Understanding the Blood Donor Problem: a Discriminant Analysis of Active, Dormant, and Non-Donors

Raymond J. Smead, Texas Tech University
John J. Burnett, Texas Tech University
ABSTRACT - The recruitment of people to donate blood has remained a serious problem. It is suggested that a weak methodological tradition may be a part of the reason why past research recommendations have proved unsuccessful. This study attempts to deal with these research weaknesses by examining three donor groups rather than two, across psychographic and demographic variables. The results indicate that sex and social consciousness were the strongest discriminating variables. Even more interesting, the relationship was not linearly related to the frequency of donation.
[ to cite ]:
Raymond J. Smead and John J. Burnett (1980) ,"Understanding the Blood Donor Problem: a Discriminant Analysis of Active, Dormant, and Non-Donors", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 622-626.

Advances in Consumer Research Volume 7, 1980     Pages 622-626

UNDERSTANDING THE BLOOD DONOR PROBLEM: A DISCRIMINANT ANALYSIS OF ACTIVE, DORMANT, AND NON-DONORS

Raymond J. Smead, Texas Tech University

John J. Burnett, Texas Tech University

ABSTRACT -

The recruitment of people to donate blood has remained a serious problem. It is suggested that a weak methodological tradition may be a part of the reason why past research recommendations have proved unsuccessful. This study attempts to deal with these research weaknesses by examining three donor groups rather than two, across psychographic and demographic variables. The results indicate that sex and social consciousness were the strongest discriminating variables. Even more interesting, the relationship was not linearly related to the frequency of donation.

INTRODUCTION

Obtaining adequate supplies of blood from donors is a problem of considerable concern for the health care industry. Voluntary donations do not meet requirements and paid donors must be used to augment supplies. Unfortunately, as Oswalt (1977) notes, the risk of hepatitis from commercial blood is at least 11 times higher than for blood supplied by volunteers; 17,000 cases of hepatitis result from blood transfusions each year and one in 20 of these die. The effort that has been directed toward creating a national system for voluntary blood donations has appeared to flounder.

Myriad studies and surveys have addressed the problem. Many have used the approaches of either (1) understanding the motivations of donors or the impediments to non-donors, or (2) identifying donors demographically or with socioeconomic variables. Proposed remedies then provide proper marketing appeals to the potential donor groups. Although effective implementation of recommendations may be part of the problem, it appears that the incorrect identification of motivations and the resulting recruitment approaches has led to the poor success in achieving an all volunteer system.

As noted by Oswalt (1977), two problems may be troubling the blood donation research tradition; (1) a possible distorted appreciation of donor motivations due to weak methodologies, and (2) a failure to understand and address the problem of donor turnover. For example, studies of donor motivations often rely on checklists or other direct questioning techniques and most often base the findings on convenience samples of donors rather than on comparisons of donors to nondonors. Oswalt (1977), in a review of the recruitment and motivation literature, notes that most of these studies find altruism/humanitarian motivation given most often as the reason for donating. However, there is reason to doubt the validity of this conclusion because of possible response biases from using self report questioning in this emotion laden area. An indication of this is given in a study by Bartel et. al. (1974) comparing direct (self) and indirect (others) probing of negative attitudes toward donation. Respondents were considerably more likely to ascribe personal and emotional reasons for not donating to other than to themselves. Other evidence comes from a study by Condie et. al. (1976) where little or no association was found between donor-ship and altruism. Their study compared donors and non-donors on a number of psychological scales including a modified version of Wrightman's altruism scale (1964), along with ad hoc measures of free-rider tendency, incentives, and cost, but did not use direct probing of motivations. They did find donors to be more responsive to social pressure and incentives and less impeded by a free-rider tendency and the cost of giving than non-don-ors. Condie et. al.'s sample (1976) was drawn from two populations of college students, who are more likely to donate, and do not represent the general population.

The donation literature's emphasis on donor recruitment may stem from a failure to appreciate the magnitude of donor turnover. Donor studies commonly cite the statistic that only three percent of individuals eligible to donate do so. However, an analysis by Miller and Weikel (1974) indicates that the true rate may be eight or nine percent. Several surveys, while not representative of the generation population, have also found the rate to be much higher (Miller and Weikel, 1974) Bettinghaus and Milkovich, 1974; Oswalt and Hoff, 1975) and have found that a large portion of people donate only once. Oswalt and Huff (1975) found 63 percent of male donors and 89 percent of female donors in their sample have become inactive.

