Te Wero tonu—the challenge continues: Māori access to medicines 2006/07–2012/13 update

He Karakia Whakatipuranga—A Blessing for Growth and Wellbeing

Manawa mai te mauri nuku
Manawa mai te mauri rangi
Ko te mauri kai au. He mauri tipua
Ka pakaru mai te Pō
Tau mai te mauri
Haumie!
Hui e Taiki e!!endnote A

Tēnā Koutou ngā mātāwaka o Aotearoa. PHARMAC is the New Zealand government agency that decides which pharmaceuticals to publicly fund.1,endnote B Under its Statement of Intent,2,endnote C PHARMAC has set three new strategic bold goals—with the first goal “to eliminate inequities in access to medicines by 2025”.

This article highlights PHARMAC’s updated information3 on Māori:non-Māori inequities in medicines access4,5 and how these inequities have changed over time. This is key data that will help the health sector prioritise, drive and monitor progress towards achieving this bold goal.2

Context

PHARMAC’s objective is to secure for eligible peopleendnote D(i) the best health outcomes reasonably achievable from pharmaceutical treatment and from within the funding provided.6 Its functions include engaging in research to meet its objective,6 which can include monitoring progress towards best outcomes.

Among eligible people in New Zealand, significant negative health disparities7 (ie, inequities8,9) exist, with Māori and Pacific peoples in particular experiencing poorer health outcomes than non-Māori/non-Pacific populations (see though endnote D(ii)). (Endnotes E10,11 and F7–9 further define and differentiate ‘equality’, ‘equity’ and ‘disparity’, and their uses.)

PHARMAC’s funding decisions assume people access funded treatments when prescribed and dispensed (according to the Pharmaceutical Schedule rules). PHARMAC takes a range of actions to support responsible and optimal use of funded treatments. Where evidence signals that people are missing out on benefitting from funded pharmaceuticals, PHARMAC can vary those actions for better access.

For better access for populations experiencing poor access/health outcomes, PHARMAC has developed its Māori Responsiveness Strategy, Te Whaioranga13 (which also helps meet Tiriti o Waitangi (Treaty of Waitangi) obligationsendnote G) and Pacific Responsiveness Strategy.14 Each strategy has community-based actions to improve access to medicines.

PHARMAC has committed to eliminate inequities in access to medicines by 2025;2 a dedicated team is driving the workplan to reach this access equity goal.

Medicines access inequities

Inequalities10,11,endnote E in health risks, disease rates, medication access and usage, and health outcomes between ethnic groups are well-described.16 While some of these inequalities are due in part to population characteristics and are unavoidable, they are also inequitable when associated with social, economic or health-system related factors that are unfair and avoidable.7–9,17 Inequity (unfair and avoidable difference) is the focus of this updated analysis.18

The evidence of inequities in health outcomes10,11,endnote E between ethnic groups in New Zealand is clear (see endnote H).19–23 Excess disease burden in Māori compared with non-Māori has been the leading cause of health loss in New Zealand, more than any disease or risk factor.24,endnote I Investing in the latest, sometimes very expensive, medicines and medical devices will not necessarily secure the best health (which includes equitable) outcomes at a population level.25 Social values26 and other issues7,27,28 such as clinical severity29 and health equity10,30 remain important. Better outcomes arise from continuing with important public health actions32–34—and having better access to, and uptake of, good healthcare. This means that everyone who needs care can and does get it35–37,endnote J—including medicines.

Previous analysis, and update

PHARMAC’s focus on best health outcomes including equity has led to developing ways to identify whether access to medicines use varies by ethnicity. In 2013 PHARMAC staff and others published a preliminary analysis,4 with an overview of medicines dispensed by prescription volumes, category and population dispensing rates for the financial year 2006/07 in Māori, Pacific peoples and non-Māori/non-Pacific peoples’ populations.endnote K The approach accounted for (i) age differences within each ethnic group, (ii) indicators of health need that combine morbidity and mortality (ie, health loss, in disability-adjusted life years (DALYs)), and (iii) breakdowns by patient numbers vs proxies for adherence. Adjusted for need, there was variable but sizeable differences in medicines dispensed to Māori compared with non-Māori, with Māori, eg, having 19–37% lower dispensings overall than non-Māori. There were however important limitations to what was preliminary analysis.

The preliminary study4 used the Ministry of Health’s New Zealand Burden of Disease Study (NZBDS) 2001,38,39 which quantified years of life lost by the New Zealand population from premature mortality and disability across many individual diseases. The NZBDS 2001 included some ethnic-specific data, using prioritised ethnicity.40 Disease burden estimates were for the year 1996. The NZBDS has been updated since (becoming the New Zealand Burden of Disease, Injury and Risk Factors Study (NZBDIRFS)).21,41,42

The earlier analysis4 has helped inform PHARMAC’s policy development for medicines funding and access. However, that analysis was preliminary and relied on disease burden estimates that had become especially outdated.38,39 To further its access equity goal, in 2015 PHARMAC commissioned UniServices (University of Auckland) to update the preliminary analysis. This present article describes the update, which extends the earlier analysis and aims for faster, more efficient routine future updates. The update used 2006/2007 and 2012/2013 dispensing claims data for publicly funded medicines and updated disease burden estimates from 2006 onwards.21 The full report is available on PHARMAC’s website (http://www.pharmac.govt.nz/tools-resources/research/maori-uptake-of-medicines/).3

The full UniServices update report3 again4 describes inequities in subsidised medicines access and persistence between Māori and non-Māori populations, and changes in access and persistence rates over time. The update also includes an overview of crude and age-standardised script rates for publicly funded medicines for key ethnic groups in New Zealand.

