Jump to: navigation, search

Amenable Mortality - deaths potentially avoidable through health care

Basic Facts

Stage of development

Evaluated

Potential or current usage

This indicator is suitable as a whole-of-system health outcome measure for use in health system performance assessment. This indicator is a measure of the extent to which available treatments are applied to diagnosed conditions. The Ministry of Health currently defines amenable mortality as deaths from those conditions for which variation in mortality rates (over time or across populations) reflects variation in the coverage and quality of health care (itself defined as preventive or therapeutic services delivered to individuals or families).

Brief overview of the measure

General description

This indicator was selected to provide information on deaths that could have been avoided through timely access to health care, furthering an understanding of the cost- effectiveness of the health system.

Rationale for selection

Amenable or avoidable mortality is a well tested indicator that has been consulted widely and is accepted in academic circles as a whole-of-system health outcome indicator. In the area of effectiveness, amenable mortality helps to answer questions around the potential for gain in health outcome and the cost effectiveness of the health system: What health gain has resulted from the increase in health expenditure? Mortality-based indicators such as this one, available from routine surveillance systems, may partly meet this information need.

Type of measure

Outcomes

Domain(s) of quality

Effectiveness

Application and interpretation of the measure

Stated intent of the measure

To measure mortality rates potentially avoidable through timely provision of appropriate health care.

Caveats - Considerations

Interpretation of the variation in amenable mortality rates across DHBs is difficult due to wide uncertainty intervals (reflecting thin data for smaller DHBs- death being a rare event in some smaller places). Some methods are available to address this caveat (i.e. using hierarchical Bayesian modelling). The residual variation could be attributed to DHB performance, provided the adjustment for compositional differences is effective. Monitoring trends in performance will be more informative than a single point-in-time analysis, although the timeliness of cause of death data is currently problematic. Unlike avoidable hospitalisations (for example), which can be monitored quarterly with a rolling six-month delay, avoidable mortality can never be a real-time indicator, although the current two to three year delay should be reducible.

Level of health care delivery/setting

All levels of health care delivery are reflected in the mortality data.

Target population

This indicator is inclusive of all ethnicities and genders for people under 75 years of age.

Stratification by vulnerable populations

Stratification by ethnicity, socio economic status and age will be useful.

Calculation of the measure

Output of calculation

Avoidable mortality rates. An example of the output is: Under-75 mortality in males, New Zealand, 2006 (see upload)

Numerator description

All deaths under 75 years of age (by age, sex, ethnicity, NZDep) from the specified cause list. 35 conditions (or groups of related conditions) classified into six super-categories: • infections • injuries • maternal and infant conditions • cancers • cardiovascular diseases and diabetes • other chronic diseases. This definition is operationalised by means of a list of condition–intervention pairs applied to deaths under age 75 years (‘premature’ deaths). To be included in the list, the specified key intervention (or package of interventions) must be shown (by RCT or observational studies) to be capable of reducing the associated mortality from the linked condition by more than 30% within five years of effective coverage. Furthermore, the intervention should have been introduced within the preceding 40 years, and the condition should account for over 0.1% of all under-75 deaths in the period of interest. Given this redefinition, amenable mortality should be able to function effectively as a whole-of-system outcome indicator. Unnecessary delays in cause-of-death coding should be reduced, so that the indicator can be produced with less than a 12–15-month delay. Use of the indicator for policy purposes should begin by examining trends and contrasts in the indicator as a whole, then drill down (for each sociodemographic subgroup and region of interest) to the six cause groups, and finally to the 35 individual conditions included within the rubric.

Numerator exclusions

Non-residents

Denominator description

Populations (by age, sex, ethnicity, NZDep) under 75 years of age

Denominator exclusions

None

Time period

12–15-months

Criteria/standard for optimal performance

Avoidable mortality should be consistently defined as per the MoH publication on the website (http://www.health.govt.nz/publication/saving-lives-amenable-mortality-new-zealand-1996-2006)

