General Research Design

In keeping with the directive allowing each state to determine the appropriate methodology in gathering information for establishing prevalence rates, we have elected not to adopt the conventional methods used in most CSAT-funded surveys-that is, personal interviews, conducted face-to-face or via telephone, with a sample of the population. Instead, for this study, as also for the one done on the Federated States of Micronesia, we have employed an indirect approach that makes use of key informants to obtain information on an entire community.

The methodology we have chosen is, admittedly, more problematic, and the epidemiological data may be challenged as corrupted. There is an obvious appeal in personal interviews done with a well-designed instrument: they are neat, simple to administer and can generate good figures.

Even so, we believe that the problems with the direct interview method in Micronesia outweigh its advantages. In many Pacific societies, personal interviews on life problem areas are not culturally appropriate and often yield information that is not reliable. Micronesians dislike talking about their own problems or those of other members of their family to outsiders. Whatever promises may be made, anonymity is impossible and confidentiality extremely rare in small island societies. In past surveys that have dealt with culturally sensitive subjects (eg, child abuse, suicide), we have had reason to suspect the reliability of information derived in direct interviews, or even with the individual's family. In these studies we have found it preferable to rely on information supplied by others in the community outside the immediate family. Hence, we have adopted the same data collection procedures for this report that we have successfully employed in other studies in Micronesia.

Another difference in approach is the way in which subjects were chosen. Rather than using random selection, as many epidemiological surveys do, we opted to survey everyone-men, women and children-in several preselected communities from each state. The communities were chosen with an eye to providing a good representation of the various sub-ethnic groups, religious backgrounds, and stages of relative acculturation as measured by position on the rural-urban residence scale. Each community selected contained between 200 and 400 persons, including children. All members of each household were screened for any alcohol and drug users, past or present, and individual forms were completed for all of those with a history of alcohol or drug use. The questionnaire form in the interviews was adapted from the core instrument prepared by the National Technical Center for Substance Abuse Needs Assessment (NTC). (The survey interview forms can be found in Appendix 3 of this report.)

Our preference for a community survey approach rather than random sampling is grounded in the social realities of a Pacific society. The community survey yields a picture of the behavioral patterns of a social group rather than an individual isolate. In a place like the Marshalls, with its enormous stress on social interaction and communal norms of behavior, we are likely to derive information that will better help us understand the etiology of the drinking and drug problem in the lives of individuals and suggest more effective forms of prevention and treatment.

The coded individual data were entered in the computer and a check was made against the projected population for 1995 to verify that the data represented at least 5 percent of the total population and that the distribution of the persons surveyed corresponded to the breakdown of the general population by state, gender, age-cohort, residence and ethnicity. Where there appeared to be significant over- or under-representation in any of these categories, adjustments were made to compensate for the differences before calculating prevalence rates and making projections on treatment.

All general population figures used for comparative purposes in this study were derived from the projected population figures for 1995 as found in the Marshall Islands Statistical Abstract for 1995. The figure given for 1995--55,575--is higher than this author's projected figure of 52,688 for two reasons: it takes no account of emigration, which has become statistically significant within the last nine years, and it has probably inflated the rate of population increase inasmuch as it uses the same 4.2 percent figure that was measured between 1980 and 1988. The government's projected population figure may fail to recognize the drop in fertility rates, while using annual growth rate figures that could have been inflated because of the undercount in the 1980 census. The author estimated the population of the Marshalls in 1997 to be 55,288. For convenience's sake, it seemed wise to use the figures for the age-sex breakdown from the official Marshall Islands projections for 1995 as provided in the Marshall Islands Abstract for 1995.

The survey instrument allowed us to generate two types of prevalence rates: lifetime prevalence and point prevalence within the last 12 months. The current prevalence rate is by far the more important and more reliable of the two, and it is this that will be presented in the tables. Where there is no indication to the contrary, the reader can assume that the twelve-month point prevalence rate is referred to in all figures. Where significant findings appear, lifetime prevalence rates will be given as well.

Selection of Representative Communities

The communities to be surveyed were chosen so as to provide a good balance of ethnicity, religion, and position on the scale of distance from modernization. The selection process aimed at picking a broad geographical range within a state.

Each community constituted a village or, if the village was too large to be surveyed in its entirety, an identifiable section of a village. (The ideal community size was established at between 200 and 400 persons of all age groups.) All households within this area were interviewed to avoid any hidden bias that might be at work in random selection.

According to the 1988 census, roughly two-thirds of the population of the Marshalls lived in the two urban centers of Majuro and Ebeye, with the remaining third distributed among the outer islands and atolls of the republic. In addition, Majuro's population was nearly twice the size of Ebeye's at that time. In this survey the sample size was set to conform to these ratios.

The communities sampled in each state and the sample size for each are given below. Additional background on the communities and their characteristics can be found in Appendix 2.

Sample Stratification

The interview data were checked against the population figures to ensure that the data represented a 5 percent sample. The interview data were then cross-checked by sex, age, ethnicity, and position on the urban-rural scale to determine whether it was representative of the percentage of the state population as presented in the FSM census. Where the data for sex and age was under-representative or over-representative, it was weighted accordingly in calculating any projections for the general population.

The study excluded all children below the age of 10. It was decided to make the cut-off point 10 rather than 15 since we believed that in using the latter age we would run the risk of eliminating boys and girls in their young teens who might be inhaling gas and glue. The results of the survey not only supported this belief but indicated that some of those in their early teens use alcohol as well.

