The NHS allocates a total of £104.3 billion to Clinical Commissioning Groups in England to fund healthcare using a set of resource allocation formulae. These formulae assess the relative need per head for health care services across CCGs using risk adjustment to make sure providers aren’t penalised for serving patients with higher expected costs and for being in a location with higher unavoidable costs of delivering services. The formulae include a set of needs and supply variables as risk adjustment, among which age and individual-level morbidity flags derived from prior diagnoses recorded at hospital admissions have been found to the most important variables explaining variations in secondary care usage1,2. However, there is concern that relying only on diagnoses recorded in a hospital setting to determine health need may be inadequate because it fails to capture health needs that are no met by the NHS1. There are likely to be individuals who have systematically lower usage of secondary care (for example, people with complex health needs living in nursing homes may not use inpatient or outpatient care) or whose needs remain unobserved because of incomplete or late diagnosis. The aim of this work is to assess the feasibility and implications of including both primary and secondary care diagnostic variables into the risk-adjustment model to predict individual-level utilisation of primary care and General and Acute hospital care and assess their relative explanatory power.
The person-based formulae currently used for resource allocation are based on the so-called utilisation approach. This approach uses regression techniques to assess the relative need per head for health care services across the country, by first estimating cost-weighted service use as a function of a set of needs and supply variables and then sterilising the effect of supply. Age and individual-level morbidity variables derived from prior diagnoses recorded at hospital have been found to the most powerful need predictors of costs1,2. Relying upon inclusion of past diagnoses from hospital-recorded diagnoses may underestimate future need for patients whose health needs are not met by the NHS hospitals1. We aim to use diagnostic information recorded in linked primary and secondary care data to assess mismatch in primary and secondary care diagnostic records and to better account for needs of patients not diagnosed in secondary care.
Our project will involve two stages.
Stage 1. Comparative study. We will use the longitudinal diagnostic information recorded for the same individual in primary care and secondary care to assess the extent to which primary care diagnoses are reflected in the secondary care diagnoses, and vice-versa, and whether the differences vary systematically by population groups (sex, age, ethnicity, area deprivation), by general practices or by hospitals.
Stage 2. Predictive modelling. We will re-estimate the risk-adjustment linear regression model (ordinary least squares) with robust standard errors underpinning the current formulae with addition of primary care diagnostic variables to predict primary care and General and Acute (inpatient spell, outpatient attendance and Accident and Emergency attendance) cost-weighted service use. We will control for as many individual general practices and area level characteristics available and assess whether inclusion of primary care diagnostic variables increases the predictive power and how it changes the coefficients on the needs, general practice and hospital variables.
Measures of cost-weighted utilisation that have used in previous work: health care utilisations in primary care (number and cost of consultations with a general practitioner and practice nurse) and General and Acute hospital care (days and costs of inpatient admissions, outpatient consultations, and Accident and Emergency attendances). Costs of primary care usage will be calculated by applying national unit costs to the number of consultations with a general practitioner and practice nurse. Costs of secondary care usage will be calculated by applying National Tariff prices to the number of inpatient admissions, outpatient consultations and Accident and Emergency attendances.
Matt Sutton - Chief Investigator - University of Manchester
Shaolin Wang - Corresponding Applicant - University of Manchester
Laura Anselmi - Collaborator - University of Manchester
Michael Anderson - Collaborator - University of Manchester
HES Accident and Emergency;HES Admitted Patient Care;HES Outpatient;ONS Death Registration Data;Patient Level Index of Multiple Deprivation;CCG Pseudonyms;Rural-Urban Classification