It is common in health care research to have to take into account the general health of individuals, particularly where people may have more than just one health condition (multimorbidity). This is necessary to better understand how different treatments or other services provided to patients affect things such as risk of a disease or death. This is commonly done using a multimorbidity "score" that varies depending on the number and type of different health conditions a patient has. However, most commonly used scores have important limitations. Examples include scores being "out of date" and less accurate in the current health context, being based on non-UK data, or being tailored to predicting a specific health outcome which might not be relevant to other work (e.g. deaths might not be relevant to usage of GP services).
In this project we aim to develop a new score based on health information relevant to UK primary care. Different scores will be developed for research looking at general practice consultations, hospital admissions, and risk of death. We expect these scores to be useful to a wide range of researchers carrying out studies using computerised health records.
In this project, we aim to develop new multi-morbidity scores for three different outcomes relevant in health services research (one year mortality, general practice consultations and unscheduled hospital admissions). We will do this using contemporary electronic health records data from the UK. Our multi-morbidity scores are based on a list of conditions relevant to characterize multimorbidity in UK primary care published by Barnett et al. (2012). It is our intention to provide transparent, simple approaches to coding multi-morbidity in CPRD that will be useful for other researchers..
This is a retrospective cohort study (including linked ONS and Basic HES data). A random sample of 300000 adults will be used for multimorbidity score development, with two separate validation cohorts used to evaluate outcomes over 1 and 5 year follow-up periods. Adjusted rate or Cox regression (as appropriate) will be used to model numbers of GP consultations, unscheduled hospitalisation and death, using model coefficients for weighting conditions in the score
We will determine medication count based on the number of long-term medications being prescribed at the index date.
We will use linked HES data to determine the occurrence of an unscheduled inpatient hospitalization during the post-exposure follow up period. A hospital admission in this period will be determined using the admission and discharge dates for hospitalizations and the start and end dates for consultant episodes. Unscheduled admissions will be those where the method of admission has an emergency admission code.
Mortality and death date will be obtained from the linked ONS data
Catherine Saunders - Chief Investigator - University of Cambridge
Silvia (Silva) Mendonca - Corresponding Applicant - University of Cambridge
Catherine Saunders - Collaborator - University of Cambridge
Duncan Edwards - Collaborator - University of Cambridge
Martin Roland - Collaborator - University of Cambridge
Rupert Payne - Collaborator - University of Bristol
HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation