Presentation is loading. Please wait.

Presentation is loading. Please wait.

Models for small area data with applications in health care

Similar presentations


Presentation on theme: "Models for small area data with applications in health care"— Presentation transcript:

1 Models for small area data with applications in health care
Nicky Best Department of Epidemiology and Biostatistics School of Public Health, Imperial College London Under the scheme, patients might be diagnosed as COPD rather than asthma so that the patient’s family can get compensation The policy => change of diagnostic practice => increase rates?

2 Outline Introduction Mapping and spatial smoothing of health data
Classifying areas and summarising geographical variations in health outcomes Modelling and mapping multiple outcomes Modelling and mapping temporal trends Hierarchical related regression models for combining individual and small area data

3 Outline Introduction Mapping and spatial smoothing of health data
Classifying areas and summarising geographical variations in health outcomes Modelling and mapping multiple outcomes Modelling and mapping temporal trends Hierarchical related regression models for combining individual and small area data

4 A brief history of disease mapping
Health indicator maps have a long history in epidemiology and public health Spot maps: Yellow fever pandemic New York (Seaman, 1798) Cholera and the Broad Street Pump (Snow, 1854) Dr. Valentine Seaman, a surgeon at the New York Hospital, assembled a map of yellow fever deaths in what is now the Lower East Side of Manhattan. At the time he created his map, physicians were embroiled in a heated debate about the source of yellow fever contagion. Unlike the majority of the medical establishment who believed the disease was being spread from person to person, Seaman argued that it was filthy conditions and direct contact with contaminated people or objects that led to transmission. In particular, he mapped yellow fever deaths in relation to the Roosevelt Street drain, which, he felt, was a likely source of contagion owing to the fact that it was “covered with numerous perishable materials…in addition to other putrid matters”. It turned out, however, that Seaman was incorrect. Instead of “putrid effluvia,” mosquitoes were responsible for the spread of yellow fever. More than 100 years after the 1798 publication of Seaman’s map in The Medical Repository, Walter Reed and the Reed Commission, working in Cuba during the Spanish American War, proved that the mosquito Aedes aegypti was responsible for spreading the disease. Nevertheless, despite Seaman’s incorrect theory on the transmission of yellow fever, his contribution to cartography remains important . More than 50 years passed before Dr. John Snow produced what is arguably the most famous map of disease ever created. Similarly inspired by an epidemic disease, Snow mapped London’s cholera deaths in relation to water pumps. He used his map to convince authorities to remove the handle to the Broad Street Pump, a likely source of contaminated water, and the epidemic subsided.

5 Spot map of cholera cases (Snow, 1854)

6 A brief history of disease mapping
Health indicator maps have a long history in epidemiology and public health Spot maps: Yellow fever pandemic New York (Seaman, 1798) Cholera and the Broad Street Pump (Snow, 1854) Chloropeth maps: Geographical distribution of mortality from heart disease, cancer and TB in England & Wales (Haviland, 1878) Cancer mortality by county in England & Wales, adjusted for age and sex (Stocks, 1936, 1937, 1939) Dr. Valentine Seaman, a surgeon at the New York Hospital, assembled a map of yellow fever deaths in what is now the Lower East Side of Manhattan. At the time he created his map, physicians were embroiled in a heated debate about the source of yellow fever contagion. Unlike the majority of the medical establishment who believed the disease was being spread from person to person, Seaman argued that it was filthy conditions and direct contact with contaminated people or objects that led to transmission. In particular, he mapped yellow fever deaths in relation to the Roosevelt Street drain, which, he felt, was a likely source of contagion owing to the fact that it was “covered with numerous perishable materials…in addition to other putrid matters”. It turned out, however, that Seaman was incorrect. Instead of “putrid effluvia,” mosquitoes were responsible for the spread of yellow fever. More than 100 years after the 1798 publication of Seaman’s map in The Medical Repository, Walter Reed and the Reed Commission, working in Cuba during the Spanish American War, proved that the mosquito Aedes aegypti was responsible for spreading the disease. Nevertheless, despite Seaman’s incorrect theory on the transmission of yellow fever, his contribution to cartography remains important . More than 50 years passed before Dr. John Snow produced what is arguably the most famous map of disease ever created. Similarly inspired by an epidemic disease, Snow mapped London’s cholera deaths in relation to water pumps. He used his map to convince authorities to remove the handle to the Broad Street Pump, a likely source of contaminated water, and the epidemic subsided.

