PhD Research Project: Statistical methods to use when imputing missing outcome data in cluster rand

In cluster randomised controlled trials (cRCTs), intact social units such as schools, worksites or medical practices - rather than individuals themselves - are randomly allocated to intervention and control conditions, while the outcomes of interest are then observed on individuals within each cluster. Such trials are becoming increasingly common in the fields of health promotion and health services research. The main consequence of the cluster-randomised design is that participants cannot be assumed to be independent due to the similarity of participants from the same cluster. This similarity or correlation of outcomes is quantified by the intra cluster correlation coefficient (ICC). Participant attrition or loss to follow-up (and missing outcome data) is a common occurrence in randomised trials, and the standard approach for dealing with the resulting missing values is imputation.

Many standard imputation and multiple imputation (MI) strategies may not be appropriate to impute missing data from cRCTs since they assume independent outcome data.

The proposed research plan would be to undertake a review of the statistical literature on methods for imputing missing data in cRCTs; followed by an audit of recently published cRCTs to determine what imputation methods are commonly in used and whether or not they allow for the clustered design in the imputation process. Data is available from a cluster randomised controlled trial which aimed to compare a psychologically informed health visitor intervention based on cognitive behavioural or person centred principles vs. usual care in new mothers with Postnatal Depression (Morrell et al 2009). The Edinburgh Postnatal Depression Scale (EPDS) was the main outcome measure. New mothers were asked to complete the EPDS at 6 weeks and 6 months postnatally: 3349 responded at 6 weeks and 2875 at 6 months.
This project will describe and compare different methods of imputing missing 6 month EPDS outcome data, with multiple imputation strategies which account for the intra-cluster correlation with the standard imputation strategies and a complete case analysis.

The project would then involve some computer simulation and analysis to compare the different methods of imputation to evaluate the influence of cluster size, number of clusters, degree of intracluster correlation, and variability among cluster follow-up rates.

How to apply

Please apply through our online postgraduate application system including the Scholarship Application section where you need to tick the ’University Scholarships’ box. The form will ask you to summarise your research proposal in less than 800 words. If you are unsure about what to put in this section, please contact your prospective supervisor. Please name your supervisor and select their department (ScHARR) through the online form.

Deadline: 5pm 1st February 2017

Funding Notes

Funding

The Faculty of Medicine, Dentistry & Health Doctoral Academy Scholarships cover Home/EU fee and RCUK rate stipend for three years. Overseas students may apply but will need to fund the difference between the Home and Overseas fee from another source.

Proposed start date: October 2017

Candidates must have a first or upper second class honors degree in mathematics or statistics

References

Morrell C.J., Warner R., Slade P., Dixon S., Walters S., Paley G., Brugha T. Psychological Interventions for Postnatal depression: Cluster Randomised Trial and Economic Evaluation (The PoNDER Trial). Health Technology Assessment, 2009; 13(30):1–176.