reported as such. Regardless of the reason, it is appropriate to include these individuals in the analysis of intake from program meals, because virtually all of them received a program meal. [ When these individuals are excluded from the analysis, the average daily intake from a home- delivered program meal relative to the RDA increases, as does the percentage of home-delivered participants obtaining at least one-third of the RDA from their intake of program meals (see Volume III, Appendix F).]

Macronutrient Content of Participants’ Intake from the Program Meal. The typical Title III congregate and home-delivered meal program participant’s intakes of total fat, saturated fat, and sodium per program meal exceed the levels recommended by the Dietary Guidelines, whereas the intake of carbohydrate from the program meal is somewhat below recommended levels. The intake of dietary cholesterol from the program meal is well within the recommended levels.

The mean intake of carbohydrate as a percentage of total food energy from a program meal is 49.4 percent for congregate participants and 48.6 percent for home-delivered participants (Table III.2). For both congregate and home-delivered participants, the mean percentages of food energy from carbohydrates are below the NRC’s recommendation of 55 percent. Both congregate and home-delivered participants’ intakes of total fat from program meals as a percentage of food energy are above the 30 percent recommended level (34.7 percent and 34.3 percent, respectively). Title III congregate and home-delivered meal program participants’ intake of saturated fat from program meals as a percentage of total calories is 12 percent, which is above the recommendation of 10 percent.

Title III congregate participants’ intake of dietary cholesterol per program meal is 87 mg. Intake of cholesterol per program meal for home-delivered participants is somewhat lower, at 71 mg. Congregate participants’ intake of sodium from program meals is, on average, 1,162 mg; sodium intake from program meals for home-delivered participants is considerably lower, at 951 mg. There are no federal ENP program regulations that quantify sodium or cholesterol contents per program meal. Applying the one-third RDA rule to the NRC recommendations as a desirable target for program meals for cholesterol and sodium indicates that participants’ intake of cholesterol from the program meal is below the maximum recommended level of 100 mg, but intake of sodium from the program meal is above the maximum recommended level of 800 mg.

TABLE III.2

PARTICIPANTS’ DAILY INTAKE OF MACRONUTRIENTS, SODIUM, AND DIETARY CHOLESTEROL PER PROGRAM MEAL

Dietary Component	Title III Congregate Meal Participants	Title III Home-Delivered Meal Participants
Carbohydratea
Mean Percentage of Food Energy (Calories)	49.4	48.6
Median Percentage of Food Energy (Calories)	48.6	47.3
Distribution of Intake as a Percentage of Food Energy (Calories)
Less than 45 percent	39	44
45 to 55 percent	35	34
56 to 65 percent	17	13
More than 65 percent	9	9
Total Fata
Mean Percentage of Food Energy (Calories)	34.7	34.3
Median Percentage of Food Energy (Calories)	35.0	34.0
Distribution of Intake as a Percentage of Food Energy (Calories)
Less than 20 percent	9	11
20 to 30 percent	26	29
31 to 35 percent	19	20
36 to 40 percent	21	15
41 to 50 percent	18	18
More than 50 percent	6	8
Saturated Fata
Mean Percentage of Food Energy (Calories)	12.2	12.3
Median Percentage of Food Energy (Calories)	11.9	11.7
Distribution of Intake as a Percentage of Food Energy (Calories)
Less than 5 percent	4	5
5 to 10 percent	39	37
11 to 15 percent	39	37
16 to 20 percent	14	14
More than 20 percent	3	6
Proteina
Mean Percentage of Food Energy (Calories)	19.4	21.1
Median Percentage of Food Energy (Calories)	18.6	19.8
Distribution of Intake as a Percentage of Food Energy (Calories)
Less than 5 percent	*	*
5 to 15 percent	30	23
16 to 25 percent	58	55
More than 25 percent	12	21
Sodium
Mean Intake (mg Per Day)	1,162	951
Median Intake (mg Per Day)	1,062	901
Distribution of Intake
Less than 800 mg per day	29	43
801 to 1,000 mg per day	17	13
More than 1,000 mg per day	54	44
Dietary Cholesterol
Mean Intake (mg Per Day)	87	71
Median Intake (mg Per Day)	78	66
Distribution of Intake
Less than 100 mg per day	68	76
101 to 133 mg per day	17	14
More than 133 mg per day	16	10
Unweighted Sample Size	1,039	815

Source: Elderly Nutrition Program Evaluation, Participant survey, weighted tabulations.