7 Female cancer (Haviland 1878)

8 Female lung cancer SMR 1921-30
(Stocks, 1939)

9 A brief history of disease mapping
Health indicator maps have a long history in epidemiology and public health Spot maps: Yellow fever pandemic New York (Seaman, 1798) Cholera and the Broad Street Pump (Snow, 1854) Chloropeth maps: Geographical distribution of heart disease, cancer and TB in England & Wales (Haviland, 1878) Cancer rates by county in England & Wales, adjusted for age and sex (Stocks, 1936, 1937, 1939) National and international disease atlases, e.g Atlas of Cancer Incidence in England & Wales (Swerdlow & dos Santos Silva, 1993) Atlas of Mortality in Europe 1980/81 & 1990/91 (WHO, 1997) Dr. Valentine Seaman, a surgeon at the New York Hospital, assembled a map of yellow fever deaths in what is now the Lower East Side of Manhattan. At the time he created his map, physicians were embroiled in a heated debate about the source of yellow fever contagion. Unlike the majority of the medical establishment who believed the disease was being spread from person to person, Seaman argued that it was filthy conditions and direct contact with contaminated people or objects that led to transmission. In particular, he mapped yellow fever deaths in relation to the Roosevelt Street drain, which, he felt, was a likely source of contagion owing to the fact that it was “covered with numerous perishable materials…in addition to other putrid matters”. It turned out, however, that Seaman was incorrect. Instead of “putrid effluvia,” mosquitoes were responsible for the spread of yellow fever. More than 100 years after the 1798 publication of Seaman’s map in The Medical Repository, Walter Reed and the Reed Commission, working in Cuba during the Spanish American War, proved that the mosquito Aedes aegypti was responsible for spreading the disease. Nevertheless, despite Seaman’s incorrect theory on the transmission of yellow fever, his contribution to cartography remains important . More than 50 years passed before Dr. John Snow produced what is arguably the most famous map of disease ever created. Similarly inspired by an epidemic disease, Snow mapped London’s cholera deaths in relation to water pumps. He used his map to convince authorities to remove the handle to the Broad Street Pump, a likely source of contaminated water, and the epidemic subsided.

10 Female lung cancer incidence 1968-85
(Swerdlow and dos Santos Silva, 1993)

11 Age-standardised mortality from IHD, 1980-81 (WHO)

12 Recent developments in disease mapping
Development of Geographical Information Systems (GIS) Geographically indexed relational database Computer program to map and analyze spatial data Increasing availability of geo-referenced data Ability to geocode, use GPS Disease outcomes, demographics, environmental quality, health services Development of statistical methods Sophisticated techniques for separating signal from noise Ability to account for spatial (and temporal) dependence Methods for cluster detection and classification of areas Interest in mapping health events at small-area scale

13 Small area health data in the UK
Administrative geography in UK includes Postcodes (10-15 households) Census Output Areas (COA; ~300 people) Electoral wards (~500 to 2000 people) Local authority districts, Health authority districts (10’s of thousands) Postcoded data on mortality, births/still births, congenital anomalies, cancer incidence, hospital admissions Population and socio-economic indicators from Census (COA) Increasing availability of modelled environmental data at fine geographical resolution (grids) Limited access to geographical identifiers for certain individual-level cohorts (e.g. Millenium Cohort, British Household Panel Survey) and health surveys (e.g. Health Survey for England)

14 Small area health data in Spain
Administrative geography in Spain is divided into: 17 regions 52 provinces ~8000 municipalities, ranging from small villages to large cities Census tracts (finer sub-division in large cities) Geocoded (place of residence) data on births, mortality (national), cancer incidence (regional; ~26% population), hospital discharge administrative data (national; public hospitals) Small area (municipality) data on population and socio-economic indicators from Census National Statistics Institute As part of Spain’s universal health system, discharge data following hospital admissions are recorded in the Ministry of Health and Consumer Affairs’ Minimum Basic Data Set (MBDS). This official database, which contains information drawn from the discharge reports issued by all 447 acute-care hospitals nationwide, is regarded as being representative of the national population.

15 Examples of recent disease atlases and health-related maps for Spain
Atlas of cancer mortality and other causes of death in Spain (López-Abente et al., 1996) Maps Age-adjusted Rates and Standardised Mortality Ratios (SMR) at province level

16

17 Atlas of cancer mortality at municipality level in Spain 1989-1998 (López-Abente et al., 2007)
Maps (Bayesian) smoothed relative risks of mortality and probability of excess risk, at municipality level

18 Also produced maps of mortality from selected causes other than cancer, e.g. Influenza

19 …… + contextual maps of socioeconomic variables and environmental hazards

20 Outline Introduction Mapping and spatial smoothing of health data
Classifying areas and summarising geographical variations in health outcomes Modelling and mapping multiple outcomes Modelling and mapping temporal trends Hierarchical related regression models for combining individual and small area data

21 Why map small area disease rates?
Interest in mapping geographical variations in health outcomes a the small area scale Highlight sources of heterogeneity and spatial patterns Suggest public health determinants or aetiological clues Small scale UK: electoral ward or census output area SPAIN: municipality less susceptible to ecological (aggregation) bias better able to detect highly localised effects

22 Why smooth small area disease rates?
Typically dealing with rare events in small areas Ai Yi is the observed count of disease in area Ai Ei is the expected count based on population size, adjusted for age, sex, other strata …., Relative risk usually estimated by SMRi = Yi / Ei Standard practice is to map SMRs BUT sparse data need more sophisticated statistical analysis techniques

