MattSutton

Use of health surveys in resource allocation: 

Use of health surveys in resource allocation Matt Sutton Senior Research Fellow University of Glasgow m.sutton@clinmed.gla.ac.uk Health Survey's User Group Friday 23 January 2004 DoH, Skipton House, London

Colleagues on previous work: 

Colleagues on previous work Scotland: “Derivation of an adjustment to the Arbuthnott formula for socioeconomic inequities in health care” with Alex McConnachie England: “Allocation of Resources to English Areas Report” with Hugh Gravelle, Stephen Morris, Alastair Leyland, Frank Windmeijer, Chris Dibben and Mike Muirhead [http://www.isdscotland.org/isd/info3.jsp?pContentID=842&p_applic=CCC&p_service=Content.show&]

Outline: 

Outline Overview of resource allocation formulae Direct use of survey information Indirect use of survey information experience in Scotland experience in England

Overview of resource allocation formulae: 

Overview of resource allocation formulae Purpose - Allocate national resources to health care organisations to: Scotland: “promote equitable access to health care” Wales: “ensure more equitable access to health services in accordance with health needs” England: “contribute to the reduction of avoidable inequalities in health”

Structure of resource allocation formulae: 

Structure of resource allocation formulae Population size Adjusted for demography Adjusted for additional need factors Adjusted for additional cost factors Market Forces Factor Additional costs of remoteness and rurality

Three basic approaches to allocation: 

Three basic approaches to allocation 1. Based on relationships between population characteristics and use of health care 2. Based on actual prevalence of ill-health 3. Based on relationships between population characteristics and prevalence of ill-health (predicted prevalence)

1. Adjustments based on average costs: 

1. Adjustments based on average costs “Rich” £200 “Poor” £400

2. Adjustments based on actual prevalence: 

2. Adjustments based on actual prevalence Region A 5% Region B 7% Cost per prevalent case = £3000

3. Adjustments based on predicted prevalence: 

3. Adjustments based on predicted prevalence “Rich” 4% “Poor” 8% Cost per prevalent case = £3000

Use of health surveys: 

Use of health surveys Wales: Direct use of survey to make adjustments for demography and additional need Scotland, England & NI: Indirect use of surveys to improve adjustments for additional need

Direct method: 

Direct method Obtain region-specific prevalence rate estimates for each age-group using a health survey Apply region-specific prevalence rate estimates to age profile of resident population to obtain estimated numbers of cases Calculate each region’s share of national cases Obtain national budget for the condition Derive regional budgets by multiplying national budget by the regional shares of cases

Practical issues for direct method: 

Practical issues for direct method Reliability of survey results at regional level Representativeness of survey results Updateability of survey results Differences in reporting behaviour between regions

Conceptual issues for direct method: 

Conceptual issues for direct method Choice of prevalence measure, e.g. circulatory disease Symptom-based measures – Rose Questionnaire Self-reported measures – longstanding illness Doctor-diagnosed measures Breadth of definition – IHD, CVD or CVC Converting prevalence into need for health care resources Share of prevalent cases implies: All “non-cases” have zero need for health care resources All “cases” have same need for health care resources

Alternative measures of circulatory disease: 

Alternative measures of circulatory disease

Indirect methods: 

Indirect methods Model risk of being a prevalent case as a function of individual-level (age/gender) and area-level characteristics Apply risk equation to small-area data to obtain prevalence rate estimates for each small area Model relationship between use of services and prevalence rate estimate(s) to obtain relative needs index for each area

Approach in Scotland: 

Approach in Scotland Adjustment for additional need based on single composite needs variable: “Arbuthnott Index” Standardised Mortality Ratio, 0-64 years Proportion claiming income support, 65+ years Standardised rate of unemployment benefit claims Proportion of households with multiple deprivation

Indirect methods - Scotland: 

Indirect methods - Scotland Original work assumed linear relationship between Arbuthnott Index and use of health care services Work on adjustment involved: Non-linear modelling of relationship between Arbuthnott Index and prevalence estimates from Scottish Health Survey Modelling of effect of fitted prevalence on use of care Simultaneously testing for unmet need (whether high deprivation or low deprivation areas had levels of use that departed significantly from prevalence-use relationship)

Data - Scotland: 

Data - Scotland 1995 & 1998 Scottish Health Surveys 1995 = 7,932 individuals aged 16-64 years 1998 = 12,939 individuals aged 2-74 years Respondents sampled from 451 of 717 areas Standardised prevalence rates calculated for six longstanding illnesses

Circulatory disease prevalence and deprivation: 

Circulatory disease prevalence and deprivation

Modelling effect of prevalence on use of health care: 

Modelling effect of prevalence on use of health care

Relative need profiles under different models: 

Relative need profiles under different models

Relative needs by deprivation decile: 

Relative needs by deprivation decile

Approach in England: 

Approach in England Additional needs modelled using a large number of potential indicators Particular concerns raised about previous review’s ability to avoid ‘unmet need’

Use of surveys - England: 

Use of surveys - England Individual-level tests of unmet need Unmet need tests in small area levels of health care use model risk of morbidity as function of area characteristics augment set of potential need indicators with predicted morbidity indices examine effects on other coefficients

Data - England: 

Data - England Health Survey for England, 1994-2000 Total respondents = 122,500 Binary measures of health care use since 1998 Individuals sampled from 5,893 of 8,414 electoral wards Records linked to a range of population, utilisation and supply variables

Individual-level tests of unmet need: 

Individual-level tests of unmet need

Morbidity models: 

Morbidity models

Augmentation of model with morbidity indices: 

Augmentation of model with morbidity indices

Impact on relative need indices: 

Impact on relative need indices

Summary - I: 

Summary - I Health surveys are increasingly being used in resource allocation availability of data concerns about unmet need in activity-based formulae Direct methods practical issues conceptual issues

Summary - II: 

Summary - II Indirect methods allow for non-linear relationships between deprivation and need inform selection of need variables provide non-linear combinations of need variables to augment data-set permit tests of unmet need