Methods

As with Metcalfe et al 2013,4 the UniServices update was an observational secondary analysis of medicines access and persistence (defined later) at a population level. It linked community prescription medicines dispensing claims data with primary health care organisation (PHO) enrolment data and burden of disease estimates (linking with anonymised person codes).

Data were obtained from prescription medicine dispensing claims for the financial years 2006/07 and 2012/13 in the New Zealand Pharmaceuticals Collection (patient-level dispensing of medicines listed on the New Zealand Pharmaceutical Schedule with demographic data).

The UniServices updated analyses3 included two analytical cohorts of medicines/people of most direct policy relevance:

X. medicines/people for people aliveendnote L on 30 June 2013 who were dispensed 1+ subsidised medicine between 1 July 2012 and 30 June 2013;

Y. medicines/people for people alive during all the seven years 1 July 2006 to 30 June 2013 who were dispensed 1+ subsidised medicine during both the 12-month period 1 July 2006 to 30 June 2007 AND the 12 months 1 July 2012 to 30 June 2013 (thus alive at the end of the two time periods, and restricted to the medicines/people cohort of existing medicines that were subsidised both between 1 July 2006 and 30 June 2007 and that continued to be subsidised at 1 July 2012). Medicines/people Cohort Y represents people dispensed medicines that were listed in both time periods but whose subsidy status or funding rules may have changed.

Endnote M provides more detail on the medicines/people cohorts.

Obtained medicines dispensing data (see endnote N) were linked, via ICD10 codes of relevant presumed/known medical condition(s), with data on disease burden for Māori and non-Māori populations obtained from the NZBDIRFS 2006–2016 report (published in 2013).21,endnote O Disease burden in NZBDIRFS is total population DALY losses, which combine incidence/prevalence and case severity (morbidity and years lost from premature death).

Outcome measures involved transformations of ratios of rate ratios for the medicines use and burden of disease data respectively, applying to incidences and denominating populations to derive counts of excess and deficit age/disease burden-adjusted scripts (Cohort X above).3,4,endnotes P,Q

Analysis over time in the UniServices update transformed rate ratios by time (Cohort Y above), using Keppel et al’s methodology for measuring change in absolute and relative disparities44 comparing 2012/13 with 2006/07.3 For this article, subsequent analysis was undertaken for statistical significance between the two time periods, using the Bucher method.45,46

Analysis then further disaggregated script excesses/deficits by access and persistence (access defined as a person being dispensed their first prescription for each item in the 12-month year; and persistence defined as a person continuing treatment with receiving subsequent dispensings in the year—see endnote R for further information); the further calculations are described in Metcalfe et al4 and the UniServices update.3

Please refer to the full UniServices report3 for further methodological detail, including relevant burden of disease data, data linking, and calculations. Subsequent analysis for statistical significance between the two time periods is in the Statistical Appendix to this article.

Key findings

The key findings of both sets of analyses are summarised below (noting these results are but part of the full UniServices analysis,3 which should be referred to at www.pharmac.govt.nz/tools-resources/research/maori-uptake-of-medicines). After adjusting for age and burden of disease, pervasive inequalities remained:

• For 2012/13, translating the relative dispensing rates to a deficit or excess of dispensing, there was a shortfall of 1,126,300 pharmaceutical treatments for Māori—ie, treatments that Māori did not receive. This shortfall comprised 41% of the treatments that could be expected to be dispensed to Māori had they been dispensed at rates equitable to non-Māori, when accounting for relative burden of disease data.
  
• Of the 1,126,300 shortfall in 2012/13 for Māori, approximately 608,800 (54%) represented lost opportunities for Māori to access treatment (ie, be dispensed a first prescription for that item in the 12-month year, ‘access’), and the remaining 46% represented unexpected gaps in ongoing treatment, with people not getting continued medicine they’d previously accessed (‘persistence’).endnote R
  
• Over time, inequalities had continued. For the cohort of medicines available in 2006/2007, between 2006/2007 and 2012/2013 the overall disease burden-adjusted inequalities in medicine dispensings between Māori and non-Māori for Cohort Y nominally widened by 6%, but this trend was not statistically significant (comparing the Māori vs non-Māori age/disease burden-adjusted standardised rate ratio (RR) overall in 2012/2013 against that in 2006/2007, ie, 0.594/0.629=0.944=-6% relative change, 95% uncertainty interval (UI) 0.552–1.615).endnote S Realistically, certain inequity in access to medicines for Māori, however, remained (2012/13 Māori vs non-Māori age/disease burden-adjusted standardised RR 0.594, 95% UI 0.407–0.867). See the Statistical Appendix to this article for details on the uncertainty estimations.
  

Figure 1 shows trends and large variability in the rate ratios for individual therapeutic groups over time.