The problem of turnover may be somewhat confounded by the fact that the number of meaningful donor categories has never been settled. Most prior research has tended to either exam donors alone or compare donors with nondonors. Yet, the findings presented by Miller and Weikel (1974) indicate that the donor category is not homogeneous and active donors tend to differ from dormant donors across several demographic variables.

This study applies the above two considerations by comparing the three donor groups--active, dormant, and non--on a number of demographic, socioeconomic and psychographic variables. These variables, listed in Table 1, include many items used in other studies, but do not include direct probing of motivations and impedances, instead, psychographic variables from the adoption literature (Gorman, 1967; Robertson, 1971) and from preventive health behavior studies (Kirscht et. al., 1966; Douglas, 1971; Zaltman and Vertinsky, 1971; Biblo, 1972) were used to compare the three groups.

TABLE 1

VARIABLES ANALYZED

Discriminant analysis was used to identify factors to explain why donors donate or become dormant, i.e., factors differentiating the three groups. This analytical model is an appropriate multivariate technique for examining a nominal dependent variable such as donor class.

METHOD, FINDINGS, AND DISCUSSION

Data for this study were obtained from a middle-sized southwestern city. Families in this panel are largely lower-middle and middle class. The sample was not completely representative, but deviated from the general population profile in the direction of what has been found to be the donor profile - above average in income and education and underrepresenting non-whites (Bettinghaus and Milkovich, 1975). Out of the 250 questionnaires mailed, 173 responses were obtained. After deleting subjects with item nonresponse on key variables, 161 usable observations were obtained for the discriminant analysis. Donors were classified as dormant if they had not given in the preceding year. There were 22 active, 54 dormant, and 85 nondonors.

The psychographic variables were operationalized through items found in the AIO library developed by Wells and Tigert (1971). The altruism variable was adapted from several sources and customized for this study (Wrightsman, 1964). All the statements were measured with a six point scale ranging from 1 (does not describe me at all) to 6 (describes me perfectly). Two of the scales--altruism and social consciousness--each had items which did not seem to consistently reflect the dimension. Accordingly, the items were factor analyzed using an oblique rotation to determine if the original scales would be reproduced. An oblique rotation was used because it was expected that the two scales would not be independent. The factor loadings did not reproduce the original scales, but produced three factors which were labeled social involvement, duty consciousness and cynicism. Items used in the altruism, social consciousness and social involvement scales, along with the other variables entering into the step-wise discriminant program are shown in Table 2. Items with factor loadings ranging from .62 to .22 were added together to form the new scale. The use of the low loadings was prompted by the exploratory nature of this study. Items were not differentially weighted so as to be consistent with the way the original scales were formed. Both the original and the new variables were included in the discriminant runs.

Discriminant analysis finds linear combinations of variables (dimensions) such that observations positioned on the dimensions have minimal separation within groups and maximal separation between groups. When more than one dimension is produced, groups are positioned in the multidimensional space or map. Results may be interpreted in two ways. First, discriminant analysis yields a function for each group that gives likelihoods of an observation being a member of each group. The functions can be used to classify observations as belonging to each group. This can be done with observations where the true group membership is known so as to evaluate the effectiveness of the analysis or where membership is not known so as to predict membership. The classification functions are shown in Table 3. Optimal classification is not the major purpose of this study. No interpretation is made of the functions.

Second, an analysis can be made by interpreting the dimensions of the map with respect to group location on the map. This is the primary purpose of our study. It requires an understanding of what the dimensions mean and has the same problem of interpretation as factor analysis. Interpretation is particularly difficult with the data set used here because the variables are not all of the same "level". Researchers using factor analysis virtually always use variables from a single conceptual domain such as attitude items. This data set mixes demographic variables with psychological scales and requires a speculation as to what factor underlies both classes of variables rather than simply a condensing of variables within a consistent set.