23 Why smooth small area disease rates?
SMR represents estimate of ‘true’ (underlying) risk in an area, Ri, i.e. Ri = SMRi Statistical uncertainty about estimate based on assuming Poisson sampling variation for data Yi ~ Poisson(Ri Ei) SE(Ri) = SE(SMRi)  1 / Ei SMRi very imprecise for rare diseases and small populations precision can vary widely between areas

24 Why smooth small area disease rates?
SMRi in each area is estimated independently ignores possible spatial correlation between disease risk in nearby areas due to possible dependence on spatially varying risk factors leads to problems of multiple significance testing

25 Is the variability real or simply reflecting unequal expected counts ?
Map of SMR of adult leukaemia in West Midlands Region, England (Olsen, Martuzzi and Elliott, BMJ 1996;313: ). Is the variability real or simply reflecting unequal expected counts ? Have the red highlighted areas truly got a raised relative risk?

26 Methods for smoothing disease maps
These problems may be addressed by spatial smoothing of the raw data Idea is to “borrow information” for neighbouring areas to produce better (more stable, less noisy) estimates of the risk in each area Similar principle to scatter plot smoothers, moving average smoothers…. Many methods available

27 Methods for smoothing disease maps
Ad hoc, local smoothing algorithms e.g. spatial moving averages, headbanging algorithm quick and simple to implement can be very sensitive to ad hoc choice of weights etc. no uncertainty estimates (standard errors) Trend surface analysis e.g. kriging, polynomial/spline smoothing estimation of ‘smoothing parameters’ based on trade-off between fit and smoothness can be sensitive to choice of penalty for trade-off standard errors usually available

28 Methods for smoothing disease maps
Random effects models e.g. empirical Bayes, hierarchical Bayes data-based estimation of model parameters that control smoothing full power of statistical modelling available: standard errors, prediction, probability calculations, inclusion of covariates more complex to understand and implement

29 Bayesian Approach Use probability model to obtain smoothed risk estimate Ri in area i that is a compromise (weighted average) of observed area-level risk ratio (Yi/ Ei) local or regional mean relative risk (m) Weights depend on the precision of the SMR ( 1 / Ei) in area i and the variability (heterogeneity) of the true risks across areas (v) local or regional mean relative risk (m) Aim is to estimate posterior probability distribution of the unknown model parameters (Ri, m, v) conditional on the data (Yi/ Ei)

30 Bayesian disease mapping model
Typical Bayesian disease mapping model: Yi ~ Poisson(Ri Ei), log (Ri) ~ Normal (m, v) Hierarchical Bayesian model also requires specification of a (prior) probability distribution for m and v These are often taken to be ‘non-informative’ Empirical Bayes involves 2-step process: Estimate m and v empirically from observed data Ignore uncertainty in estimates of m and v and plug these values into the Bayesian model above

31 Software Estimation of Bayesian hierarchical models requires computationally intensive simulation methods (MCMC) Implemented in free WinBUGS and GeoBUGS software: Free software INLA (Rue et al, 2008) implements fast approximation: Empirical Bayes smoothing implemented in Rapid Inquiry Facility (RIF):

32 (B) smoothed by Bayesian methods (A) unsmoothed SMR
Map of occurrences of adult leukaemia in West Midlands Region, England (B) smoothed by Bayesian methods (A) unsmoothed SMR (Olsen, Martuzzi and Elliott, BMJ 1996;313: )

33 Comparison of estimation methods
Expected count

34 Including spatial dependence in disease risk
Ri are typically spatially correlated because they reflect, in part, spatially varying risk factors Incorporation of spatial dependence in the distribution of the Ri’s Conditional Autoregressive (CAR) model log (Ri ) ~ Normal (mi , vi) mi = k Rk / ni = average risk in neighbouring areas vi = v / ni → variance inversely proportional to number of neighbours Besag, York, Mollie (1991) Annals of the Institute of Statistics and Mathematics, 43: 1-59

35 Childhood leukaemia incidence in London, 1986-1998
Non-spatial smoothing (posterior mean Ri) Raw data (SMR) <0.5 >2.0 RR Spatial smoothing (posterior mean Ri)

36 Mapping uncertainty Mapping the mean posterior value of Ri does not make full use of the posterior distribution Relative Risk, Ri Map posterior SD Map Probability (Ri > 1) Note – this is not the same as a classical p-value

37 Posterior SD of relative risk estimates
Posterior mean relative risk Posterior sd of relative risk RR SD <0.5 >2.0 <0.2 >1.2

38 Posterior probability that relative risk > 1
Posterior mean relative risk Posterior probability that relative risk > 1 RR Prob <0.5 >2.0 <0.25 >0.75

39 Atlas of cancer mortality at municipality level in Spain 1989-1998 (López-Abente et al., 2007)

40

41 Outline Introduction Mapping and spatial smoothing of health data
Classifying areas and summarising geographical variations in health outcomes Modelling and mapping multiple outcomes Modelling and mapping temporal trends Hierarchical related regression models for combining individual and small area data

