Background

In the field of health outcomes research, disease registries have been useful to collect clinical data, track patients over time and provide information for determining disease progression, safety and effectiveness. A registry is defined as “an organized system that uses observational study methods to collect uniform data (clinical and other) to evaluate specified outcomes for a population defined by a particular disease, condition or exposure, and that serves a predetermined scientific, clinical or policy purpose(s)” [1]. While the importance of clinically focused patient registries is well-defined, supplementing these registries with economic and quality of life (QOL) data provide an opportunity to assess cost-effectiveness of various interventions [2]. In contrast to patient registries, clinical trial outcome databases capture summary-level information from published randomized control trials (RCTs) and observational studies which can be used to perform network and model-based meta-analyses of safety and efficacy parameters [3]. The growth of the pharmacoeconomic literature over the past few decades has prompted several organizations to create databases or registries to compile key variables in published economic evaluations. Examples of current economic databases include the National Health Service (NHS) Economic Evaluations Database [4], Tufts CEA registry [5, 6], Tufts Global Health DALY (Disability-Adjusted Life Year) Registry [7], Pediatric Economic Database Evaluation Project [8], European Network of Health Economic Evaluation Databases (EURONHEED) [9] and the WHO Cost-Effectiveness and Strategic Planning (WHO-CHOICE) [10] databases. These registries are broad in their scope and represent economic evaluations conducted on a variety of healthcare, pharmaceutical and device interventions. Some of the aforementioned databases include different types of economic evaluations [4], while others are specific to cost-utility analysis [5] or cost per DALY studies [7]. However, there are only a few examples of comprehensive condition-specific health economic model registries, such as the Mount Hood Diabetes Challenge Network [11, 12]. Therefore, we sought to supplement the clinical literature on hormonal contraceptives (HC) with literature on all published complete and partial economic evaluations. Hormonal contraceptives were selected as a focus for a pharmacoeconomic registry for several reasons.

Globally, around 99 million pregnancies were unintended in 2010–2014 [13]. In the United States, nearly 45% of all pregnancies were unintended in 2011 [14]. Unintended pregnancies represent a major economic burden to the US public insurance programs, estimated to be $21 billion in 2010 [15]. In developing countries, women having an unmet need of contraceptives contribute to 84% of all unintended pregnancies, whereas the remaining occur among contraception users [16]. The reasons for contraception failure depend upon the inherent efficacy of the method, non-adherence, imperfect use, frequency of intercourse and age of the woman [17]. However, consistent use of contraceptives, still results in around 5% of unintended pregnancies in the USA [18]. Around 151 million women of reproductive age use hormonal contraceptive pills worldwide [19]. Recognizing the large public health implications, the FDA, the pharmaceutical industry and academia have focused efforts towards understanding the role of drug-drug interactions (DDIs) on HC failure [20]. Challenged by a lack of data on the clinical effect of DDIs associated with HCs, efforts are underway to study the impact of DDIs on HC failure from a multi-disciplinary perspective, including pharmacometrics, pharmacoepidemiology and pharmacoeconomics [21]. In order to evaluate the financial implications of unintended pregnancies resulting from drug-drug interactions with HCs, information was collated from all complete and partial economic evaluations on HCs. The present paper describes how the pharmacoeconomic clinical registry was developed and implemented, detailing its strengths and limitations with the aim of facilitating the creation of other condition specific pharmacoeconomic clinical registries. The registry structure and content are also presented along with best practice recommendations.

Methods

This study was registered with the International Prospective Register of Systematic Reviews (PROSPERO CRD 42019118036) in January 2019. An expert panel consisting of researchers with expertise in pharmacoeconomics and outcomes research was convened. Based on the specific clinical question of pharmacoeconomics of HCs, the general approach to develop the registry was outlined.

Environmental scan and study selection

A keyword-based search strategy was developed to identify relevant articles from PubMed and EMBASE. The search strategy was limited to consider women of reproductive age using hormonal contraceptives for the purpose of contraception, with the comparator being any other method of contraception or non-use of contraception. Studies assessing the impact of family planning interventions where the costs of contraceptives were aggregated, as well as studies that were economic evaluations of HCs for the management of other conditions such as heavy menstrual bleeding and emergency contraception were excluded. Details of the search strategy and keywords used can be found in Additional file 1. The search was limited to original research studies published in English prior to and including December 2019. Review articles, letters, editorials, commentaries and conference abstracts were excluded. Once duplicate citations were removed, all remaining titles and abstracts were screened by MS and categorized as being relevant, potentially relevant or not relevant. The potentially relevant full-text articles were then reviewed with AW and judged for relevance to be included in the registry. Additionally, hand-searching reference list of articles retrieved were screened to identify relevant economic evaluations. The relevant articles included complete economic evaluations namely, cost-effectiveness analyses (CEA), cost-utility analyses (CUA), cost-benefit analyses (CBA) and cost-minimization analysis (CMA). Partial economic evaluations included cost of illness or cost of unintended pregnancy (COUP), cost analyses (CA) and quality of life (QOL) studies. Studies classified as COUP provided country-wide estimates of the costs attributable to unintended pregnancy. The CA studies provided estimates of cost of the intervention without considering outcomes and the QOL studies measured health state utility values of unintended pregnancy. The Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines for developing literature search were followed.