Notes: Tabulations are weighted to be representative of a cross-section of participants receiving Title III meals on a given day. The Dietary Guidelines recommend that intake of (1) total fat should be 30 percent or less of food energy, and (2) saturated fat should be 10 percent or less of food energy. The National Research Council recommends (1) a carbohydrate intake of 55 percent or more of food energy, (2) a total sodium intake of 2,400 mg or less daily, and (3) a total cholesterol intake of 300 mg or less daily. Applying the one-third standard to the sodium and cholesterol recommendations implies that the intake per program meal should not exceed 800 mg of sodium or 100 mg of cholesterol.

^aExcludes 89 participants who received but did not consume a program meal during the 24-hour recall period (23 congregate participants who attended the meal site and usually eat a program meal but did not that day because of medical tests or other reasons; 66 home-delivered meal program participants who received a program meal but chose not to eat the meal during the recall period, saving it for another time).

* = Less than 0.5 percent.

Percentage of Total Daily Dietary Intake Provided by the Program Meal. Participants consume substantial proportions of their total daily intake of nutrients from Title III program meals on days when they either attend the congregate meal site or receive home-delivered meals. Title III congregate and home-delivered participants’ average intake from program meals ranges between 36 and 51 percent of their total daily intake of the 18 nutrients examined (Table III.3). For example, the typical congregate participant gets 44 percent of his or her daily intake of food energy (calories) from the program meal; the comparable figure for home-delivered participants is 39 percent. Congregate participants, on average, get 49 percent of their total daily intake of protein from the program meal, compared with 47 percent for home-delivered participants. Averaging the mean percentages of intake from program meals across the 18 nutrients shows that, for congregate participants, approximately 45 percent of the total daily intake of these nutrients (not as a proportion of the RDAs) is derived from program meals. Home-delivered participants’ percentage of total daily intake of these nutrients from program food is slightly higher, at 47 percent.

These findings on mean percentage of daily intake from program meals suggest that program meals are an important part of daily nutrient intake for a large number of participants. This is confirmed by other evidence in the characteristics survey. For example, when asked how important the meal program is as a source of food, nearly half (45 percent) of congregate participants reported that the program is their major source of food.

TABLE III.3

PERCENTAGES OF PARTICIPANTS’ TOTAL DAILY INTAKE FROM ALL PROGRAM MEALS

	Title III Congregate Meal Participants			Title III Home-Delivered Meal Participants
Nutrient	Mean	Median		Mean	Median
Food Energy	43.6	43.4		39.4	40.1
Protein	49.1	49.6		46.8	49.3
Vitamin A	50.3	49.0		48.7	51.4
Vitamin C	44.7	40.5		42.6	39.8
Vitamin D	46.9	45.8		44.5	43.9
Vitamin E	47.3	47.6		43.8	43.8
Thiamin	39.3	37.0		37.0	36.3
Riboflavin	41.5	40.0		39.1	38.4
Niacin	43.6	43.0		40.0	41.3
Vitamin B6	43.8	41.6		41.7	41.7
Folate	39.6	37.5		36.2	33.1
Vitamin B12	49.1	48.4		46.5	45.4
Calcium	45.7	45.3		44.6	45.1
Iron	40.2	39.0		35.6	34.9
Phosphorous	45.8	44.8		43.7	45.5
Potassium	46.3	45.6		43.3	44.4
Magnesium	42.9	41.5		39.8	40.2
Zinc	47.5	46.9		44.5	46.1
Sodium	47.6	46.2		42.5	42.6
Dietary Cholesterol	49.2	47.0		45.9	44.8
Unweighted Sample Size	1,039	1,039		815	815

Source: Elderly Nutrition Program Evaluation, Participant survey, weighted tabulations.

