PIP Data for Health Economic Modelling: An Alternative Data Source
Up to now, the reimbursement of new innovative pharmaceuticals was merely based on registration data (efficacy, safety and quality parameters). Nevertheless, increasing health care costs have become a major concern for health care decision-makers resulting in the implementation of new cost containment measures that lead to additional data requirements for new pharmaceuticals, which relates to the use of innovative medication in real daily practice. The most important new data requirements are: effectiveness, cost-effectiveness and budgetary impact. Effectiveness offers a picture of the actual value of an innovation in daily practice. Effectiveness can also be measured through measuring the level of compliance/ adherence with a specific therapy. Adherence to therapies is a primary determinant of treatment success .
Cost-effectiveness data from a state of the art health economic analysis should allow reliable, reproducible and verifiable insights into the effectiveness of a drug and the possible savings that might be achieved relative to other drugs and/or treatments. Non-compliance is a specific determinant of the effectiveness of a treatment in daily practise and is therefore important in cost-effectiveness analyses. It is revealing that estimates of the total annual health care costs in the United States resulting from patient noncom-pliance vary from $100 billion to $170 billion to $300 billion . Poor adherence attenuates optimum clinical benefits and therefore reduces the overall effectiveness of a pharmaceutical and increases the total costs due to treatment failure. As a consequence, non-compliance has a negative impact on the cost-effectiveness of treatment.
Because health economic evaluation has become critical in health care decision-making, it is important that the methodologies used in such evaluations are continuously explored and any potential for improvement in methods or data sources is considered.
In most clinical trials, economic data are not collected alongside the study. Even when they are, the data may need to be projected to populations, time periods, or settings that were not observed in the clinical study. In these cases, decision-analytic models may be used to generate some of the missing information. The model, resulting from decision analysis must correspond, as much as possible, to the real-life situation of the disease and should reflect actual treatment patterns with input values (probabilities and items of healthcare utilization) deviating as little as possible from population values . The data in a health economic model may be derived from various sources and is associated with varying degrees of uncertainty. The reliability of the input data for a model depends on the choice of the data sources (e.g. selection criteria, external validity). As a consequence, there may be a potential bias in the choice of the data sources to be used in the model. Potential sources for the input variables in a model are clinical trials, literature (e.g. meta-analysis), medical records, databases, Delphi panels, and official costs lists for resource utilisation. A limitation of these data sources is that they lack the input from the patients’ experience and perspective. Especially with regards to integrating compliance character-istics in a model the patient’s voice is required as the patients themselves are the key determinants in the outcomes of treatment in daily life and the level of adherence with a therapy.
Patient Intelligence refers to skills, technologies, applications, and practices used to help an organi-zation acquire a better understanding of its position in the health care context . Patient Intelligence may also refer to the information collected by patients. Patient Intelligence applications provide historical, current and predictive views of any given present situation regarding behaviour and intentions of persons suffering from a disorder, disease or complaint. Patient Intelligence is often aimed at the support of better decision making in the health care environment. Thus, a Patient Intelligence system can be called a decision support system (DSS), but the outcomes can also be used as an input for variables for health economic models .
The objective of this paper to explore the opportunity of integrating Patient Intelligence applications as an alternative data source for a Delphi panel and databases in health economic studies.
Current Data Sources
The use of expert opinion is appropriate in situations in which there is little or no published material in a particular area, or in which the results of a thorough literature review or meta-analysis are considered unreliable , conflicting , or insufficient to cover the requirements of a study.
Delphi panels operate in rounds, in an effort to obtain convergence of expert opinion on a particular subject The main areas of weakness are as follows: in pharmacoeconomic studies, there are no explicit criteria for the selection of experts for participation in the studies . The use of a “physician-expert panel” to estimate resource use carries the risk that respondents may give inaccurate estimates or specify the resources required for ideal care, those rather than provided in real daily practice . Physicians may adjust their estimates based on other estimations because they do not want be outliers. In addition, physicians may overestimate variables related to the success of their treatment (for example: response rates, complications, adverse events), and also underestimate variables related to health care use, in order be more efficient. The most important weakness of a Delphi panel is that the data are not real data, but only estimates.
We can distinguish between two different types of databases can be distinguished: claim databases and clinical outcomes databases . In the case of claim databases, the objective is to collect, for administrative purposes, all data on health care resources used; in clinical databases, the objective is to measure clinical outcomes for medical or scientific purposes. In both cases the databases can be considered static as the information in the databases are accumulated facts of demographics, treatments and endpoints, which may not correspond with the specific data requirements of a health economic model. A database may not be a suitable source for guiding decisions in health care since so much of the data it contains is not scientifically valid. Although databases may contain a lot of detailed information on both clinical and economic outcomes, the format of this information usually does not fit the structure of the health economic model, as the majority of the existing databases have not been developed for health economic evaluations.
Patient Intelligence Database
An online panel database (Internet access panel) consists of a group of pre-screened respondents who have agreed to participate in surveys and/ or patient feedback sessions. After completing a profiling questionnaire the respondents become “panellists.” The data which is collected in this questionnaire includes demographics, information on medical history, and current health status characteristics. A patient-specific online panel is the Patient Intelligence Panel (PIP), which gives researchers access to patients worldwide . Having on-line access to thousands of people globally who are willing to participate in research on health care and specific indications means all questions can be asked and a wide range of feedback can be obtained. This feedback can be integrated in a structured defined path into health economic models. PIP data can be collected specifically for a health economic model both for economic as well as clinical and Quality of Life outcomes and therefore perfectly fulfilling the data input requirements of the model.
