Misclassification bias in estimates of bereavement effects

Area of Science:

• Gerontology
• Sociology
• Public Health
• Biostatistics

Background:

• Prospective studies often fail to update marital status in statistical models.
• This leads to misclassification of marital status, biasing estimates of health and mortality.
• Specifically, bereavement effects associated with widowhood may be inaccurately represented.

Purpose of the Study:

• To illustrate how misclassification of marital status biases estimates of bereavement effects.
• To examine the impact of updating marital status information on health and mortality estimates.
• To assess the bereavement effect of widowhood using accurate marital status data.

Main Methods:

• Utilized data from the Longitudinal Study of Aging (LSOA), 1984-1990.
• Focused on 3,192 respondents aged 70+ who were married at baseline.
• Compared reported marital status from interviews with spouse death records using survival models.

Main Results:

• Reliance on interview data led to substantial misclassification of marital status.
• Updating marital status in hazard models significantly altered estimates of widowhood effects.
• Interview data suggested a protective effect of widowhood, while spouse death records revealed a significant bereavement effect (relative mortality risks 1.3-1.4).

Conclusions:

• Accurate and updated marital status information is essential for valid health and mortality research in older adults.
• Failure to update marital status can mask significant bereavement effects, particularly for widowhood.
• Utilizing spouse death records provides a more accurate assessment of widowhood's impact on mortality.