TABLE 2

ITEMS EMPLOYED IN SELECTED SCALES

TABLE 3

CLASSIFICATION FUNCTIONS

The SPSS Discriminant Procedure with the Wilks' lambda Criterion was used to analyze the data. As noted in Table 5 only the first discriminant function of the two possible was significant. This may be due to the small sample size employed which may also account for the small number of variables that entered the equation. Because this study was exploratory in attempting to find useful new variables from the adoption and health care literature, variables were allowed to enter the functions with an F value as low as 1.25. Of course, this magnifies the possibility that one or more variables entered is spurious. But since this is an exploratory study, we are at least as concerned with avoiding Type II errors as Type I errors. The reader should note that this second function may only be noise. The classification results (Table 4) indicate that 55 percent of the observations were correctly classified and that for each group an observation was more likely to be assigned correctly than to be assigned to each of the other groups. These results can also be compared to the proportional chance criterion, suggested by Morrison (1969) in this type of analysis. In this study the chance criterion was 42%. These results are rather good considering the possible obscurity of factors distinguishing donors from dormant and non-donors. Of course, just as with the unadjusted R2 statistic in regression analysis, the correct classification percentage will be inflated somewhat due to chance. No holdout sample was used because, given the nature of the study and the small size, we thought it more useful to employ the observations in getting better coefficient estimates and improving the power of analysis than to use split runs to get an unbiased estimate of the classification percentage and an estimate of the stability of the coefficients.

The primary dimension of the discriminant map presents a truly startling and potentially very valuable finding: the three groups are not ordered along this dimension from active to dormant to non-donor but from dormant to active to non-donor. Thus, the main differentiating factor is not simply a propensity to donate as one might have expected. Understanding exactly what the factor is, is not as easy. Sex is the strongest variable on this dimension. This is not surprising as virtually every study that has examined sex has found that donors, active or dormant, are predominantly male. Sex is the only demographic to enter, which is surprising since other studies have found donors, active and dormant, to be more commonly white and to have higher income and education and to have more prestigious jobs (Bettinghaus and Milkovich, 1975). See Figure 1.

TABLE 4

PREDICTED CLASSIFICATION

The social involvement variable was also strong and is the main basis for the speculation on the meaning of the dimension. (Table 6 shows groups means for each of the scale variables). The variable suggests a tendency to make a sacrifice for the community, to be concerned and to make an effort to right wrongs (see Table 2). At the risk of over-speculating about non-significant variables, the dormant donor may have a tendency toward a traditional sex role orientation.

Perhaps dormant donors have stopped donating because they met their responsibility to the community with either a single or a few donations. This is indicated by the frequency of donation of active and dormant donors. Eighty-eight per cent of our dormant donors have given three or less times. This seems a credible explanation of the donor turnover problem.

The first dimension suggests the existence of a second; if the first dimension is not a propensity to donate (or keep on donating) then something else must be. This is why the second function is included despite its low significance level. The second dimension is more of a riddle: what is "not safety conscious" and is "an opinion leader"? The safety conscious variable is obvious, but opinion leadership is not. Perhaps opinion leaders are not just spokesmen but also embody community virtues or the willingness to set examples. If the second dimension is propensity to become a regular donor then it may incorporate elements of lessened body impairment apprehensions together with a willingness to be responsible for community virtues. But that is quite speculative.

SUMMARY AND CONCLUSIONS

A comparison has been made of active, dormant and non-donors on some traditional demographics and on several psychographic variables never before used in this area. The sample size makes this somewhat of a pilot study, but the data does yield some intriguing findings. Getting people to donate once or twice seems to require a social involvement propensity, but a small number of acts of giving may satisfy that psychological need. Continued donation seems to require a lessened health apprehension and possibly a sense of responsibility toward community standards.

Although the interpretation remains somewhat vague, it appears that future research in this area should employ at least three donor categories and a multivariate form of analysis. The inclusion of additional variables would also be a viable research direction. Nevertheless this study has successfully identified some unique findings in the troubled area of blood donation.

TABLE 5

DISCRIMINANT FUNCTIONS

FIGURE 1

DISCRIMINANT MAP AND GROUP CENTROIDS

TABLE 6

DONOR GROUP MEANS FOR DISCRIMINANT FUNCTION SCALES

REFERENCES

Bartel, W. P., Stelzerm, W., and Higgins, J. (1974) "Attitudes Underlying Reluctance to Donate Blood," Transfusion, 15, 275-77.