Variable selection and data dictionary

A list of variables to be included in the registry was developed after careful discussions with the expert panel. It was decided to match variables to the type of study (CEA, CUA, CBA, CMA, COUP, CA, QOL). The registry therefore included general bibliographic variables and information on the funding source, country and region. The variables extracted from studies included: 1) study objective; 2) intervention and comparator; 3) target population characteristics; 4) currency and year; 5) sample size; 6) perspective; 7) source of cost data; 8) direct cost components and amount; 9) non-medical cost components and amount; 10) indirect cost components and amounts; 11) outcome measure; 12) source of outcome data; 13) time horizon; 14) discounting and adjustment of costs; 15) scenario analysis; 16) modeling method description; 17) assumptions; 18) sensitivity analysis; 19) conclusions and 20) limitations. Variables from existing registries were used as guides in the variable development process [4,5,6,7]. Based on the type of economic study, other variables such as cost-benefit ratio, average cost-effectiveness (utility) ratio, incremental cost-effectiveness (utility) ratios, and quality-adjusted life-year (QALY) were catalogued. For COUP studies, we included variables to document the total number of unintended pregnancies and costs to the payer and impact of contraceptive non-adherence on total costs. During the course of data extraction of each article, if additional variables were found to be important for the registry, they were identified and included in the registry (e.g., variables for adjustment of cost of birth to account for mistimed rather than truly unwanted births). The data dictionary was continuously revised during this process to account for the changes and can be accessed in Additional file 2.

Online data interface & data operationalization for querying

In order to facilitate entry of data into the registry by members of the expert panel and to collect the resulting data in real-time, an online data entry platform using Google Forms was created. However, to allow for querying the registry each variable was further operationalized, i.e., made either numeric (continuous or categorical) or a text entry (see data dictionary in Additional file 2).

Expert panel verification process

The verification process included two-steps. After the expert panel members received assigned articles to critique, articles were entered into the registry using the Google Form. A formalized approach to review articles that were uploaded in the registry was used to ensure quality and consistency among all the reviewers. Responses collected were then transcribed in the operationalized database in Microsoft Excel by MS. Finally, AW verified all variables within each article. Discrepancies in entry were then discussed among AW and MS until a consensus could be agreed upon.

Results

From the 1401 articles that were retrieved from the literature search, excluding duplicates, the pharmacoeconomic registry included 65 articles comprising of 22 CEA, 9 CUA, 7 CBA, 1 CMA, 14 CA, 10 COUP and 2 QOL studies. The PRISMA flow diagram is shown in Fig. 1.

Fig. 1
figure 1

PRISMA flow diagram for the inclusion of studies in the pharmacoeconomic registry of hormonal contraceptives

Full size image

Queries of the pharmacoeconomic registry

Most of the articles were published in the past 25 years except one from 1972 [22]. The complete registry can be accessed in Additional file 3. The table of cost and probabilities can be accessed in Additional file 4. Table 1 presents a summary of the CEA, CUA, CBA, CMA, CA and QOL studies. From the CEA and CUAs, ten studies were decision-tree models [27, 31, 38,39,40, 42, 48, 49, 51, 52] and eleven were Markov models [30, 32,33,34, 37, 41, 44,45,46,47, 50] and one was conducted using a systematic review and meta-analysis as part of a health technology assessment submission [28]. The remaining studies determined cost-effectiveness using real-world data or observational studies [22, 29, 35, 36, 43] or utilized a simple methodology considering costs and failure rates of each method [24,25,26] or a cost equation for clinical outcomes [23].

Table 1 Summary of economic evaluations and quality of life studies retrieved from HC pharmacoeconomic registry
Full size table

Among all the CEA, CUA, CBA, CMA, CA and QOL studies, 29 were conducted among the US population, eight in the UK, two each in Australia, Kenya, South Africa, Uganda and one each from Canada, France, Iran, Nigeria, Senegal, Sierra Leone, Sweden, Thailand, Singapore and Zambia (Table 1). Based on the country-wise distribution, in the published literature there were few economic evaluations conducted in developing countries as compared to developed countries.

