Abstract

Background Recent avoidable mortality trends in Australia suggest that health care has made a substantial contribution to reducing mortality. This study investigates if the benefits of health care have been distributed equally by comparing declines in avoidable with non-avoidable mortality over time by socioeconomic status (SES).

Methods We calculated avoidable and non-avoidable mortality rates in Australia by small areas for 1986, 1991, 1997 and 2002. We performed pooled cross-sectional trend analysis of indirectly standardized mortality rates by SES and year, modelling using Poisson regression with over-dispersion. Socioeconomic inequalities were quantified using the relative (RII) and slope (SII) index of inequality.

Results The annual percentage decline in avoidable mortality at the higher end of the socioeconomic continuum (5.0%; 95% CI: 4.7–5.4%) was larger than at the lower end (3.5%; 3.2–3.8%), with increasing relative inequality between 1986 (RII = 1.54; 1.46–1.63) and 2002 (RII = 2.00; 1.95–2.06), greater than that in non-avoidable mortality (P = 0.036). In absolute terms, avoidable deaths fell annually by 7.4 (6.9–7.8) and 8.4 (7.9–8.9) deaths per 100 000 at the higher and lower end of the spectrum, respectively, with absolute inequality decreasing between 1986 (SII = 97.8; 87.6–107.9) and 2002 (SII = 81.5; 74.6–88.5).

Conclusions Health care has contributed to decreasing the absolute SES mortality gap. However, advantaged people have obtained a disproportionate benefit of health care, contributing to widening relative health inequalities. A universal heath care system does not guarantee equality in health-care-related outcomes.

Introduction

By international standards, Australia is a healthy nation with life expectancy for males and females amongst the highest in the world.1 However, as in other countries, this high level of health has not been shared equally across socioeconomic groups and gradients in mortality have been reported consistently. Further, relative inequalities in mortality rates have tended to increase in recent decades with the percentage rate of decline in mortality rates generally greater in the least disadvantaged groups than the most disadvantaged groups.2–6

Socioeconomic status (SES) may be associated with mortality because higher SES involves access to resources—including knowledge, money, power, prestige and beneficial social connections—that help people avoid diseases and minimize their negative consequences through a variety of mechanisms.7 While these mechanisms have largely been attributed to factors outside health care (e.g. housing and employment conditions), inequalities in the quantity and quality of health care consumed may also contribute to disparities in health. This may be particularly true in recent decades where health care has made a substantial contribution to increasing life expectancy in Australia.8

For Australia, as with most OECD countries, horizontal equity—equal care for equal need—is an explicit objective of the health care system, and this equity principle underlies Australia's universal health care system, Medicare.9 Nevertheless, a universal system does not guarantee equity in health care, or indeed, health-care related outcomes. In particular, because of factors such as a mixture of public and private funding, out-of-pocket expenses, gaps in services, differences in patient expectations and professional control, those who are more socioeconomically advantaged may derive more benefit from the system than those who are less advantaged. In addition, inequalities in the benefits of health care may arise because of the differential capacity of SES groups to adapt to new knowledge about causes and management of disease. Nevertheless, there is little evidence about whether or not there has been a differential impact of health care by SES, which may, in part, account for observed patterns of socioeconomic inequalities in mortality.

One approach for assessing the contribution of health care to declining mortality rates is to examine trends in avoidable mortality. Avoidable mortality refers to premature deaths from certain conditions that are considered to be largely avoidable given timely and effective health care.10 Because declines in avoidable mortality rates may also reflect influences of factors outside health care, such as improved living conditions, the gradient in avoidable mortality is best compared to that in non-avoidable mortality, which is also likely to be affected by such factors.

Avoidable mortality trends over the past three decades in Australia suggest that health care has made substantial contributions to the reduction in mortality. This is shown in the steady decline in avoidable mortality rates with slower declines in non-avoidable mortality rates. Between 1968 and 2001, avoidable death rates fell around 70% with non-avoidable rates falling around 34%.11 The question remains as to whether individuals across the socioeconomic spectrum have benefited equally from this contribution of health care. While patterns in avoidable mortality by SES have been previously reported for Australia,8,12,13 these studies adopted a broader avoidable mortality classification than used here and the results were not analysed in a way to allow for conclusions to be made regarding the differential impact of care.

In this study, we investigate trends in avoidable and non-avoidable mortality across socioeconomic strata to determine whether there has been inequality in the impact of health care in Australia between 1986 and 2002. If it is assumed that the need for health care within SES strata is proportional to the avoidable mortality rate and there is equal benefit for equal need, then the percentage declines over time in avoidable mortality (compared with non-avoidable mortality) should be equal across SES strata and relative inequality should remain unchanged over the period. At the same time, given higher avoidable mortality rates at the lower end of the SES spectrum at the start of the period, these assumptions imply absolute inequality should fall. Alternatively, if higher SES strata have received a disproportionate benefit from health care, this will be shown in greater percentage declines in avoidable mortality (compared with non-avoidable mortality) in those strata and increasing relative inequality.

Methods

Data

We used anonymized unit record mortality data and population data from the Australian Bureau of Statistics, aggregated to the level of the Statistical Local Area (SLA). Before aggregating the death data, we classified each death as either avoidable or non-avoidable based on the International Classification of Disease (ICD) code for underlying cause of death. Avoidable deaths comprised two categories of conditions: (i) those amenable to medical care [‘medical care indicators’ (MCI)] and (ii) those responsive to health policy but that lack effective treatment once the condition has developed [‘health policy indicators’ (HPI)]. The list of avoidable conditions is shown in Table 1. Medical care indicators are conditions that are considered to have identifiable effective interventions that are administered by health care providers. They exclude preventable conditions that have a relative lack of effective treatment once the condition has developed.10 Our list is based on that developed by Nolte and McKee. Their justification for the selection of the conditions and the age limit imposed (at 74 years for most causes) is outlined in their review,10 and our modifications to the list have been outlined in a previous publication.11 The three HPI causes on our list are those consistently used in studies that include such causes in the definition of avoidable mortality.14–16 The non-avoidable category includes the remaining causes of death (e.g. metabolic disorders, most neurological disorders, diseases of the musculoskeletal system).