Bettinghaus, E. P. and Milkovich, M. B. (1975) "Donors and Nondonors: Communication and Information," Transfusion, 25, 165-69.

Biblo, Robert L. (1972) "Marketing and Enrollment Strategies for Prepaid Group Practice Plans." Marketing Pre-Paid Health Care Plans, by U. S. Department of Health, Education, and Welfare, 5-40.

Coleman, James, Katz, Elihu, and Menzel, Herbert. (1975) "The Diffusion of an Innovation Among Physicians," Sociometry, 20, 253-70.

Condie, S. J. and Maxwell, N. (1970) "Comparative Demographic Profiles: Voluntary and Paid Donors," Transfusion, 10, 84-88.

Condie, S. J., Warner, W. K. and Gillman, D. C. (1976) "Getting Blood from Collective Turnips: Volunteer Donation in Mass Blood Drives," Journal of Applied Psychology, 61, 290-294.

Douglas, Chester W. (1971) "A Socio-Psychological View of Health Behavior for Health Services Research." Health Services Research, 6, 6-14.

Ford, G. T. and Wallace, E. L. (1975), "Effects of Donor Recruitment Methods on Population Responses," Transfusion, 15, 159-164.

Gorman, Walter P. (1967) "Analyzing Consumer Markets Through Diffusion Research." Rural Sociology, 32, 456-463.

Kirscht, John R., Haefner, Don P., Kegeles, Stephen S. and Rosenstock, Irwin M. (1966) "A National Study of Health Beliefs," Journal of Health and Human Behavior, 7, 248-254.

Leibrecht, B. C., Hogan, J. M., Luz, G. A., Tobias, K. I., and Camp, F. R. (1973), A Survey Inventory of Donor and Nondonor Motivations. United States Army Medical Research Laboratory, Fort Knox.

London, P. and Hemphill, B. M. (1965) "The Motivations of Blood," Transfusion, 9, 559-568.

Metzner, Charles A. and Bashshur, Rashid L. (1967) "Factors Associated with Choice of Health Care Plans," Journal of Health and Social Behavior, 8, 156-163.

Miller, T. and Weikel M,. (1974) "Blood Donor Eligibility, Recruitment, and Retention in the United States," Transfusion, 14, 612-618.

Morrison, Donald G. (1969) "On the Interpretation of Discriminant Analysis," Journal of Marketing Research, 6, 156-163.

Myhre, B. A. (1976), "Blood - Legend Laden Liquid of Life," Transfusion, 16, 270-275.

Nie, Norman H. (1975), Statistical Package for the Social Sciences, 2nd ed. McGraw-Hill Book Company.

Oborne, D. J. and Bradley, S. (1975) "Blood Donor and Nondonor Motivation: A Transnational Replication," Journal of Applied Psychology, 60, 409-410.

Oswalt, R. M., (1977) "A Review of Blood Donor Motivation and Recruitment," Transfusion, 17, 123-134.

Oswalt, R. M., Hoff, T. E. (1975) "The Motivations of Blood Donor and Nondonors: A Community Survey," Transfusion, 15, 68-72.

Reynolds, F. D. and Darden, W. R. (1971) "Mutually Adaptive Effects of Interpersonal Communications," Journal of Marketing Research, 8, 449-454.

Robertson, Thomas S. (1971) Innovative Behavior and Communication. New York: Holt, Rinehart, and Winston, Inc.

Sherer, Peter B. (1975) "Blood Donor Motivation: A Background Paper for the NHLI Conference on Behavioral Science and Medicine. NHLI Conference Proceedings.

Tigert, D. J. and Wells, W. D. (1970) "Life Style Correlates of Age and Social Class," paper presented at the First Annual Meeting, Association for Consumer Research, Amherst.

Titmuss, R. M. (1971) The Gift Relationship: From Human Blood to Social Policy, New York: Pantheon Books.

Wells, W. D. and Tigert, D. G. (1971) "Activities, Interest, and Opinions," Journal of Advertising Research, 11, 27-35.

Wrightsman, L. (1964) "Measurement of Philosophies of Human Nature," Psychological Reports, 14, 743-751.

Zaltman, Gerald and Vertinsky, Ilan. (1971) "Health Service Marketing: A Suggested Model." Journal of Marketing, 35, 19-27.

----------------------------------------