Notes: Tabulations are weighted to be representative of a cross-section of participants receiving Title III meals on a given day. Includes 89 participants who received but did not consume a program meal during the 24-hour recall period (23 congregate participants who attended the meal site and usually eat a program meal but did not that day because of medical tests or other reasons; 66 home-delivered meal program participants who received a program meal but chose not to eat the meal during the recall period, saving it for another time).

Comparisons with Previous Studies. The findings on dietary intake from program meals and the contribution of program meals to participants’ overall total daily intake summarized in the previous sections are consistent with those reported in earlier evaluations of the Title III meals program.

Using participants’ dietary intake between 11 am and 4 pm as a proxy for their intake from program meals, Kirschner et al. (1983) found similar percentages of participants consuming one-third of the RDAs for selected nutrients as the current evaluation. [ To be comparable to the Kirschner (1983) results, the results reported in Table III.4 for participants in the current evaluation include only those participants who consumed a program meal during the recall period. Thus, the unweighted sample sizes in Table III.4 are less than those reported in the previous tables of this chapter, which included participants who received but did not consume a program meal. ] For example, the current evaluation found that 88 percent of congregate participants who consumed a program meal had intakes per program meal that provided at least one-third of the RDA for protein, compared with 87 percent of congregate participants surveyed by Kirschner et al. in 1981 (Table III.4). The current evaluation found that 81 percent of home-delivered participants received one-third or more of the RDA for protein from a program meal, compared with 82 percent of home-delivered participants in the Kirschner et al. study.

Similar to the current evaluation, three previous studies found that Title III program meals contributed substantially to participants’ total daily dietary intake (Caliendo 1980; Harrill et al. 1981; and Kohrs et al. 1978). Similar to the current evaluation findings, all three studies indicated that congregate and home-delivered meal program participants consumed an average of 40 percent or more of their total daily nutrient intake during the program meal (results not shown). [ The studies cited involved single area or local sites. In addition, the tabulations in each of these studies were based only on samples of participants that ate a program meal during the 24-hour period, whereas the tabulations reported for the current evaluation also include participants who received a program meal but did not consume it during the 24-hour period.]

TABLE III.4

PERCENTAGE OF PARTICIPANTS WHOSE DIETARY INTAKES FROM PROGRAM MEALS PROVIDED AT LEAST ONE-THIRD OF THE RDA: COMPARISON BETWEEN CURRENT AND PREVIOUS NATIONAL EVALUATIONS

(Includes Only Participants Who Consumed a Program Meal)

	Title III Congregate Meal Participants			Title III Home-Delivered Meal Program Participants
Nutrient	Current Evaluation a	Previous Evaluation b		Current Evaluation a	Previous Evaluation b
Food Energy	49	53		36	48
Protein	88	87		81	83
Vitamin A	65	55		67	50
Vitamin C	68	59		56	52
Thiamin	78	70		71	67
Riboflavin	80	78		79	75
Niacin	85	73		77	66
Calcium	58	51		65	50
Iron	74	75		65	67
Unweighted Sample Size	1,016	800		749	340

Source: Elderly Nutrition Program Evaluation, Participant survey, weighted tabulations; Kirschner et al. (1983).

Note: Participant tabulations for the current evaluation are weighted to be representative of a cross-section of participants receiving Title III meals on a given day.

^a Refers to the intake of nutrients from program meal sources during the 24 hours as reported by participants who ate a program meal during the recall period.

^b Refers to the intake of nutrients during the 11 am to 4 pm period as reported by participants who ate a program meal during the recall period. Kirschner et al. used intake during the 11 am to 4 pm period as a proxy for intake from program meal sources because they did not ask participants about meal sources. They reported that all of the surveyed meal programs and a majority of programs in general served their meals during these hours.

