The cost of child health inequalities in Aotearoa New Zealand: a preliminary scoping study

This study is positioned within a Kaupapa Māori theoretical framework, a research approach that is critically mindful of how research processes may contribute to the marginalisation of indigenous peoples [14, 15]. Kaupapa Māori methodology is driven by a Māori world view and recognises the complexity of Māori historical and contemporary realities. Understanding the current health inequities experienced by Māori children and young people requires an understanding of New Zealand’s colonial history and the erosion of the social, economic, political, and cultural base for Māori whānau and iwi (families and kinship groups/tribes), from the early 1800s onwards [3]. Kaupapa Māori methodology privileges the indigenous Māori “voice”, validating and legitimising Māori language, knowledge, culture and values. It acknowledges and challenges the power dynamics that have created and maintain the unequal position of Māori in New Zealand society, including the role played by the socio-economic determinants of health and the health system in limiting Māori health outcomes [15].

In this study, quantitative epidemiological and economic costing methods are used as tools, with questions and findings developed and interpreted within a Kaupapa Māori framework. The analysis reflects this understanding by placing Māori children as the group “at the centre” of this analysis rather than at the margins.

The focus of this study is on inequities; therefore our analysis is based on the deaths and hospitalisations that are categorised as “potentially avoidable”, using New Zealand Ministry of Health and District Health Board (DHB) definitions [16‐18]. This includes illness and injury preventable through primary care intervention, injury prevention or health promotion approaches. The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM) diagnostic coding was used for all admissions and deaths [19]. We assigned children to Māori and non-Māori groups using Ministry of Health protocols for prioritised ethnicity; that is, anyone who identified as Māori, either alone or as one of multiple ethnic groups, was included in the Māori group [20].

Datasets were obtained with permission from Statistics New Zealand (SNZ) and the New Zealand Health Information Service (NZHIS). We calculated the number and crude rates of death from avoidable causes by ethnicity and age group (age groups: 0–1 month, 1 month < 1year, 1- < 5years, 5- < 15years) using SNZ death registration data coded by NZHIS for 2003–2007. Live births from 2003–2007 (SNZ birth registration data, excluding stillbirths, coded by NZHIS) were used as the denominator to estimate neonatal (<28days) and post-neonatal (defined as 28days- < 1year) infant mortality and hospital admission rates [unpublished datasets, SNZ and NZHIS]. For mortality rates in children over one year, population estimates from New Zealand Census data were used to estimate Māori and non-Māori rates [21, 22].

Avoidable deaths were assigned by their primary ICD-10AM code into ICD-10AM chapter groupings. The number of avoidable deaths that would have occurred had Māori had the same rate as non-Māori in each age band was computed, with the difference between the actual number of deaths and the estimated deaths representing excess avoidable deaths. The resulting number of years of life lost (YLL) by Māori children was computed, assuming that all would have lived to the 2005 life expectancy of their non-Māori same-sex counterparts [23].

We obtained the number and crude rates of hospitalisations from potentially avoidable causes by ethnicity and age group (Māori, non-Māori; age groups: 0- < 1year, 1- < 5years, 5- < 15years) from hospitalisation data (National Minimum Dataset, NZHIS) for the same period, 2003–2007. The number of avoidable hospitalisations that would have occurred if Māori had the same rate as non-Māori in each age band and diagnostic grouping (ICD-10AM chapter) was computed, with the difference between the actual number of hospitalisations and the estimated hospitalisations representing excess avoidable hospitalisations.

We assumed that injuries were as avoidable for Māori as for non-Māori children, thus that any difference in injury rates was inequitable, and that the incidence of injury and primary care-responsive illness was at least as high in Māori as non-Māori children [24, 25]. Primary care utilisation rates were estimated based on Primary Health Organisation (PHO) service utilisation data for visits for enrolled populations during routine hours for the years 2007–8 [unpublished datasets, used with permission of District Health Boards NZ]. Total utilisation rates were compared, as the quality of diagnostic coding in primary care precluded sub-group analysis. The number and rate of general practice (first point of contact personal and family primary medical care services) and primary care nursing visits were calculated for Māori and non-Māori children by age group (0- < 5years, 5- < 15years). The number of consultations that would have occurred if Māori had the same rate as non-Māori were computed by age group, with the difference between the actual number of visits and the estimated number representing “excess” or “under-utilisation.”

Similar methods were applied to Accident Compensation Corporation (ACC) injury claims, pharmaceutical and laboratory claims and outpatient datasets, estimating excess or under-utilisation for total claims/visits. Laboratory and pharmaceutical data have only recently been linked to NHI numbers (and thus able to be disaggregated by ethnicity), so we used the most recent years available (2006–8), as NZHIS advised NHI linking was then over 95% complete [personal communication, NZHIS].

