Chapter 7 Decision analysis for screening tests for GDM: Case identification . 151
7.4 Decision tree structure of GDM screening tests: case identification
In all screening test strategies, when the screening test result is positive, the standard procedure is to confirm the result by further diagnostic tests, as mentioned in section 4.7. In other words, only patients whose tests return positive will undergo further testing.
In the four screening test strategies used in this study (not including ‘no screening test’) there also exists the need for further testing when test results are found to be over the accepted threshold, known as the negative dominant strategy (NDS). All guidelines that recommend screening tests for GDM employ the NDS.
In this economic evaluation, decision trees are designed to assess the five strategies of GDM. Decision trees for case identification were built based on the combinations of test approaches that include the negative dominant strategy (NDS) and the positive dominant strategy (PDS) as mentioned in section 3.4. This economic evaluation of GDM screening tests was applied to both of these approaches in order to build decision trees. However, decision tree maps in previous economic evaluations for GDM were applied only to NDS, which offers a second diagnostic test to people who initially test positive in their screening tests.
This model introduces a new approach by applying PDS to the economic evaluation of GDM. This section illustrates and discusses the differences between the decision trees in both NDS and PDS strategies.
As mentioned, the four different screening test strategies in this model (not including ‘no screening’) have different approaches that include a one step approach, a two step approach, universal screening and selective screening. However in the construction of decision trees for the selective screening approach in previous economic evaluation studies the true disease stage branch precedes the risk factor branch. With such a construction, risk factor screening cannot function as a screening tool. Therefore to improve on previous studies and present risk factors as a screening tool, the decision trees in this economic model begin with risk factor screening for the selective screening approach. Risk factor screening can function as a screening tool and is able to benefit the selective screening approach. The details of the decision tree structure are outlined below.
Decision node: The tree begins with a decision node which indicates the five strategies of the screening test and represents the decision being addressed in the model.
Chance node: In universal screening tests, there are two possible chance nodes: with disease (D+) or without disease (D-). Branches of chance nodes present possible outcomes of probability for the prevalence of gestational diabetes. For selective screening, the initial chance nodes are for risk factor status, considered as risk factor positive (R+) and negative (R-). Following the risk factor chance nodes is the true disease stage with and without risk factors. Subsequently chance nodes indicate the screening and diagnostic stages. Branches of chance nodes illustrate the possible positive (T+) or negative (T-) test results. The series of chance nodes show the following points of uncertainty. The basic statistics for the clinical diagnosis tests are indicated in each branch of possible outcomes of a screening and diagnostic test. Sensitivity is the probability of a positive test result if disease is present, and specificity the probability of negative screening if gestational diabetes is absent.
Terminal node: Each pathway ends in a terminal node, which represents the possible outcome of each pathway. The decision trees terminate at case identifications of GDM with 4 possible outcomes; True Positive (TP), False Negative (FN), False Positive (FP) and True Negative (TN) as shown in Table 7.3. Mothers with TP and FP receive treatment during gestation.
Table 7.3 The details and signs of terminal nodes of screening tests for GDM Results at terminal nodes Details
True positive GDM with treatment
False negative GDM without treatment
False positive No GDM with treatment
True negative No GDM without treatment
7.4.1 Case identification: decision tree structure for GDM screening tests in the negative dominant strategy (NDS)
Figure 7.1 illustrates the NDS decision tree for the various screening test strategies. NDS is defined as a test strategy in which positive test results are dominated by negative test results, and in which patients with negative screening results receive no further tests. In other words this means that all pregnant women are asked to undertake a follow up diagnostic test after screening positive. For example, SIGN 2001 proposed a two-step universal screening NDS approach. In this approach there is only one TP result following positive results in both test 1 (screening test) and test 2 (diagnostic test) in the disease branch. The probabilities along each of the TP pathways represent the probabilities of cases detected for each strategy.
In another example, SIGN 2010 proposed the use of selective screening, whereby the decision tree commences with the true risk factors stage. This model considers risk factor screening to be a selective screening tool. Therefore, if a patient is negative for risk factors the decision pathway subsequently terminates; those with disease being classified as FN and those without as TN.
Figure 7.1 Decision tree model for screening tests for GDM (negative dominant strategy)
7.4.2 Case identification: decision tree structure for GDM screening tests in the positive dominant strategy (PDS)
Another test combination approach which needs to be considered for screening tests is the PDS. PDS is defined as a test strategy in which negative test results are dominated by positive test results. In this strategy patients with positive screening results do not receive additional tests, and all patients that have negative screening tests undertake follow up diagnostic tests. Figure 7.2 illustrates the PDS decision tree for the five different strategies. For example adapting SIGN 2001 to a PDS approach, it can be seen that there are two TP results, one following a positive screening test in the disease branch and the second resulting from a negative screening test and positive diagnostic test, in the same disease branch. Therefore, to calculate the probability of cases detected for this strategy, the probabilities for both TP pathways were added together.
Similarly, considering selective screening in terms of PDS, in which risk factors are considered to be a screening test tool, patients that test positive for risk factors receive no further tests and the decision pathway is terminated. In the arm for positive risk factors, patients with disease are classified as TP while mothers without disease are FP.
Figure 7.2 Decision tree model for screening for GDM (positive dominant strategy)