2. Social Contacts and Activities from Meal Program Sources

In addition to providing nutritious meals, a major goal of the ENP is to reduce the social isolation of elderly persons. The congregate meals program, through provision of group dining, recreation, and other activities, affords opportunities for social interaction and companionship. Although much more limited, the home-delivered meals component also provides an opportunity for an important social contact: the contact between the participant and a staff person or volunteer who delivers the program meal to the participant’s home.

Under a broad definition of "social contacts," congregate participants average 95 social contacts per month, and home-delivered meal program participants average 100. [ Social contacts were defined to include talking on the telephone; visiting or being visited by relatives, friends, or neighbors; attending church or religious services; attending clubs or other organizations; attending congregate meal sites for meals and/or recreation services; receiving home-delivered meals from the meal program; and receiving other home or community-based long-term care services, such as personal care, homemaker, home health, and attending adult day care programs. ] Program sources represent, on average, approximately 25 percent of Title III congregate and home-delivered participants’ total social contacts per month (Table III.5). [ Program sources of social contacts refer to attending congregate meal sites for meals or recreation, and having contact with staff or volunteers who deliver the home-delivered meal.] For 11 percent of congregate participants and 13 percent of home-delivered participants, social contacts afforded by the meals program account for 50 percent or more of their total monthly activities and social contacts.

For home-delivered participants, social contacts from program sources are exclusively contacts they have with program staff when the meal is delivered to them. These contacts tend to be limited: 75 percent of home-delivered meal program participants reported that the delivery person leaves immediately, whereas 25 percent reported that the delivery person spends some time to talk with or check on them (not shown). Regardless of the length of the contact, home-delivered meal program participants value it highly. For example, when asked to mention the things they like about the meals program, 59 percent of home-delivered participants reported that they like the contact with the delivery person, and 98 percent reported that the meal delivery person is usually pleasant.

Congregate participants avail themselves of the opportunities for social interaction and activities provided by the meal program. Ninety percent reported that they typically spend more than one hour at the meal site when they attend. Forty-seven percent reported participating at least once a week in recreation activities provided at the meal program (not shown).

TABLE III.5

PARTICIPANTS’ TOTAL MONTHLY SOCIAL CONTACTS FROM PROGRAM SOURCES

(Percentages, Unless Stated Otherwise)

Social Contacts	Title III Congregate Meal Participants	Title III Home-Delivered Meal Participants
Total Number Per Month from All Sources
Mean	95.4	100.5
Median	74.1	85.9
Proportion from Program Sources (Percent Distribution)
1 to 10 percent	16	11
11 to 20 percent	28	33
21 to 30 percent	21	21
31 to 40 percent	16	14
41 to 50 percent	9	8
51 to 75 percent	8	10
More than 75 percent	3	3
Mean	27.4	28.6
Median	22.6	23.6
Unweighted Sample Size	1,040	818

Source: Elderly Nutrition Program Evaluation, Participant survey, weighted tabulations.

Note: Tabulations are weighted to be representative of a cross-section of participants receiving Title III meals on a given day. Social contacts include talking on the telephone; visiting friends, relatives, or neighbors; attending church or religious services; attending clubs; attending congregate meal sites; and having contact with program person who delivers home-delivered meal and with providers of personal care services, such as home health, homemaker, and adult day care. Program social contact sources refer to attending congregate meal sites for meals or recreation, and having contact with the person who delivers the home-delivered meal.

This section examines the contribution of the Title III meal program to participants’ daily dietary intakes and social contacts by comparing participants and a matched comparison group of program-eligible nonparticipants on the mean values of these outcomes, using multivariate regression methods. [ Appendix F contains tables showing the simple differences in mean values of 24-hour dietary intakes and number of monthly social contacts for Title III congregate participants and congregate-eligible nonparticipants, as well as home-delivered participants and home-delivered eligible nonparticipants, controlling for race and ethnicity, gender, income, and disability via constructed weight variables. Two sets of nonparticipant weight variables were derived, one corresponding to each Title III participant group. Their derivation is described in Appendix C on weighting. The simple differences in mean values, controlling for participant-nonparticipant differences in the above-mentioned demographic characteristics via the weight variables, are essentially the same as the regression-adjusted results reported in the text.] As a context for assessing the findings, it is important to understand how the comparison group of eligible nonparticipants was identified, as well as how the statistical comparisons between participants and eligible nonparticipants were conducted.