42 Classifying areas with excess risk
Richardson et al (2004): Simulation study investigating use of posterior probabilities in disease mapping studies Classify an area as having an elevated risk if [Prob (Ri > 1)] > 0.8 High specificity (false detection < 10%) Sensitivity 60%-95% for Ei of 5-20 and true Ri of

43 Childhood leukaemia in London
Posterior mean RR Posterior prob(RR>1) <0.5 >2.0 <0.8 >0.8

44 Comparison of SaTScan and Bayesian classification rule
SaTScan (Kuldorff, Location of most likely cluster Bayesian: Probability of excess risk Most likely cluster; p<0.001 2nd most likely cluster; p = 0.2 < 0.8 ≥ 0.8

45 Summarising geographic variation
Often interested in providing overall summary measure of variability between areas, e.g. to compare variability of different outcomes to quantify how much variation can be explained by covariates Percentile Ratio: Ratio of outcomes (relative risks) in areas ranked at the qth and (100-q) th percentiles e.g. 90th Percentile Ratio, PR90 = R95%/R5% Posterior distribution of PR90 easily calculated from MCMC output

46 Relative survival from colon cancer, England
Data Survival/censoring times for all 7007 cases of colon cancer diagnosed in England in 1995 and followed for 5 years (provided by B Rachet, LSHTM) Covariates: sex, age at diagnosis, clinical stage, deprivation score, Health Authority (95 area, cases per HA) Population mortality rates by age and sex for England and Wales, Questions of interest Is there evidence of differences between Health Authorities in relative survival that may indicate differences in effectiveness of care received? Relative survival measures difference between age/sex-adjusted mortality rate in general population and in patients with disease of interest How do these geographical differences change when we adjust for socioeconomic deprivation and clinical stage of cancer?

47 Relative survival from colon cancer, England
ykit ~ Poisson(mkit) (subject k, area i, time interval t) log(mkit – Ekit) = log nkit + at + bxki + Hi Standard model for relative survival Area spatial effect

48 Relative survival from colon cancer, England
ykit ~ Poisson(mkit) (subject k, area i, time interval t) log(mkit – Ekit) = log nkit + at + bxki + Hi Area spatial effect Standard model for relative survival Without adjustment for deprivation and clinical stage After adjustment for deprivation and clinical stage Relative excess mortality <0.85 ≥1.15 PR90 = 1.95 (95 % CI ) PR90 = 1.83 (95 % CI )

49 Ranking and classifying extreme areas
Interest in ranking areas for e.g. policy evaluation, ‘performance’ monitoring Rank of a point estimate is highly unreliable Would like to measure uncertainty about rank Straightforward to calculate posterior distribution of ranks (or any function of parameters) using MCMC Obtain interval estimates for ranks Can also calculate posterior probability that each area is ranked above a particular percentile

50 Rank (posterior mean and 95% CI) of the 95 Health Authorities
Without adjustment for deprivation and clinical stage After adjustment for deprivation and clinical stage Upper quartile Upper quartile Rank Rank

51 Posterior probability that HA is ranked in top 5%
Without adjustment for deprivation and clinical stage After adjustment for deprivation and clinical stage 0.0 >0.5

52 Outline Introduction Mapping and spatial smoothing of health data
Classifying areas and summarising geographical variations in health outcomes Modelling and mapping multiple outcomes Modelling and mapping temporal trends Hierarchical related regression models for combining individual and small area data

53 Joint spatial variation in risk of
multiple diseases Disease 1 Disease 2 RR Specific component 2 Specific component 1 Shared component Knorr-Held and Best (2001)

54 Statistical model Y1i ~ Poisson(R1i E1i); log R1i = Si + U1i
Si ~ spatial model (shared component of risk) U1i ~ spatial model (component of risk specific to disease 1) U2i ~ spatial model (component of risk specific to disease 2) Extends to >2 diseases (Tzala and Best, 2006) Extends to shared variations in space and time (Richardson et al, 2006) © Imperial College London

55 Joint variation in COPD and lung cancer in GB
COPD SMR Lung cancer SMR Best and Hansell (2009)

56 Modelled risk estimates
Shared risk COPD specific risk Shared risk interpreted as mainly reflecting geographical variations in community-level smoking behaviour COPD specific risk interpreted as reflecting smoking-adjusted variations in COPD mortality

57 Joint variation in relative survival of colon and breast cancer by English Health Authority
Shared spatial patterns of relative survival may reflect variations in effectiveness of health care system Observed 5-year relative survival: Breast Observed 5-year relative survival: Colon < 65% 65% to 70% 70% to 75% 75% to 80% >80% < 20% 20% to 30% 30% to 40% 40% to 50% >50%