Table 1

Avoidable causes of death

Cause of death Age range ICD-8 ICD-9 ICD-10 
Medical care indicators     
    Intestinal infections 0–14 000–009 001–009 A00–A09 
    Tuberculosis 0–74 010–019 010–018, 137 A15–A19, B90 
    Other infections (diphtheria, tetanus, poliomyelitis 0–74 032, 037, 040–043 032, 037, 045 A36, A35, A80 
    Whooping cough 0–14 033 033 A37 
    Septicaemia 0–74 038 038 A40–A41 
    Measles 1–14 055 055 B05 
    Malignant neoplasm of colon and rectum 0–74 153–154 153–154 C18–C21 
    Malignant neoplasm of skin (excl. melanoma) 0–74 173 173 C44 
    Malignant neoplasm of breast 0–74 174 174 C50 
    Malignant neoplasm of cervix uteri 0–74 180 180 C53 
    Malignant neoplasm of cervix uteri and body of uterus (excl. overlap with above codes) 0–44 182 179, 182 C54–C55 
    Malignant neoplasm of testis 0–74 186 186 C62 
    Hodgkin's disease 0–74 201 201 C81 
    Leukaemia 0–44 204–207 204–208 C91–C95 
    Disease of the thyroid 0–74 240–246 240–246 E00–E07 
    Diabetes mellitus 0–49 250 250 E10–E14 
    Epilepsy 0–74 345 345 G40–G41 
    Chronic rheumatic heart disease 0–74 393–398 393–398 I05–I09 
    Hypertensive disease 0–74 400–404 401–405 I10–I13, I15 
    Cerebrovascular disease 0–74 430–438 430–438 I60–I69 
    G45 
    All respiratory diseases (excl. pneumonia/influenza) 1–14 460–466 460–479 J00–J06 
  490–519 488–519 J20–J99 
    Influenza 0–74 470–474 487 J10–J11 
    Pneumonia 0–74 480–486 480–486 J12–J18 
    Peptic ulcer 0–74 531–533 531–533 K25–K27 
    Appendicitis 0–74 540–543 540–543 K35–K38 
    Abdominal hernia 0–74 550–553 550–553 K40–K46 
    Cholelithiasis and Cholecystitis 0–74 574–575 574–575.1 K80–K81 
    Nephritis and Nephrosis 0–74 580–584 580–589 N00–N07, N17–N19, N25–N27 
    Benign prostatic hyperplasia 0–74 600 600 N40 
    Maternal deaths All 630–678 630–676 O00–O99 
    Congenital cardiovascular abnormalities 0–74 746–747 745–747 Q20–Q28 
    Perinatal deaths (excl. stillbirths) All 760–779 760–779 P00–P96, A33–A34 
    Misadventures to patients during surgical and medical care (incl. complications) All E930–E936 E870–876, E878–E879 Y60–Y69, Y83–Y84 
    Asthma 0–74 493 493 J45–J46 
    Ischaemic heart disease 0–74 410–414 410–414 I20–I25 
Health policy indicators     
    Malignant neoplasm of trachea, bronchus and lung 0–74 162 162 C33–C34 
    Chronic liver disease and cirrhosis 0–74 571 571 K70, K71.7, K73–K74, K76.0 
    Motor vehicle accident 0–74 E810–E823 E810–E825 V02–V04, V06.1, V09.0–V09.3, V12–V14, V19.0–V19.6 V20–V79 V80.3–V80.5, V81.0–V81.1, V82.0–V82.1, V83.0–V83.3, V84–V88, V89.0, V89.2–V89.9 
Cause of death Age range ICD-8 ICD-9 ICD-10 
Medical care indicators     
    Intestinal infections 0–14 000–009 001–009 A00–A09 
    Tuberculosis 0–74 010–019 010–018, 137 A15–A19, B90 
    Other infections (diphtheria, tetanus, poliomyelitis 0–74 032, 037, 040–043 032, 037, 045 A36, A35, A80 
    Whooping cough 0–14 033 033 A37 
    Septicaemia 0–74 038 038 A40–A41 
    Measles 1–14 055 055 B05 
    Malignant neoplasm of colon and rectum 0–74 153–154 153–154 C18–C21 
    Malignant neoplasm of skin (excl. melanoma) 0–74 173 173 C44 
    Malignant neoplasm of breast 0–74 174 174 C50 
    Malignant neoplasm of cervix uteri 0–74 180 180 C53 
    Malignant neoplasm of cervix uteri and body of uterus (excl. overlap with above codes) 0–44 182 179, 182 C54–C55 
    Malignant neoplasm of testis 0–74 186 186 C62 
    Hodgkin's disease 0–74 201 201 C81 
    Leukaemia 0–44 204–207 204–208 C91–C95 
    Disease of the thyroid 0–74 240–246 240–246 E00–E07 
    Diabetes mellitus 0–49 250 250 E10–E14 
    Epilepsy 0–74 345 345 G40–G41 
    Chronic rheumatic heart disease 0–74 393–398 393–398 I05–I09 
    Hypertensive disease 0–74 400–404 401–405 I10–I13, I15 
    Cerebrovascular disease 0–74 430–438 430–438 I60–I69 
    G45 
    All respiratory diseases (excl. pneumonia/influenza) 1–14 460–466 460–479 J00–J06 
  490–519 488–519 J20–J99 
    Influenza 0–74 470–474 487 J10–J11 
    Pneumonia 0–74 480–486 480–486 J12–J18 
    Peptic ulcer 0–74 531–533 531–533 K25–K27 
    Appendicitis 0–74 540–543 540–543 K35–K38 
    Abdominal hernia 0–74 550–553 550–553 K40–K46 
    Cholelithiasis and Cholecystitis 0–74 574–575 574–575.1 K80–K81 
    Nephritis and Nephrosis 0–74 580–584 580–589 N00–N07, N17–N19, N25–N27 
    Benign prostatic hyperplasia 0–74 600 600 N40 
    Maternal deaths All 630–678 630–676 O00–O99 
    Congenital cardiovascular abnormalities 0–74 746–747 745–747 Q20–Q28 
    Perinatal deaths (excl. stillbirths) All 760–779 760–779 P00–P96, A33–A34 
    Misadventures to patients during surgical and medical care (incl. complications) All E930–E936 E870–876, E878–E879 Y60–Y69, Y83–Y84 
    Asthma 0–74 493 493 J45–J46 
    Ischaemic heart disease 0–74 410–414 410–414 I20–I25 
Health policy indicators     
    Malignant neoplasm of trachea, bronchus and lung 0–74 162 162 C33–C34 
    Chronic liver disease and cirrhosis 0–74 571 571 K70, K71.7, K73–K74, K76.0 
    Motor vehicle accident 0–74 E810–E823 E810–E825 V02–V04, V06.1, V09.0–V09.3, V12–V14, V19.0–V19.6 V20–V79 V80.3–V80.5, V81.0–V81.1, V82.0–V82.1, V83.0–V83.3, V84–V88, V89.0, V89.2–V89.9 

Note: In Australia, the 8th Revision of the International Classification of Disease (ICD-8) was used to code deaths registered between 1968 and 1978, ICD-9 for 1979–98, and ICD-10 for deaths registered from 1999 onwards.