• The overall increase in the apparent gap seemed due to a further deterioration in ‘access’, while relative persistence had improved—so in 2012/2013 the proportion of Māori receiving their first prescription (compared with expected had they received prescriptions at the same rate as non-Māori) had decreased compared with in 2006/2007, but those who did so were staying on their medicines for perhaps a little longer (relative to non-Māori) compared with in 2006/2007;endnote T however, confirmatory statistical testing awaits (adapting the methods in Statistical Appendix).
  
• Much caution is needed interpreting these results, due to many complexities, caveats and limitations3—and further uncertainty calculations are awaited (ie, uncertainty limits around multiple point estimates and rate ratios, see Statistical Appendix). Nevertheless, important apparent inequities in disease burden-adjusted script rates continue to exist for cardiovascular disease, asthma and COPD, mental health (particularly the management of anxiety and depression), diabetes, cancer and bacterial infections.
  

Figure 1: Change and variability in rate ratios for Māori:non-Māori age/disease burden-adjusted script rates, Cohort Y, 2006/07 and 2012/13 (loge scale).

source: UniServices report3 Table 6.

Key:

Interpretation:

• At a rate ratio of 1.0 (depicted as loge(1.0)=0.0), Māori and non-Māori have equal rates of medicines use (ie, scripts received), after adjusting for population, age and disease burden.
  
• The higher the rate ratio (towards the top of the graph), the greater the extent that medicines use in Māori exceeds that of non-Māori after adjusting for population, age and disease burden.
  
• The lower the rate ratio (towards the bottom of the graph), the greater the extent that medicines use in Māori trails that of non-Māori, adjusted for population, age and disease burden.
  
• Rate ratios are depicted logarithmically, ie, on a natural logarithm (loge) scale, where, eg, -1 = Māori age standardised/disease burden-adjusted rate is 37% that of non-Māori (depicted loge (RR 0.37)=-1.0), -0.7 = Māori rate is half that of non-Māori (loge (RR 0.50)=-0.69), 0.4 = Māori rate is two-thirds non-Māori (loge (RR 0.67)=-0.41), 0.0 = Māori rate equals non-Māori (loge (RR 1.0)=0.0), +0.4 = Māori rate is 50% higher than non-Māori (loge (RR 1.5)=+0.41), etc.
  

Hence, eg:

• Total medicines – M:nM age/disease-adjusted script rate ratio RR 0.63 in 2006/07=37% shortfall for Māori overall; RR 0.59 in 2012/13=41% shortfall; thus a small worsening of the already sizeable shortfall, but numerical counts only of medicines and not tested for statistical certainty; depicted as loge(0.63)=-0.462, loge(0.59)=-0.528.
  
• Cardiovascular medicines – 2006/07 adjusted RR 0.56=44% shortfall, 2012/13 RR 0.58=42%, thus a small improvement in the still sizeable shortfall, but numerical counts only of medicines and not tested for statistical certainty; depicted as loge(0.56)=-0.579, loge(0.58)=-0.545.
  
• Respiratory medicines – 2006/07 adjusted RR 0.75=25% shortfall, 2012/13 RR 0.68=32%, thus a worsening in the shortfall, but numerical counts only of medicines and not tested for statistical certainty; depicted as loge(0.75)=-0.288, loge(0.68)=-0.386.
  
• etc.
  

Please refer to the full report including its online appendices and to the Statistical Appendix to this article for further detail.

Discussion

Tēnā rā Koutou. E toru ngā tino mātāpono o Te Tiriti o Waitangi. Ko te noho rangapū-partnership. Ko te mea whakaurunga-participation. Ko te whakamaru-protection. Kei a tātou tonu katoa te wero kia mahikaha, kia mahi tino mōhio, me te mahitahi. (Greetings. The three key principles of te Tiriti o Waitangi—partnership, participation and protection—help guide all of our work. The challenge continues for us to work harder, work smarter, and work together.)

Caveats

As stated in the UniServices update report, there are many important complexities, caveats and limitations to the analysis, and caution is needed interpreting its results. At least 29 of these limitations are outlined in depth over 10 pages (pages 55 to 64 of the update report itself). These include major influencers such as:endnote U

• the standard population age structure used (technical but important);
  
• diluted gaps by including other groups with high disease burden and low access in the non-Māori comparator (eg, Pacific peoples);
  
• use of prioritised ethnicity;
  
• how to interpret gaps themselves (unnecessary overuse with wastage by the non-Māori comparator? true under-use by Māori? Māori experiencing harm from over-use of suboptimal regimens?); and
  
• at a medicine-specific level, how changes in persistence relate to optimal treatment durations, and how changes in prescribing relate to changes in standard treatment pathways;
  

alongside many other caveats. Further caveats (not stated in the UniServices update) include:

• ageing of cohort Y (by excluding patients who die);
  
• possible bias from numerator/denominator mismatch using PHO populations as numerators but Statistics New Zealand census population denominators—affecting pharmaceutical estimates and age-specific ethnic proportions.
  

As well, although the UniServices update is mainly based on Metcalfe et al’s4 methodology, direct comparison of findings from the Metcalfe et al publication and the updated analysis is invalid for several reasons, including different populations, different burden of disease methods, different age standards and different medicine-disease linkages.endnote V Cohort Y in the updated analysis instead provides valid internally-consistent comparison over time.