While most studies compared the cost-effectiveness of different methods of contraception, some studies addressed more specific issues (Table 1). Seven studies compared the impact of immediate postpartum or post-abortion insertion of long-acting reversible contraceptives (LARC) with delayed insertion [36, 38, 40, 48, 55, 60, 67]. Three studies examined the cost-savings from increasing access to oral contraceptives by making available over-the-counter (i.e., without a prescription) or allowing pharmacists to prescribe them [50, 51, 71]. Ten studies considered the impact of enhancing access to contraceptives in Africa utilizing novel delivery methods such as community health workers and self-injection [45, 49, 52, 56, 58, 65, 66, 69, 70, 72]. Among the studies conducted in US, several focused on populations that faced disparities in access to contraceptives. The majority of studies focused on Medicaid-eligible or low-income women [24, 26, 38, 40, 51, 54, 55, 57, 60, 62, 63, 67, 71, 74]. Three studies considered adolescent contraceptive provision [26, 38, 60] and one focused on active-duty military women [39]. Two studies considered the underinsured, quantifying the impact of expansion of contraceptive coverage through Affordable Care Act (ACA) [42, 46].

Except for one CBA that determined the willingness-to-pay for contraception [56], all studies considered direct medical costs in their analyses. Direct non-medical costs were considered among six studies [22, 29, 45, 49, 52, 62] and included overhead costs and capital costs, utilization of auxiliary services and social welfare program costs. Indirect costs were considered in eleven studies and included waiting and travel times, productivity losses and personnel training times [22, 35, 45, 47, 49, 52, 64, 70, 77,78,79]. No study considered the impact of intangible costs. Fifteen studies considered the cost of side effects, out of which seven studies mentioned disaggregated costs of each side effect [23, 24, 30, 32, 34, 44, 50].

Quality of life, in the form of QALY was considered in six of the nine CUAs with the remaining considering DALYs averted (Table 1) [45, 49, 52]. Methodological improvements have occurred over time in the conduct of CUAs, especially with regards to the source of health utility values. In the absence of published estimates, the earliest CUA employed utility values for pregnancy outcomes and other health states that were elicited from a convenience sample of female members of the research team and an advisory panel and these estimates were used in a later CUA [44, 48]. Recognizing the paucity of reliable utility values for unintended pregnancy, Schwarz et al. [75] conducted a study to assess the potential impact of unintended pregnancy on non-pregnant women. These values were used in subsequently conducted CUAs [46, 47, 50, 51]. In an effort to supplement future CUAs in pregnancy and contraception, Lundsberg et al. measured utility values from a sample of pregnant women for varying contexts such as intention, timing and wantedness (Table 1) [76]. However, the literature is lacking utility values of various pregnancy contexts and outcomes from developing country settings.

Table 2 summarizes the COUP studies. Most of these studies provided country-wide estimates of the burden of unintended pregnancies in the USA [80, 81], Norway [77], Sweden [78], UK [82], Spain [83], South Africa [79], Canada [84] and Russia [85], however, one study estimated these costs for a US employer-sponsored health insurance plan [86]. Imperfect contraceptive adherence contributed to over half of these costs, thus making a case for greater LARC adoption [77, 78, 80, 83, 84].

Table 2 Cost of unintended pregnancies by country
Full size table

Purposes of pharmacoeconomic registry and best practice recommendations

The completed HC pharmacoeconomic registry can serve various purposes to guide future research and policy as mentioned in Table 3. We have used cost implications of DDIs with HCs as an example to demonstrate how the registry could be used to inform this decision problem. This project was undertaken to systematically document the key variables in all published pharmacoeconomic literature pertaining to HCs. Inclusion of all types of economic articles, including QOL allowed for the creation of an up-to-date compendium, that could aid future researchers and funding agencies in identifying existing research and prioritize future research. Costs, probabilities and utility values documented from each study, including their original references, allows for the collection of input variables for subsequent pharmacoeconomic analyses and budget impact models. Additionally, the registry catalogues the various published model structures, methodologies and assumptions which inform the development of subsequent models.

Table 3 Various purposes for developing a pharmacoeconomic clinical registry and how querying the registry can inform decision making
Full size table

Developing the registry in an iterative manner and using a trial-and-error approach allowed us to identify some of the methodological best practices for the development of a pharmacoeconomic registry which are summarized in Table 4. We recognized the importance of dissemination and transparency of the registry and therefore decided to make it open-access to allow researchers to further improve and use the registry for their own research. Therefore, early planning and engagement with project collaborators, funding agencies, and journals should be undertaken to ensure open availability of the registry. The registry should be well documented by defining each variable in columns and how to read data across rows using an accompanying data dictionary. A core team comprising of experts in fields such as pharmacy, pharmacoeconomics and outcomes research must be assembled to provide continuous feedback on the development and validity of the registry. Variables extracted should be operationalized to allow for querying to the extent desired. Finally, the database must be extensively queried to identify potential transcription errors and areas for improvement. For example, a testing exercise may reveal that a new variable may need to be added in the registry, in that case, the change has to be then made for all prior entered studies.

