Hugh Gravelle

How much of the relation between population mortality and unequal distribution of income is a statistical artefact

The absolute income hypothesis—that holding other factors constant, the higher an individuals income the better is their health—is supported by a considerable body of evidence.1 2 3 However, according to the more recent relative income hypothesis, an individuals health is also affected by the distribution of income within society. Someone with a given income would have worse health if he or she lived in a society with greater inequality of income than in a society in which income is more equally distributed.4 Several recent papers examining the relation between population mortality and income inequality seem to support the relative income hypothesis.5 6 7 8 9 10 11 They suggest that greater inequality is associated with higher population mortality and that this relation persists even when account is taken of the average income of the population. However, some scepticism has been expressed about the relative income hypothesis.12 To quote one of the papers cited above, the “mechanisms underlying the association between income distribution and mortality are poorly understood.”7 There may be a very simple explanation for some, or all, of the reported associations between inequality of income and population health used to support the relative income hypothesis. They may be, at least partly, a statistical artefact caused by using population data rather than individual data. A positive correlation between population mortality and income inequality can arise at aggregate level even if inequality has no effect on the individual risk of mortality. Thus, we do not need the relative income hypothesis to explain the observed associations between population health and income inequality—the absolute income hypothesis will serve. The absolute income explanation can be illustrated with the help of the figure (the mathematical argument is presented in the Appendix 1). In this, the individual risk of mortality depends …

Impact of case management (Evercare) on frail elderly patients: controlled before and after analysis of quantitative outcome data

Objectives To determine the impact on outcomes in patients of the Evercare approach to case management of elderly people. Design Practice level before and after analysis of hospital admissions data with control group. Setting Nine primary care trusts in England that, in 2003-5, piloted case management of elderly people selected as being at high risk of emergency admission. Main outcome measures Rates of emergency admission, emergency bed days, and mortality from April 2001 to March 2005 in 62 Evercare practices and 6960-7695 control practices in England (depending on the analysis being carried out). Results The intervention had no significant effect on rates of emergency admission (increase 16.5%, (95% confidence interval −5.7% to 38.7%), emergency bed days (increase 19.0%, −5.3% to 43.2%), and mortality (increase 34.4%, −1.7% to 70.3%) for a high risk population aged >65 with a history of two or more emergency admissions in the preceding 13 months. For the general population aged ≥65 effects on the rates of emergency admission (increase 2.5%, −2.1% to 7.0%), emergency bed days (decrease −4.9%, −10.8% to 1.0%), and mortality (increase 5.5%, −3.5% to 14.5%) were also non-significant. Conclusions Case management of frail elderly people introduced an additional range of services into primary care without an associated reduction in hospital admissions. This may have been because of identification of additional cases. Employment of community matrons is now a key feature of case management policy in the NHS in England. Without more radical system redesign this policy is unlikely to reduce hospital admissions.

Measuring quality of care with routine data: avoiding confusion between performance indicators and health outcomes

Abstract Objective: To investigate the impact of factors outside the control of primary care on performance indicators proposed as measures of the quality of primary care. Design: Multiple regression analysis relating admission rates standardised for age and sex for asthma, diabetes, and epilepsy to socioeconomic population characteristics and to the supply of secondary care resources. Setting: 90 family health services authorities in England, 1989-90 to 1994-5. Results: At health authority level socioeconomic characteristics, health status, and secondary care supply factors explained 45% of the variation in admission rates for asthma, 33% for diabetes, and 55% for epilepsy. When health authorities were ranked, only four of the 10 with the highest age-sex standardised admission rates for asthma in 1994-5 remained in the top 10 when allowance was made for socioeconomic characteristics, health status, and secondary care supply factors. There was also substantial year to year variation in the rates. Conclusion: Health outcomes should relate to crude rates of adverse events in the population. These give the best indication of the size of a health problem. Performance indicators, however, should relate to those aspects of care which can be altered by the staff whose performance is being measured. Key messages The NHS executive has proposed that admission rates for asthma, diabetes, and epilepsy could be used at health authority level as indicators of the quality of primary care There is considerable year to year variation in the ranking of health authorities by admission rates for these conditions, even when rates are aggregated. This makes it hard to interpret a single years data: a 3 year average would be more reliable Morbidity, socioeconomic characteristics, and secondary care supply are important confounding factors that explain between a third and a half of the variation in admission rates across health authority areas Performance indicators should relate to aspects of care that can be controlled by decision makers. Confounding factors have a clear impact on admission rates and must therefore be taken into account if such rates are to be used as performance indicators

