JRSM >
Volume 99, Number 2 >
Pp. 81-89
1 Lecturer in Primary Care, Department of Primary Care and Social Medicine,
Imperial College Faculty of Medicine, London W6 8RP
2 Professor of Primary Care, Department of Primary Care and Social Medicine,
Imperial College Faculty of Medicine, London W6 8RP
3 Public Health Information Analyst, London Health Observatory, London
4 Lead Analyst, London Health Observatory, London
5 Lecturer in Medical Statistics, Medical Statistics Unit, Research and
Development Directorate, University College Hospitals NHS Trust, London,
UK
Correspondence to: Dr Sonia Saxena E-mail: s.saxena{at}imperial.ac.uk
 | SUMMARY |
|---|
 TOP SUMMARY INTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Design: Cross sectional analysis at primary care trusts level using routine data from multiple sources.
Setting: All 31 primary care trusts in London with a total resident population of 7 million patients.
Main outcome measures: Age-standardized hospital admission rates for asthma, diabetes, heart failure, hypertension and chronic obstructive pulmonary disease.
Results: Admission rates varied widely for the conditions examined across the 31 primary care trusts. In 2001, age adjusted admission rates for asthma varied from 76 to 189 per 100 000 and for diabetes from 38 to 183 per 100 000. There was a significant association between higher admission rates and measures of underlying ill health and material deprivation but not quantitative measures of primary care service provision. Provision of specialist chronic disease services in primary care for diabetes but not for asthma were significantly associated with reduced admission rates. There was no association of prescribing levels in primary care trusts with admission rates for any of the conditions examined.
Conclusions: Although hospital admission for some chronic diseases is potentially avoidable and rates of hospital admission for these conditions are possible indicators of the quality of care, they should be interpreted in conjunction with measures of population composition and deprivation. Failure to do this may result in primary care trusts and general practitioners being criticized for aspects of health care utilization that are not under their direct control.
| INTRODUCTION |
|---|
|
TOPSUMMARYINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Work from the USA has suggested that hospital admission rates are a marker of poor primary care.3 Hence, there has emerged the notion of a preventable or avoidable admission, which has been used to indicate poor quality of care in primary care.4 A number of initiatives have tried, both in the UK and elsewhere, to increase the management of chronic diseases in primary care and reduce hospital admission rates.5 Since 1990, the UK government has introduced numerous targets for the National Health Service aimed at improving access to high quality primary care and specialist services and reducing waiting times for hospital treatment.6,7 Health services have been extensively reorganized to shift responsibilities from the secondary care sector to primary care. In England's NHS, Primary Care Trusts are now responsible for a number of activities including planning and commissioning services, managing budgets and demonstrating health improvement by meeting centrally set targets that will rank and compare primary care trusts performance nationally.8 In the most recent change, the new general practitioner contract9 sets out quality indicators that reward individual practices for achieving targets in managing key chronic diseases that account for a large proportion of morbidity and mortality in the UK and which are also expensive to treat.7,10
The notion of avoidable admissions, however, rests on the assumption that provision of good primary care alone can drive down hospital admission rates. There are a number of other reasons, however, why chronic disease may be harder to manage in certain areas. The distribution of chronic conditions may vary widely within the population, for example, in urban areas where there are higher percentages of resident South Asians, one would expect to see a higher prevalence of diabetes and coronary heart disease.11,12 Mortality from coronary heart disease and chronic obstructive pulmonary disease is higher in deprived areas and disease severity is greater among disadvantaged groups.13,14 Differential access to care and distribution of services may also affect hospital admission rates15 and in some areas care at home may not be feasible for reasons unrelated to health status or provision.16 Hence, different primary care trusts populations have different health needs and basing the measurement of primary care trusts performance on admissions must allow for this variation and, some argue, attempt to direct resources to tackle these inequalities. Previous UK studies suggested that many practices are starting from very different baselines with deprivation, poor health and underdeveloped care accounting for variation in admission rates to hospital.17-19,20
We aimed to test the hypothesis that higher levels of underlying ill health in the population, material deprivation and lower levels of primary care service provision are each associated with increased rates of potentially avoidable hospital admissions in primary care trusts in London. We selected London for study because it has an ethnically and socio-economically diverse population.21 We selected five key conditions: asthma, diabetes, heart failure, hypertension and chronic obstructive pulmonary disease because these are conditions that contribute significantly to the healthcare burden in the UK22 and that previous research has identified as conditions where the risk of hospital admission was influenced by the quality of their treatment in primary care and which were thus potentially avoidable.3
| METHODS |
|---|
|
TOPSUMMARYINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Census We obtained data from the 2001 Census23 supplied by the Office for National Statistics. The data included population characteristics including age and sex profiles, proportion of elderly living alone, proportion of lone parents, percentage ethnic minority residents and deprivation scores for all the output areas (formerly known as `enumeration districts') in London.
