Time Horizon

A time horizon is the period over which outcomes, costs, and consequences are assessed in a research study, particularly in health economic evaluations, clinical trials, and decision analytic models. The choice of time horizon determines which events and costs are captured in the analysis and can substantially affect the conclusions about whether an intervention is effective or cost effective. A time horizon may range from days or weeks in an acute care trial to a lifetime perspective in a cost effectiveness model comparing treatments for a chronic disease. The time horizon should be long enough to capture all important differences in outcomes and costs between the interventions being compared.

The time horizon decisively shapes the results and conclusions of both clinical and economic evaluations. A treatment that appears more expensive over one year may prove cost saving over a lifetime if it prevents costly complications. Similarly, a therapy that shows modest short term benefits may demonstrate substantial long term advantages when followed over a longer period. The National Institute for Health and Care Excellence (NICE), for example, asks for a time horizon long enough to reflect all important differences in costs or outcomes between the technologies compared; it considers a lifetime horizon usually appropriate when technologies affect costs and outcomes over a patient's lifetime, and needed when they lead to differences in survival or benefits that last for the rest of a person's life. Choosing too short a time horizon risks underestimating the benefits of prevention, early intervention, and treatments that alter disease trajectories, potentially leading to suboptimal resource allocation decisions.

The time horizon is specified at the design stage of the study or model. In a clinical trial, the time horizon is determined by the follow up period and may be extended through post trial observational studies. In an economic model, the time horizon defines how many cycles the model runs and over what total duration outcomes are projected. For decision analytic models such as Markov models, the cycle length (for example, one month or one year) and the total number of cycles together determine the time horizon. When the time horizon exceeds one year, both costs and health outcomes are typically discounted to reflect the time preference for benefits received sooner rather than later, using annual discount rates specified by the relevant health technology assessment body (for example, NICE in England uses 3.5% a year for both, and Canada's Drug Agency, formerly CADTH, uses 1.5%). Sensitivity analyses examine how the results change under different time horizons, providing insight into whether the conclusions are resistant to this choice.

The time horizon should be explicitly chosen and justified in any study that tracks outcomes or costs over time. Use a short time horizon (weeks to months) for acute conditions where all relevant outcomes occur quickly, such as treatment of an uncomplicated infection. Use an intermediate time horizon (one to five years) for conditions where the intervention's effects are expected to play out over a defined period, such as a rehabilitation program. Use a lifetime time horizon for chronic diseases, preventive interventions, and any comparison where important cost or outcome differences are expected to accrue beyond the trial observation period. For NICE, a horizon shorter than a lifetime could be justified only if there is no differential mortality effect and the differences in costs and outcomes relate to a relatively short period.

A common mistake is selecting a time horizon that is too short to capture all meaningful differences between interventions, leading to biased conclusions. For example, evaluating a diabetes prevention program over only one year misses the long term reductions in cardiovascular events and renal failure that constitute its primary value. Another error is failing to discount future costs and outcomes when the time horizon exceeds one year, or using an inappropriate discount rate. Researchers sometimes fail to conduct sensitivity analyses around the time horizon, leaving uncertainty about whether their conclusions depend on this modeling assumption. In clinical trial design, an inadequate follow up period may miss late treatment effects, including both delayed benefits and late appearing adverse events.

A cost effectiveness analysis compares a new cholesterol lowering drug to standard statin therapy for patients at high cardiovascular risk. In the five year clinical trial, the new drug reduces major cardiovascular events by 15%. A Markov model extends the time horizon to a lifetime: per patient, the drug's extra lifetime costs of $35,400 are partly offset by $28,000 of savings from avoided hospitalizations and procedures, a net incremental cost of $7,400, and the accumulated event reductions add 0.40 QALYs, yielding an incremental cost effectiveness ratio of $18,500 per QALY gained (costs and QALYs discounted at 3.5% a year). A scenario analysis using only the five year trial horizon, in which the drug costs a net $3,100 more per patient and adds 0.05 QALYs, produces an ICER of $62,000 per QALY. Against the willingness to pay threshold of $50,000 per QALY used in the analysis, the lifetime analysis finds the drug cost effective and the five year scenario does not, demonstrating how the time horizon decisively changes the economic conclusion.