Identification of low health literacy-risk patients in GP clinical systems

Problem

Health literacy is the skills to access understand and use information for health. Low health literacy is associated with poorer health, more long-term conditions (LTCs), more difficulties managing LTCs, and earlier death. Low health literacy is a problem for 60% of working-age (16 to 65 years) English people, with higher levels in areas of socio-economic deprivation. People rarely volunteer information about low skills due to stigma, and health professionals find it difficult to identify people with low health literacy. This is a particular issue for General Practices, which provide over 90% of patient contacts in the NHS.

Approach

To enable automatic calculation, within GP clinical systems, of the risk of a patient having low health literacy. A logistic regression analysis was undertaken on a national skills dataset of English working age adults (aged 16 years to 65 years) containing individual-level health literacy, socio-demographic data and Lower Super Output Area (LSOA) of residence. Look-up tables were produced that predicted the risk of an individual falling below a ‘health literacy competency threshold’ using age, sex, ethnicity (White British vs. other), preferred language (English vs other), and LSOA of residence. An algorithm was written to run the calculations in the EMIS GP Clinical System in General Practices in one inner-city CCG with high levels of ethnic diversity and socio-economic deprivation: Lambeth, London. The estimates required minimum patient data on LSOA (from patient postcode), age, and ethnicity; other missing data were imputed using the mean value for the LSOA of residence.The algorithm accuracy was tested by comparing the results of the automated method with those from the manual look-up tables for a sample of patients. One record was randomly selected for each cell in the predicted risk table (n=513); in addition, a further 500 records were randomly selected from the remaining records.

Findings

The algorithm was run on all records in the Lambeth Datanet (LDN) (March 2104 data extraction) involving 270,799 patients. Of these records, 184,492 were in the age range. Of this total, the algorithm could be calculated on 135,400 patients (73.39%). The algorithm showed 100% accuracy in predicting the individual-level look-up table risk score.

Consequences

This method provides a way to automatically calculate possible individual-level risk of an individual patient falling below the health literacy competency threshold. Further work is required to assess the feasibility and accuracy of the method in other CCGs, particularly those with a less ethnically-diverse population and lower socio-economic deprivation, to assess validity, and to explore with GPs and patients how such a method could be used to improve GP services for patients with low health literacy.