Data source

Numerator: Mortality data (national collection) Denominator: Statistics NZ

Method of extraction

Death estimate Extract all deaths in individuals aged 0–74 years from the National Minimum Dataset Mortality Collection, by month and year of registration, from 1996 to 2006 (the most recent year available for cause of death). Exclude deaths of non-residents. Fields extracted for each death should include: month and year of registration of death, month and year of birth, sex, ethnicity, domicile (DHB), country of birth (New Zealand or overseas), NZDep score and decile (meshblock level), and underlying cause of death. Deaths can be classified into cause categories, and ‘avoidability’ categories, based on ICD-9 or ICD-10 code (the former applies from 1996 to 1999 and the latter from 2000 onwards). Population estimate Population data can be sourced from Statistics New Zealand. Spline interpolation based on Census population (1996, 2001 and 2006 censal years) should be applied to estimate population by ethnicity, by DHB (based on 2001 boundaries), and by NZDep (based on meshblock-level 2001 NZDep deciles from 1996 to 2001, and 2006 NZDep deciles from 2002 to 2006) for inter-censal years. In this way, population counts are derived by single year of age, sex, ethnicity, NZDep decile and DHB, for each calendar year from 1996 to 2006. Estimation of rates Amenable mortality rates should be age-standardised using the direct method, with the WHO world population as the standard. Rates by NZDep quintile should be double standardised by age and ethnicity, whereas rates by ethnic group do not need to be standardised for deprivation. This is because ethnicity is a confounder of the deprivation–mortality association, but deprivation (or socioeconomic position) is a mediator, not a confounder, of the ethnicity–mortality association. For ethnic standardisation of mortality rates by deprivation quintile, weights were constructed by multiplying the age weights by 0.141, 0.052, and 0.807 for Māori, Pacific, and non-Māori/non-Pacific age and ethnic-specific mortality rates respectively (based on the ethnic distribution in the 2001 Census). Note that rates are not standardised for variation in the distribution of the Asian population (because this has varied widely over time), but this should not greatly affect the double (age- and ethnic-) standardised rates.

Key issues and challenges for data management

Monitoring trends in DHB performance may be more informative than a single point-in-time analysis, although the timeliness of cause of death data is currently problematic. Unlike avoidable hospitalisations (for example), which can be monitored quarterly with a rolling six-month delay, avoidable mortality can never be a real-time indicator, although the current two- to three-year delay should be reducible to as little as 12 to 15 months, as is already the case in Australia. - See more at: https://www.hqmnz.org.nz/measures/staying-healthy/amenable-mortality#sthash.C9PR7qD4.dpuf

Appraisal of the measure

Availability of evidence to support application of the measure

Measure is formulated on and underpinned by an evidence based clinical practice guideline., Measure is formulated on and underpinned by evidence from a published systematic review, meta-analysis, or other peer-reviewed synthesis of clinical evidence relating to the area of focus., The measure has been reviewed using the Sieve Tool and a report is available., The measure has been cited in one or more peer-reviewed journals, applying or evaluating the properties of the measure., A formal consensus procedure involving experts in relevant clinical and/or methodological sciences has been completed and documented., The measure has been developed or endorsed by an organization that promotes rigorous development and use of clinical performance measures (at an international, national, regional or local level)., The measure has been developed or endorsed by an organisation seeking to improve clinical effectiveness as part of a continuous quality improvement cycle (at an international, national, regional or local level).

Evidence of feasibility and reliability of implementation

Reliability - The measure has been demonstrated to be reliable (i.e. free from random error)., Interpretation - The measure allows unambiguous interpretation of better or worse performance., Data extraction - Data collection specifications for the measure are well defined., Data sources - Required data elements for the measure can be obtained from existing data sources., Availability of data - Required data elements for the measure can be gathered during routine practice activities, IT software - Existing IT software is sufficient for data collection., Adaptability - Measure is able to be adapted for use in multiple care settings, Validity - The measure has been demonstrated to be valid (i.e. it measures what it purports to).

Development approach

1 The three most recently developed and widely used AM lists were identified: • Australia and New Zealand Atlas of Avoidable Mortality • Health of Nations (Nolte and McKee) • AMIEHS provisional list. 2 These lists were mapped against each other to identify candidate condition–intervention dyads. 3 Each condition-intervention dyad was filtered through the criteria developed above. 4 Those condition-intervention pairs passing the filter were included in a draft interim list. 5 An ad hoc expert panel reviewed the criteria, the cross-mapping, the filtering and the draft list. 6 The draft list was amended based on discussion with the expert panel (largely conducted by email), yielding the final interim list.

Other items

Additional information

Avoidable-mortality-rates.docx (18.0 KB)

Owner details

Reference number

599

Date of entry to library

2012-05-29 11:22:43

Owner (Organisation name)

Health Quality and Safety Commission, Indicators project team

Owner (Email contact)

richard.hamblin@hqsc.govt.nz

Creator (Organisation name)

Health Quality and Safety Commission

Creator (Email contact)

richard.hamblin@hqsc.govt.nz