More detailed level analysis at an individual medicine level provides some evidence of both good and relatively less good access to some more commonly prescribed medicines. With individual medicines and therapeutic subgroups (many hundreds), there is much subtlety and variation in the gaps and their changes by time. These data are available in the full report and its associated data tables online,3 and deserve further investigation, including pharmacoepidemiological research incorporating clinical event data and discussion with relevant parts of the health sector.

The research cannot provide disease-burden adjusted information for other ethnic groups and others, as burden of disease data is unavailable for ethnicities other than Māori, nor other groups, eg, those suffering socioeconomic deprivation. This misses likely large inequities in other groups beyond Māori, while diluting the true extent of Māori medicines inequities compared with, say, New Zealand European people. For example, Pacific peoples are recorded as non-Māori, which means the comparison between Māori and non-Māori would most likely show greater gaps if Pacific peoples’ data were excludable from the non-Māori data.

Implications

The overall burden-adjusted approach3,4 therefore complements and adds to, rather than replaces, other research into disparities in prescription medicines access.5 Nonetheless, these updated apparent inequities in medicines access and use were linked to chronic conditions responsible for ~88% of the burden of disease in New Zealand; and given the magnitude and extent of the observed inequities (and lack of countervailing evidence, with consistency with other datasets and studies), plausibly, apparent inequities have not only existed but also persisted for government-funded pharmaceuticals in New Zealand.

PHARMAC’s funding decisions should not create or worsen barriers to people accessing medicines, and PHARMAC acts to support optimal and equitable prescribing and uptake (part of PHARMAC’s responsible use of medicines statutory function). However, the causes of these apparent inequities are likely to be complex and systemic.47 Addressing the complex barriers to accessing medicines and optimising their use requires a whole of sector approach.47

Once adjusted for burden of disease, inequalities become one (or a combination) of three factors:

1. true disparities (inequity in access or persistence, ie, Māori not receiving sufficient of a medicine if at all compared with non-Māori, thus lost health gain opportunities);
   
2. wastage (the non-Māori comparator group is receiving excess medicines, unnecessarily, without real gains but with near-inevitable side effects); or
   
3. harm (Māori receiving excess medicines of lesser benefit and/or greater adverse effects, and thus experience harm, ie, net health loss via opportunities foregone, compared with the non-Māori comparator group receiving better or ‘gold standard’ treatments; eg, Māori receiving more of older antipsychotics and/or depot antipsychotics, but less of newer antipsychotics and/or oral antipsychotics, than non-Māori, or higher rates of inhaled beta-agonist asthma asthma relievers but lower rates of inhaled corticosteroid preventers4).
   

Inequities in healthcare and outcomesendnote W borne by Māori and other New Zealanders, including medicines access, are unacceptable (Martin Luther King Jr saying “Of all the forms of inequality, injustice in health is the most shocking and inhuman…”48). Health inequities are inconsistent with principles of social justice and human rights, including indigenous rights as reaffirmed by te Tiriti o Waitangi49 and the United Nations Declaration on the Rights of Indigenous People (UNDRIP).50–52 This is where the lack of improvement in top-line medicines access for Māori signals that the broader health systemendnote X as a whole has yet to take all the “necessary steps” for indigenous people to attain equal standards of health (as per UNDRIP article 24(2),50,51 supported by New Zealand52) (see endnote Y).

Human life and potential is wasted when not everyone gets the healthcare they are entitled to—when every person in New Zealand should have the same access to the funded medicines they need; as a society, we lose opportunities when people don’t get to live, thrive and participate.53

Future directions for PHARMAC with this work will be:

(a) Further quantitative research and tools development

The multi-factorial nature of medicines access inequities suggests that we will need multi-agency approaches to provide the range of solutions and interventions to improve equity of access. Recognising this, in addition to the report, PHARMAC has commissioned UniServices to develop two additional tools of use to funders, policymakers and others within the health sector. These are:

• an updateable process using the New Zealand Universal List of Medicines (NZULM) to link community, cancer and hospital medicines listed on the Pharmaceutical Schedule (Sections B and Section H Part II) to the NZBDIRFS data through the Anatomical Therapeutic Chemical (ATC) classification system; and
  
• a geospatial analysis of variation in access to pharmaceuticals, adjusted for disease burden, by DHB areas, including the creation of an interactive map to visualise this data.
  

Other research activities PHARMAC is considering include:

• improving the validity and reliability of DALY-adjusted dispensing measurement, ie, the epidemiology/pharmacoepidemiology (see section 9 of the full UniServices report3);
  
• commissioning or otherwise securing more comprehensive burden of disease data for New Zealand tailored to PHARMAC’s needs, eg, including Pacific and perhaps Asian peoples as discrete ethnic populations; using updated prescription data to 2017/18; and using varying standard populations54 for age-standardisation;
  
• working with PHARMAC therapeutic group managers for individual medicines within individual therapeutic group levels, to identify particular gaps and needs for future research;
  
• seeking objective advice from PHARMAC’s clinical advisers for specific areas;
  
• as above, pharmacoepidemiological research at individual medicine level, incorporating clinical event data and discussion with relevant parts of the health sector.
  