58 Posterior Prob that shared difference > 0
Difference in relative survival in each HA compared to England as a whole Posterior Prob that shared difference > 0 Shared difference < -30% -15% to -30% -15% to 15% 15% to 30% >30% < 0.2 0.2 – 0.8 > 0.8

59 Difference in relative survival in each HA compared to England as a whole
Difference specific to breast cancer Difference specific to colon cancer < -30% -15% to -30% -15% to 15% 15% to 30% >30% < -30% -15% to -30% -15% to 15% 15% to 30% >30%

60 Cancer-specific spatial residuals
Diet-related cancers in Greece oesophagus stomach colorectal pancreas prostate bladder Cancer-specific spatial residuals Spatial common factor 1 2 3 4 5 Cut-points based on quintiles of distribution of factor values and of residuals across all cancers Tzala and Best (2006)

61 Outline Introduction Mapping and spatial smoothing of health data
Classifying areas and summarising geographical variations in health outcomes Modelling and mapping multiple outcomes Modelling and mapping temporal trends Hierarchical related regression models for combining individual and small area data

62 Extensions of disease mapping to space time modelling
Noisy data in each area Noise model: Poisson/Binomial Latent structure: Space + Time + (Residuals) + joint Bayesian estimation Inference

63 Basic space-time model set-up
Yit ~ Poisson(Rit Eit); log Rit = Si + Tt + Uit Si ~ spatial CAR model (common spatial pattern) Tt ~ random walk (RW) model (common temporal trend) Uit ~ Normal(0, v) (space-time residual reflecting idiosyncratic variation) Extends to shared variations of 2 outcomes in space and time © Imperial College London

64 Space-time variations in Male and Female lung cancer incidence (Richardson et al, 2006)
Lung cancer, with its low survival rates, is the biggest cancer killer in the UK Over one fifth of all cancer deaths in UK are from lung cancer (25% for male and 18% for female) Major risk factor is smoking. Smoking time trends different for men/women: uptake of smoking started to decrease in cohorts of men after 1970, while for women the levelling off was later, after 1980 Other risk factors include exposure to workplace agents, radon, air pollution … Interested in similarity and specificity of patterns between men and women

65 Space-time analysis of Male and Female lung cancer incidence
Male/Female lung cancer incidence in Yorkshire:81-85, 86-90, 91-95, (Richardson, Abellan, Best, 2006)

66 Shared and specific patterns and time trends
Time trend for male RRs in 10 wards Shared component Female/Male differential Time trend for female RRs

67 Detection of space-time interaction patterns

68 Detection of space-time interaction patterns
Noisy data in each area Noise model: Poisson/Binomial Latent structure: Space + Time + (Residuals) + joint Bayesian estimation Inference

69 Detection of space-time interaction patterns
Noisy data in each area Noise model: Poisson/Binomial Latent structure: Space + Time + Interactions Any patterns? + + joint Bayesian estimation Inference

70 Detection of space-time interaction patterns
Study the persistence of patterns over time Interpreted as associated with stable risk factors, environmental effects, socio-economic determinants Highlight unusual patterns, via the inclusion of space time interaction terms, which are modelled by a mixture model Unusual patterns in some areas may be linked to recording changes, emerging environmental hazards, impact of new policy or intervention program, … a general tool for surveillance ?

71 Detection of space-time interaction patterns
Yit ~ Poisson(Rit Eit); log Rit = Si + Tt + Uit Si ~ spatial CAR model (common spatial pattern) Tt ~ random walk (RW) model (common temporal trend) Uit ~ Normal(0, v) (space-time interaction; idiosyncratic variation) © Imperial College London

72 Detection of space-time interaction patterns
Yit ~ Poisson(Rit Eit); log Rit = Si + Tt + Uit Si ~ spatial CAR model (common spatial pattern) Tt ~ random walk (RW) model (common temporal trend) Uit ~ q Normal(0, v1) +(1-q) Normal(0, v2); v2 > v1 (mixture model to characterise ‘stable’ and ‘unstable’ patterns over time) Compute posterior probability, pit, that interaction parameter Uit comes from the Normal(0, v2) component Classify area as ‘unstable’ if pit > 0.5 for at least one time, t (simulation study → 10% false positive rate; 20% false negative rate) © Imperial College London

73 Detecting unusual trends in congenital anomalies rates in England (Abellan et al 2008)
Annual postcoded data on congenital anomalies (non chromosomal) recorded in England for the period 1983 – 1998 Annual postcoded data on total number of live births, still births and terminations 136,000 congenital anomalies  84.5 per 105 birth-years Congenital anomalies are sparse:  Grid of 970 grid squares with variable size, to equalize the number of births and expected cases per area Variations could be linked to socio-economic or environmental risk factors or heterogeneity in recording practices Interest in characterising space time patterns © Imperial College London

74 Congenital anomalies in England, 1983-1998
Spatial main effect: evidence of spatial heterogeneity, linked to deprivation and maternal age Temporal main effect: downward trend around 1990 reflects implementation of “minor anomalies” exclusion policy