1. Research Methods

Our objective in assessing the impacts of participation in the Title III program is to answer the question, "What would participants’ dietary intake and socialization be had these individuals never participated, and how do these outcomes compare with outcomes of participating?" The purpose of a comparison group of eligible nonparticipants is to represent what would happen to participants in the absence of the program. The comparison sample of nonparticipants should ideally be as similar as possible to the sample of participants, except for program participation and random variation.

The preferred type of comparison group is achieved under an experimental research (or random assignment) design. Under this design, program-eligible older individuals who want to participate in the program would be randomly assigned to either a "treatment" group, which receives nutrition services from the program, or to a "control" group, which does not. If the randomization of program eligibles is executedproperly, then at the time of the randomization, the control group would not differ in any systematic or unmeasured way from the treatment group, on average. Under this design, subsequent observed differences in the mean values of outcomes between participants and nonparticipants can be attributed to participation in the program with known statistical confidence.

Random assignment was not possible in the current evaluation. The absence of random assignment of individuals to "program participant" or "program nonparticipant" status created significant challenges in assessing the effects of the program. In response, we selected a sample, from the Health Care Financing Administration (HCFA) Medicare Beneficiary File, of nonparticipants in the same locations as participants, in which the nonparticipants were matched with participants in terms of key variables. This was the next best alternative to a randomized control group of nonparticipants. Multivariate regression techniques were then used to compare program participants and eligible nonparticipants on dietary intakes and social contacts, controlling for characteristics that could be related to both program participation and the outcomes studied.

a. Identifying Program-Eligible Nonparticipants

We used the HCFA Medicare Beneficiary File to identify program-eligible nonparticipants. [ During the design phase of the project, we considered using as the source of the comparison group eligible individuals who were on program waiting lists. Individuals on waiting lists, who would like to receive program services but cannot because of program resource constraints, seemed more likely to be similar to program participants on both measured and unmeasured characteristics than a nonparticipant group selected from the community. However, there are problems with using program waiting lists: (1) not all meal programs, particularly congregate ones, have waiting lists; and (2) individuals on waiting lists may be different from those receiving meals because of the criteria sites use to determine who gets on the list, and, once on the list, who gets served first. For these and other reasons, MPR and the evaluation 's technical advisory group concluded that using the Medicare Beneficiary File was preferable to a combined strategy of using program waiting lists when available and screening from the general population when not.] Samples of congregate- and home-delivered-eligible nonparticipants were selected from the file in the zip code areas covered by the meal sites and delivery routes selected for the participant samples. MPR requested andobtained from HCFA the names and addresses of all Medicare beneficiaries with addresses in these zip codes; the names of elderly beneficiaries were randomized within each zip code. Medicare beneficiaries were then selected for screening in the order in which they appeared on the sorted lists. Potential nonparticipants were screened by telephone for age, income, and disability status and for program participation to make sure they were not participating in either the congregate or home-delivered program. The screened sample was then stratified by income and disability status. Random samples were selected in a manner that ensured the participant and nonparticipant samples would have approximately the same distribution of income and disability status. [ The six income/disability cells were (1) poor, nonfrail; (2) near poor, nonfrail; (3) nonpoor, nonfrail; (4) poor, frail; (5) near poor, frail; and (6) nonpoor, frail. Poor refers to income less than 100 percent of the DHHS poverty guidelines; near poor refers to income between 100 percent and 200 percent of the DHHS poverty guidelines, and nonpoor refers to income greater than 200 percent of the DHHS poverty guidelines. Individuals were classified as disabled if they had either mobility or self-care impairments, and as not impaired if they did not. Cells 1 through 3 define congregate eligible, and cells 4 through 6 define home-delivered eligible older people.]