Death data were aggregated on the basis of the SLA in which the person was residing at the time of death. SES was assigned to each SLA using the Socioeconomic Indexes for Areas (SEIFA).17 SEIFA are summary area measures of socioeconomic conditions based on aggregated census data (e.g. percentage of low-income families, percentage of early school leavers), produced by the Australian Bureau of Statistics. There are four versions of the SEIFA, based on 1986, 1991, 1996 and 2001 census data, and each version has several indexes. Of these, the Index of Disadvantage (formerly known as the Index of Relative Socioeconomic Disadvantage) was chosen for this study as it is the most general index and uses the same underlying variables in all versions of the SEIFA. The Index of Disadvantage is an ordinal ranking of disadvantage.

The geographic boundaries (Australian Standard Geographical Classification) upon which the SLAs are based change approximately every two years, thus SLAs vary across time in both the SEIFA (census) and mortality unit record data. To match the geographic areas in the mortality data with those used in the SEIFA, years in which the mortality data were coded using the same version of the Australian Standard Geographical Classification as used in the SEIFA were selected for analysis—1986, 1991, 1997 and 2002. There were between 1314 (1997) and 1335 SLAs (1991) in each of these years for which there was a SEIFA code. Due to substantial changes in boundaries, no attempt was made to match areas over time. Analysis was restricted to mortality below the age of 75 years in each of these years, including deaths registered in the year of death (∼95% of the deaths) or in the following year. There was a small percentage of records within SLAs for which there was no matching SEIFA index—2.2% in 1986, 0.2% in 1991, 0.8% in 1997 and 0.9% in 2002. These records were excluded from the analysis. The mean population size (people under 75 years) of the SLAs was 12 564 (min = 4, max = 238 518; interquartile range = 2231–11 236).

Analysis

To broadly describe mortality trends by SES for each time period, we aggregated SLAs into quintiles, each containing approximately 20% of the population. Rates were directly standardized for age (using 5-year age groups) and sex using 2001 Australian population data. They were not analysed separately by age group and sex in order to maximize statistical power and simplify the presentation of results. Incidence rate differences (RD) and rate ratios (RR) were calculated to describe the change in rates between 1986 and 2002 and the difference in rates between the lowest (Q1) and highest (Q5) quintiles in each year.

To investigate our hypotheses regarding trends in avoidable compared with non-avoidable mortality by SES, we performed pooled cross-section trend analysis using the SLA-level data. We used two models. The first model examined relative inequalities, and the second, absolute inequalities. For both models, we assumed deaths were Poisson distributed within an area; however, to allow for random variation in rates between areas we modelled for over-dispersion. Prior to the modelling procedure, we used indirect standardization to calculate the expected deaths for each SLA in each year using the age and sex-specific rates for 2002 as the standard. The observed and expected deaths for each SLA in each year were then used in the analysis.

The measure of SES included in these models was the cumulative proportion of the population in the area (SLA), ranked from lowest to highest SES (i.e. the fractional rank). Advantages of this SES measure are that it includes all the available data, allows the SES data to be treated as continuous and avoids the problem of changing socioeconomic group sizes over time. The coefficient for this SES measure is used to calculate the relative index of inequality (RII) (Model 1) and the slope index of inequality (SII) (Model 2). Originally described by Pamuk,18 and later modified by Mackenbach and Kunst,19 the RII is the ratio of the rate of mortality that is predicted for the lower end of the socioeconomic continuum to the rate of mortality for the higher end. In Model 1, it is equal to the exponential of the negative regression coefficient for SES.20 Similarly, the regression coefficients for SES from Model 2 were used to estimate the SII which can be interpreted as absolute differences in standardized mortality ratios predicted for the lower and higher ends of the socioeconomic continuum. To represent the SIIs on a rate scale, we multiplied the coefficients by the 2002 national rates for avoidable and non-avoidable mortality.

For both the relative (Model 1) and absolute (Model 2) models, we modelled avoidable and non-avoidable mortality separately. In Model 1, we used negative binomial regression to model the annual rate of mortality decline by SES by regressing observed deaths on the covariates using a log link, with the log of expected deaths as an offset. The covariates included Year (modelled as a continuous variable) and SES; we also included an SES × Year interaction term to examine change in inequality over time. We compared the change in inequality for avoidable mortality with that for non-avoidable mortality by comparing the SES × Year coefficients from the avoidable and non-avoidable models using the seemingly unrelated estimation (SUEST) procedure in STATA.21 For Model 2, we used an identity link, weighting by the expected number of deaths. We used quasi-likelihood estimation for Poisson regression with a scale parameter to model for over-dispersion as we were unable to use an identity link and incorporate weights to model rates using negative binomial regression.

We used Stata 9.0 statistical software for all analyses except for the absolute models, which were fitted in R version 2.3.0 using the GLM function.

Results

Descriptive statistics

Avoidable and non-avoidable mortality rates by SES quintile for each year are shown in Table 2. Across the four time periods, there was a steady decline in avoidable and non-avoidable mortality rates in every quintile, with the exception that non-avoidable mortality rates changed very little between 1986 and 1991 and actually rose during this time in Q1 (low SES). Both absolute declines (RD) and relative declines (RR) in avoidable mortality rates were much greater than in non-avoidable rates. Percentage declines over time showed a socioeconomic gradient for both avoidable and non-avoidable mortality, ranging from 44% in the lowest quintile (Q1) to 55% in the highest (Q5) for avoidable mortality, and 8% (Q1) to 23% (Q5) for non-avoidable mortality. This pattern of greater percentage declines in avoidable than non-avoidable mortality is also reflected in the decline in avoidable mortality rates as a proportion of total mortality for each SES quintile, as shown in Figure 1.

