CLOC Home

Summer Workshop and Data Release

Overview and Methodology

Data and Codebooks

Research Team

Research Projects and Publications

Frequently asked Questions




Skip to:
History of CLOC - Design and Method - Data Analysis - Other Questions

HISTORY OF CLOC

Who were the original investigators?

Dr. Camille Wortman (now of SUNY-Stony Brook) was the principal investigator, and Dr. Ronald Kessler (now at Harvard University) was the co-investigator on the original National Institute on Aging (NIA) grant that was funded in 1985. This project was conducted as part of a program of research on stress in later life. The initial plan was to monitor members of the American Changing Lives (ACL) sample for mortality, but this was not feasible. Hence, the Changing Lives of Older Couples (CLOC) project was initiated. Since then, other co-investigators have included Dr. James House, who was the overall program principal investigator, and James Lepkowski.

When were the data collected?

The initial baseline data were collected between 1987-88. Subsequent waves of data were collected over the next five years, until 1994.

Is the CLOC study just about bereavement?

Although the primary objective of the CLOC project is to study prospectively the course and predictors bereavement, many related research questions can be investigated with the CLOC data. The CLOC includes rich data on personality, affective experience, social integration, interpersonal relationships, mental health, cognitive functioning, and physical health.

How should I cite the support that makes the CLOC possible?

The CLOC data analysis and data dissemination project was made possible by a grant from the National Institute on Aging, R01-AG15948-01A1 (Randolph M. Nesse, PI). The original data collection was made possible by a grant from the National Institute on Aging, R01-AG610757 (Camille Wortman, PI).

DESIGN AND METHOD

What is the design of the CLOC study?

The CLOC study is a prospective study of a two-stage area probability sample of 1,532 married individuals from the Detroit Standard Metropolitan Statistical Area. This sample includes 423 couples in which both members served as respondents (i.e., N=846 respondents). Baseline data were obtained using face-to-face interviews, conducted from June 1987 through April 1988. Spousal loss was monitored by reading daily obituaries in three Detroit-area newspapers and by using monthly death record tapes provided by the State of Michigan. Deaths were confirmed using the National Death Index. Of the 319 respondents who lost a spouse during the study, 86% (n=276) participated in at least one of three follow-up interviews, which were conducted 6 months (Wave 1), 18 months (Wave 2), and 48 months (Wave 3) after the death. Controls from the original sample of 1,545 were selected to match the widowed persons along the dimensions of age, race, and sex. The matched controls were re-interviewed at roughly the same time as the corresponding widowed person.

When were Wave 1 conducted?

Wave 1 interviews took place 5-8 months after the date of spousal death. However, this follow-up interview may have occurred years after the original baseline interview, given that the deaths occurred over a very broad time period. The "GAP" variable (see below) indicates the length of time between the baseline and Wave 1 interviews. Researchers should be sure to include this measure as a control in their multivariate analyses; the baseline indicators are most temporally close from wave 1 indicators for persons whose spouses died relatively soon after the baseline interview.

How were sample members selected?

The sample was drawn from three counties in metropolitan Detroit using a two-stage area probability sampling method: To be eligible for the study, respondents had to be English-speaking members of a married couple in which the husband was aged 65 or older. This age stipulation was intended to increase the base rate of widowhood, as older persons are more likely to experience spousal loss in the imminent future. Eligibility requirements also included the criteria that participants be non-institutionalized and capable of participating in a 2-hour long interview.

What are the response rates and attrition?

Approximately 65% of those contacted at baseline for an interview participated, which is consistent with the response rate from other Detroit studies. This response rate yielded 1,532 participants including 1,058 women and 474 men. Of the 319 participants who lost a spouse during the study, 86% (n=276) participated in at least one of the three follow-up interviews. The primary reasons for non-response were refusal to participate (38%) and ill health or death at follow-up (42%).

How many subjects lost a spouse?

A total of 319 respondents from the original sample of 1,532 lost a spouse during the course of the study. Of these 319 ever-widowed, 276 persons (86%) participated in at least one of the three follow-up interviews. Specifically, 250 participated in wave 1, 210 participated in wave 2, and 106 participated in wave 3.

Why are there twice as many women as men?

Women were oversampled in an effort to maximize the number of baseline respondents who would go on to experience spousal loss in the near future. Thus, multivariate analyses should use weighted data to account for this oversampling.

Do all of the men in this study also have data collected from their spouse?

No. We have spousal data for 423 of the 474 men interviewed at baseline, and for 423 of the 1058 women interviewed at baseline.

Who are the matched controls?

Individuals from the original sample of 1,532 respondents were selected to match the widowed persons along the dimensions of age, sex, and race. The matched controls were re-interviewed at the three follow-up interviews at roughly the same time as their corresponding widowed persons. Some of the controls participated in multiple waves, but most participated in only one. Controls are not the same from wave to wave, and they can differ from the MacBat controls as well.

What is MacBat?

Funding from the MacArthur Foundation made it possible to gather additional health-related data at Baseline and Waves 1and 2 on a subset of 432 respondents. These respondents included all respondents in couples where the husband was 70 or older, and all CLOC respondents who became widowed or served as a control. Data collected from these respondents included assessments of gait and movement, memory, cognitive functioning, height, weight, eyesight, and chemical and endocrine measurements based on blood and urine samples.