75 Congenital anomalies: Space-time interactions
Most areas are stable (cluster 1) Some have a change around where modifications in the classification of anomalies occurred (clusters 2 and 3) Identified one very unusual time profile due to a change of local recording practice

76 Outline Introduction Mapping and spatial smoothing of health data
Classifying areas and summarising geographical variations in health outcomes Modelling and mapping multiple outcomes Modelling and mapping temporal trends Summary Hierarchical related regression models for combining individual and small area data

77 Summary Smoothing of small area risks is important to help separate ‘signal’ (spatial pattern) from ‘noise’ Allows meaningful inference even when data are sparse Achieved by ‘borrowing’ information from neighbouring regions Bayesian hierarchical modelling provides formal method for carrying out this ‘borrowing of information’ Provides rich output for statistical inference (estimation, quantification of uncertainty, hypothesis testing) But, depends on “structural” assumptions built into the model (e.g. spatial dependence) Computationally intensive © Imperial College London

78 Summary Bayesian approach extends naturally to allow:
Adjustment for covariates (see later) Joint mapping of 2 or more health outcomes Joint modelling of spatial and temporal variation

79 Benefits of Joint Analysis of related health outcomes
Joint analysis of two related health outcomes is of interest in several contexts: Epidemiology: quantify ‘expected’ variability linked to shared risk factors and tease out specific patterns Health planning: assess the performance of the health system, e.g. for health outcomes linked to screening policies Data quality issues: uncover anomalous patterns linked to a data source shared by several outcomes

80 Benefits of Space Time Analysis for (non-infectious) health outcomes
Study the persistence of patterns over time Interpreted as associated with stable risk factors, environmental effects, distribution of health care access … Highlight unusual patterns in time profiles via the inclusion of space-time interaction terms Time localised excesses linked to e.g. emerging environmental hazards with short latency Variability in recording practices Increased epidemiological interpretability Potential tool for surveillance

81 Outline Introduction Mapping and spatial smoothing of health data
Classifying areas and summarising geographical variations in health outcomes Modelling and mapping multiple outcomes Modelling and mapping temporal trends Hierarchical related regression models for combining individual and small area data

82 Introduction Models and applications discussed so far have focused on:
describing geographical and temporal patterns in health outcomes partitioning sources of variation into e.g. systematic and idiosyncratic, spatial and temporal, shared and specific, … Growing interest in trying to explain geographical variations at level of areas and individuals Build regression models linking health outcomes and explanatory variables

83 Regression models for small area data
Standard regression for individual-level outcomes Individual exposure Aggregate exposure Individual outcome yij xij Zi, Xi Aggregate exposure Yi Aggregate outcome Ecological regression Zi, Xi Aggregate outcome Individual exposure Aggregate exposure Individual outcome yij xij Yi Hierarchical Related Regression (HRR; Jackson et al, 2006, 2008a,b) Zi, Xi

84 Case study: Socioeconomic inequalities in health
Jackson, Best and Richardson (2008b) Geographical inequalities in health are well documented One explanation is that people with similar characteristics cluster together, so area effects are just the result of differences in characteristics of people living in them (compositional effect) But, evidence suggests that attributes of places may influence health over and above effects of individual risk factors (contextual effect) economic, environmental, infrastructure, social capital/cohension Question Is there evidence of contextual effects of area of residence on risk of limiting long term illness (LLTI) and heart disease, after adjusting for individual-level socio-demographic characteristics

85 Data and Methodological Issues
Surveys typically contain sparse individual data per area so difficult to estimate contextual effects Can’t separate individual and contextual effects using only aggregate data (ecological bias) Improve power and reduce bias by combining data using new class of multilevel models developed by BIAS Our goal: data synthesis using Individual-level data Health Survey for England, Area-level (electoral ward) data 1991 census small-area statistics Hospital Episode Statistics

86 Data sources INDIVIDUAL DATA AREA (WARD) DATA
Health Survey for England Self-reported limiting long term illness Self reported hospitalisation for heart disease age and sex ethnicity social class car access income etc. AREA (WARD) DATA Census small area statistics Carstairs deprivation index (area-level material deprivation) Ward codes made available under special license Individual-level Health outcomes Contextual effect Individual predictors

87 Multilevel model for individual data
yij = disease (1) / no disease (0) xij = non-white (1) / white (0) Zi = deprivation score b c m,v2 logit pij = ai + b xij + c Zi ai ~ Normal(m, v2) yij ~ Bernoulli(pij), person j, area i ai xij yij b = relative risk of disease for non-white versus white individual c = contextual effects ai = “unexplained” area effects Zi person j area i

88 Results from analysis of individual survey data: Heart Disease (n=5226)
Area deprivation No car Social class IV/V Non white Univariate regression Multiple regression

89 Results from analysis of individual survey data: Limiting Long Term Illness (n=1155)
Area deprivation Female Non white Doubled income Univariate regression Multiple regression

90 Comments CI wide and not significant for most effects
Some evidence of contextual effect of area deprivation for both heart disease and LLTI Adjusting for individual risk factors (compositional effects) appears to explain contextual effect for heart disease Unclear whether contextual effect remains for LLTI after adjustment for individual factors Survey data lack power to provide reliable answers about contextual effects What can we learn from aggregate data?

