A total of 384 evaluations met the inclusion criteria (see Appendix). Of the 318 Australian evaluations, 94 were reported in the original ACE-Prevention Report and also published in a journal article (used as the primary source), 137 were published only in the Report. Beyond the original ACE-Prevention Report, 15 more evaluations arose from an obesity report and another 72 arose from 23 peer-reviewed articles. All 66 NZ evaluations came from 25 peer-reviewed articles arising from the BODE3 Programme.
We excluded three interventions deemed no longer relevant: circumcision of men to prevent HIV infection (not relevant in NZ and Australia),(16) and two cardiovascular disease polypill evaluations with a price of AU$5000 per year(16) – which is far higher than current pricing.
Characteristics of included evaluations
Table 1 shows the characteristics of evaluations, by country. The majority of Australian evaluations were published before 2015 (77.7%) and used 2003 as the base-year, while the majority of NZ evaluations were published after 2015 (92.4%) and all used 2011 as the base-year. All NZ evaluations used a lifetime perspective, and only 4.1% of the Australian evaluations had less than a lifetime perspective – and these were mostly mental health and communicable diseases interventions. All evaluations used a 3% discount rate. The majority of Australian evaluations were targeted (8.2%) or partially targeted (56.6%), while the NZ evaluations were mostly population-wide (72.7%). The large majority of evaluations in both countries (92.4%) were preventive. Over half (57.3%) of evaluations across both countries were for interventions that persisted over the remainder of the population’s life span (e.g. tax interventions) with the second most common intervention duration being one-off or up to 1 year (20.8% overall). Over half of Australian evaluations were related to cardiovascular disease (29.6%), overweight & obesity (14.2%) and diet (13.5%). The NZ evaluations were more concentrated by domain, with 81.8% being from one of three domains (cancer, dietary salt and tobacco).
Table 1
Characteristics of included evaluations
|
Australia
|
New Zealand
|
Total
|
|
N
|
%
|
|
%
|
|
%
|
Total evaluations
|
318
|
|
66
|
|
384
|
|
Year published:
|
|
|
|
|
|
|
2010-2014
|
247
|
77.7%
|
5
|
7.6%
|
252
|
65.6%
|
2015-2018
|
71
|
22.3%
|
61
|
92.4%
|
132
|
34.4%
|
Base-year in model
|
|
|
|
|
|
|
2000-04
|
207
|
65.1%
|
0
|
0%
|
207
|
53.9%
|
2005-09
|
40
|
12.6%
|
0
|
0%
|
40
|
10.4%
|
2010-15
|
71
|
22.3%
|
66
|
100%
|
137
|
35.7%
|
Time horizon
|
|
|
|
|
|
|
10y to <lifetime
|
13
|
4.1%
|
0
|
0%
|
13
|
3.4%
|
Lifetime
|
95
|
95.9%
|
66
|
100%
|
371
|
96.6%
|
Discount rate (annual)
|
|
|
|
|
|
|
3%
|
318
|
100%
|
66
|
100%
|
384
|
100%
|
Other
|
0
|
0%
|
0
|
0%
|
0
|
0%
|
Degree of targeting
|
|
|
|
|
|
|
Population-wide
|
112
|
35.2%
|
48
|
72.7%
|
160
|
41.7%
|
Partially targeted
|
180
|
56.6%
|
14
|
21.2%
|
194
|
50.5%
|
Targeted
|
26
|
8.2%
|
4
|
6.1%
|
30
|
7.8%
|
Intervention duration
|
|
|
|
|
|
|
One-off or up to 1 year
|
64
|
20.1%
|
16
|
24.2%
|
80
|
20.8%
|
1-5 years
|
9
|
2.8%
|
1
|
1.5%
|
10
|
2.6%
|
6-20 years
|
20
|
6.3%
|
1
|
1.5%
|
21
|
5.5%
|
Persistent
|
172
|
54.3%
|
48
|
72.7%
|
220
|
57.3%
|
Not specified
|
53
|
16.7%
|
0
|
0%
|
53
|
13.8%
|
Type of intervention
|
|
|
|
|
