THIS PAGE IS UNDER CONSTRUCTION AND WILL BE UPDATED SOON – IF YOU HAVE IDEAS ON HOW TO IMPROVE IT – REACH OUT TO US
In this BMJ publication, there is a worked example of how to estimate TNT, which may be helpful if you want to learn the method.
TNT can be expressed as:
- the clinician time needed to improve the outcome for one person (TNTNNT)
- the clinician time needed to provide the intervention for all eligible in a population (absolute TNT)
- the proportion of the total clinician time available for patient care needed to implement the intervention for everyone eligible (relative TNT)
TNT estimates are based on assumptions about:
- how much clinician time it takes to provide the recommendation to each person
- the fraction of the population that is eligible for the recommendation
- the numbers needed to treat (NNT) to improve the outcome for one person
- the total available clinician time in the relevant context
- the category of healthcare personnel that will perform the work
The clinician time needed for each person
The first step is to estimate how much clinician time it takes to implement the recommendation for each person. To do this, the recommendation often needs to be divided into several “sub-interventions” in a care-cascade.
Imagine for example a recommendation to screen for, diagnose, and treat type 2 diabetes. Such a recommendation could consist of the following “sub-interventions”: screening for risk factors for type 2 diabetes with a questionnaire; measuring blood glucose in people scoring above a certain cut-off at the questionnaire; interpreting the blood glucose test and inform the individual; give advice on lifestyle to all people above a certain cut-off on the blood glucose test; prescribe medicine for all people above other cut-offs on the test; follow-up; deal with incidental findings; and so on.
The time it takes for the clinician to provide each of the “sub-interventions” recommended for each person should be estimated as accurately as possible. If a lot of uncertainty exists about how much clinician time a recommendation will take to implement – it might be sensible to make estimates for a worst- and best-case scenario – as well as for the most likely scenario.
The fraction of the population eligible
The next step is to estimate the eligible fraction of the population. Note that this is most often different for each of the “sub-interventions” (see above) – i.e., while the whole adult population may be eligible for screening for risk factors for type 2 diabetes with a questionnaire, only the fraction of the population that turn out to be at high risk through this questionnaire may be eligible for testing of blood glucose, and only the fraction of the population that are above a certain cut-off at the blood glucose test may be eligible for further interventions, and so on.
The numbers needed to treat
To estimate the TNT to improve the outcome for one more person (TNTNNT), we need to know the number-needed-to-treat (NNT) for the recommended intervention.
Note that TNTNNT do not reflect the certainty or relevance of the chosen outcome, and TNTNNT estimates are therefore not directly comparable across interventions. In other words, a high TNTNNT may well be acceptable when the outcome is a patient-relevant outcome with high-certainty evidence, while it may not be if the outcome is a surrogate outcome with low-certainty evidence of a beneficial effect.
Note that NNTs needs to be estimated for each “sub-intervention” – i.e., we need to estimate how many people have to be screened, how many people have to be tested, how many people have to be given lifestyle advice, and how many people have to receive medicine, and so one, for one person to experience an improved outcome from the whole cascade of sub-interventions.
The available clinician time
Information about the available clinician time for the relevant category of healthcare personnel within a certain country or region is often available at health authorities’ websites. Estimates of the available clinician time should be based not only on the number of clinicians available but also on information of the average time spent on clinical work per clinician. The clinician time should then be converted into full-time employments – which should then be converted into hours per year face-to-face with patients. Sometimes such information is available from health authorities. If it is not, assumptions have to be made. In our estimates, we have assumed that each clinician works 40 hours a week for 47 weeks during a year, of which 60% are spent face-to-face with patients. This results in a total time for care of 1128 hours per year per full-time employed clinician.
The category of healthcare personnel
Note that time from several different categories of healthcare personnel is often needed throughout the cascade (i.e., health coaches may screen for type 2 diabetes with a questionnaire, nurse assistants may take the blood glucose test, diabetes nurses may give lifestyle advice, and general practitioners may interpret the blood glucose test, prescribe medicine and follow-up). TNTs for each relevant category of healthcare personnel should be estimated separately.
Estimate TNTNNT
TNTNNT is the clinician time needed to improve the outcome for one person.
TNTNNT is estimated by adding up the clinician time needed for all “sub-interventions” to improve the outcome for one person. If time from several categories of healthcare personnel is needed, then these should be estimated separately.
As an hypothetical example, assume a recommendation to screen for, diagnose, and treat disease/risk factor X, and that:
- we need to screen 1000 people (5 minutes for nurses per individual – 5000 minutes of nurse time) to identify 200 people with risk for disease/risk factor X,
- these 200 people then needs further work-up with a diagnostic test (5 minutes for nurse assistants per individual – 1000 minutes of nurse assistants time),
- the diagnostic test needs to be interpreted and discussed with each individual with a positive screening test (10 minutes for general practitioners (GPs) per individual – 2000 minutes of GP time),
- of the 200 people tested, 40 will be diagnosed with disease/risk factor X and require treatment and follow-up (20 minutes for GPs per individual – 800 minutes of GP time AND 40 minutes for nurses per individual – 1600 minutes of nurse time),
- of the 40 people diagnosed with disease/risk factor X, 1 individual will experience improved outcome Y as a result of the earlier diagnosis.
The time needed to improve outcome Y for one person (TNTNNT) would then be:
- Nurse assistant time: 1000 minutes = 17 hours
- Nurse time: 5000 minutes + 1600 minutes = 110 hours
- GP time: 2000 minutes + 800 minutes = 47 hours
Estimate absolute TNT
The absolute TNT is the clinician time needed to provide the intervention for all eligible in a specific population.
We continue with the hypothetical example above. Assume that 50% of the general population is eligible for screening – and that we want to estimate the total time needed for each relevant category of healthcare personnel in a GP practice of 2000 individuals. If we make the same assumptions as above, this would mean that:
- 1000 people in the GP practice (50% of 2000 people) would be eligible for screening (5 minutes for nurses per individual – 5000 minutes)
- of those 1000, 200 would have a positive screening test and need further diagnostic work-up (5 minutes for nurse assistants per individual – 1000 minutes of nurse assistants time AND 10 minutes for GPs per individual – 2000 minutes of GP time)
- of those 200, 40 will be diagnosed with disease/risk factor X and require treatment and follow-up (20 minutes for GPs per individual – 800 minutes of GP time AND 40 minutes for nurses per individual – 1600 minutes of nurse time)
The absolute TNT for a GP practice of 2000 individuals would then be:
- Nurse assistant time: 1000 minutes = 17 hours
- Nurse time: 5000 minutes + 1600 minutes = 110 hours
- GP time: 2000 minutes + 800 minutes = 47 hours
Estimate relative TNT
The relative TNT is the proportion of the total clinician time available for patient care needed to implement the intervention for everyone eligible.
Relative TNT is estimated by dividing the absolute TNT by the total time available for patient care per clinician per year. We continue with the hypothetical example above. Assume that each clinician have a total time face-to-face with patients of 1128 hours per year (see above), and that there is one GP, one nurse and one nurse assistant in each GP practice of 2000 individuals.
The proportion of the total clinician time available for patient care needed to implement the intervention for everyone eligible in a GP practice of 2000 individuals (the relative TNT) would then be:
- 2 % of the available nurse assistants time (17 of 1128 hours)
- 10 % of the available nurse time (110 of 1128 hours)
- 4 % of the available GP time (47 of 1128 hours)
Note that the relative TNT can also be estimated as a proportion of the available clinician time in a whole country or region.