91 Aggregate versions of individual predictors
Area-level data AREA (WARD) DATA Census small area statistics Carstairs deprivation index population count by age and sex proportion reporting LLTI proportion non-white proportion in social class IV/V proportion with no car access PayCheck (CACI) mean & variance of household income Hospital Episode Statistics number of admissions for heart disease Contextual effect Aggregate health outcomes & denominators Aggregate versions of individual predictors

92 Standard ecological regression model
Yi = number with disease Ni = population Xi = proportion non-white Zi = area deprivation score M,V2 B C Yi ~ Binomial(qi, Ni), area i logit qi = Ai + BXi + CZi Ai ~ Normal(M, V2) Ai B = association between disease prevalence and proportion non-white C = contextual effects Ai = “unexplained” area effects Zi Yi Xi Ni area i

93 Comparison of individual and ecological regressions: Heart Disease
Area deprivation Ecological No car Social class IV/V Non white

94 Comparison of individual and ecological regressions: Limiting Long Term Illness
Area deprivation Ecological Female Non white Doubled income

95 Ecological bias Ecological bias (difference between individual and aggregate level effects) can be caused by: Confounding confounders can be area-level (between-area) or individual-level (within-area). → include control variables and/or random effects in model Non-linear covariate-outcome relationship, combined with within-area variability of covariate No bias if covariate is constant in area (contextual effect) Bias increases as within-area variability increases …unless models are refined to account for this hidden variability

96 Standard ecological regression model
Yi = number with disease Ni = population Xi = proportion non-white Zi = area deprivation score M,V2 B C Yi ~ Binomial(qi, Ni), area i logit qi = Ai + BXi + CZi Ai ~ Normal(M, V2) Ai B = association between disease prevalence and proportion non-white C = contextual effects Ai = “unexplained” area effects Zi Yi Xi Ni area i

97 Integrated ecological regression model
Yi = number with disease Ni = population Xi = proportion non-white Zi = area deprivation score Average of the individual probabilities of disease, pij, in area i m,v2 b c Yi ~ Binomial(qi, Ni), area i qi =  pij(xij,Zi,ai, b, c)fi(x)dx ai ~ Normal(m, v2) ai b = relative risk of disease for non-white versus white individual c = contextual effects ai = “unexplained” area effects Zi Yi Xi Ni area i

98 Combining individual and aggregate data
Multilevel model for individual data Integrated ecological model b c m,v2 m,v2 b c ai ai xij yij Zi Zi Yi Xi person j Ni area i area i

99 Combining individual and aggregate data
Hierarchical Related Regression (HRR) model (Jackson, Best, Richardson, 2006, 2008a,b) b c m,v2 Joint likelihood for yij and Yi depending on shared parameters ai, b, c, m, v2 ai xij yij Zi Yi Xi person j Ni area i

100 Combining individual and aggregate data
Hierarchical Related Regression (HRR) model (Jackson, Best, Richardson, 2006, 2008a,b) b c m,v2 Joint likelihood for yij and Yi depending on shared parameters ai, b, c, m, v2 Estimation carried out using R software (maximum likelihood) or WinBUGS (Bayesian) ai xij yij Zi Yi Xi person j Ni area i

101 Comparison of results from different regression models: Heart Disease
Individual Area deprivation Standard ecological Integrated ecological No car HRR Social class IV/V PR95 = 10.1; 95% CI(5.3, 18.1) PR95 = 4.2; 95% CI(3.6, 5.1) Non white

102 Comparison of results from different regression models: Limiting Long Term Illness
Individual Area deprivation Standard ecological Integrated ecological Female HRR PR75 = 2.7; 95% CI(1.7, 4.1) PR75 = 2.9; 95% CI(2.4, 3.7) Non white Doubled income

103 Comments Integrated ecological model yields odds ratios that are consistent with individual level estimates from survey Large gains in precision achieved by using aggregate data Significant contextual effect of area deprivation for LLTI but not heart disease More unexplained between-area variation (PR95) for heart disease than LLTI Little difference between estimates based on aggregate data alone and combined individual + aggregate data Individual sample size very small (~0.1% of population represented by aggregate data) In other applications with larger individual sample sizes and/or less informative aggregate data, combined HRR model yields greater improvements (simulation study)

104 Strengths of HRR approach……
Aims to provide individual-level inference using aggregate data by: Fitting integrated individual-level model to alleviate one source of ecological bias Including samples of individual data to help identify effects Uses data from all geographic areas (wards, constituencies), not just those in the survey Improves precision of parameter estimates Improves ability to investigate contextual effects