|
|
Prevention
|
298
|
93.7%
|
57
|
86.4%
|
355
|
92.4%
|
Treatment
|
20
|
6.3%
|
4
|
6.1%
|
24
|
6.3%
|
Missing
|
0
|
0%
|
5
|
7.6%
|
5
|
1.3%
|
Domain
|
|
|
|
|
|
|
Cancer
|
27
|
8.5%
|
8
|
12.1%
|
35
|
9.1%
|
Alcohol
|
16
|
5.0%
|
0
|
0%
|
16
|
4.2%
|
Cannabis or other illicit drugs
|
5
|
1.6%
|
0
|
0%
|
5
|
1.6%
|
Communicable disease
|
7
|
2.2%
|
5
|
7.6%
|
12
|
3.1%
|
Cardiovascular disease
|
94
|
29.6%
|
1
|
1.5%
|
95
|
24.7%
|
Diabetes
|
13
|
4.1%
|
0
|
0%
|
13
|
3.4%
|
Diet
|
43
|
13.5%
|
0
|
0%
|
43
|
11.2%
|
Injury
|
1
|
0.3%
|
5
|
7.6%
|
6
|
1.6%
|
Mental illness
|
8
|
2.5%
|
0
|
0%
|
8
|
2.1%
|
Other NCD
|
26
|
8.2%
|
0
|
0%
|
26
|
6.8%
|
Overweight & obesity
|
45
|
14.2%
|
1
|
1.5%
|
46
|
12.0%
|
Physical activity
|
16
|
5.0%
|
0
|
0%
|
16
|
4.2%
|
Salt (dietary)
|
3
|
0.9%
|
32
|
48.5%
|
35
|
9.1%
|
Tobacco
|
14
|
4.4%
|
14
|
21.2%
|
28
|
7.3%
|
HEALTH GAIN
|
HALYs per 1000 total population
|
< 0.10
|
122
|
28.4%
|
8
|
12.1%
|
130
|
33.9%
|
0.10 - 1
|
86
|
27.0%
|
2
|
3.0%
|
88
|
22.9%
|
1 - 10
|
77
|
24.2%
|
30
|
45.5%
|
107
|
27.9%
|
>10
|
28
|
8.8%
|
22
|
33.3%
|
50
|
13.0%
|
Missing†
|
5
|
1.6%
|
4
|
6.1%
|
9
|
2.3%
|
HALYs per person in target population
|
<01
|
28
|
8.8%
|
23
|
34.9%
|
51
|
13.3%
|
01-099
|
0
|
0%
|
20
|
30.3%
|
20
|
5.2%
|
0.1-0.99
|
1
|
0.3%
|
3
|
4.5%
|
4
|
1.0%
|
Missing†
|
289
|
90.9%
|
20
|
30.3%
|
309
|
80.5%
|
INCREMENTAL HEALTH EXPENDITURE
|
Net cost* per 1000 total population
|
< US$0 [Cost saving]
|
103
|
32.4%
|
46
|
69.7%
|
149
|
38.8%
|
US$0 to $10,000
|
99
|
31.2%
|
9
|
13.6%
|
108
|
28.1%
|
Cost > US$10,000
|
67
|
21.1%
|
8
|
12.1%
|
75
|
19.5%
|
Missing†
|
49
|
15.4%
|
3
|
4.6%
|
52
|
13.5%
|
Net cost* per target population
|
< US$0 [Cost saving]
|
1
|
0.3%
|
32
|
48.5%
|
33
|
8.6%
|
US$0 to $1000
|
24
|
7.6%
|
10
|
15.2%
|
34
|
8.9%
|
Cost > US$1000
|
4
|
1.3%
|
2
|
3.0%
|
6
|
1.6%
|
Missing†
|
289
|
90.9%
|
22
|
33.3%
|
311
|
81.0%
|
COST per HALY or Incremental cost-effectiveness ratio
|
Cost saving
|
97
|
30.5%
|
47
|
71.2%
|
144
|
37.5%
|
US$0 to $50,000 per HALY
|
127
|
39.9%
|
17
|
25.8%
|
144
|
37.5%
|
>US$50,000 per HALY
|
82
|
25.8%
|
1
|
1.5%
|
83
|
21.6%
|
Dominated
|
3
|
0.9%
|
0
|
0%
|
3
|
0.8%
|
Missing†
|
9
|
2.8%
|
1
|
1.5%
|
10
|
2.6%
|
*2016 US$ |
† Most studies reported results for HALYs and costs for either a total population perspective (e.g. for all eligible people in Australia) or a per capita perspective – but not both (although we were able to sometimes calculated both if sufficient data was reported in the paper). |
The majority of evaluations across both countries (97.7%) were able to have results expressed for the total population, whereas only a minority (19.5%) were able to have results expressed per capita of a target population. The distribution of HALYs gained and net costs per 1000 of the total population, and per capita of the target population, are shown in Table 1. Most (71.2%) of the NZ evaluations were cost-saving, but only 30.5% of the Australian ones were. Conversely, 26.7% of the Australian evaluations had either an ICER >US$50,000 (beyond the rule of thumb of Gross Domestic Product (GDP) per capita being a threshold beyond which interventions are deemed not cost-effective) or were dominated (i.e. performed worse than the comparator), compared to only 1.5% of the NZ evaluations.