What happened to the baseline blood and urine samples?

The data based on blood and urine samples are complete except at Baseline, where most data is missing because specimens were ruined when a freezer failed.

Are the controls for the MacBat the same as for the widow group?

Not necessarily. Some individuals serve as controls for both samples; others serve either as a control for the CLOC sample only or as a control for the MacBat sample only.

Why are there so few controls at wave 1?

Controls were not available for all bereaved subjects at the 6-month follow-up because funding was cut from the proposed budget, making it impossible to collect data on the optimal number of controls. However, funding was later reinstated, enabling the researchers to obtain data on a larger number of controls at the Wave 2 interview.

How many subjects have complete data for all waves?

The CLOC has complete (i.e., three waves) data for 106 widowed persons.

What are clones?

Clones are the 13 individuals who initially participated in a follow-up interview as a control subject, but who subsequently experienced spousal loss, and then entered the study as a bereaved subject. More precisely, "clones" are cases that participated as both widowed AND control for the SAME wave (even though the interviews may have occurred at very different points in time). The ID numbers of the clones are 2292, 2309, 2603, 2673, 2710, 2797, 2984, 3031, 3320, 3346, 3357 and 3375. The "cloned" data (data from these person's interview as a control subject) have been removed from all datasets except for the WidowControl (with clones) file. All data for clones also exists as a separate data file, though only the most advanced user will want to use this data. In most cases, analyses will not utilize the cloned records; however, some analysts using a sophisticated case-control type of design may wish to explore the cloned data. In sum: When in doubt, leave clones out. (and for your benefit, nearly all available datasets have already left them out!)

How many subjects lost a spouse?

A total of 319 respondents from the original sample of 1,532 lost a spouse during the course of the study. Of these, 250 participated in wave 1,210 participated in wave 2, and 106 participated in wave 3. (From Table)

DATA ANALYSIS

How many variables are there?

There are 4,438 variables in the CLOC-MACBAT Merged data set.

What is the content of the CLOC variables?

A major goal of the CLOC project is to investigate how loss of a spouse influences psychological and social adjustment among older Americans. Hence, the CLOC data were designed to assess social, behavioral, psychological, and physical reactions to bereavement. Included in this assessment are instruments that measure grief reactions, mental health, physical functioning and health behaviors, personality, social support, and interpersonal relationships. The entire list of variables can be found in the Codebook entitled, "Contents of the CLOC-MACBAT merged dataset." (See below)


What are the different data sets?

CLOC Data - The Core Dataset: contains data from all respondents, at all four waves (BL, W1, W2, W3). This is the complete dataset with 1532 records. This dataset contains the MacBat biomedical indicators but does not contain the clones.

CLOC Baseline Dataset: contains data from the baseline interview only (n=1532).

Couples Dataset: contains data from 423 couples (846 persons). At baseline, data were collected from both the husband and wife of 423 couples. This dataset includes all available data, including the MacBat data, from each of the four waves (BL, W1, W2, W3) for these 846 persons only. The first set of variables on the line represent responses from the wife, whereas the second set of variables on the line represent responses from the husband. "v" represents wife, "s" represents husband.

Widow-Control Dataset: contains all available data (BL, W1, W2, W3) except clones from those respondents who participated in a follow-up interview, either as a widowed person or as a control (n=545).

Widow-Control-Clones Dataset: contains all available data (BL, W1, W2, W3) including clones from those respondents who participated in a follow-up interview, either as a widowed person, a control, or both (n=558).

Clones-only dataset: contains all available data (BL, W1, W2, W3) from those respondents who served as both a control and a widow at the same wave (n=13).

Where are the codebooks?

The codebooks can be found on our website: www.cloc.isr.umich.edu/codebooks.htm.

What do the variable numbers represent?

The variable numbers simply reflect the wave when the data were obtained. The first digits reflect the wave.

V1-1001 Baseline survey data

V1001+ Wave 1 survey data

V4001+ Wave 2 survey data

V7001+ Wave 3 survey data

V20000+ MacBat Data

The latter digits reflect the question content. You can find comparable measures of items across the multiple waves by looking at the latter 2-3 digits of the variable number. You can also view the documentation found in the codebook.pdf file which also lists which specific variables are available at each wave.

What is the couples' dataset?

The couples' dataset is a cross-sectional source of data, consisting of the Baseline responses of both members of 423 couples. Results from a hierarchical linear modeling analysis indicated that there was no couple level effect on mortality. Hence, individuals can be treated as independent sources of data when mortality is the dependent variable of interest.

Which data set should I use?

The data set you use will likely depend on your research question. If you are interested in factors that affect marital dynamics, for example, you may wish use the couples' data set in order to have responses from both members of the couple. If you are interested in examining the change in grief reactions over time, you may wish to use a data set in which you only select those persons who participated as a widowed subject for each wave of data collection. If you are interested in understanding whether widowhood fundamentally changes the psychology of older adults, you may wish to use the widow/control data set, which consists of the responses of both widows and controls at each wave. When in doubt, start with the CLOC.data (the core dataset) and begin to subselect the sample you are most interested in. Yes, almost any analysis can be performed with the core dataset.