(b) Behavioural and health systems research

The causes of inequities are complex, and solutions lie beyond simply the funding of medicines or simply the health system. There are likely barriers to equity at multiple levels,5,47 including:

• patient/population factors as access barriers to healthcare (including accessing appointments, delayed access), related to costs, transport, family structure, expectations, beliefs, etc;
  
• health system factors with structural barriers such as how care is organised (eg, accessing appointments, wait times, after-hours advice and access, completing referrals); and
  
• health professional factors leading to differential treatment, with inability of providers and health systems to address all groups’ needs equitably (institutional and professional bias, cultural competency,55 health literacy involving health professionals (ie, beyond patients/whānau),56 knowledge and skills, adherence, etc.)
  

—all in the context of inequities in wider underlying structural and systems57 (including institutional and professional bias), social and economic determinants of health.10,16–18,22,30,47,55,57–70

More broadly, PHARMAC’s newly established Access Equity team will lead further work better understanding what barriers Māori, and other under-served groups including those with relatively poor health outcomes,7–9 face in accessing and using medicines, including down to the level of particular medicines or therapeutic groups. Such work could include how population factors, health professional factors and health system factors interact to produce inequities,65–70 alongside a behavioural science with medicines/health system focus. This work would aim beyond simply patient and whānau behaviours; it would extend to, importantly, prescriber and other health sector provider behaviours and systems effects47 too, and their interactions65–70—where such research remains comparatively sparse, and yet has great potential to advance Māori health.

c) Implementing PHARMAC’s equity bold goal

The aim is a robust evidence-base and policy work programme that will focus on:

• Identifying key points of intervention and prioritising these by their amenability and potential to address inequities;
  
• Reviewing programmes that have successfully reduced health inequities and identify why;
  
• Opportunities to work with PHARMAC’s partners (Whānau Ora collectives and other sector partners) to develop locally-based programmes to reduce inequities;
  
• Working with system-level partners (clinical, consumer, Māori, other groups experiencing disparities, national health bodies) to identify gaps and influence policy and practice barriers;
  
• Better understanding the barriers to funded medicines being prescribed and used optimally, eg, commissioning further research; and
  

Better ways to monitor and evaluate PHARMAC’s progress over time.

PHARMAC will also continue to implement Te Whaioranga, its Māori responsiveness strategy,13 and its Pacific Responsiveness Strategy,14 both having access equity at their heart.

(d) Implications for the wider health sector and beyond

Although this analysis is about access inequities for medicines, these inequities’ causes and responses will be those that apply to generic healthcare inequities—to solve medicines access inequities, alongside the other healthcare inequities. Healthcare disparities comprise health system factors, health professional factors and patient/population factors—so that any inequity in healthcare access, quality or outcomes is ultimately the result of a complex interaction of factors.47 These factors are themselves complex and entrenched—as a result of historical and contemporary social, political, cultural and economic processes. Hence we need a systemic (and in fact multi-sectoral) approach.47

Eliminating inequities, in access to/use of medicines or health inequities more broadly, will therefore require efforts and partnerships beyond accessing and use of medicines alone, covering wider aspects across the whole health sector and afar.71,72 This includes public policy, regulators and professional quality assurance organisations, universities and other training providers, personal skills and engaging patients and whānau, community action, and health services themselves.10,47

Leadership and commitment by the health system, health organisations and health practitioners is required, with the expectation that all New Zealanders will have equity of health outcomes.73

Everyone working in the health system—government agencies like PHARMAC,74 the Ministry of Health,73 district health boards75 and primary health organisations, the Health Quality & Safety Commission (HQSC), doctors, nurses, pharmacists, public health, pharmaceutical suppliers, others—has a role to play to reduce these inequities and make sure funded medicines reach all the people who need them. Good engagement and partnership is essential with tāngata whenua13,22 and other populations experiencing poor health outcomes7 and variations in medicine use.

Policymakers and funders need to look for ways to allocate Vote Health funds so that population groups are not unduly burdened by pharmaceutical co-payments5 or access to primary care itself due to cost (eg, 15% Pacific and 14% Māori adults reporting they’re unable to pick up prescriptions due to cost, 22% Māori adults at times not visiting a GP due to cost) or other factors.76

Organisations should commit to, fund and be accountable for63,73,74,75 medicines equity targets. This will need expertise, support, guidance, collaboration and engagement with affected communities and others in the health system. Setting systems and organisational performance indicators and targets73 may substantially improve equity of access, as with, eg, the Health Targets for childhood immunisation.77

Health services must make sure primary care and pharmacy services are redesigned and set up in ways accessible, available and acceptable to all. This may include rethinking the physical location of pharmacies and primary care services and considering alternative ways for patients to receive both medicines and advice about taking the medicine in ways that work for them.

Interventions to help achieve medicines access equity should partner with those most affected. Cultural competence for both prescribers and pharmacists needs to continue to improve,55,78 alongside ensuring people’s experience of the health system overall is culturally safe.79,endnote Z Unwarranted variation in medicines prescribing and access should be routinely reported on at the DHB level, as a quality improvement concern.