105 …..and limitations of HRR approach
Integrated individual-level model relies on large contrasts in the predictor proportions across areas e.g. limited variation in % non-white across constituencies: (median 2.7%, 95th percentile 33) Our estimates may not be completely free from ecological bias (Jackson et al, 2006) If individual level data too sparse, may be overwhelmed by aggregate data

106 Data requirements for HRR models
Individual data requires geographical (group) identifiers for individual data Aggregate data requires large exposure contrasts between areas requires information on within-area distribution of covariates Important to check compatibility of different data sources when combining data

107 Thank you for your attention
Acknowledgements: Sylvia Richardson, Juanjo Abellan, Virgilio Gomez-Rubio, Chris Jackson Training courses in Bayesian Analysis of Small Area Data using WinBUGS and INLA, London, July See for details

108 References Abellan JJ, Richardson S and Best N. Use of space-time models to investigate the stability of patterns of disease. Environ Health Perspect 116(8), (2008), Best N and Hansel A. Geographic variations in risk: adjusting for unmeasured confounders through joint modelling of multiple diseases. Epidemiology, 20(3), (2009), Best, N.G., Richardson, S. and Thomson, A. A comparison of Bayesian spatial models for disease mapping. Statistical Methods in Medical Research 14, (2005), Jackson C, Best N and Richardson S. Hierarchical related regression for combining aggregate and individual data in studies of socio-economic disease risk factors. J Royal Statistical Society Series A: Statistics in Society 171 (2008a) , Jackson C, Best N and Richardson S. Studying place effects on health by synthesising individual and area-level outcomes. Social Science and Medicine, 67, (2008b), Jackson C, Best N and Richardson S. Improving ecological inference using individual-level data. Statistics in Medicine, 25, (2006), Knorr-Held, L. and Best, N.G., A shared component model for detecting joint and selective clustering of two diseases, Journal of the Royal Statistical Society, Series A 164, (2001), Richardson, S., Thomson, A., Best, N.G. and Elliott, P. Interpreting posterior relative risk estimates in disease mapping studies. Environmental Health Perspectives 112, (2004), Richardson, S., Abelan, J.J.,and Best, N. Bayseian spatio-temporal analysis of joint patterns of male and female lung cancer in Yorkshire (UK). Statistical Methods in Medical Research 15, (2006), Tzala, E. Best, N. Bayesian latent variable modelling of multivariate spatio-temporal variation in cancer mortality. Stat Methods Med Res 17, (2008),

109 Spanish disease atlases and other related resources
López-Abente G, Pollan M, Escolar A, Errezola M, Abraira V Atlas of cancer mortality and other causes of death in Spain Madrid: Fundación Científica de la Asociación Española contra el Cáncer. Maps of SMRs and Directly Standardised Rates at province level Benach J, Yasui Y, Borrell C, Rosa E, Pasarin MI, Benach N, Español M, Martinez JM, Daponte A Atlas of mortality at small area level in Spain Barcelona: Universitat Pompeu Fabra. Empirical Bayes smoothing of relative risks at municipality level (aggregated some municipalities to reduced sparseness). All of the following produce maps of Bayesian smoothed relative risks at municipality level: López-Abente G, Ramis R, Pollán M, Aragonés N, Pérez-Gómez B, Gómez-Barroso D, Carrasco JM, Lope V, García-Pérez J, Boldo E, García-Mendizabal MJ Atlas of cancer mortality at municipality level in Spain Área de Epidemiología Ambiental y Cáncer del Centro Nacional de Epidemiología, ISCIII. Botella P, Zurriaga O, Posada de la Paz M, Martinez-Beneito MA, Bel E, Robustillo A, Ramalle E, Duran E, Sanchez-Porro P National-Provincial Atlas of Rare Diseases Martinez-Beneito MA, Lopez-Quilez A, Amador A, Melchor I, Botella P, Abellan C, Abellan JJ, Verdejo F, Zurriaga O, Vanaclocha H, Escolano M Atlas of Mortality in the Valencian Region Benach J, Matinez JM, Yasui Y, Borrell C, Pasarin MI, Español E, Benach N Atlas of mortality at small area level in Catalonia Barcelona: Universitat Pompeu Fabra / Fundació Jaume Bofill / Editorial Mediterrània

110 Spanish disease atlases and other related resources
DEMAP group, Andalusian School of Public Health, Granada. Produced interactive mortality atlas for Andalucia, and socioeconomic indices at municipality level for Spain. See MEDEA: Research network on Epidemiology and Public Health, working on socioeconomic and environmental inequalities in health at small area level. See VPM Atlas Project (Atlas de Variaciones en la Práctica Médica). Studying and mapping variations in provision and usage of health care at small area level in 16 of the 17 regions of Spain, using data on hospital discharges. See

111

112 Smoothing of the RRs of hot spots (4 contiguous areas with average expected counts ≈ 5) for different spatial models Richardson et al (EHP, 2004) True RR = 3 True RR = 2


Download ppt "Models for small area data with applications in health care"

Similar presentations


Ads by Google