Parameter values and acceptable incremental cost-effectiveness ratios

Where possible probability data used in the model were derived from published evidence. The literature search strategy (details are summarised in appendix 1) reflects a realistic balance between comprehensiveness and manageable workload. For example, only main data bases were searched, articles were restricted to the Anglo- American and European literature and hand searches were kept to the tw o sources felt to be most relevant. Most o f the information required was particular to certain settings and thus o f questionable generalisability. The lack o f information for the model made it necessary to consider data derived from descriptive, and often small studies, or calculated rather than observed, based on several assumptions. Guidance on how to assess the validity o f such data is much less clear than for randomised controlled trials^^^. Appraisal required i) the application o f basic epidemiological principles^^^ to eliminate flawed studies; ii) the identification o f particular study weaknesses and the

biases to which they may be susceptible; and iii) the acceptance o f uncertainty o f best estimates, which was explored in a sensitivity analysis.

I made the primary assessment o f the literature. If I encountered difficulties other team members were independently consulted and disagreements were resolved by consensus. It would have been preferable to have more than one regular reviewer to increase objectivity^®^. This however was not judged to be a priority for the use o f restricted project resources.

Relatively few sources o f routine health service data were useful for the model because o f the level o f data aggregation and the incomplete recording o f ethnic group. It was beyond the scope o f this study to perform primary data analysis.

The search for grey literature and unpublished data was not as systematic as the formal literature search because o f restricted resources: instead it had to be purposeful and guided by practical considerations o f easily accessible material. This might have biased the choice o f literature towards locally available data, though a particular effort was made to try and obtain material from outside London (see appendix 1).

Information from a wide range of experts (see acknowledgements) was obtained through informal interviews, correspondence and telephone contact. Experts were particularly helpful in quickly communicating specialist knowledge, commenting on current practice and validating the parameter values obtained from other sources (face validity). It was important to consult not only a multidisciplinary panel but also to ensure that more than one representative from each profession was heard to obtain a balanced range o f views. Patients were not consulted as their views were not judged to be suitable to inform probability values, and utility parameters were excluded from the model (see above 8 .2 .1).

Clearly such probability estimates carry the risk o f bias. Policy decisions, however, cannot be avoided because o f lack o f data. This imperfect approach, because explicit and evaluated by sensitivity analysis does not invalidate the technique o f decision analysis, but rather the process o f exposing uncertainties is one reason for its use.

In this study only the cost o f the screening programme to the NHS were included. Adopting a societal perspective, although theoretically the superior approach*^^, was not judged to be justified in terms o f the additional research time required. Furthermore restriction to a health service perspective made the inevitable use o f charge and payment data as substitutes for opportunity costs more acceptable as charges are real costs from the point o f view o f the programme^^\ Cost data were assembled from various sources. As comparability o f such data is notoriously difficult, estimates were subjected to a sensitivity analysis.

ICERs were computed for all districts in the UK (appendix table A7.1) to inform policy makers about the cost-effectiveness o f moving from a selective to a universal policy. However, to facilitate interpretation, and provide a more specific and directly useful policy framework, the magnitude o f ICER values likely to be acceptable to UK policy makers had to be estimated. Criteria about what constitutes acceptable values for Hb-pathy screening are empirical and depend on competing health care priorities and the political context. To date there have been no attempts to agree nationally on the magnitude o f such values. There was hardly any comparable information from other studies o f antenatal screening available to inform the estimates for reproductive choice, and little is known about the view o f patients or the public about the worth o f reproductive choice compared to other health related benefits, as methods o f obtaining meaningful answers are experimental^^^'^^V Our estimates should therefore be viewed as tentative and have been used to illustrate rather than to ‘prescribe’ how value judgements together with cost-effectiveness data can be used to inform policy

decisions.

As neonatal screening reduces mortality, comparison with estimates o f the worth o f life years gained from economic analyses o f various life saving interventions was attempted. However, generalisability o f economic studies outside the context o f their particular setting is questionable. Neonatal screening also substantially reduces morbidity, which is not accounted for when only its effect on life years gained is considered.

To gauge societal values attached to the objectives o f Hb-pathy screening, examples o f litigation charges for deficient screening have been reviewed. However, they reflect

the individual circumstances of each case and their usefulness has thus been very limited. To take account o f the considerable uncertainty about the acceptable values attached to ICERs a wide range was included in the sensitivity analysis.

8.3

Results