b. Multivariate Analysis of Differences Between Participants and Eligible Nonparticipants

Despite efforts to identify a group of eligible nonparticipants who were comparable to participants across several critical individual characteristics related to outcomes, the characteristics of the two samples differed. [ If we consider only the socioeconomic characteristics of the unweighted samples of participants and nonparticipants, Title III program participants tend to be older, to have completed fewer years of formal education, to be more racially mixed, and to be more likely to live alone and be female. Program participants ' incomes tend to be much lower than nonparticipants ' incomes, and a far greater percentage of Title III participants take part in other food assistance programs. That participants are, on average, older is partially a product of the study design: reflecting the age eligibility rules for receipt of Medicare, the Medicare Beneficiary File consists of elderly persons age 65 and older, whereas program participants are age 60 and older (and may be under age 60 if married to a participant age 60 or older). ] Consequently, our analyses used statistical methods to control for differences in the characteristics of participants and nonparticipants that affect outcomes and may be correlated with program participation.

Basic OLS Regression. The basic approach used for estimating differences in mean values on dietary intake and socialization outcomes between program participants and nonparticipants involves estimation of a linear model of the form:

(1)BOLD ITAL {Y SUB {i}~=~ b SUB {k}`` X SUB {ki}~ +~ c CMP SUB {i}~ +~ d HDMP SUB {i}~ +~ e SUB {i}``,}

where:

I="individuals"

Y_i =the outcome of interest (for example, intake of food energy relative to the RDA)

X_ki =a matrix of person-specific characteristics that are thought to affect the outcome of interest [ The following person-specific characteristics ( X ki ) were used in the OLS regressions: age; gender; minority status; mobility limitations, as measured by the number and types of activities of daily living and instrumental activities of daily living that the person had difficulty doing; income; self-reported health status; whether the person has hypertension, high blood cholesterol, or had a stroke; whether the person takes vitamin supplements; educational attainment; marital status; whether the person lives with other family members; physical activities in the prior month; self-reported appetite; and whether the person owns a microwave. We also included indicators for whether the person was on a general diabetic diet or one designed for low-salt, low-cholesterol, low-sugar, low-fat, high-fiber, or lactose-free intake.]

b_k =a vector of unknown regression coefficients (parameters) to be estimated, each of which shows the change in the outcome variable resulting from a unit change in the corresponding regressor variable X_ki

CMP_i = a binary variable that equals "1" if the ith individual participates in the Title III congregate meals program, "0" otherwise

HDMP_i = a binary variable that equals "1" if the ith individual participates in the Title III home-delivered meals program, "0" otherwise

NP_i = a binary variable that equals "1" if the ith individual is a nonparticipant, "0" otherwise (omitted binary variable)

c = an unknown regression coefficient to be estimated that measures the difference in the outcome between congregate participants and the omitted category nonparticipantsd = an unknown regression coefficient to be estimated that measures the difference in the outcome variable between home-delivered meal program participants and the omitted category nonparticipants

e_i = a stochastic error term

The combined unweighted sample of Title III congregate participants, home-delivered participants, and nonparticipants was used to get consistent estimates of the regression coefficients. This process was repeated separately for the 24-hour intake of food energy, for each of the other nutrients, and for the total number of social contacts per month. [ As will become apparent during the discussion of selection bias in the next section, estimation of the single linear equation implicitly assumes that any existing unmeasured differences between program participants and eligible nonparticipants do not systematically affect the outcomes being studied.]

