Univariate sensitivity analysis. The overall estimate (top row) gives the projection with the default parameter values (underlined). The additional number of appointments, number on antiretroviral treatment (ART) and life-years saved are given in lower rows with one parameter changed at a time
| Number of appointments per month (1000s) | Number of ART (1000s) | Cumulative life-years saved (100000s) year | |||||
| Pre-ART | Starting ART | Follow-up on ART | |||||
| Year | 2010 | 2010 | 2010 | 2010 | 2030 | 2010 | 2030 |
| Overall estimate (with default parameters) | 41 | 4 | 18 | 110 | 667 | 1 | 62 |
| Extra with | – | – | – | – | – | – | |
| Initiation rule | – | – | – | – | – | – | – |
| Symptomatic | Ref | – | – | – | – | – | – |
| WHO-CD4 | 0† | 2 | 8 | 48 | 541 | 0 | 55 |
| Interval between scheduled appointments (pre-ART) | – | – | – | – | – | – | – |
| 12 months | −15 | 0 | −1 | −6 | −28 | 0 | −1 |
| 6 months | Ref | – | – | – | – | – | – |
| 3 months | 32 | 0 | 0 | 2 | 21 | 0 | 2 |
| Interval between scheduled appointments (post-ART) | – | – | – | – | – | – | – |
| 12 months | 0 | 0 | –9 | n/a* | n/a* | n/a* | n/a* |
| 6 months | Ref | – | – | – | – | – | – |
| 3 months | 0 | 0 | 17 | n/a* | n/a* | n/a* | n/a* |
| Effect of ART | – | – | – | – | – | – | – |
| Worst | 0 | 0 | 0 | 0† | –184 | 0 | –14 |
| Mid | Ref | – | – | – | – | – | – |
| Best | 0 | 0 | 1 | 0† | 111 | 0 | 9 |
| ANC referral and VCT uptake | – | – | – | – | – | – | |
| Poor ANC referral and low testing uptake | −14 | 0 | −2 | −13 | −119 | 0 | −6 |
| Good ANC referral but low testing uptake | Ref | – | – | – | – | – | – |
| Good ANC referral and high testing uptake | 18 | 1 | 4 | 25 | 200 | 0 | 16 |
| Effectiveness at finding diagnosed: | – | – | – | – | – | – | – |
| Few found (∼10% after 5 years) | −10 | −1 | −7 | −41 | −21 | 0 | −8 |
| Some found (∼50% after 5 years) | Ref | – | – | – | – | – | – |
| Many found (∼90% after 5 years) | 13 | 2 | 8 | 51 | 42 | 0 | 11 |
| Behaviour change: | – | – | – | – | – | – | – |
| No behaviour change | 17 | 1 | 6 | 38 | 193 | n/a* | n/a* |
| Best estimate (behaviour change in 1990s) | Ref | – | – | – | – | – | – |
| Behaviour change in 1990s + intervention from 2005 | 0† | 0 | 0† | 0† | −309 | n/a* | n/a* |
| Overall least estimate‡ | 14 | 3 | 12 | 71 | 158 | n/a* | n/a* |
| Overall greatest estimate‡ | 120 | 11 | 55 | 332 | 1996 | n/a* | n/a* |
| With same behaviour change assumed: | – | – | – | – | – | – | – |
| Overall least estimate‡ | 14 | 2 | 11 | 70 | 336 | n/a* | n/a* |
| Overall greatest estimate‡ | 119 | 11 | 55 | 331 | 1990 | n/a* | n/a* |
| With same effect of ART and behaviour change assumed | – | – | – | – | – | – | – |
| Overall least estimate‡ | 14 | 3 | 12 | 72 | 505 | n/a* | n/a* |
| Overall greatest estimate‡ | 120 | 11 | 54 | 327 | 1779 | n/a* | n/a* |
↵* These values are not meaningful because the model does not recognise that regularly monitoring on ART could prolong survival or that life-years are saved when incidence declines and infections are averted.
↵† Here the simulation output has been replaced by 0 because the actual output represented a small value (<5% of the overall estimate in the top row), which should not be interpreted as a meaningful difference in these stochastic simulations.
↵‡ The overall estimate for the least and greatest estimates make opposite sets of assumptions chosen (on the basis of the rest of the table) to give the outer-bounds for the projections in 2030 with the following exceptions: in all scenarios, behaviour change in the 1990s is assumed to have reduced incidence, and patients on ART are monitored every 6 months (in the model it is assumed that this has no influence on ART outcomes).
VCT, voluntary counselling and testing.