Figure 1

Avoidable deaths as a proportion of total deaths by year for each SES quintile (Q)

Figure 1

Avoidable deaths as a proportion of total deaths by year for each SES quintile (Q)

Table 2

Avoidable, non-avoidable and total deaths by year, and age- and sex-adjusted mortality rates by year and SES quintile (Q), for Australian population aged <75 years

 1986 1991 1997 2002   
Population 14 914 474 15 737 506 17 574 653 18 501 662   
Avoidable deaths 34 757 31 715 27 877 23 847   
Non-avoidable deaths 23 407 25 825 27 295 26 417   
Total deaths 58 164 57 540 55 172 50 264   
 1986 1991 1997 2002   
Population 14 914 474 15 737 506 17 574 653 18 501 662   
Avoidable deaths 34 757 31 715 27 877 23 847   
Non-avoidable deaths 23 407 25 825 27 295 26 417   
Total deaths 58 164 57 540 55 172 50 264   
 Mortality rates, per 100 000 (95% CI) Rate difference (2002–1986) Rate ratio (2002/1986) 
Avoidable mortality        
    Q1 (Low) 275 (269–281) 254 (249–260) 195 (191–200) 155 (151–159) −120 (−128 to −113) 0.56 (0.54–0.58) 
    Q2 272 (265–278) 229 (224–235) 175 (170–179) 143 (139–146) −129 (−136 to −122) 0.53 (0.51–0.54) 
    Q3 266 (260–272) 220 (215–225) 173 (168–177) 134 (131–138) −132 (−139 to −125) 0.50 (0.48–0.52) 
    Q4 244 (238–250) 200 (194–205) 149 (145–153) 114 (110–117) −130 (−136 to −123) 0.47 (0.45–0.49) 
    Q5 (High) 210 (205–216) 174 (169–179) 122 (119–126) 95 (92–99) −115 (−121 to −109) 0.45 (0.43–0.47) 
    Total 254 (251–257) 216 (214–219) 163 (161–165) 128 (127–130) −126 (−129 to −123) 0.50 (0.50–0.51) 
    Rate difference (Q1–Q5) 65 (57–73) 80 (73–88) 73 (67–79) 60 (54–65) – – 
    Rate ratio (Q1/Q5) 1.31 (1.26–1.35) 1.46 (1.41–1.52) 1.60 (1.54–1.66) 1.63 (1.56–1.70) – – 
Non-avoidable mortality        
    Q1 (Low) 183 (178–187) 211 (206–216) 183 (179–188) 168 (164–173) −15 (−21 to −8) 0.92 (0.89–0.96) 
    Q2 180 (175–185) 180 (175–184) 161 (156–165) 152 (148–156) −28 (−34 to −22) 0.84 (0.81–0.88) 
    Q3 175 (170–180) 177 (172–182) 165 (161–170) 145 (141–149) −30 (−36 to −24) 0.83 (0.80–0.86) 
    Q4 160 (156–165) 158 (154–163) 152 (148–156) 132 (128–135) −28 (−35 to −23) 0.83 (0.82–0.85) 
    Q5 (High) 148 (144–153) 145 (140–149) 131 (127–134) 114 (110–117) −34 (−40 to −30) 0.77 (0.79–0.85) 
    Total 170 (167–172) 175 (172–177) 159 (157–160) 142 (141–144) −28 (−30 to −25) 0.84 (0.82–0.85) 
    Rate difference (Q1–Q5) 35 (28–41) 66 (59–72) 52 (47–58) 54 (49–60)   
    Rate ratio (Q1/Q5) 1.23 (1.18–1.28) 1.45 (1.40–1.51) 1.40 (1.35–1.46) 1.48 (1.42–1.54) – – 
 Mortality rates, per 100 000 (95% CI) Rate difference (2002–1986) Rate ratio (2002/1986) 
Avoidable mortality        
    Q1 (Low) 275 (269–281) 254 (249–260) 195 (191–200) 155 (151–159) −120 (−128 to −113) 0.56 (0.54–0.58) 
    Q2 272 (265–278) 229 (224–235) 175 (170–179) 143 (139–146) −129 (−136 to −122) 0.53 (0.51–0.54) 
    Q3 266 (260–272) 220 (215–225) 173 (168–177) 134 (131–138) −132 (−139 to −125) 0.50 (0.48–0.52) 
    Q4 244 (238–250) 200 (194–205) 149 (145–153) 114 (110–117) −130 (−136 to −123) 0.47 (0.45–0.49) 
    Q5 (High) 210 (205–216) 174 (169–179) 122 (119–126) 95 (92–99) −115 (−121 to −109) 0.45 (0.43–0.47) 
    Total 254 (251–257) 216 (214–219) 163 (161–165) 128 (127–130) −126 (−129 to −123) 0.50 (0.50–0.51) 
    Rate difference (Q1–Q5) 65 (57–73) 80 (73–88) 73 (67–79) 60 (54–65) – – 
    Rate ratio (Q1/Q5) 1.31 (1.26–1.35) 1.46 (1.41–1.52) 1.60 (1.54–1.66) 1.63 (1.56–1.70) – – 
Non-avoidable mortality        
    Q1 (Low) 183 (178–187) 211 (206–216) 183 (179–188) 168 (164–173) −15 (−21 to −8) 0.92 (0.89–0.96) 
    Q2 180 (175–185) 180 (175–184) 161 (156–165) 152 (148–156) −28 (−34 to −22) 0.84 (0.81–0.88) 
    Q3 175 (170–180) 177 (172–182) 165 (161–170) 145 (141–149) −30 (−36 to −24) 0.83 (0.80–0.86) 
    Q4 160 (156–165) 158 (154–163) 152 (148–156) 132 (128–135) −28 (−35 to −23) 0.83 (0.82–0.85) 
    Q5 (High) 148 (144–153) 145 (140–149) 131 (127–134) 114 (110–117) −34 (−40 to −30) 0.77 (0.79–0.85) 
    Total 170 (167–172) 175 (172–177) 159 (157–160) 142 (141–144) −28 (−30 to −25) 0.84 (0.82–0.85) 
    Rate difference (Q1–Q5) 35 (28–41) 66 (59–72) 52 (47–58) 54 (49–60)   
    Rate ratio (Q1/Q5) 1.23 (1.18–1.28) 1.45 (1.40–1.51) 1.40 (1.35–1.46) 1.48 (1.42–1.54) – – 

The RD and RRs comparing Q1 with Q5 in each year show significant absolute and relative socioeconomic inequality, respectively, in both avoidable and non-avoidable mortality. Rate differences show that absolute inequality in avoidable mortality rose between 1986 and 1991 before falling thereafter. A similar pattern was seen for non-avoidable mortality. In contrast, relative inequality in avoidable mortality rose steadily over time, with RRs ranging from 1.31 in 1986 to 1.63 in 2002. For non-avoidable mortality, RRs increased between 1986 and 1991 from 1.23 to 1.45, with little change thereafter, the rise in the RR in 1991 being driven by the rise in non-avoidable death rates in Q1 (with rates remaining stable in the other quintiles).