Table 4 Methodological best practices and associated considerations for developing a pharmacoeconomic clinical registry
Full size table

Discussion

The article presents the development of a pharmacoeconomic registry using a case study of economic evaluations and QOL studies conducted on HCs for the purpose of contraception. The best practice recommendations for curating this registry and various cases for use are also presented. The registry included 65 studies, covering a variety of contraceptive interventions across various countries and settings.

One of the major questions that guided the development process was, “How would a potential researcher utilize this resource?”. Therefore, emphasis was placed on ensuring transparency, open availability and wide dissemination throughout the process. Building credibility into the resource requires documenting every step undertaken to enable reproducibility of the final outcome [87]. In this regard, there are two questions worth considering: 1) Was the original study described in a transparent manner? and 2) Does the registry adequately reflect the information contained within the study? In order to address the first question, while standard quality assessment checklists and reporting standards such as the consolidated health economic evaluation reporting standards (CHEERS) guided our variable selection process, we did not formally assign a score for each study [88]. This is because, our registry was not restricted to full economic evaluations, but also included partial economic analysis such as CAs and cost of illness studies (COUP), elements of which may not be fully met by criteria mentioned in quality assessment checklists. Additionally, while all the studies essentially aimed to answer the question of the cost-effectiveness or savings that access to contraceptives can provide in various settings, the methodology and outcome measure used varied considerably among the studies, especially those conducted pre-2000 (Table 1). We, therefore, leave it to the researcher to decide whether estimates or conclusions derived from a particular study from the registry should be weighted differently as compared to another study for their individual research purposes.

Other well-known registries are good resources which document economic evaluations conducted on non-disease specific interventions [4, 5, 7]. A query using the term “contraceptive” on the Tufts CEA registry showed 15 studies, of which four were on contraceptives used for contraception, with the other studies focused on contraceptives used for other clinical conditions such as uterine fibroids, endometriosis and dysmenorrhea. The Tufts DALY study revealed one study and the NHS EED had 50 studies of which 20 were relevant to our clinical question. However, the Tufts CEA registry is not completely publicly accessible, and the NHS EED has discontinued updates and lacks a querying functionality. Therefore, we decided to create this specific, open-access database that can be queried. This is also in response to calls from the health economics community advocating for greater transparency and openness in health economic evaluations [11, 89,90,91,92,93]. The features of our registry seem to address the barriers to utilization of cost-effectiveness models and meet the desirable features of such a database as assessed by Teerawattananon et al. during interviews of policymakers, technical advisors and researchers in Bangladesh, India and Vietnam [94].

Our registry offers a resource for researchers working in the field of family planning and reproductive health, which is an important component towards fulfilling the United Nations sustainable development goal 3 [95]. It was therefore not aimed at synthesizing information from methodologically similar studies, as is common in systematic reviews, rather the registry aimed to provide an all-encompassing overview of economic studies conducted among various methodologies. Systematic reviews on economic evaluations of contraceptives have been conducted by Mavranezouli [96] and Lynch et al. [97], with one considering family planning interventions in general in low and middle income countries conducted by Zakiyah et al. [98] Economic evaluations addressing contraceptives especially in developing country settings is lacking [96]. This remains an important challenge, since results of economic evaluations from developed countries may not be transferable to developing country settings, given the vastly different payment models (payer versus out of pocket) and contraceptive use patterns, among other differences [19, 99]. While the number of studies conducted in the US was the most among other countries in our registry, a recent analysis revealed a lack of cost-utility analysis addressing the Healthy People 2020 priority areas, which includes reproductive and sexual health [100].

Our study had some limitations that can be addressed in future iterations of the registry. The CAs that were retrieved for the registry were only those identified from the database literature search, hand-searching was not extensively conducted to identify other cost analyses. We did not include budget impact analyses and cost-consequences analyses in our registry. Other sources from which economic studies obtain their probabilities such as clinical trials or observational studies, or reports from the Guttmacher Institute [16], UN [19] and National Survey of Family Growth [101] were not included as they were not in the scope of our current registry.

ConclusionGliklich RE, Dreyer NA, Leavy MB. Registries for evaluating patient outcomes: a user’s guide. 3rd ed. Rockville (MD): Agency for

The current paper provides a description of the development of a pharmacoeconomic clinical registry using a case study of economic evaluations conducted on hormonal contraceptives. The registry is a comprehensive database which provides key variables from all published economic literature relating to HCs and is hoped to serve as a tool that could aid in identifying global trends and avenues for future research and economic models. The development of best practices to appropriately design and execute a pharmacoeconomic clinical registry can provide helpful information to researchers and clinical practitioners for comparative effectiveness decision making. We invite the health economics community to make use of this resource for their own research purposes.