Selected examples of ANZ-HILT outputs
Figure 1 demonstrates a histogram output from ANZ-HILT showcasing a selection of interventions. The left-hand panel of ANZ-HILT allows the user to select permutations of: domain (three here: cancer, diet and tobacco), actual interventions (seven here), currency and year to show dollars in, range of publication years that the evaluation was published in, and the outcome variable to plot (HALYs here). The tabs: across the top allow the user to view instructions, population denominator (here per 1000 of the total population), or cost-effectiveness plane (see Figure 2 below); and beneath toggle between table and ‘plot top 10’ options (plot shown here). A hover-over with one’s computer mouse allows the user to see the following for each evaluation: expected number of HALYs / discount rate / time horizon of follow-up / comparator / and intervention duration and/or frequency. For example, for the 10% per annum tobacco tax intervention, hovering over the bar will cause the following text to appear: “53,200 / 0.03 / Lifetime / Business-as-usual (no tax increases from 2011-to 2025) / 14 years of tax increases, then persistent”. Further information can be found in the table tab, e.g. the actual lower and upper uncertainty limits.
Turning to the substantive patterns in Figure 1, there is an over 1000-fold variation from 23 per 1000 (95% uncertainty interval (UI): 18 to 29) HALYs gained over the remainder of life for NZ population aged 35+ years in 2011 for 25% of salt in processed food being replaced by potassium and magnesium salts (where that intervention is ‘left on’ for the remainder of the population’s lifespan), down to 0.019 per 1000 (95% UI: 0.011 to 0.029) for all stage III colon cancer patients diagnosed in 2011 being assisted by cancer care coordinators to navigate more quickly and with higher coverage to surgery and chemotherapy. With this (and other intervention combinations), differences in the timespan and target groups of interventions often account for differences in the magnitude of health gains.
This variability in intervention conceptualisation noted, we can make some inferences from the selected examples shown Figure 1:
-
25% salt substitution with potassium and magnesium salts (NZ) and food taxes across saturated fat, excess salt, sugar-sweetened beverages and sugar in processed foods (Australia) leads to similarly large health gains.
-
Interventions such as ‘Tick’ logos on health food and dietary advice for those with high blood pressure have much smaller health gains when summed up across the population.
The health system expenditure impacts of the interventions shown in Figure 1 are correlated – but with large cost-savings (due to future disease rates being considerably lower) for interventions with large health gains (Supplementary Figure 1). Intervention impacts can also be presented in terms of HALYs gained per capita in the target population (Supplementary Figure 2).
Figure 2 is an example of a cost-effectiveness plane output for five evaluations, with two overlays: the text pop-ups that appear as the user hovers over each point are shown for all five interventions; the black dashed line is a super-imposed threshold line at about GDP per capita per HALY. The plane now allows the user to simultaneously see (often massive) variation between interventions in all of health gain, cost and cost per HALY gained. For this example, we present results in 2016 US$. For example, the tobacco retail outlet reduction intervention is in the southeast quadrant with substantial health gains (7 HALYs per 1000 over the remaining lifespan of the population) and cost-savings (US$89,100 per 1000). On effectiveness and efficiency grounds alone, this intervention should be considered for implementation – but there are other considerations such a political will and societal preferences that are not captured in ANZ-HILT. The Helicobacter pylori screening programme (to detect infection that is then treated with antibiotics, reducing stomach cancer incidence rates years into the future) is in the northeast quadrant, but beneath the willingness to pay line (black dashed line) suggesting it is cost-effective at a GDP per capita threshold. The computed tomography (CT) screening of heavy smokers is also in the northeast quadrant costing US$42,000 per QALY gained (i.e. 33.9/0.81) – about the threshold GDP per capita per QALY gained.