Methodology

The development of our tool has undergone full academic peer review, and has been published with BMJ Open. You can obtain a free copy of our paper, and its pre-publication history by clicking here.

Update: Version 1.1 

Version 1+ of PsyMaptic makes a number of critical improvements in prediction over our previously launched Beta version (0.5). For those of you with limited time, a summary of the changes to this update is provided below. All readers are recommended to read our detailed Release Notes, which contain complete details on the changes made to Version 1+.

Please note: Version 1.1 predictions are identical to Version 1.0 predictions. Version 1.1 releases these at local authority, county and national levels. County level predictions were not available in Version 1.0.  Major changes in the Version 1+ upgrade include:

1. Updated denominator data in the prediction models from 2009 mid-year census estimates to the population estimated from the 2011 Census of Great Britain.

2. Change in the prediction algorithm. Version 0.5 took into account the age, sex and ethnic structure of each Local Authority [LA], as well as population density (people per hectare), which is strongly associated with the incidence of first episode psychotic disorders. The new model used in Version 1.0 still uses these variables, but also includes the level of deprivation in each LA, estimated using the “Extent of Deprivation” from the Index of Multiple Deprivation 2010. The extent of deprivation measures the proportion of people in each LA living in the 20% most deprived areas in England. For full details see our detailed methodology update.

3. Improved validation data. Our previous version of PsyMaptic was tested in East Anglia by seeing how closely the model predicted the expected number of cases in the region, compared with what was actually observed over a 1.5 year period. Data were provided from the SEPEA study of first episode psychosis in people aged 16-35 years old. The new model was still validated against SEPEA data, but observed data were available for a 3.5 year period, meaning that we have more confidence in the predictive validity of the model.

4. Data by ethnic group. Our validation procedure suggests that predictions were sufficiently precise to allow the publication of predicted numbers of new cases of first episode psychosis in each Local Authority, per year, by the following variables: overall, by sex, by broad age group (16-64, 16-35, 36-64 years) and by the following ethnic groups: white British, black Caribbean, black African, Pakistani, Bangladeshi, mixed ethnicities, all other ethnic groups (including Indian, Chinese, Arabic and other backgrounds).

5. Data at County Level. We have now published prediction maps at two levels, local authority (as before) and counties. County maps are based on the same predictions (at local authority level), but simply summed for counties with new 95% prediction intervals. Data for Wales, London Boroughs, Unitary Authorities and Metropolitan Districts  are unchanged in the County maps.

An overview of our new modelling procedure is included in a peer reviewed article.

Future versions of PsyMaptic will release data at different levels of geography, to include PCTs and CCGs. Presently, prediction into these boundaries is not straight-forward and we are working on methods to improve this. Future versions will be conditional on continued research funding.