System wise, there are good frameworks for how the medicines access equity bold goal could be achieved. These include the Ministry of Health’s Equity of Healthcare for Māori framework,73 and others including a Māori implementation framework (He Pikinga Waiora).80

There is also participatory/co-design actively involving all stakeholders (eg, end users/customers/patients and whānau, citizens, partners, iwi/hapū/marae, health providers, funders and agencies),81,82 and the importance of Māori leadership.73,83

Information underpins all of this.5 For tāngata whenua, health service and outcomes data are taonga, and Māori researchers and health providers can be kaitiaki in partnership, with good data governance including dissemination and accountability for progress84–87 within kaupapa Māori research principles.88–93

For the health sector, high-quality ethnicity data is needed to measure and monitor healthcare and outcomes for ethnic groups and identify health inequities; implementing the revised Ethnicity Data Protocols—collecting ethnicity data accurately, appropriately, and often—will be crucial for everyone.94

For researchers, issues of age standards need to be promoted both in the New Zealand health sector and wider, including using the age structure of the groups experiencing the greatest disadvantage.54 Using equal explanatory power study designs (ie, sampling equal number of participants from groups experiencing poorer health outcomes and the comparator population)95,96 and equal explanatory analysis and equity focus reporting (which involves reporting on the equity gap as well as by ethnicity)97 can help gain at least the same depth and breadth of information for smaller disadvantaged groups—for fairer comparisons in policy and funding decisions.

Finally, the current analyses derive from an administrative dataset (the New Zealand Pharmaceuticals collection) and therefore have limited ability to capture relevant clinical variables and system-related barriers. For example, the evolution of data repositories generated by electronic transmission of prescriptions and eDispensing would allow, via anonymised data linkage, better understanding of primary and secondary non-adherence. Linking clinical encounters, or indeed prescriptions, to SNOMED/Read codes would provide more accurate mapping of dispensings to NZBDIRFS categories.

Conclusion

Inequities in access to medicines are unacceptable, and PHARMAC is committed to eliminating these inequities, as a priority. The findings in the Updated Variation in Medicines Use by Ethnicity report3 provide a good evidence-base5 to inform PHARMAC’s access equity activity and commitment for 2025, and people and the health sector in general.

PHARMAC will be working with its partners in the health sector, tāngata whenua and others to better identify barriers and underlying causes of these inequities and act to improve use of medicines—narrowing and eliminating the gaps.

Nō reira!! Kei te mau tonu tātou i te wero, kei a tātou ngā kaimahi hauora katoa. Ko te wero tonu, kia hikina te hauora Māori kia tae orite ki te Hauora-a-tauiwi i te tuatahi. Kei te werohia tonu te wero nei mo ake tonu atu!! Kia mau!! (Therefore!! The challenge remains for all of us in the health system. The challenge of equitable health outcomes of Māori with non-Māori is the first challenge. The challenge is ever-present. Seize the opportunity!!) Everyone in the health sector has a role.

Note: The full Auckland UniServices update report is available at http://www.pharmac.govt.nz/tools-resources/research/maori-uptake-of-medicines/

Endnotes are available here.

Statistical Appendix

Calculating 95% uncertainty limits for Māori and non-Māori age/disease burden/population-adjusted script rates and rate ratios

Context and overall method

The UniServices analysis on Variation in medicines use by ethnicity: a comparison between 2006/7 and 2012/13 (http://www.pharmac.govt.nz/tools-resources/research/maori-uptake-of-medicines/)3 provided point estimates of inequities in Māori:non-Māori script rates, adjusted for age, population and disease burden, but did not calculate uncertainty. It is thus limited by not assessing for random error (chance) and other uncertainty; such was not required in PHARMAC’s commissioning of the research in 2015. The following supplementary analysis retrofits and retrospectively calculates confidence limits and uncertainty limits (akin to confidence intervals) for overall age standardised rates and year-specific rate ratios for Cohort Y, and relative change over time.

In particular, the UniServices analysis reported a nominal 6% change in relative uptake (Māori:non-Māori script rates, adjusted for age, population and disease burden) for Cohort Y overall over the six years 2012/13 vs 2006/07. This is from the calculated ratio of rate ratio (RR) of 0.944 when comparing the 2012/13 rate ratio (RR 0.594 overall M:nM age standardised disease burden-adjusted scripts) with the 2006/07 rate ratio (RR 0.629), where 0.594/0.629 = the 0.944 RR = the 6% relative reduction (1 (ie, equipoise) minus 0.944). The following analysis thus includes retrospectively calculating uncertainty limits for that 0.944 ratio of rate ratios, to examine chance or non-sampling error as a possible likely reason for the 6% change.

To assess uncertainty, the datasets in the analysis (scripts, burden of disease DALYs) were treated as distinct entities otherwise not directly comparable, and were thus combined using methods for indirect comparison.98 This approach is common to economic analysis, with the use of model simulation etc. to assess uncertainty.

So separate to the UniServices report,3 we have calculated 95% confidence or uncertainty limits for age-standardised rates,99 and used the Bucher method98,100 to calculate 95% uncertainty limits (ULs), using the following three steps:

1. Firstly, extracting or calculating standard errors for both Māori and non-Māori age-standardised rates for overall scripts and for overall disease burden for each time period;
   
2. Then for each time period, calculating and combining rate ratios (RRs) for Māori:non-Māori (M:nM) age-standardised rates for overall scripts and for overall disease burden (disability-adjusted life years lost (DALYs)), with sample-based confidence limits (CLs) and ULs for scripts and disease burden respectively;
   
3. Then calculating and combining the M:nM script:disease burden RR and confidence/uncertainty intervals comparing the 2012/13 period with 2006/07, using standard errors.
   

