Frank Popham

Background Teenage pregnancy rates across the UK and other English-speaking high-income countries have fallen across recent decades. Britain’s rates fell from 46.9 pregnancies per 1,000 women aged under 18 in 1998 to 15.8 per 1,000 women in 2019. The reasons are not well understood, hampering research and the development of policy. We used systematic review methods to identify plausible hypothesised causes of this decline in pregnancy rates. Methods From the sexual health literature, we constructed an initial tree diagram of interacting causal ‘branches’. Branches included teenage pregnancy interventions and wider environmental changes which may have influenced pregnancy rates. From this causal tree we developed a search strategy, combining terms to sensitively search for ‘adolescent’ or ‘teenage’ populations with a ‘pregnancy’ or ‘conception’, alongside terms relating to ‘contraception’ or action to ‘reduce’ risk or ‘prevent’ pregnancies. We searched four databases for reviews published since 1990 assessing evidence for each of the branches’ contributions to falling rates within the UK and in similar high-income European or English-speaking countries. We extracted data on the strength of evidence for existing branches and assessed evidence for additional branches. Reviews were graded according to risk of bias. Evidence across reviews was synthesised – prioritising high-quality reviews that used systematic search and synthesis methods – to give overall assessments of the certainty of 1) effectiveness of an exposure in reducing pregnancy rates, using modified Bradford Hill criteria; and 2) reach across whole populations of the exposure of interest. We used these combined scores to assess the causal contribution of each branch to observed falling rates. Results We identified 89 reviews from 631 articles found through searches and citations. 62 were rated high-quality. Included reviews presented evidence assessing the contribution of 20 causal branches. Evidence for teenage pregnancy interventions was more common than evidence for environmental changes. We eliminated two branches which had been extensively reviewed but for which evidence was consistent with a null effect and added seven novel branches. We retained 13 plausible contributors to falling teenage pregnancy rates in the updated causal tree. Two branches – changing contraceptive technologies and policies promoting increased educational involvement – were rated as having ‘strong evidence’. Two branches were not assessed in any reviews and were retained out of caution; the remaining nine included branches presented ‘weak evidence’ of their effect on rates. Conclusion The evidence-based causal tree provides a basis for both further investigation of hypothesised causes and development of effective interventions.

Background: Health surveys provide a rich array of information but on relatively small numbers of individuals and evidence suggests that they are becoming less representative as response levels fall. Routinely collected administrative data offer more extensive population coverage but typically comprise fewer health topics. We explore whether data combination and multiple imputation of health variables from survey data is a simple and robust way of generating these variables in the general population. Methods: We use the UK Integrated Household Survey and the English 2011 population census both of which included self-rated general health. Setting aside the census self-rated health data we multiply imputed self-rated health responses for the census using the survey data and compared these with the actual census results in 576 unique groups defined by age, sex, housing tenure and geographic region. Results: Compared with original census data across the groups, multiply imputed proportions of bad or very bad self-rated health were not a markedly better fit than those simply derived from the survey proportions. Conclusion: While multiple imputation may have the potential to augment population data with information from surveys, further testing and refinement is required.

The ‘inclusion health’ agenda aims to draw attention to health disadvantages accompanying experiences putatively characterised by social exclusion, such as homelessness, problem substance use, or imprisonment. However, its increasing prominence has surfaced conceptual uncertainties and potential tensions with other understandings of health inequalities. We undertook a discourse analysis of how recent health inequalities policy documents describe, explain, and make recommendations relating to inclusion health. Using the UK as a case study, and with reference to public health accounts of multi-level governance theory, we selected five recent health inequalities policy reviews covering Scotland, UK, European Union, and the World Health Organisation. All documents referred to some inclusion health concerns, though their relative emphasis differed between documents. Terms like inclusion, exclusion, and vulnerability were commonly used, but ill-defined and often ambiguous. Explanatory discourses were diverse, with a particular focus on intergenerational cycles and disproportionate exposure to risk, with a varying emphasis on individual versus structural factors. Few documents provided coherent explanatory accounts for the relationship between the issues of interest to inclusion health, their associations with poor health, and other axes of inequality. Our results suggest that health inequalities policymaking in a multi-level context may benefit from comprehensive conceptual frameworks which encompass diverse forms of social stratification, advantage, and disadvantage, and acknowledge potential tensions and trade-offs between different understandings. This may necessitate further theoretical and empirical work for inclusion health on its definitions, bounds, and how its scope of interest interacts with other forms of social and health inequality.

This article presents two ways of quantifying confounding using logistic response models for binary outcomes. Drawing on the distinction between marginal and conditional odds ratios in statistics, we define two corresponding measures of confounding (marginal and conditional) that can be recovered from a simple standardization approach. We investigate when marginal and conditional confounding may differ, outline why the method by Karlson, Holm, and Breen recovers conditional confounding under a “no interaction”-assumption, and suggest that researchers may measure marginal confounding by using inverse probability weighting. We provide two empirical examples that illustrate our standardization approach.