Modelled trends in avoidable and non-avoidable mortality

The results of Model 1 (Table 3) show that at the lower end of the SES continuum avoidable mortality rates declined annually by 3.47% (95% CI: 3.15–3.80%) and non-avoidable rates by 0.80% (0.43–1.17%). At the higher end of the continuum, avoidable rates fell by 5.03% (4.73–5.33%) per annum, while non-avoidable rates fell 1.50% (1.17–1.83%). At both ends of the SES continuum, the declines in avoidable mortality rates were greater than in non-avoidable mortality rates (P < 0.001 for both).

Table 3

Results of Model 1—relative inequalities: parameter estimates for mortality rates regressed on SES and year of death (1986–2002) using negative binomial regression

Variable Estimate (β) SE (β) P-value (β = 0) 
Avoidable    
    SES −0.434 0.029 <0.001 
    Year −0.035 0.002 <0.001 
    SES × Year −0.016 0.003 <0.001 
Non-avoidable    
    SES −0.440 0.035 <0.001 
    Year −0.008 0.002 <0.001 
    SES × Year −0.007 0.003 0.027 
Variable Estimate (β) SE (β) P-value (β = 0) 
Avoidable    
    SES −0.434 0.029 <0.001 
    Year −0.035 0.002 <0.001 
    SES × Year −0.016 0.003 <0.001 
Non-avoidable    
    SES −0.440 0.035 <0.001 
    Year −0.008 0.002 <0.001 
    SES × Year −0.007 0.003 0.027 

Notes: (i) n = 5311 SLAs and (ii) SES is represented by the cumulative density for SLAs weighted by population, with the SLAs ordered from low to high SES.

Comparing trends at the two ends of the SES continuum showed the declines in avoidable mortality were greater at the higher than the lower end, as revealed in the negative SES × Year interaction (P < 0.001). This resulted in a rise in relative inequality over time: at the start of the period, in 1986, the avoidable mortality rate at the lower end of the continuum was 50% higher than at the higher end (RII = 1.54; 1.46–1.63) and by 2002 it was twice as high (RII = 2.00; 1.95–2.06). Similarly, the declines in non-avoidable mortality were greater at the higher SES end (P = 0.027), with inequality in non-avoidable mortality rates rising over time [RII in 1986 = 1.55 (1.45–1.66); in 2002 = 1.74 (1.68–1.80)]. However, the SUEST procedure comparing the SES × Year interaction terms across the models showed that the rise in relative inequality in avoidable mortality was significantly greater than that for non-avoidable mortality (P = 0.036).

The results of the model of absolute rates (Model 2, Table 4) show that at the lower end of the SES continuum avoidable mortality fell by 8.38 (7.87–8.90) deaths/100 000 per year and non-avoidable mortality by 1.71 (1.22–2.20). At the upper end, avoidable mortality fell 7.37 (6.89–7.84) deaths/100 000/year and non-avoidable mortality fell 2.15 deaths (1.69–2.61).

Table 4

Results of Model 2—absolute inequalities: parameter estimates for mortality rates regressed on SES and year of death (1986–2002) using quasi-Poisson regression

Variable Estimate (β) SE (β) P-value (β = 0) 
Avoidable    
    SES −0.758 0.040 <0.001 
    Year −0.065 0.002 <0.001 
    SES × Year 0.008 0.002 0.021 
Non-avoidable    
    SES −0.420 0.032 <0.001 
    Year −0.012 0.002 <0.001 
    SES × Year −0.003 0.003 0.295 
Variable Estimate (β) SE (β) P-value (β = 0) 
Avoidable    
    SES −0.758 0.040 <0.001 
    Year −0.065 0.002 <0.001 
    SES × Year 0.008 0.002 0.021 
Non-avoidable    
    SES −0.420 0.032 <0.001 
    Year −0.012 0.002 <0.001 
    SES × Year −0.003 0.003 0.295 

Notes: (i) n = 5311 SLAs and (ii) SES is represented by the cumulative density for SLAs weighted by population, with the SLAs ordered from low to high SES.

Comparing trends at the two ends of the SES continuum showed the decline in avoidable mortality/100 000 population was significantly greater at the lower than the higher end, as shown by the positive SES × Year interaction (P = 0.021). Consequently, there was a decrease in absolute inequality in avoidable mortality over time, with the SII in 1986 of 97.75 (95% CI: 87.60–107.91) deaths/100 000 falling to 81.51 (74.55–88.47) deaths by 2002. In contrast, the trends in non-avoidable mortality over time were similar at the lower and higher ends of the SES continuum (P = 0.295), with the SII in 1986 equal to 60.00 (51.00 to 69.04) deaths/100 000 and in 2002 equal to 67.06 (59.59–74.55) (formal comparison of the SES × Year interaction effects for avoidable and non-avoidable models, P = 0.015). Note that similar results were obtained for Models 1 and 2 when year was modelled as indicator variables rather than as a continuous variable.

Subanalyses

We performed three subanalyses. First, to examine any possible effect modification by sex, we repeated Model 1 separately for males and females (standardizing only by age) and compared coefficients across the models. While there was greater inequality in both avoidable and non-avoidable mortality in males than females, the SES × Year interactions were not significantly different across sexes for either avoidable (P = 0.557) or non-avoidable (P = 0.736) mortality, indicating that trends over time in inequality were similar for males and females.

