Interpreting intentions: Research and measurement challenges in studies of unintended pregnancy among Latina adolescents in the United States and young married women in Bangalore, India

Delineating the role of pregnancy intentions on the causal pathway to pregnancy is essential for public health professionals and demographers interested in understanding fertility patterns and preventing unintended pregnancy. However, substantial methodologic and conceptual obstacles hinder progress towards this end. The accurate measurement of pregnancy intentions presents arguably the most significant challenge, as intentions are not observable, can be complex and nuanced, and may change over time.;This dissertation investigated several questions surrounding the measurement and significance of pregnancy intentions among two longitudinal cohorts of youth: Latino teenagers living in San Francisco and married women in urban South India. Analyses among Latino teenagers revealed that the feelings some teenagers held about the prospect of pregnancy were complex and ambivalent. While 86% of participants reported definitely not wanting pregnancy in the next six months, only 43% indicated that they would be very unhappy if pregnancy occurred. Boys and girls generally held similar intentions, though boys provided higher estimates of the likelihood of pregnancy.;Although the probability of pregnancy generally increased with higher intention level for each of the four measures examined, risk of pregnancy was elevated only at the highest response levels, categories into which only small proportions of teens fell. Among girls and boys respectively, 73% and 68% of pregnancies occurred among those who definitely did not want to get pregnant. Furthermore, intentions were dynamic for some teens over time, with 18% of responses to the wantedness measure and 40% of responses to happiness measure changing between six month assessments.;In analyses among Latina adolescents, pregnancy wantedness was an important, independent risk factor for pregnancy, not a mediating variable between other risk factors and pregnancy. According to a causal attributable risk calculation, an intervention that could reduce all wantedness to "definitely not" wanting pregnancy, even without modifying other risk factors, could be expected to reduce pregnancy rates in this population by 16%. Few teens (14%) actually expressed any wantedness for pregnancy; thus, the large majority (84%) of pregnancies would likely occur even if pregnancy wantedness were removed.;A psychometric analysis of the London Measure of Unplanned Pregnancy (LMUP) using Item Response Modeling among young married Indian women found that the reliability of the scale was acceptable (0.69). Average overall scores correlated well with each item's response categories, providing evidence of internal structure validity. That pregnant women and women of higher parity had lower overall scores offered support for the external structure validity of the LMUP.;Analyses also detected several potential problems with the use of the LMUP in India. For instance, the middle response categories to four items received low endorsement, and the item on contraceptive use added little to the precision of the scale. Four items functioned differentially by parity, casting doubt on the validity of using a single scale across all women. Furthermore, the low stability of LMUP scores over one year raises questions about the reliability of retrospective accounts of pregnancy intentions in this setting.;This dissertation provided a unique perspective into pregnancy intentions by using prospective data collected on multiple occasions and utilizing improved, multi-item intentions measures. In view of the enormous health, economic and environmental consequences of unintended pregnancy around the world, improving our understanding of how women form intentions and developing more sophisticated, culturally-appropriate prospective measurement tools are high research priorities.