The use of uncertainty limits for disease burden (rather than simple sample-based confidence limits) was as used in the Global Burden of Disease Study (GBDS),101 to account for added uncertainty from modelling—ie, accounting for additional non-sampling error, with both measurement error from model instability in the input non-fatal health loss (YLD) component of disease burden inputs, and model specification error from Rx/disease mapping.

(This is where, internationally (including for New Zealand), the GBDS42,101,102 now reports 95% uncertainty intervals (UIs) rather than confidence intervals (CIs). Unlike confidence intervals, UIs capture uncertainty from multiple modelling steps, as well as from sources such as model estimation and model specification, rather than simply from sampling error alone. Uncertainty associated with estimation of mortality and years of life lost (YLLs) due to premature mortality reflects sample sizes of data sources, adjustment and standardisation methods applied to data, parameter uncertainty in model estimation, and uncertainty within all-cause and cause-specific mortality models. For estimation of prevalence, incidence, and years of life lived with disability (YLDs), UIs incorporate variability from sample sizes within data sources, adjustments to data to account for non-reference definitions, parameter uncertainty in model estimation, and uncertainty associated with establishment of disability weights. The GBDS assumes that because direct information about the correlation between uncertainty in YLLs and YLDs has been scarce, uncertainty in age-specific YLDs is assumed independent of age-specific YLLs or death rates.101)

Equations

Direct age-standardised rate ratios (RRs) and their standard errors calculated as:99

RR = (ASR1)/(ASR2),

95% CI for RR = ((ARS1)/(ASR2))1±Z/χ, where

Z is standardised normal deviate (1.96 for 95% CIs),

χ (ie, variance) = (ASR1 – ASR2)/√((SEASR1)2 + (SEASR2)2);

SEASR is the standard error for an age-standardised rate;

algebraically, SE = (95% CI or UI) / Z

Bucher method RR for indirect comparison,98,100

where

RRc = RRa × RRb (= exp((ln(RRa)+ln(RRb))):

95% CI or UI = exp (Σ((lnRRa,lnRRb,..) ±Z√(Σ(var(ln(RRa)), var(ln(RRb)),var…))

= exp (Σ((lnRRa,lnRRb,..) ±Z√(Σ(SERRa2, SERRb2, …))

where

ln is natural logarithm loge, exp is natural exponential base e, var(ln(RR)) = SE, var(RR) = SE2, Z = 1.96

Calculations

The above three steps were calculated and combined as follows:

1. Age-standardised rates with standard errors

Using standard methods for direct age standardisation,99

• ASRM,s,1 Māori direct age-standardised overall scripts in 2006/07
  

= 7154.9 per 1000 population age-standardised scripts,

standard error (SE) ±273.6:1,000

• ASRnM,s,1 non-Māori age-standardised overall scripts in 2006/07
  

= 6057.5:1,000 age-standardised scripts, SE ±116.7:1,000

• ASRM,s,2 non-Māori age-standardised overall scripts in 2012/13
  

= 8517.8:1,000 age-standardised scripts, SE ±299.7:1,000

• ASRnM,s,2 non-Māori age-standardised overall scripts in 2012/13
  

= 7685.2:1,000 age-standardised scripts, SE ±140.1:1,000

2. Rate ratios with standard errors and uncertainty limits

Using the Bucher method RR for indirect comparison,98,100 and age-standardised rates data from Appendices F and G of the UniServices analysis (at http://www.pharmac.govt.nz/assets/2018-02-26-Maori-uptake-of-medicines-appendices.xlsx),

• s1 rate ratio (RR) Māori:non-Māori (M:nM) overall age-standardised scripts in 2006/07
  

s1 = ASRM,s,1/ASRnM,s,1

= 7,154.9/6,057.5 per 1,000

RR = 1.1812, 95% CI 1.1809–1.1814, standard error (SE) ±0.00011626

• s2 RR M:nM overall age-standardised scripts in 2012/13
  

s2 = ASRM,s,2/ASRnM,s,2

= 8,517.8/7,685.2 per 1,000

RR = 1.1083 (1.1082–1.1085), SE ±0.00005646

• d1 RR M:nM overall age-standardised DALYs in 2006/07,
  

d1 = ASRM,d,1/ASRnM,d,1

RR = 1.741, 95% UI 1.300–2.331, SE ±0.1938;

(where sample error-only 95% CI is 1.7017–1.7811, SE ±0.0148)

• d2 RR M:nM overall age-standardised DALYs in 2012/13,
  

d2 = ASRM,d,2/ASRnM,d,2,

RR = 1.741, 95% UI 1.301–2.329, SE ±0.1931;

(where sample error-only 95% CI is 1.7018–1.7810, SE ±0.0116)

where:

• s2’s age distribution is proxied by 2006/07 age distribution
  
• d1’s by 2013 New Zealand Burden of Disease, Injury and Risk Factors Study (NZBDIRFS)42,102 standard errors (proportional to point estimates) to total disease then calculated for Māori and non-Māori
  
• d2’s proportional standard errors for calculating 95% confidence limits are proxied by 2006 NZBDIS21,41,103 standard errors (proportionate to point estimates) for total disease for Māori and non-Māori (adjusted for RR 1.754), where Māori in 2006 experienced 207,150 DALYs (sample error-only SE 2,323), non-Māori 747,426 (sample error-only SE 5,320).
  