Exclusion of Patients from Pay-for-Performance Targets by English Physicians

BACKGROUND In the English pay-for-performance program, physicians use a range of criteria to exclude individual patients from the quality calculations that determine their pay. This process, which is called exception reporting, is intended to safeguard patients against inappropriate treatment by physicians seeking to maximize their income. However, exception reporting may allow physicians to inappropriately exclude patients for whom targets have been missed (a practice known as gaming). METHODS We analyzed data extracted automatically from clinical computing systems for 8105 family practices in England (96% of all practices), data from the U.K. Census, and data on practice characteristics from the U.K. Department of Health. We determined the rate of exception reporting for 65 clinical activities and the association between this rate and the characteristics of patients and medical practices. RESULTS From April 2005 through March 2006, physicians excluded a median of 5.3% of patients (interquartile range, 4.0 to 6.9) from the quality calculations. Physicians were most likely to exclude patients from indicators that were related to providing treatments and achieving target levels of intermediate outcomes; they were least likely to exclude patients from indicators that were related to routine checks and measurements and to offers of treatment. The characteristics of patients and practices explained only 2.7% of the variance in exception reporting. We estimate that exception reporting accounted for approximately 1.5% of the cost of the pay-for-performance program. CONCLUSIONS Exception reporting brings substantial benefits to pay-for-performance programs, providing that the process is used appropriately. In England, rates of exception reporting have generally been low, with little evidence of widespread gaming.

Follow up of people aged 65 and over with a history of emergency admissions: analysis of routine admission data

Abstract Objective To determine the subsequent pattern of emergency admissions in older people with a history of frequent emergency admissions. Design Analysis of routine admissions data from NHS hospitals using hospital episode statistics (HES) in England. Subjects Individual patients aged  65,  75, and  85 who had at least two emergency admissions in 1997-8. Main outcome measures Emergency admissions and bed use in this “high risk” cohort of patients were counted for the next five years and compared with the general population of the same age. No account was taken of mortality as the analysis was designed to estimate the future use of beds in this high risk cohort. Results Over four to five years, admission rates and bed use in the high risk cohorts fell to the mean rate for older people. Although patients  65 with two or more such admissions were responsible for 38% of admissions in the index year, they were responsible for fewer than 10% of admissions in the following year and just over 3% five years later. Conclusion Patients with multiple emergency admissions are often identified as a high risk group for subsequent admission and substantial claims are made for interventions designed to avoid emergency admission in such patients. Simply monitoring admission rates cannot assess interventions designed to reduce admission among frail older people as rates fall without any intervention. Comparison with a matched control group is necessary. Wider benefits than reduced admissions should be considered when introducing intensive case management of older people.

Income, Income Inequality and Health: What can we Learn from Aggregate Data?

It has been suggested that, especially in countries with high per capita income, there is an independent effect of income distribution on the health of individuals. One source of evidence in support of this relative income hypothesis are analyses of aggregate cross section data on population health, per capita income and income inequality. We examine the empirical robustness of cross-section analyses by using a new data set to replicate and extend the approach in a frequently cited paper. We find that the estimated relationship between income inequality and life expectancy is dependent on the data set used, the functional form estimated and the way in which the epidemiological transition is specifed. The association is never significant in any of our models. We argue there are serious methodological difficulties in using aggregate cross sections as means of testing hypotheses about the effect of income, and its distribution, on the health of individuals.