Second, we investigated whether or not the patterns of increasing relative inequalities in avoidable mortality compared with non-avoidable mortality were similar for both those conditions that are treatable (MCI) and those conditions that are largely avoidable through primary prevention only (HPI). Given that ischaemic heart disease (IHD) accounts for approximately half of the total MCI deaths, we also calculated the RIIs for IHD separately. All three categories of death showed similar patterns, however the rise in relative inequality over time in mortality rates was lower for MCI than for IHD and HPI (P < 0.001 for both comparisons).

Finally, we also re-analysed the data including only those people who lived within capital city statistical subdivisions (63–64% of population in each year; number of SLAs ranged from 525 to 585) because of the potential for confounding due to remoteness. The pattern of results for the city sub-sample is similar to those for the entire sample, although the rise in inequality over time in avoidable mortality, while significant (P = 0.014), was lower than for the total sample (while non-avoidable mortality trends did not differ significantly between city and total sub-samples).

Discussion

Our findings indicate that individuals across the socioeconomic spectrum have benefited from health care, as shown by the greater declines in avoidable mortality than in non-avoidable mortality over time at both the low and high ends of the spectrum. However, the declines in avoidable mortality (compared with non-avoidable) were greater at the higher end of the SES spectrum resulting in increasing relative inequality. This suggests that, of those in need of heath care, higher SES individuals were more likely to have benefited from it than those of lower SES. Although methodologies and time periods vary considerably, the findings from our study are generally consistent with those reported for other countries.22–26 At the same time, absolute declines in avoidable mortality were greater at the lower end of the spectrum resulting in decreasing absolute inequality over time, reflecting the greater need for care in this population. Thus, while there is evidence of inequality in the individual benefits of health care, the overall population impact in terms of improved survival is greater at the lower end of the socioeconomic spectrum.

Limitations to the study

As has been noted in other studies using the avoidable mortality approach, there are several limitations to consider when interpreting these findings. First, as originally discussed by Rutstein and his colleagues27 who introduced the concept of avoidable mortality, the chain of events that leads to death may be long and complex, thus the partitioning of deaths into those that are avoidable, and those that are not, is an inexact science. Second, over the period covered by this study, there were changes in ICD coding that may affect trends. However, the effect of such changes on overall trends have been examined previously and are not expected to have a major impact on this study, particularly given trends in individual causes of death were not examined (with the exception of IHD).28 Third, as the avoidable mortality concept only captures deaths and not other health outcomes, it offers an incomplete assessment of the benefits of health care. Fourth, there is the criticism that the findings may just reflect differences in incidence. However, avoidable causes are those for which death is largely avoidable once the condition has developed (with the exception of the three HPI causes) and differences in incidence rates also reflect, at least in part, differences in preventive health care and health policy initiatives (HPI).

Other limitations of this study include, first, the potential for biased estimates as the numerator data and denominator data were obtained from different sources—mortality and census data. We attempted to minimize the extent of this bias by selecting years in which data from the two sources used identical SLA coding. Second, with regard to the SES area-based measure used, while such an index is a multidimensional indicator that may measure aspects of SES not captured by individual-level indicators,29 it is less precise than individual-based measures and is likely to result in an underestimation of the true extent of socioeconomic inequality.30,31 Third, to preserve power and simplify the presentation of results, we aggregated results by age and sex. While we adjusted for these factors, the combined results could have potentially obscured important age and sex differences. Nevertheless, a subanalysis showed no effect modification by sex.

Conclusions

The relative inequality in the benefits of health care may be because those of lower SES are less likely to use formal health care services than higher SES individuals with similar needs. There may also be differences in the quality of services provided across SES groups. There is evidence that socioeconomically disadvantaged people in Australia are less likely to attend screening,34–36 undergo elective or discretionary surgery,37–40 be admitted to a private hospital,38 or have lengthy general practice consultations,41,42 while the findings on frequency of such consultations by SES are mixed.43–46 Inequality in avoidable mortality rates may also reflect the differential capacity of SES groups to adapt to new health knowledge, whether this be delivered through formal or informal channels (e.g. 47,48). Certainly, the role of informal channels of health care in improving health and creating inequalities cannot be overlooked. Because mass behaviour change may occur in response to information flowing through the mass media, and in an uncoordinated or informal way through health-professionals, health care ‘needs to be understood broadly, not just as a domain of professional practice, nor as a bundle of commodities to be “delivered” but rather as an institution in which the whole of society participates’.49

The rises in relative inequality over time were larger for conditions avoidable largely through HPI and for IHD than for other treatable conditions (MCI). However, the distinction between categories should be interpreted cautiously and it should not be assumed that non-medical health care is driving the inequalities: declines in HPI deaths may reflect advances in medical care, just as many conditions amenable or preventable through medical care (MCI and IHD) are also preventable through HPI. For example, the declines in IHD reflect health policy interventions through reductions in smoking and dietary risk factors as well as medical care such as management of abnormal blood pressure, emergency department care and surgical coronary artery procedures.50,51 Another example is the reduction in road traffic deaths, which reflect policy initiatives such as national laws regarding seat-belt use and alcohol control, as well as improvements in trauma care for road accident victims.52 In addition, we cannot rule out inequalities in health care for particular conditions within MCI, as these were not examined separately.

Whether due to direct medical care or HPI, formal or informal channels, health care has clearly benefited people across the socioeconomic spectrum. While the population impact has been largest in those who have the most to gain—those at the lower end of the spectrum—those at the higher end have obtained a disproportionate benefit resulting in widening relative health inequalities. This pattern of inequalities is consistent with the theory put forward by Phelan and Link53 that inequalities are driven by our increasing capacity to control disease and death in combination with existing socioeconomic inequalities—as we make advances in health care, the benefits are distributed according to SES-related resources such as money and power.

The fact that there are inequalities in deaths that are potentially avoidable through health care should be of particular concern. While not all determinants of the distribution of health may be avoidable, the inequitable distribution of health care and health knowledge is one that is amenable to change. While this article does not address what it is about health care in Australia that results in inequalities, it does indicate that having a ‘universal’ system does not guarantee equality in health-care related outcomes. If these outcomes of the less advantaged members of society were to improve to the levels of the more advantaged members, not only would this make for a fairer society, but also it would improve the overall level of the population's health.

Acknowledgements

We are very grateful to the three anonymous reviewers and the Associate Editor for their constructive comments on an earlier version of this paper. This study has received financial support from National Health & Medical Research Council Project Grant No. 268013.