Note that the standard errors for the 2006 NZBDIS DALY estimates,21,41,103 for total disease for Māori and non-Māori, are based solely on sampling error-derived 95% confidence intervals. By contrast, the standard errors for the 2013 NZBDIRFS DALY estimates,42,102 for total disease for total population (ie, not stratified by ethnicity for Māori and non-Māori), are based in sampling and nonsampling error-derived uncertainty intervals. This means that available standard errors for DALYs in 2006 are necessarily smaller than available standard errors for DALYs in 2013; standard errors in the 2006 NZBDIS relate to 95% confidence limits, whereas the bigger standard errors in the 2013 NZBDIRFS related to less confident uncertainty limits.

3. Ratio of rate ratios, with 95% uncertainty limits

Using the Bucher method again,98,100

Calculation 1: rate ratio for M:nM disease burden-adjusted age-standardised scripts in 2006/07:

RR1 = s1/d1 = 1.18/1.74 = 0.629

95% UI = exp(ln(RR1) ±Z√(Σ(SE(s1)2,(SE(d1)2)))

= exp(ln(0.629) ±1.96√((00011626)2+(0.1938)2)))

= 0.430 to 0.920

(And where corresponding 95% CI (ie, sample error only) is similarly calculated substituting new SEs in the above equation, ie 95% CI = 0.611 to 0.648)

Calculation 2: rate ratio for M:nM disease burden-adjusted age-standardised scripts in 2012/13:

RR2 = s2/d2 = 1.083/1.74 = 0.594

95% UI = exp(ln(RR2) ±Z√(Σ(SE(s2)2,(SE(d2)2)))

= exp(ln(0.594) ±1.96√((00005646)2+(0.1931)2)))

= 0.407 to 0.867

(With corresponding sample error-only 95% CI = 0.518 to 0.608)

Calculation 3: rate ratio 2012/13 vs. 2006/07 for M:nM disease burden-adjusted age-standardised scripts:

RR3 = RR2/RR1 (= (s2/d2)/(s1/d1)) = 0.594/0.629 = 0.944,

95% UI = exp(ln(RR3) ±Z√(Σ(SE(s1)2,(SE(d1)2,(SE(s2)2,(SE(d2)2)))

= exp(ln(0.944) ±1.96√((00011626)2+(0.1938)2)+(00005646)2+(0.1931)2)))

= 0.552 to 1.615

(With corresponding sample error-only 95% CI = 0.910 to 0.980)

Interpretation and extended use

For each of the individual years 2006/07 and 2012/13, Cohort Y’s rate ratios for M:nM disease burden-adjusted age-standardised scripts were statistically significant.

• For 2006/07, with the rate ratio for M:nM disease burden-adjusted age-standardised scripts of 0.63, 95% UI 0.43 to 0.92, the overall adjusted rate in Māori was 37% less than expected vs. non-Māori (calculated from 1 minus 0.63).
  
• For 2012/13, with the rate ratio for M:nM disease burden-adjusted age-standardised scripts of 0.59, 95% UI 0.43 to 0.92, the overall adjusted rate in Māori was 41% less than expected vs. non-Māori (calculated from 1 minus 0.59)
  

However, Cohort Y’s relative differences in overall adjusted scripts over time were not statistically significant.

• With the ratio of rate ratios 2012/13 vs 2006/07 for M:nM disease burden-adjusted age-standardised scripts of 0.944, 95% UI 0.552 to 1.615, the relative change over the 6 years was -5.6%, with a plausible range (95% UI) of -61.5% to +44.8% (calculated from 1 minus 0.944, 1 minus 1.615, 1 minus 0.552)
  

Hence, although Cohort Y’s overall differences were significant for individual years, the magnitude of the overall difference did not change significantly over the six years. We were unable to exclude chance and accepted modelling artefacts, with uncertainty limits, causing any nominal 6% “deterioration” in Cohort Y’s M:nM inequity over time. The 6% gap could have plausibly improved by half, or deteriorated by 3/5ths. Simply, there was no improvement in the overall pattern over the six years, but likewise no good evidence that any “deterioration” was real and overt.

(Confining analysis to sampling error, ie, just confidence limits, did provide statistically significant deterioration, with a range around the 6% relative worsening of 2 to 10%, but this excluded additional nonsampling modelling error, so is not reasonably valid.)

The above approaches can be used to assess uncertainty in PHARMAC’s and others’ future monitoring of disease burden-adjusted script inequities, including one-year prevalence by therapeutic subgroup and major pharmaceuticals, access vs persistence, etc.

Because of the suitability of sampling-only error-derived standard errors for pharmaceutical usage (with 95% confidence limits), but not for burden disease (which require additional nonsampling error, to derive 95% uncertainty limits), note we would be more confident of detecting changes in pharmaceutical usage over time, but less so detecting changes in disease burden and consequent DALY-adjusted pharmaceutical usage.

‍