Waiting times and socioeconomic status: Evidence from England

Abstract

Waiting times for elective surgery, like hip replacement, are often referred to as an equitable rationing mechanism in publicly-funded healthcare systems because access to care is not based on socioeconomic status. Previous work has established that that this may not be the case and there is evidence of inequality in NHS waiting times favouring patients living in the least deprived neighbourhoods in England. We advance the literature by explaining variations of inequalities in waiting times in England in four different ways. First, we ask whether inequalities are driven by education rather than income. Our analysis shows that education and income deprivation have distinct effects on waiting time. Patients in the first quintile with least deprivation in education wait 9% less than patients in the second quintile and 14% less than patients in the third-to-fifth quintile. Patients in the fourth and fifth most income-deprived quintile wait about 7% longer than patients in the least deprived quintile. Second, we investigate whether inequalities arise “across” hospitals or “within” the hospital. The analysis provides evidence that most inequalities occur within hospitals rather than across hospitals. Moreover, failure to control for hospital fixed effects results in underestimation of the income gradient. Third, we explore whether inequalities arise across the entire waiting time distribution. Inequalities between better educated patients and other patients occur over large part of the waiting time distribution. Moreover we find that the education gradient becomes smaller for very long waiting. Fourth, we investigate whether the gradient may reflect the fact that patients with higher socioeconomic status have a different severity as proxied through a range of types and the number of diagnoses (in addition to age and gender) compared to those with lower socioeconomic status. We find no evidence that differences in severity explain the social gradient in waiting times.

Introduction

Waiting times are a major health policy issue in many OECD countries. Average waiting times can reach several months for common procedures like cataract and hip replacement (Siciliani & Hurst, 2004). For example, in England patients waited on average more than 200 days before being admitted for a hip operation in 1997 and more than 100 days in 2007 (Cooper, McGuire, Jones, & Le Grand, 2009). They generate dissatisfaction for patients and the general public. Waiting times postpone and therefore reduce patients’ benefits. They may deteriorate patients’ health status, prolong suffering and generate uncertainty.

In the absence of other rationing mechanisms, waiting times help to bring into equilibrium the demand for and the supply of healthcare by deterring patients with small benefit from demanding treatment (Cullis et al., 2000, Lindsay and Feigenbaum, 1984, Martin and Smith, 1999). Other rationing mechanisms also exist. For example, co-payments might be an alternative to contain moral hazard (Zweifel, Breyer, & Kifmann, 2009, chap. 6). However, co-payments are often perceived as inequitable as poor patients may be deterred from seeking care. In contrast, waiting times are perceived as equitable, as the patients’ costs or disutility generated by waiting do not depend on their ability to pay (while the loss of utility generated by co-payments does). Waiting times should only depend on the need or severity of the patient (i.e. on clinical factors) and not whether the patient has high or low-income, or whether the patient is highly educated. This is in line with National Health Services which provide access to care on the basis of need and not on the basis of ability to pay.

There is some evidence that this may not be the case. Cooper et al. (2009) finds evidence of inequality in NHS waiting times favouring patients living in advantaged neighbourhoods of England (see Section Related literature for a more detailed review of the literature). Therefore, the non-monetary price (waiting time) paid by the different patients is not the same and individuals with higher socioeconomic status pay (wait) less.

This study focuses on hip replacement, a common elective procedure performed on more than 30,000 patients every year in England. The number of hip replacements has been steadily increasing in the past and likely to increase in the future driven by ageing of the population (Stargardt, 2008). To measure waiting times we use administrative data from Hospital Episodes Statistics (HES), which includes patients treated by the publicly-funded National Health Service (NHS) in England. We use HES data for year 2001. We choose this year because our focus is on ‘explaining’ the gradient between socioeconomic status and waiting times. Since data on education and income are available from the Census year of 2001, we choose the administrative data which best match with the Census. We use the skills (education) sub-domain and income domain from the Indices of Multiple Deprivation. Using more recent administrative data (say for year 2010) together with the 2001 census data would reduce the precision of the measurement of income and education because their distribution may have changed over time. Moreover, data on the latest Census are not yet available.

Building on the work by Cooper et al. (2009), we advance the literature by explaining variations of inequalities in waiting times in four different ways. First, we ask whether inequalities are driven by education rather than income. Although income and education are positively correlated (i.e. income-deprived areas are likely to be education-deprived too), there is a substantial number of patients coming from areas with low education and medium income, and from high education and medium income. Our analysis indeed shows that education deprivation and income deprivation have distinct effects on waiting time. More precisely, OLS results suggest that patients in the first quintile with least deprivation in education wait 9% less than patients in the second quintile and 14% less than patients in the third-to-fifth quintile. Moreover, patients in the fourth and fifth most income-deprived quintile wait about 7% longer than patients in the least deprived quintile.

Second, we investigate whether inequalities arise “across” hospitals or “within” the hospital. More precisely, we investigate whether inequalities arise because richer and more educated patients live in areas where hospitals have lower waiting times (across hospitals) or because “within the hospital” patients with higher socioeconomic status are able to obtain shorter waiting times. To investigate this issue we compare the results when we control for supply using hospital fixed effects and when we do not. Waiting times may vary considerably between hospitals, due to variations in capacity, practice style, efficiency and other local factors that are not related to socioeconomic status. If hospitals with short waits tend to be located in urban areas where income-deprived people are more concentrated, omitting hospital effect might underestimate the social gradient in waiting time. The analysis provides evidence that most inequalities occur within hospitals rather than across hospitals. The education gradient exhibits only a small reduction: patients from the least deprived quintile wait 12.8%–13.6% less (as opposed to 15.4%–15.7% less) than patients from the bottom three quintiles. Without hospital fixed effects, we find no income gradient. When controlling for hospital fixed effects, we find that patients in the bottom quintile of the income domain (the most disadvantaged) wait 7.5% longer than those in the highest quintile. This is consistent with the hypothesis that patients from areas most deprived in income are more likely to be treated in hospitals with short waiting times, which in turn may be due to the fact that hospitals with lower waiting times are generally located in urban areas where income-deprived people are concentrated (Noble et al., 2004). In summary, failure to control for hospital heterogeneity results in underestimation of the income gradient.

Third, we explore whether inequalities arise across the entire waiting time distribution. Previous work only identifies an average effect: we do not know whether inequalities persist along the whole distribution. To investigate this issue we use duration analysis in addition to OLS. Although the latter provides easily interpretable results, the former is more appropriate for comparing the duration of states, like the time waited. Duration analysis allows investigating differences in waiting times over the whole distribution of time waited enlarging the scope of inequality analyses. Moreover, it allows for modelling non-normal dependent variables relaxing some parametric assumptions of the linear models. Kaplan–Meier survival curves and estimated hazard functions show that inequalities between better educated patients and other patients occur over large part of the waiting time distribution. Moreover we find that the education gradient becomes smaller for very long waiting.

Fourth, part of the gradient may reflect the fact that patients with higher socioeconomic status may have a different severity compared to those with lower socioeconomic status. Patients are typically prioritised on the waiting list, with more severe patients waiting less (Gravelle and Siciliani, 2008a, Gravelle and Siciliani, 2008b), and severity being (positively) correlated with deprivation. The lack of adequate controls on severity might potentially generate biased results and underestimate the gradient between waiting times and socioeconomic status: for example we could find that there is no gradient (everyone waits the same) when we do not control for severity but that there is one after controlling for severity. We explore whether the gradient is sensitive to the use of accurate controls for patients’ severity using a range of types and the number of diagnoses (in addition to age and gender). We find no evidence that differences in severity (in addition to age and gender) explain the social gradient in waiting times.