The age matrix used in our tables has been adapted from the standard US matrix so as to conform better to the life-cycle of the Marshallese societies studied. The age cohorts used here are: 10-14, 15-19, 20-29, 30-44, 45-64, and 65+. A breakdown into five-year age cohorts in the teens is helpful since during these years many young islanders begin using drugs and alcohol. On the other hand, the age of 18 is not a significant boundary marker for Marshallese as it is for Americans. In many cases the most intense period of alcohol and drug use comes during the 20s. The next 15 years of life, between the ages of about 30 and 44, mark young adulthood and are a transitional time for many male islanders. Often a heavy drug user or drinker will modify his intake or cease altogether as he approaches the age of responsible maturity. By the age of 45 or so, a man is expected to attain full maturity and exercise control over the impulses that may have dominated his life as a young man.

The urban-rural spectrum can be divided into two categories: town dwellers (ie, in Ebeye or Majuro) and inhabitants of the outer atolls. Town dwellers are those who live in or near the port towns of Majuro and Ebeye, the commercial centers that enjoy a relatively modern living standard not found elsewhere. The coral atolls, which often lie hundreds of miles away from the towns, have the simplest life-style and the fewest amenities of modern life. Their contacts with the population centers are tenuous and infrequent.

Interview Methods

The field investigator, a Micronesian with long work experience in the FSM National Government, selected between two to four persons to serve as key informants for each community. The informants were Marshallese residing in the community who were familiar enough with the families to possess detailed information on all the members of the households and were willing to do so on the guarantee that they and the information they furnished would be held strictly in confidence. Church ministers and older persons with a high position in the community were excluded in favor of younger adults who might be more knowledgeable about the behavior of that segment of the population engaged in drug use. An attempt was made to include at least one female informant to ensure adequate coverage of women in the community.

Working with the field investigator, the informants completed a sheet on each household listing all the members of the household, their age and sex, their religion and ethnicity. Next to each name the informants indicated whether that person had ever used drugs in his/her life. This preliminary survey of the community served as a screen to identify individuals about whom more detailed information was to be collected.

Once the household survey forms were finished, an individual interview sheet was completed for any individual known to have been using alcohol or drugs at any time. In addition to basic biodata-sex, age, marital status, educational background, occupational status, and travel abroad-the sheet recorded detailed information on the type of drugs used, the frequency and extent of use, the seriousness of the problem, and the kind of treatment sought, if any. The interview protocol used was a heavily modified and abbreviated form of the core instrument designed by NTC.

It should be noted that the names were retained on these interview sheets, as well as in the computer files, so that any additional information that might be found from additional sources such as court records, police files, and case reports, might be added to the individual's record. We felt that names were necessary if we were ever to compile some "thick data" on individual users that might help us determine key factors that put persons at risk for substance abuse problems. Once the data collection was completed, however, the names were deleted from the files to maintain confidentiality.

Although we were confident that a key informant methodology was far more suited to the Marshalls than direct interviews, the reliability of third-person interview data remained a serious question. As a check on the reliability of our data, therefore, we conducted direct personal interviews with a small sample (10-12 persons) from each community. These persons were a convenience sample, but selected to include a male and female from each of the major age cohorts employed in this study. A random preliminary check was then made to determine the extent of discrepancies between the direct interviews and the third-person interviews. In this preliminary comparison of the interview results for ten persons, the correspondence of the data was strikingly close. There were no discrepancies at all in the reports on the type of drugs used and surprisingly little on the amount consumed; the greatest variation appeared in the reported frequency of drug use, with the third-person interview reporting a lower frequency than the direct personal interview.

Data Processing and Analysis

As the interviews in each community were completed, they were checked for completeness and consistency by the field investigator. When this was done, the survey results were entered on the computer in a dBASE IV file, with a field for each of the questions asked. A computer record was established for all individuals surveyed, even those who had no history of drug use of any kind, so that statistical tables could be more easily generated.

When all the computer entries for a state were completed, the computer entries were checked for keying errors through the use of the EpiInfo 6 statistical frequency function. When any errors were corrected in the dBASE IV file and the record numbers for each community were checked against the original interview forms, we began generating tables on the use of each drug. Tables showing age-sex distribution of current drug users were first generated, and then tables indicating the frequency and amount of the drug consumed.

After the tables were reviewed, the decision was made as to what correlations should be examined. Numbers and percentages for such correlations were generated from the dBase IV file, but further statistical operations to determine the confidence level and p-value were performed through the EpiInfo 6 program. These were used in the tables and narrative only to the degree that was deemed appropriate.

Projections of total current drug users on the island were made on the basis of the sex-age breakdowns of the survey data, since age and sex correlated more strongly with drug use than any other factors. Each age-sex group of persons surveyed was compared with the same group in the general population to derive the percentage of the sample before projections were made for the users in this category.

Collection of Other Data

Despite the reporting requirements that are built into most US federal program grants, data collection and maintenance remains uneven in the Marshalls. This hampered us in our attempt to gather data on social indicators of alcohol and drug abuse. Figures on alcohol imports by quantity and type were not available, and figures on the dollar amount of imports into the Marshalls could be obtained for only a few years. Deaths due to alcohol-related illness, accidents, homicide and suicide were recorded by the state departments of health services, but the criteria used in determining these deaths may have varied from one hospital to another. Moreover, the number of suicides recorded by the states was fewer in almost every case than the number generated by the author from the data-base he has maintained on suicide cases for the past twenty years.

In the end, we used whatever reliable data we could get on social indicators of drug and alcohol abuse. Where possible, we attempted to get figures for the past five years. Social indicators for which reasonably good data was found are: alcohol-related deaths (as recorded by the hospitals); arrests for alcohol-related crimes as a percentage of total arrests; and suicides occurring while under the influence of alcohol or drugs as percentage of total suicides.