The estimation results are summarized in Tables III.6 and III.7. Consider the comparisons for congregate and congregate-eligible nonparticipants first. For each outcome, we present the adjusted mean value for Title III congregate participants and then the adjusted mean value for congregate-eligible nonparticipants, as well as the difference in the mean values on the outcome between the two groups expressed as a percentage of nonparticipants’ mean outcome. We also indicate whether the difference is statistically significant at conventional levels, assuming a one-tailed test. [ One-tailed hypothesis tests are used in the analyses of all outcomes because the expectation a priori is that program participation improves dietary intake and socialization opportunities. The sharper hypothesis test in each case is that the differences in mean values between participants and nonparticipants are positive (or negative, in the case of dietary cholesterol, sodium, or fat intake, given meals are to conform to the Dietary Guidelines ), as opposed to the differences simply being different from zero.] ^, [ Standard errors and significance tests were adjusted to account for the complex sample design. There are two factors, working in opposite directions, that affect the adjustment. Because we are using a clustered sample in which individuals are selected from a limited set of nutrition projects selected at an earlier stage, there should be an upward adjustment in the standard errors. On the other hand, the sample is also stratified, in that a certain number of projects were selected in each census region on the basis of the elderly population in that region, which leads to a downward adjustment. A separate adjustment factor or "design effect " was calculated for the 18 nutrients studied and for congregate and home-delivered meal participation. The adjustments ranged from a 56 percent decline in variance to a 126 percent increase in variance, with the average adjustment being a 22 percent increase in variance. For the subgroup analysis, the design effect adjustments are somewhat smaller than the adjustments we obtained for the overall impact regressions. See also the discussion of design effects in Volume III, Appendix D.] The adjusted mean value for a particular outcome is calculated by inserting the weighted mean values of Title III congregate participants for the values of each covariate and multiplying the mean of each covariate by the corresponding estimated regression coefficient. The weighted mean values for congregate participants are used for deriving the adjusted mean outcomes for both congregate participants and congregate-eligible nonparticipants. The difference in mean values for a specific outcome is the estimated value of coefficient c in the linear model. A similar process is used to calculate the adjusted mean outcomes for home-delivered participants and eligible nonparticipants. The only differences are that the mean values for the covariates used in the equation are the weighted mean values of Title III home-delivered participants, and the difference in mean values for a specific outcome is the estimated value of the coefficient d in the linear model.

TABLE III.6

REGRESSION-ADJUSTED COMPARISON OF MEAN DAILY NUTRIENT INTAKES AS A PERCENTAGE OF THE RDA FOR PARTICIPANTS AND ELIGIBLE NONPARTICIPANTS

(Regression-Adjusted Means)

	Title III Congregate Meal				Title III Home-Delivered Meal
Nutrient	Participants	Nonparticipants	Percent Difference		Participants	Nonparticipants	Percent Difference
Food Energy	78.8	71.6	10.0***		70.7	67.5	4.8**
Protein	122.7	111.9	9.6***		111.7	105.7	5.6**
Vitamin A	150.7	119.8	25.8***		141.8	117.6	20.6**
Vitamin C	165.7	140.3	18.1***		148.0	139.0	6.4*
Vitamin D	107.0	84.4	26.8***		109.4	83.5	30.9***
Vitamin E	91.0	76.5	19.0***		74.2	72.5	2.4
Thiamin	140.7	131.3	7.1**		129.8	126.4	2.7
Riboflavin	141.5	122.7	15.3***		137.5	119.3	15.3***
Niacin	139.5	129.4	7.8***		125.0	122.1	2.4
Vitamin B6	97.3	88.4	10.2***		89.8	84.1	6.8***
Folate	140.3	129.0	8.7***		131.1	122.6	6.9**
Vitamin B12	203.4	185.6	9.6		213.5	180.2	18.5***
Calcium	92.6	74.9	23.7***		91.1	72.8	25.1***
Iron	130.5	125.3	4.2*		124.5	119.6	4.1
Phosphorous	135.2	117.3	15.2***		126.6	112.5	12.6***
Potassium	130.7	112.1	16.6***		116.3	105.3	10.5***
Magnesium	87.3	75.4	15.8***		79.2	70.9	11.8***
Zinc	75.2