A recent PIP research study included 300 patients diagnosed with depression and taking different antidepressants. They were asked to quantify the severity of side effects encountered when taking antidepressants, identify each side effect occurring, and rate this side effect on severity on a scale of 1 to 10 (10 being most severe)(see Fig. 1). By quantifying such an emotional patient point of view, this feedback can be integrated into a health economic model. The key message here is that cost-effectiveness is not just a function of the incidence of side effects from a registration trial, but it is the perception of the side effect by the patients, which is driving the results of a cost-effectiveness of a new innovation in real life, and therefore the appropriate input for a cost-effectiveness model.
A PIP study can be designed to measure the impact of a particular disease or condition on clinical and patient-specific outcomes, and to document the outcomes associated with different treatments or settings of care in a quantitative matter. Patients can be followed prospectively and data are collected on disease severity and clinical outcomes, as well as resource use, functional status and quality of life as reported by the patient. PIP data reflect the current treatment patterns without influencing the treatments or interventions and consequently the PIP study is fully naturalistic without any intervention with real practice (e.g. no randomisation) and has a high external validity.
The PIP data can yield real-life data for the comparator in the health economic model, which are based on data from daily practice. The PIP may also yield more statistically solid safety data with high external validity because of the large sample size of the PIP. The large sample size of the PIP may also allow the identification of any type of covariance, which could not be proven in a clinical trial because of lack of power. As a consequence a PIP study has the power for the development of statistically solid multiple regression equations with high external validity, which can be incorporated in a health economic model.
A KOL validation process can be integrated after statistically relevant outcomes of the PIP patient’s feedback. It is prudent to include KOLs in the survey conception, because this allows online surveys to be validated from a health care profes-sional point of view, and therefore add another dimension to the outcomes.
A comparison of the PIP data set versus the Delphi panel results in favour of the PIP data set, as the Delphi panel is based on estimates only and the PIP database on real data. In addition, the Delphi panel suffers from methodological limitations, as described above, which may lead to a bias in the estimates. A comparison of the PIP data set versus static databases is more comprehensive. At first glance the most important similarity is that both data sets are based on actual data and thereby overcoming the limitations of the Delphi panel, but a lot of differences between the PIP data set and database can be noticed: The PIP dataset is not limited by power constraints as static databases, and clinical databases, which usually have a limited number of patients. Claims databases have usually a large number of “patients”, but the absolute number is fixed. To the contrary, the sample size of the PIP data set can be adjusted based on a priori sample size calculations in order to show statisticaly significant results.
The prospective design of PIP data set allows the “a priori definition” of all economic and clinical variables, which are required for the health economic model, and which will fulfill the technical requirements of the model. Contrary, database studies are usually retrospective, which means that not all data may be collected or data may not fully correspond with the requirements of the model. Finally, most databases, especially claim databases, do not contain clinical outcomes (QALYs or Quality of Life, patient satisfaction, PROs, or non-compliance), which are essential for a cost-effectiveness study. The PIP data design allows the collection of all relevant clinical outcomes, including outcomes from a patient perception, which cannot be derived from a database, whereas these outcomes are the ideal input data for a cost-effectiveness model because of its high representativeness and external validity.
A problem associated with a database can be the lack of consensus on defining criteria for a pathology, and the overlap of symptoms, as was mentioned before. The PIP data set can be used to exactly define the patients, which need to be included (or excluded) from the health economic model, including criteria for co-morbidity or risk factors. The databases usually contain limited sociodemographic data, but the PIP data set can collect all relevant socio-demographic information, which is relevant for the cost-effectiveness model.
With the growing importance of modelling studies for health economic evaluations, a new area for research has been created. In order to obtain objective and reproducible results from those studies it is important to have standardised methods of evaluation contained in accepted guidelines on methodology. This article showed that when integrating the patients’ voice in the models, a more holistic outcome will be the result corresponding with the concept of cost-effectiveness requiring a high external validity and outcomes representing real life. The patients’ voice can be considered the optimal data source for a health economic model as it has the highest representativeness of the effectiveness of a treatment in real-life. Specifically for perception sensitive factors in health economic models, like quality of life (QALYs), adherence, side-effect severity and discontinuation rational, the patients’ voice should be integrated as the patient is sole source for outcomes related to the patients’ experience with pharmaceutical therapy. Finally the use of Patient Intelligence research suits perfectly with the concept of evidence-based medicine (EBM), which means that clinical encounters should be supported by scientific conclusions based on real data as much as possible.
1 Pateriya LP, Jha A, Munjal S. Application of information technology to overcome non-compliance of drugs.https://www.indianjournals. com/glogif t2k6-1/theme_1ar ticle%201.htm. [Accessed April 4, 2009].
2 Cost of Patient Non-Compliance. Available from: http://www.alignmap.com/wp-content/ uploads/2006/03/Costs%20Of%20Patient%20 Noncompliance.pdf. [Accessed August 7, 2010].
3 Weinstein MC, Fineberg HV. Clinical decision analysis. Philadelphia: WB Saunders Co., 1980.
4 Van Dongen. Are you patient intelligent? Pharmaceutical Marketing Journal 2010;1:42-3.
5 Van Dongen Nadine. Let’s be effective, let the patients talk; does patient intelligence have an effect on improvements in quality in the healthcare environment? Dovepress, 2009.
6 Jariath N, Weinstein J. The Delphi methodology (part two): a useful administrative approach. CJONA 1994;7:7-20.
7 Jones J, Hunter D. Consensus methods for medical and health services research. BMJ 1995;311:376-80.
8 Evans C. The use of consensus methods and expert panels in pharmacoeconomic studies: practical applications and method-ological shortcomings. Pharmacoeconomics 1997;12:121-9.
9 Nuijten MJ. The selection of data sources for use in modelling studies. Pharmacoeconomics 1998;13:305-16.