We reviewed the available literature on costing health inequities and assessed relevant economic methods. This review suggests that there are significant unresolved challenges in valuing child health and disability that continue to be debated in the literature. Many of the assumptions of neoclassical economics do not hold for children and are indifferent to concepts of equity [26‐32]. Monetising the value of a child’s life and discounting future benefits (used by economists to reflect the assumption that future benefits are less valued than benefits enjoyed in the present) remain contentious in the health sector and do not fit easily within a kaupapa Māori framework. Re-interpreting cost benefit analysis theory within such a framework is an interesting direction for future kaupapa Māori research. In the absence of this, we chose to use “cost of illness” methods [33‐37] in this preliminary study, with a limited prevalence-based perspective. This restricts estimates to a range of publicly funded health sector costs, the cost of premature mortality, “out of pocket” care-giver expenditure on primary care, and loss of days of work for care-givers, due to excess avoidable hospital admissions of their children. We estimated an annual cost based on the total costs of the “inequity excess” for the period 2003–7.

Costing data for hospital admissions was based on the National Pricing Project (NPP) 2008–9 dataset [unpublished data, Ministry of Health]. The direct hospital cost savings (or extra costs) if Māori children had avoidable admission rates equal to non-Māori children was computed using the weighted median Diagnostic Related Group (DRG) costs for each ICD-10AM category and age group. ACC injury claims cost data, Pharmaceutical Management Agency of New Zealand (PHARMAC) pharmaceutical claims and NZHIS laboratory services claims for 2007–8 were used to cost those services. Māori:non-Māori differences in the use of ACC, laboratory and pharmaceutical services were costed using the weighted median costs of non-Māori claims in each age group. Paediatric outpatient visits (including mental health) were assigned the Weighted Inlier Equivalent Separations (WIES) inter-district flow cost (2008–9) for a paediatric medical outpatient visit of $343.23.

We estimated the difference in primary care costs using the “notional” Ministry of Health general medical subsidy amount per visit and assumed a nursing visit total cost of $30 [38, 39].

For primary care, we conservatively assumed the median “out of pocket” expenditure for children under six in 2007–2008 as zero, that children aged six and over were all subsidised $15 for general practitioner (family doctor) and nursing visits, and that caregivers “out of pocket” payments represented the difference between that and the notional visit cost [38‐40]. A conservative estimate of the cost of work days lost (measured as wages) was based on loss of two days work (the weighted median length of stay in this study) for each “excess” admission, for one parent at the 2007–2008 gross median wage ($NZ35,000) [41].

We used non-Māori male and female life expectancy from SNZ Life Tables 2005–7 [23] to derive the excess avoidable “Years of Life Lost” (YLL). Various discounting rates were then applied to test the impact of different time preferences, using discount rates applied in recent New Zealand cost of illness studies [33‐36].

Two methods were used to assign costs to YLL. Firstly, we assumed a “base case” scenario of a mean life expectancy at birth of 78 years, a 3% discount rate, and estimated the total cost of YLL by dividing the “base case” Value of a Statistical Life (VoSL) of $NZ3.352million [35, 36] by the Statistical Lives lost (“Method 1”). We assumed that the VoSL included lost production costs.

We also computed costs applying standard economic methods used by O’Dea & Tucker and others [34, 35] (“Method 2”) to allow comparability with other New Zealand cost of illness studies. Using this methodology, the VoSL has the same value for both genders, but given life expectancy is gender-dependent, the value of a “year of life lost” varies depending on age and gender of the avoidable deaths. The annualised value of the VoSL is calculated using the equation VosL = Sum [VoSLY/(1 + d)^L, where d is the discount rate and L the life expectancy. Therefore the VoSLY = VoSL*d/[1-(1 + d)^−L, where d = the discount rate, and L is life expectancy.

There were a total of 871,094 hospital admissions for children <15 years in the period 2003–2007. The crude rate ratio for total Māori:non-Māori admissions was 0.98 (95% CI 0.975–0.984). 36% of hospitalisations of children in this age group were classified as “potentially avoidable”. Māori:non-Māori rate ratios for potentially avoidable hospitalisation rates were significantly greater than one in all age groups. Respiratory diseases made the greatest contribution to the “excess”, with diseases of the ear, digestive system and skin, and injuries also being important. In children aged 5–15years the highest rate ratio was for circulatory diseases (including rheumatic heart disease), where the Māori:non-Māori rate ratio was 3.33 (95% CI 2.93–3.79). Overall, summing the “excess” avoidable hospitalisations, there were a total of 15,376 “excess” Māori avoidable admissions during 2003–7 or 3,075 each year.

Hospital outpatient utilisation rates for Māori children were consistently lower than for non-Māori children for all age groups in the period July 2006-December 2008 (Table 1). The overall Māori:non-Māori rate ratio was 0.864 (95% CI 0.862–0.867). If Māori children had the same outpatient consultation rate as non-Māori there would have been 23,373 more consultations by Māori children annually. Māori children were nearly 30% less likely than non-Māori to attend outpatient mental health services. If Māori had the same consultation rate as non-Māori there would have been 5740 more consultations by Māori children per year in mental health services.