Are there created indexes?

Yes, see the codebook for details of scale construction and alphas.

Are labels provided?

Labels for each of the constructed variables are provided. In general, higher values reflect greater levels of the construct that is being measured.

What scales are most frequently used?

The scales that you use will depend on the research question you choose to pursue. However, in some cases multiple measures are used to assess the same underlying construct. Thus far, the more common measures include the 11-item CES-D, the Bradburn Positive Well-Being scale, the satisfaction with health scale, and the functional health index.

What are the most frequently used control variables?

Control measures typically consist of demographic factors known to affect mortality. The following are often used: age, gender, income, education, and race. In addition, you may wish to control for other design features such as the length of time between baseline responses and wave 1 interviews. This variable is referred to as the "GAP" variable.

What variable should be used to identify widows?

"Widow" means that the participant is one of 319 persons who lost their spouse over the course of the entire study. To identify those individuals who lost a spouse who also participated in a subsequent wave, select cases in which V1003=11 (CLOC Wave 1); v4003=21 (CLOC Wave 2); v7003=31(CLOC Wave 3).

What is the "Gap" variable?

This lists the number of months from time from baseline interview to the wave 1 interview.

What is the weight and how was it created?

Variable number 903 is the final centered weight that adjusts for sampling and response rates.

How was missing data handled?

Missing values were imputed for each composite. Missing values were coded as either "8" or "9".

How is missing data coded?

Usually "8" is "don't know", and "9" is "not applicable".

Are there any subjects who are missing?

Yes, subject 1151 is missing in all datasets.

What process was used to impute missing values?

We used a Stochastic Proportional Missing Value Imputation procedure. First, the proportion of responses for each category for that variable was calculated. Then a randomizing procedure was used to fill the missing value with a random integer with frequencies proportionate to the frequency of that integer for the variable as a whole. This procedure does not use any characteristics of the individual and therefore tends to increase variance but it does not introduce any systematic error in a way that would require adjustment for degrees of freedom

What does the "S" mean as a prefix for many variables?

An "S" is used to refer to data obtained directly from one's spouse. Remember that for 846 persons, we have both their OWN answer to every question administered at Baseline, and their SPOUSE'S response to every question administered at the baseline interview. The "V" variables denote one's own response, while the S" variable denotes one's spouse's response.

What is a match ID?

A match ID identifies respondents at the couple level, matching up their own ID with that of their spouse.

What health measures are available?

A broad array of health measures and health behaviors are included in both the CLOC data set and the MACBAT data set. These data include self-rated health, health symptoms, conditions, ADLs, smoking, drinking, diet, and sleep patterns. Consult the codebooks for specific information on variables.

How can I tell if and when a subject died?

Look at the "Date of Death" or "Vital Status 2001".

How were grief scales computed?

They were based on the individual grief variables with the missing valued imputed, indicated by an "I" suffix. A description is on the web page and in the codebook.

What are the variables with the "I" suffix?

The "I" signifies imputed variables for grief items. We wanted variables with complete data for these core items so all missing cells were imputed and new variables were created. These variables were used to create the grief scales and subscales.

Where did the grief items come from?

The grief items were drawn from several widely used grief scales, including the Bereavement Index, Present Feelings about Loss, and Texas Revised Inventory of Grief. A summary of these scales can be found in the article: Carr, D. & Utz, R. 2002. Late-life widowhood in the United States: New directions in research and theory. Ageing International, 27.

OTHER QUESTIONS

What papers have been published?

The CLOC research team is an active group; several papers are in progress, under review, and forthcoming on a continual basis. You should regularly check the projects and publications page on this site. As soon as a paper is published, we make a machine-readable linkage available from the web page.

Why are there grief scale scores for control respondents at Wave 3?

At Wave 3, a "grief equivalent" set of 10 questions was asked of married controls. Unlike the widowed respondents, these questions are not asked in reference to any specific event. The time frame is the past month. These items allow a comparison of grief scores in widows with similar items in controls. The ten questions are as follows:

V7301 During the past month, have you felt afraid of what lies ahead for you?
V7302 Have you felt extremely anxious or unsettled during the last month?
V7303 During the past month have you felt worried about how you would manage your day-to-day affairs?
V7304 Has life seemed empty?
V7305 During the past month, did you feel amazed at your strength?
V7307 During the past month, did you feel proud of how well you were managing?
V7313 In the past month, have you felt anger toward God?
V7314 Have you felt empty inside, like an important part of you is missing?
V7315 In the past month, have you felt that life has lost its meaning?
V7316 In the last month, have you had any regrets about anything that happened between you and your (husband/wife)?

Complete scales were created for anxiety (alpha=.695) and despair (alpha=.688) in controls. The anger scale was a single item in controls. The overall grief scale (alpha=.718) in controls was derived from 7 items rather than the 19 used in the widowed population.

 


University of Michigan - Institute for Social Research - Survey Research Center
Research Center for Group Dynamics - Inter-University Consortium for Political and Social Research