Conflicts of interest: None declared.

KEY MESSAGES

  • Life expectancy has been steadily increasing in Australia, as in other developed countries, which can be partly explained by advances in health care.

  • People across the socioeconomic spectrum have benefited from these advances in health care.

  • Health care has contributed to decreasing the absolute mortality gap, but advantaged people have obtained a disproportionate benefit, contributing to widening relative health inequalities.

  • Having a universal health care system does not guarantee equality in the benefits of health care.

References

1
World Health Organization
World Health Statistics 2005
 , 
2005
Geneva
WHO Press
2
Burnley
IH
Inequalities in the transition of ischaemic heart disease mortality in New South Wales, Australia, 1969–1994
Soc Sci Med
 , 
1998
, vol. 
47
 (pg. 
1209
-
22
)
3
Burnley
IH
Rintoul
D
Inequalities in the transition of cerebrovascular disease mortality in New South Wales, Australia 1969–1996
Soc Sci Med
 , 
2002
, vol. 
54
 (pg. 
545
-
59
)
4
Draper
G
Turrell
G
Oldenburg
B
Health inequalities in Australia: Mortality. Health Inequalities Monitoring Series No. 1
AIHW Cat No. PHE 55
 , 
2004
Queensland University of Technology and the Australian Institute of Health and Welfare
5
Hayes
LJ
Quine
S
Taylor
R
Berry
G
Socio-economic mortality differentials in Sydney over a quarter of a century, 1970–94
Aust N Z J Public Health
 , 
2002
, vol. 
26
 (pg. 
311
-
17
)
6
Turrell
G
Mathers
C
Socioeconomic inequalities in all-cause and specific-cause mortality in Australia: 1985-1987 and 1995-1997
Int J Epidemiol
 , 
2001
, vol. 
30
 (pg. 
231
-
39
)
7
Link
BG
Phelan
J
Social conditions as fundamental causes of disease
J Health Soc Behav
 , 
1995
, vol. 
Spec No
 (pg. 
80
-
94
)
8
National Health Performance Committee 2004
National report on health sector performance indicators 2003
 , 
2004
Canberra
Australian Institute of Health and Welfare
9
Australian Institute of Health and Welfare
Australia's health 2006
AIHW cat. no. AUS 73
 , 
2006
Canberra
AIHW
10
Nolte
E
McKee
M
Does health care save lives? Avoidable mortality revisited
 , 
2004
London
The Nuffield Trust
11
Korda
RJ
Butler
JRG
Effect of healthcare on mortality: Trends in avoidable mortality in Australia and comparisons with Western Europe
Public Health
 , 
2006
, vol. 
120
 (pg. 
95
-
105
)
12
Hayen
A
Lincoln
D
Moore
H
Thomas
M
Trends in potentially avoidable mortality in NSW
N S W Public Health Bull
 , 
2002
, vol. 
13
 (pg. 
226
-
36
)
13
Population Health Division
The health of the people of New South Wales:Report of the Chief Health Officer, 2004
2004
Sydney
NSW Department of Health
14
Albert
X
Bayo
A
Alfonso
JL
Cortina
P
Corella
D
The effectiveness of health systems in influencing avoidable mortality: a study in Valencia, Spain, 1975–90
J Epidemiol Community Health
 , 
1996
, vol. 
50
 (pg. 
320
-
25
)
15
Holland
WW
European Community Atlas of ‘Avoidable Death’
Commission of the European Communities Health Services Research Series No. 9
 , 
1993
, vol. 
Vol II
 
2nd edn
Oxford
Oxford University Press
16
Nolte
E
Scholz
R
Shkolnikov
V
McKee
M
The contribution of medical care to changing life expectancy in Germany and Poland
Soc Sci Med
 , 
2002
, vol. 
55
 (pg. 
1905
-
21
)
17
Australian Bureau of Statistics
Information Paper: Census of Population and Housing – Socio-Economic Indexes for Areas, Australia, 2001
Cat No. 2039.0
 , 
2003
18
Pamuk
ER
Social class inequality in mortality from 1921 to 1972 in England and Wales
Popul Stud (Camb)
 , 
1985
, vol. 
39
 (pg. 
17
-
31
)
19
Mackenbach
JP
Kunst
AE
Measuring the magnitude of socio-economic inequalities in health: an overview of available measures illustrated with two examples from Europe
Soc Sci Med
 , 
1997
, vol. 
44
 (pg. 
757
-
71
)
20
Hayes
LJ
Berry
G
Sampling variability of the Kunst-Mackenbach relative index of inequality
J Epidemiol Community Health
 , 
2002
, vol. 
56
 (pg. 
762
-
65
)
21
Weesie
J
Seemingly unrelated regression and the cluster-adjusted sandwich estimator
STATA Technical Bulletin
 , 
1999
, vol. 
52
 (pg. 
34
-
46
)
22
Mackenbach
JP
Stronks
K
Kunst
AE
The contribution of medical care to inequalities in health: differences between socio-economic groups in decline of mortality from conditions amenable to medical intervention
Soc Sci Med
 , 
1989
, vol. 
29
 (pg. 
369
-
76
)
23
Marshall
SW
Kawachi
I
Pearce
N
Borman
B
Social class differences in mortality from diseases amenable to medical intervention in New Zealand
Int J Epidemiol
 , 
1993
, vol. 
22
 (pg. 
255
-
61
)
24
Phelan
JC
Link
BG
Diez-Roux
A
Kawachi
I
Levin
B
‘Fundamental causes’of social inequalities in mortality: a test of the theory
J Health Soc Behav
 , 
2004
, vol. 
45
 (pg. 
265
-
85
)
25
Song
YM
Byeon
JJ
Excess mortality from avoidable and non-avoidable causes in men of low socioeconomic status: a prospective study in Korea
J Epidemiol Community Health
 , 
2000
, vol. 
54
 (pg. 
166
-
72
)
26
Tobias
M
Jackson
G
Avoidable mortality in New Zealand, 1981–97
Aust N Z J Public Health
 , 
2001
, vol. 
25
 (pg. 
12
-
20
)
27
Rutstein
DD
Berenberg
W
Chalmers
TC
Child
CG
3rd
Fishman
AP
Perrin
EB
Measuring the quality of medical care. A clinical method
N Engl J Med
 , 
1976
, vol. 
294
 (pg. 
582
-
88
)
28
Korda
RJ
Butler
JRG
The impact of health care on mortality: Time trends in avoidable mortality in Australia 1968–2001
Working Paper Number 49
 , 
2004
National Centre for Epidemiology and Population Health
pg. 
41
 