Discounting and decision making in the economic evaluation of health-care technologies

Discounting costs and health benefits in cost-effectiveness analysis has been the subject of recent debate - some authors suggesting a common rate for both and others suggesting a lower rate for health. We show how these views turn on key judgments of fact and value: on whether the social objective is to maximise discounted health outcomes or the present consumption value of health; on whether the budget for health care is fixed; on the expected growth in the cost-effectiveness threshold; and on the expected growth in the consumption value of health. We demonstrate that if the budget for health care is fixed and decisions are based on incremental cost effectiveness ratios (ICERs), discounting costs and health gains at the same rate is correct only if the threshold remains constant. Expecting growth in the consumption value of health does not itself justify differential rates but implies a lower rate for both. However, whether one believes that the objective should be the maximisation of the present value of health or the present consumption value of health, adopting the social time preference rate for consumption as the discount rate for costs and health gains is valid only under strong and implausible assumptions about values and facts.

Doctor Behaviour under a Pay for Performance Contract: Treating, Cheating and Case Finding?

The UK National Health Service introduced a pay for performance scheme for primary care providers in 2004/5. The scheme rewarded providers for the proportion of eligible patients who received appropriate treatment. Eligible patients were those who had been reported by the provider as having the relevant disease minus those they exception reported as not suitable for treatment. Using rich provider level data, we find that differences in reported disease rates between providers, and differences in exception rates both between and within providers, suggest gaming. Faced with ratio performance indicators, providers acted on denominators as well as numerators.

The practice of discounting in economic evaluations of healthcare interventions

OBJECTIVES Discounting of costs in health-related economic evaluation is generally regarded as uncontroversial, but there is disagreement about discounting health benefits. We sought to explore the current recommendations and practice in health economic evaluations with regard to discounting of costs and benefits. METHODS Recommendations for best practice on discounting for health effects as set out by government agencies, regulatory bodies, learned journals, and leading health economics texts were surveyed. A review of a sample of primary literature on health economic evaluations was undertaken to ascertain the actual current practice on discounting health effects and costs. RESULTS All of the official sources recommended a positive discount rate for both health effects and costs, and most recommended a specific rate (range, 1% to 8%). The most frequently specified rates were 3% and 5%. A total of 147 studies were reviewed; most of these used a discount rate for health of either 0% (n = 50) or 5% (n = 67). Over 90% of studies used the same discount rate for both health and cost. While 28% used a zero rate for both health and cost, in 64% a nonzero rate was used for both. Studies where the health measure was in natural clinical units (direct) were significantly more likely to have a zero discount rate. CONCLUSION The finding that 28% of studies did not discount costs or benefits is surprising and concerning. A lower likelihood of discounting for benefits when they are in natural units may indicate confusion regarding the rationale for discounting health effects.

Inequality in the geographical distribution of general practitioners in England and Wales 1974-1995

Objectives: To compare geographical inequality in the distribution of general practitioners (GPs), other resources and mortality around 1995 in England and Wales; to measure trends between 1974 and 1995 in inequality of GP distribution; to examine the implications of different need adjustments and inequality measures on the degree of geographic inequality; and to analyse the impact of policies (increased supply, area inducements and entry regulation) on inequality. Methods: Measurement of relative inequality (decile ratio, Gini coefficient, Atkinson index) and absolute inequality (standard deviation) in the ratio of GPs to need-adjusted population in former Family Practitioner Committee/Family Health Services Authority areas each year from 1974 to 1995; and relative inequality across areas in the distributions of income, other resources and standardised mortality ratios (SMRs) around 1995. Regression of 1995 GP/population ratios on 1974 ratios. Application of equalising net advantages location model to GP distribution. Results: Inequality in the distribution of GPs in 1995 was less than inequality in other primary care resources, but greater than inequalities in disposable income, SMRs, primary school expenditure, and hospital and community health services expenditure. The decile ratio shows little change between 1974 and 1995. Gini and Atkinson inequality indices indicate some reduction in inequality between 1974 and 1980, but little change thereafter. The standard deviation of need-adjusted provision increased over the period. Areas that had the lowest GP provision in 1974 tended to have the lowest in 1995. Conclusions: The choice between relative and absolute inequality measures and, to a lesser extent, the method of adjusting for need affect conclusions about the trend in inequality. Both types of measure and most need adjustments suggest that the policies adopted did not lead to a reduction in inequality over the period. Interactions between policies may reduce their overall effectiveness.