Table 2.
Parameters in Computer Simulation*
Parameter | Data Source (Reference) | Strength of Evidence
|
Does Data Source Meet Strict Evidence Criteria?
|
Parameter Distribution if Data Source Used | Parameter Distribution if Data Source Not Used | |||||
---|---|---|---|---|---|---|---|---|---|---|
Study Design | Internal Validity | External Validity | Study Design Level 1 | Internal Validity Good | External Validity High | All 3 Criteria | ||||
Mortality rate in absence of HIV | Observational; life tables (12) | 1 | Good | Low | Yes | Yes | No | No | Point estimates, variable | Uniform (0.5X, 1.5X estimates) |
| ||||||||||
Mortality rate attributable to HIV | Observational; 1 study in similar population (13) | 1 | Good | Low | Yes | Yes | No | No | Normal (0.19, 0.06) | Uniform (0, 0.38) |
| ||||||||||
Impact of HIV treatment on mortality | Observational; 13 studies pooled from similar populations (14) | 1 | Good | High | Yes | Yes | Yes | Yes | Normal (0.15, 0.02) | NA |
| ||||||||||
Probability of taking HIV medications | Observational (15) | 1 | Fair | Low | Yes | No | No | No | Never used† | NA |
Observational (16) | 1 | Good | High | Yes | Yes | Yes | Yes | Normal (0.75, 0.02) | NA | |
Observational (17) | 1 | Good | Low | Yes | Yes | No | No | Never used† | NA | |
Observational (18) | 1 | Good | Low | Yes | Yes | No | No | Never used† | NA | |
| ||||||||||
Effectiveness of DOT | Randomized, controlled trial; 1 study in dissimilar population (19) | 1 | Good | Low | Yes | Yes | No | No | Normal (0.46, 0.01) | Uniform (0, 2) |
| ||||||||||
Utility with HIV | Observational; 1 study in similar population (20) | 1 | Poor | Low | Yes | No | No | No | Normal (0.87, 0.04) | Uniform (0.5, 1.0) |
| ||||||||||
Decrement in utility with HIV treatment | Observational; 1 study in similar population (unpublished) | 1 | Poor | Low | Yes | No | No | No | Normal (0.05, 0.01) | Uniform (0, 0.5) |
| ||||||||||
Annual cost of DOT | Expert opinion | 3 | Poor | Low | No | No | No | No | Never used† | NA |
Observational; 1 study in dissimilar population (21) | 2-2 | Good | Low | No | Yes | No | No | Point estimate $4600 | Uniform ($200, $36 500) | |
Observational; 1 study in dissimilar population (22) | 2-2 | Poor | Low | No | No | No | No | Never used† | NA | |
Observational; 1 study in dissimilar population (23) | 2-2 | Poor | Low | No | No | No | No | Never used† | NA | |
| ||||||||||
Annual cost of nondrug HIV care | Observational; 1 study in similar population (24) | 1 | Poor | High | Yes | No | Yes | No | Normal ($9000, $300) | Uniform ($200, $20 000) |
Observational; 1 study in similar population (25) | 1 | Poor | High | Yes | No | Yes | No | Never used† | NA | |
| ||||||||||
Annual cost of HIV drugs | Observational; 1 study in similar population (25) | 1 | Good | High | Yes | Yes | Yes | Yes | Normal ($10 300, $700) | NA |
Individual data sources were eligible to inform parameter estimations if their strength of evidence met or exceeded criteria in 3 separate domains (study design, internal validity, and external validity). If no data sources met strength of evidence criteria, a uniform distribution across a wide plausible range was substituted. When more than 1 study met strength of evidence criteria, the parameter’s distribution was based on the study with the most precise statistical estimate. Normal = normal distribution (mean, standard deviation); uniform = uniform distribution (lower bound, higher bound). DOT = directly observed therapy; NA = not applicable because data source either was never used or met all 3 of the strict evidence criteria.
“Never used” because results were statistically less precise than those of another study or studies with equal or superior grades of evidence in all 3 domains. It is also possible that a particular data source could be statistically more precise but have lower strength of evidence than alternative data sources, although this situation did not arise with our example. In accord with our decision rules, we would have used the statistically more precise data source when evidence criteria were more inclusive and the statistically less precise data source when evidence criteria were less inclusive.