29
Macintyre
S
Ellaway
A
Kawachi
I
Berkman
LF
Neighborhoods and health: An overview
Neighborhoods and Health
 , 
2003
New York
Oxford University Press
(pg. 
20
-
42
)
30
Finkelstein
MM
Ecologic proxies for household income: how well do they work for the analysis of health and health care utilization?
Can J Public Health
 , 
2004
, vol. 
95
 (pg. 
90
-
94
)
31
Taylor
RJ
Mamoon
HA
Morrell
SL
Wain
GV
Cervical screening by socio-economic status in Australia
Aust N Z J Public Health
 , 
2001
, vol. 
25
 (pg. 
256
-
60
)
32
AIHW
de Looper
M
Magnus
P
Australian health inequalities 2: trends in male mortality by broad occupational group
 , 
2005
Canberra
Australian Institute of Health and Welfare
33
Mishra
GD
Ball
K
Dobson
AJ
Byles
JE
Do socioeconomic gradients in women's health widen over time and with age?
Soc Sci Med
 , 
2004
, vol. 
58
 (pg. 
1585
-
95
)
34
Perkins
JJ
Sanson-Fisher
RW
Byles
JE
Tiller
K
Factors relating to cervical screening in New South Wales, Australia
Health Place
 , 
1999
, vol. 
5
 (pg. 
223
-
33
)
35
Siahpush
M
Singh
GK
Sociodemographic predictors of pap test receipt, currency and knowledge among Australian women
Prev Med
 , 
2002
, vol. 
35
 (pg. 
362
-
68
)
36
Siahpush
M
Singh
GK
Sociodemographic variations in breast cancer screening behavior among Australian women: results from the 1995 National Health Survey
Prev Med
 , 
2002
, vol. 
35
 (pg. 
174
-
80
)
37
Brameld
KJ
Holman
CD
The use of end-quintile comparisons to identify under-servicing of the poor and over-servicing of the rich: a longitudinal study describing the effect of socioeconomic status on healthcare
BMC Health Serv Res
 , 
2005
, vol. 
5
 pg. 
61
 
38
Glover
J
Harris
K
Tennant
S
A Social Health Atlas of Australia
 , 
1999
Second
Adelaide
Public Health Information Development Unit, University of Adelaide
39
Hall
SE
Holman
CD
Inequalities in breast cancer reconstructive surgery according to social and locational status in Western Australia
Eur J Surg Oncol
 , 
2003
, vol. 
29
 (pg. 
519
-
25
)
40
Hall
SE
Holman
CDAJ
Wisniewski
ZS
Semmens
J
Prostate cancer: socio-economic, geographical and private-health insurance effects on care and survival
BJU International
 , 
2005
, vol. 
95
 (pg. 
51
-
58
)
41
Britt
HC
Valenti
L
Miller
GC
Determinants of consultation length in Australian general practice
Med J Aust
 , 
2005
, vol. 
183
 (pg. 
68
-
71
)
42
Furler
JS
Harris
E
Chondros
P
Powell Davies
PG
Harris
MF
Young
DY
The inverse care law revisited: impact of disadvantaged location on accessing longer GP consultation times
Med J Aust
 , 
2002
, vol. 
177
 (pg. 
80
-
83
)
43
Parslow
R
Jorm
A
Christensen
H
Jacomb
P
Factors associated with young adults’ obtaining general practitioner services
Aust Health Rev
 , 
2002
, vol. 
25
 (pg. 
109
-
18
)
44
Turrell
G
Oldenburg
BF
Harris
E
Jolley
D
Utilisation of general practitioner services by socio-economic disadvantage and geographic remoteness
Aust N Z J Public Health
 , 
2004
, vol. 
28
 (pg. 
152
-
58
)
45
van Doorslaer
E
Massiera
C
Income-related inequality in the use of medical care in 21 OECD countries
Towards High-Performing Health Systems: Policy Studies
 , 
2004
Paris
OECD
(pg. 
109
-
65
)
46
Young
AF
Dobson
AJ
Byles
JE
Determinants of general practitioner use among women in Australia
Soc Sci Med
 , 
2001
, vol. 
53
 (pg. 
1641
-
51
)
47
Bower
C
Miller
M
Payne
J
Serna
P
Promotion of folate for the prevention of neural tube defects: who benefits?
Paediatr Perinat Epidemiol
 , 
2005
, vol. 
19
 (pg. 
435
-
444
)
48
Merom
D
Phongsavan
P
Chey
T
Bauman
A
Long-term changes in leisure time walking, moderate and vigorous exercise: Were they influenced by the national physical activity guidelines?
J Sci Med Sport
 , 
2006
, vol. 
9
 (pg. 
199
-
208
)
49
Powles
J
Eckersley
R
Dixon
J
Douglas
B
Healthier progress: historical perspectives on the social and economic determinants of health
The Social Origins of Health and Well-being
 , 
2001
Cambridge
Cambridge University Press
(pg. 
3
-
24
)
50
Kuulasmaa
K
Tunstall-Pedoe
H
Dobson
A
, et al.  . 
Estimation of contribution of changes in classic risk factors to trends in coronary-event rates across the WHO MONICA Project populations
Lancet
 , 
2000
, vol. 
355
 (pg. 
675
-
87
)
51
Tunstall-Pedoe
H
Vanuzzo
D
Hobbs
M
, et al.  . 
Estimation of contribution of changes in coronary care to improving survival, event rates, and coronary heart disease mortality across the WHO MONICA Project populations
Lancet
 , 
2000
, vol. 
355
 (pg. 
688
-
700
)
52
Atkinson
L
Merry
G
Advances in neurotrauma in Australia 1970–2000
World J Surg
 , 
2001
, vol. 
25
 (pg. 
1224
-
29
)
53
Phelan
JC
Link
BG
Controlling disease and creating disparities: a fundamental cause perspective
J Gerontol B Psychol Sci Soc Sci
 , 
2005
, vol. 
60
 (pg. 
27
-
33
)