Deprivation The Department for Environment, Transport and the Region's Index of Multiple Deprivation for 2000 was used to ascribe a deprivation score to each primary care trusts calculated by averaging the ward scores in each district after they have been population weighted. This measure describes the primary care trusts as a whole, taking into account the full range of ward scores across a primary care trusts. The advantage of the average of ward score measure is that it describes the wards while retaining the fact that the more deprived wards may have more extreme scores, which can be obscured if ranks are used.
Deaths Data from Deaths Registration collected by the Office for National Statistics were used to calculate condition-specific mortality rates for each of the five relevant conditions. Rates were age-standardized per 100 000 persons from the average number of deaths per year during 1999-2001 as a proportion of the number of resident patients in 2001. This was used as a proxy measure of underlying ill health in the primary care trusts.
Prescriptions The Department of Health provided aggregated data from the Prescription Pricing Authority's Prescribing Analysis and Cost scheme data24 on prescriptions dispensed. We obtained prescribing information on all primary care trusts in London for the calendar year 2000/2001 on groups of drugs for respiratory disease, diabetes and cardiovascular drugs by selecting relevant codes from the British National Formulary (Appendix A).
|
Primary care Information on the total number of general practitioners within each primary care trusts, the average list size, the number of GPs with lists greater than 2500 patients and those offering health promotion clinics for asthma and diabetes were obtained from the National Database for Primary Care Groups and Trusts, developed by the National Primary Care Research and Development Centre at the University of Manchester.
Hospital admissions Data from the Hospital Episode Statistics25 were used to calculate hospital admission rates per 100 000 resident population for the financial year 2001/2002 for each of the 31 London primary care trusts defined above. Included were the total numbers of first (elective and emergency) finished consultant episodes during the calendar year 2000-2001 for five chronic diseases: asthma (International Classification of Diseases 10 codes J45 & J46), diabetes (E10-14), heart failure (I50), hypertension (I10-15) and chronic obstructive pulmonary disease (J42-44).26 Age-standardized rates were calculated using the total number of patients of all ages admitted in 2001 divided by the total number of patients of all ages resident at the primary care trusts in 2001.
Statistical analysis
Age-standardized admission rates, prescribing rates, population and
practice characteristics are summarized using medians and interquartile ranges
(Tables 1 and
2). Because admission rates
were not normally distributed, we used median quantile
regression27 to
investigate the association between age-standardized admission rates and
practice, prescribing and population factors, based on data for all 31 primary
care trusts. We restricted our analysis to univariate associations because
with 31 units of observation, we could adjust only for one or two parameters,
the most important of these being age.
|
|
| RESULTS |
|---|
|
TOPSUMMARYINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
|
|
|
|
Factors associated with admission rates
Table 3 shows quantile
regression coefficients describing the relationship between age-standardized
admission rates and population, GP and prescribing characteristics of the
primary care trusts. Underlying mortality (as a crude proxy for burden of
illness) was significantly associated with increased hospital admission rates
for chronic obstructive pulmonary disease (regression coefficient 4.7, 2.3 to
7.2). Deprivation was associated with increased hospital admission rates for
all the conditions studied. Lone parenthood, which has been used as an
indicator of the presence of vulnerable groups in the population, showed a
significant association with higher admission rates for diabetes (regression
coefficient 27.0, 5.5 to 48.4) and chronic obstructive pulmonary disease
(regression coefficient 34.8, 11.1 to 58.5). Percentage of elderly living
alone was strongly negatively associated with hospital admissions rates for
asthma, heart failure, hypertension and chronic obstructive pulmonary disease
(Table 3). primary care trusts
with higher ethnic minority populations had significantly higher hospital
admission rates for heart failure (regression coefficient 0.99, 0.64-1.35) and
hypertension (regression coefficient 0.27, 0.09-0.44).
|
Although for asthma there was some evidence of a decrease in admission rates with larger list sizes, in general practice supply factors such as total number of GPs and proportion of practices with higher list sizes were not significantly associated with hospital admission rates. The provision of specialist services for diabetes management in primary care was significantly associated with decreased hospital admission rates for diabetes (regression coefficient 71.6, 72.6 to 70.7). However, the same was not true for asthma management. There were no significant associations between prescribing rates and admission rates for any of the disorders studied.