There were 291,502 Māori and 1,322,560 non-Māori ACC claims for children < 15years during the 2003–2007 period. Māori rates were significantly lower than non-Māori, with an overall rate ratio of 0.680 (95% CI 0.678–0.683). There were 26,442 fewer claims per year by Māori children in the period than if claims had been made at the non-Māori rate.

A mean number of 353,734 Māori and 1,374,972 non-Māori children were enrolled in Primary Health Organisations (PHOs) during 2007–8; fewer in total than the SNZ population estimates for the same period, and a significantly lower proportion of Māori children than in official population estimates. If Māori children had consulted GPs at the same rate as non-Māori, there would have been 40,041 more GP consultations annually. Nursing consultations showed the inverse; although total nursing consultations were much fewer than GP consultations, Māori utilisation rates were 1.68 times greater than non-Māori, resulting in 17,194 consultations more than expected if Māori utilisation had been equivalent to the non-Māori rate.

There were over 7.9 million pharmaceutical claims for children <15 years during 2007–8; only 21.7% of these were for Māori <15years. Māori rates of claims were significantly lower than non-Māori, representing an “under-utilisation” of 198,108 claims per year.

In 2007–8 there were 1.28 million laboratory claims for children <15years; 12.9% of these were recorded as being for Māori children. The overall Māori:non-Māori rate ratio was 0.454 (95% CI 0.452–0.455); rate ratios in children <1years were even lower. There would have been 101,922 more claims by Māori children each year if Māori had the same rate of laboratory use as non-Māori.

In the period 2003–7 there were 299,421 live births and 2,345 deaths in children aged 0- < 15yrs. 85% of these deaths were “potentially avoidable”. If Māori avoidable mortality rates had equalled non-Māori rates for each age group in the period, there would have been 333 fewer Māori deaths (nearly 67 deaths per year); 56% of these would be in the 28 day- <1year age group (Table 2).

Māori avoidable mortality rates were significantly higher than non-Māori in all age groups except for the first month of life. The biggest contributor to mortality in this age group was deaths related to perinatal conditions (mainly resulting from prematurity), where Māori infant mortality rates were non-significantly higher than non-Māori.

In the age group 28days-<1year, 41.5% of the avoidable deaths were assigned “Sudden Unexpected Death in Infancy” (SUDI) codes. SUDI rates were significantly higher for Māori and especially Māori males in this age group; respiratory diseases (J00-99 codes) were the other main contributor to Māori infant deaths under one year.

There were significantly higher rates in Māori for “external causes” (accident, injury and assault codes V01-Y98) in all age groups, with the highest rates and rate ratios seen in infants 28 days - <1year. In girls 1–5years, the Māori:non-Māori rate ratio for injury/external causes was 3.35 (95% CI 2.00–5.62), and this accounted for nearly all the excess deaths. In 5–15year-old girls, nearly two thirds of the “excess” mortality was injury-related. In 5–15year-old males, injury/external causes again contributed to the majority of “excess” deaths, although the highest Māori:non-Māori rate ratios were for the ICD-10AM “A–B” and “E” codes (infectious and parasitic diseases, and endocrine, nutritional and metabolic diseases).

The years of life lost (YLL) to excess avoidable mortality in each age/sex group was computed, based on the difference between the midpoint of age of death in each age group and the non-Māori life expectancy for their same sex counterpart. The total value was 5210 life years lost per year. We examined the data using a range of discount values, based on those used in recent NZ cost of illness studies [33‐37, 42‐44]. Applying a discount rate of 3% reduces the life years lost to 38% of their present value; if it is increased to 8%, it is only 16% of present value (831 life years) (Table 3).

The Method 1 “base case” scenario assumes a mean life expectancy at birth of 78 years, YLL discounted at 3%, and uses the New Zealand Ministry of Transport VoSL at June 2008 prices ($3,352,000) [36, 42]. The YLL at 3% discounting are equivalent to the value of 25.7 “Statistical Lives” annually, computing to a cost of $NZ86.18 million annually at present value.

Method 2 results in similar values to Method 1 at zero discounting ($NZ223.3million per year), but as the VoSLY increases in Method 2 with increases in the discount rate, at 3% this computes to $NZ224 million per annum.

The cost of work days lost by caregivers is based on the median length of stay for avoidable hospital admissions. This computes to $269 for each of the 3075 annual “excess” avoidable admissions, a total cost of $827,175 per year. Māori caregivers “out of pocket” payments for primary care were estimated to be $245,523 less than expected, primarily due to lower GP consultation rates for tamariki Māori.

Table 4 shows the total annualised cost estimate for each method.

We examined the impact of different discount rates, using both methods. Method 1 results in a rapid decrease in present value as the discount rate increases. However using O’Dea & Tucker’s method there is little variability in the total value at differing discount rates (Table 5).

We also varied the “value of a statistical life year” (using values derived from previous studies [34‐37]); predictably this has a significant impact on the final result, as the cost of “excess” avoidable mortality is a large proportion of the total costs (Table 6). The final range of estimates shows the total cost is highly sensitive to the method used, and the discount rate and VoSL applied.