Collecting resource use data is one of the most important aspects of a health economic evaluation conducted alongside a trial. Examples of health and social care resource use include hospital length of stay, medication, GP visits, operations and diagnostic tests. It may also include resources outside of healthcare settings, such as contact with the police, benefits payments and days off work.
When thinking about what resource use data to collect and how, it may help to consider the following questions.
Who will be the key audience for the results of this evaluation?
When thinking about what costs to collect it can be pertinent to think about who the main stakeholders might be: who would be responsible for funding and delivering the intervention and who could benefit from cost savings. For the majority of health economic evaluations, funding for the intervention or new technology comes from a healthcare budget; therefore, data collection usually focuses on healthcare resource use and costs. Some interventions may also have an impact on other funders’ resource use, such as for crime, social care, sexual health or substance misuse treatment.
What trial resources are available to collect resource use data?
The availability of administrative resources often determines how resource use data is collected in a trial. Medical records provide a relatively easy way to obtain healthcare resource use data if a simple data download from a healthcare system is possible.
However, if a simple download is not possible, a member of trial staff may have to trawl through patient files to collect the data. In some instances, data from patient files is not available, or its use is not efficient. In this case, questionnaires completed by the patient (or carer, or clinical staff) and patient diaries are the most common way of collecting resource use data.
What kind of questionnaire return rate is expected for this patient group?
The decision whether to use questionnaires or patient files to collect resource use data can also be driven by expected return rates of questionnaires. Although some patient groups, particularly those engaged in their treatment, can have good follow up and questionnaire return rates, other groups are notorious for poor questionnaire return rates over the full duration of a trial, for example: people who regularly change address, such as young people; the homeless; and people in contact with criminal justice agencies. Using patient files for particularly pertinent costs (for example, inpatient stays) may mean less detailed data for the analysis, but with the advantage that a larger percentage of patients can be included, which may reduce bias from loss to follow-up.
Is there anything about the patient group that might make completing questionnaires challenging?
The reliability of self-reporting of health and social care use can vary by age and mental health. In particular, some patients may lack the cognitive ability to complete questionnaires or patient diaries. In these instances, data collection may require patient files, researcher interviews or completion of questionnaires by carers.
If using questionnaires, are there any templates available?
Resources to assist with the design of resource use questionnaires include: the DIRUM database (www.dirum.org/); and published economic evaluations in the same, or a similar, clinical area. You can also seek advice from other researchers working in the same area.
Are unpaid carers affected by the intervention?
Informal care from families and friends can be an important consideration, as it can account for 80% of the cost of care across a range of disease areas. It can be challenging to collect reliable data and costing on informal care use, so guidance from a health economist may need to be sought on the best way to do this.
What non-routine data are needed to cost the intervention?
Most of the data, such as the number of treatment sessions attended, are often collected routinely as part of the trial, but more precise data, such as the duration of the intervention, may require additional questions. In some instances, a time-and-motion study may help to obtain more precise estimates of the resources required. Recording resource use for ‘treatment as usual’ also requires careful consideration, for example: relative cost-effectiveness may be affected if patients in the control group access similar services to the intervention elsewhere.
For more information visit the DIRUM website: www.dirum.org