| DISCUSSION |
|---|
|
TOPSUMMARYINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Comparison with other studies
Our findings support those of earlier studies in general practice in London
showing hospital admission rates were associated with prevalence of chronic
illness and deprivation. These studies called for routine standardization of
hospitalization rates for population and hospital
characteristics.18,28
Other studies from different healthcare systems have also reported that
failing to consider patient characteristics and health status may lead to
false conclusions that care is of poor
quality3 but suggest
that organizations that focus on better integration between primary and
secondary care and patient self management cost less and result in fewer
hospitalizations.5
Our study provides an overview of variations in admissions across London but cannot examine differences relating to the severity of illness of patients being admitted. We focused on crude outcomes and included both emergency and elective admissions, unlike one study where residents from deprived areas were found more likely to be admitted as an emergency for common cancers,29 which may indicate poorer prognosis or less social support to keep patients out of hospital. One possible interpretation of our finding could be that admission rates for all chronic diseases were lower in primary care trusts with a higher proportion of elderly living alone. This may reflect a higher level of independence, with accompanying support structures, in the elderly population in affluent areas. Another study showed unusually high rates of hospital admission for asthma in a deprived area of east London that were strongly associated with smaller practice size and higher rates of night visiting. This study predated the NHS Direct service and other changes in out-of-hours care, which may be affecting the admission rates described in the current study.20
Strengths
Ours is among the first large UK population-based studies to examine the
variation in admission rates for diseases where admission is potentially
avoidable at primary care trusts level. The study incorporates a wide range of
data from multiple routinely-collected but not routinely-available sources
including the latest census estimates. The five conditions examined are of
central importance to the Department of Health's agenda for health improvement
and the methodology of using multiple data sources can be applied to conduct
future research.
Limitations
There is a small bias in ascribing primary care supply factors to
particular primary care trusts because some patients may be registered with a
GP in one primary care trusts but be resident in another neighbouring primary
care trusts. Additionally, several explanatory factors are crude measures of
population deprivation. For example, the deprivation scores are based on
average ward scores for primary care trusts and are not related to individual
measures of deprivation for those admitted to hospital. Ill health measures
were not available by primary care trusts; using primary care trusts mortality
rates for the key conditions was a crude measure of health status.
Furthermore, the relatively small numbers of asthma deaths meant that it was
not possible to compute asthma mortality.
Using large routinely collected data is subject to a number of potential biases, including those inherent in cross-section designs. In this study we also did not validate data quality. For example, hospital activity data can be affected by misclassification of diseases such as respiratory infections and chronic obstructive pulmonary disease which may be misclassified as asthma, as well as differential levels of completeness of data between hospital trusts.26 Our study did not include data from patients admitted to private hospitals and this may further bias our results by under-counting admissions in more affluent primary care trusts. Although much of the data in our study is population-based, the purpose was to examine differences among the 31 primary care trusts in London and thus the statistical analyses possible were limited to examining simple associations between outcomes.
Implications and future research
A key finding of the study is that higher hospitalization rates across
primary care trusts in London are associated with population profiles and
measures of deprivation. This suggests that additional resources are needed if
primary care trusts in deprived areas are to meet the challenges of the UK
government's National Service Frameworks for diabetes and coronary heart
disease.30,31
However, hospital admissions do not exclusively reflect the quality of primary
care or outpatient services and are sensitive to many changes in the ways that
services are organized and delivered.
Service use has also been dependent on patients' health-seeking behaviour. This may change with a move away from a largely responsive service in primary care to a more planned target-driven one with centrally determined targets. A higher proportion of GPs' income under the new contract will be derived from meeting preventive targets,9 which should have implications for prescribing and referring patients for specialist care. Those GPs who are best able to achieve these targets will be those working in practices with stable and compliant patient populations and better IT facilities which may further increase the wide variation in hospital admission rates seen in this study. Emphazising earlier prevention and optimizing management may reduce mortality rates but could conversely result in higher apparent levels of morbidity in the population because of improved case detection. This is also likely to impact on hospitalization rates and other tools for assessing healthcare service performance. Future studies in this field will need to examine the effect of recent large-scale policy changes within primary care including those occurring as a result of the new GP contract. Finally, following on from its implementation in the USA, the UK government intends to introduce ambulatory case management in the UK. Whether this leads to a reduction in admission rates for the conditions for which it is being implemented and reduces the degree of variation in admission rates reported in this paper remains to be seen.
| CONCLUSIONS |
|---|
|
TOPSUMMARYINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
| Acknowledgments |
|---|
| REFERENCES |
|---|
|
TOPSUMMARYINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Bottle, P. Aylin, and A. Majeed Identifying patients at high risk of emergency hospital admissions: a logistic regression analysis. J R Soc Med, August 1, 2006; 99(8): 406 - 414. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
