Pharmaceutical policies: effects on rational drug use
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
To determine the effects of pharmaceutical policies on rational drug use.
Background
This is the protocol for a set of reviews. The objectives of these reviews are to determine the effects of pharmaceutical policies on rational drug use. The following reviews have been published:
-
Pharmaceutical policies: effects of financial incentives for prescribers (Sturm 2007)
-
Pharmaceutical policies: effects of restrictions on reimbursement (Green 2010)
-
Pharmaceutical policies: effects of reference pricing, other pricing, and purchasing policies (Aaserud 2006)
-
Pharmaceutical policies: effects of cap and co‐payment on rational drug use (Austvoll‐Dahlgren 2008)
Most countries face large increases in expenditures on pharmaceuticals. Expenditures on drugs account for between 7% and 22% of spending on healthcare in OECD (Organisation for Economic Co‐operation and Development) countries (OECD 2002). Increasing expenditures on drugs puts pressure on policy makers to control drug costs and ensure that this money is well spent. Pharmaceuticals make an important contribution to people's health. However, drugs are frequently not used appropriately. Improving the use of drugs can improve health outcomes and, in many circumstances, can result in large savings without adverse health consequences. On the other hand, cost‐containment strategies can have unintended effects on health and costs.
For example, when the effects of the New Hampshire Medicaid program's three‐prescription‐per‐month payment limit in the USA were examined in a time‐series analysis (Soumerai 1987), an intended reduction from 1.1 to 0.7 prescriptions per patient per month (‐30%) was found. However, a follow‐up study (Soumerai 1991) found an associated, unintended, significant increase in admissions to a nursing home compared with controls (relative risk = 1.8) and a non‐significant increased risk of hospitalization (relative risk = 1.2). Similarly, a review of policies to control pharmaceutical expenditure found that policies such as user charges may reduce efficient as well as inefficient drug use, and may do more harm than good (Bloor 1996a; Bloor 1996b).
Another review of pharmaceutical policies in developing countries found little valid evidence of either intended or unintended effects because of a lack of well‐designed evaluations (Ratanawijitrasin2001). Other reviews of pharmaceutical policies have all found that commonly used interventions, including formularies, administrative restrictions and price controls, have seldom been rigorously evaluated. These include reviews of studies of state drug reimbursement policies in the USA (Soumerai 1993), international policies to control pharmaceutical expenditure (Bloor 1996a; Bloor 1996b; Bloor 1996c), interventions to improve the rational use of drugs (Grand 1999), and reference‐based pricing schemes (Ioannides‐Demos 2002).
A wide variety of pharmaceutical policies may and have been used with the intention of improving different aspects of rational drug use, such as the costs of drug use and appropriateness regarding efficacy and safety. Previous reviews of the effects of alternative policies have been limited in scope and have not been kept up to date. All have identified some evidence that can help guide policy decisions, while noting important limitations in the available research. Our aim is to support informed decisions about pharmaceutical policies and to guide future evaluations by preparing an up‐to‐date, comprehensive summary of what is known from well‐designed research about the effects of alternative policies for improving rational (appropriate and efficient) drug use.
"Rational drug use" can be defined as use of those drugs that
‐Are most effective,
‐Have the least serious and fewest side effects, and
‐Cost the least (including drug costs and other health service costs)
Decisions about what is most "rational" may depend on judgement regarding trade‐offs, for example between marginal increases in effectiveness and higher costs; and there is always some degree of uncertainty regarding each of these dimensions. We will therefore report the effects of drug policies on each of these dimensions and not attempt to make judgements about trade‐offs between these different outcomes, which is the responsibility of policy makers.
We have organised pharmaceutical policies into 13 categories of policies. We will prepare reviews for each group of policies using the same methods, as described in this protocol, and subsequently prepare an overview of these reviews, as described at the end of the methods section in this protocol. The 13 categories of policies are listed below. Policies will be assigned to one of these pragmatic categories. The examples given for each category are not intended to be comprehensive.
1. Registration and classification policies: Policies that affect decisions about the registration or licensing of drugs. Registration (licensing) is defined as mandatory approval by a government agency before a drug can be sold, offered for sale, distributed or possessed for the purposes of sales, distribution or other use. Included in this category are policies regarding the registration of new drugs, deregistration, restrictions on registered drugs, essential drug lists and changes in classification (e.g. from prescription to over the counter).
2. Patent and profit policies: Policies that regulate patents for drugs and the profits of drug manufacturers.
3. Marketing policies: Policies that regulate marketing by drug manufacturers, including direct‐to‐consumer advertising.
4. Sales and dispensing policies: Policies that regulate who can sell drugs (for example sales by physicians, pharmacies, outside of pharmacies) and regulate ownership, location and numbers of pharmacies, policies targeted at dispensing behaviour, such as dispensing regulations, regulation of marketing by retailers, financial incentives for pharmacies and other dispensers, generic substitution by pharmacies and import/trade regulations.
5a. Prescribing policies (financial incentives): Policies that intend to affect prescribing by means of financial incentives. Included in this category are management of drug budgets by prescribers, indicative prescribing schemes, financial incentives and disincentives for prescribers such as pay‐for‐performance if they specifically aim at prescribing or drug utilization. (Sturm 2007)
5b. Prescribing policies (educational or regulatory policies targeting prescribers): Policies that regulate who can prescribe drugs and other policies targeted at prescribers. Included in this category are monitoring and enforcement of restrictions, generic prescribing, programs to implement treatment guidelines, system‐wide policies regarding monitoring drug safety, and legislated or mandatory continuing education or quality improvement specifically targeted at prescribing.
6. Policies that regulate the provision of drug insurance: Policies that determine who can provide drug insurance, who receives it, who pays for it and who makes decisions on reimbursement, e.g. decentralisation of decision making. Included in this category are private versus public insurance, non‐profit versus for‐profit, and tax‐based versus fee‐based.
7. Policies that determine which drugs are reimbursed: Policies that determine which drugs are eligible for third‐party payment. Included in this category are "positive lists" (formularies) that specify which drugs are eligible versus "negative lists" that specify which drugs are not eligible, system‐wide policies requiring Drugs and Therapeutic Committees, rules for determining which drugs are included or excluded (e.g. based on economic analyses), and rules for reassessment after a specified period.
8. Restrictions on reimbursed drugs: Policies that restrict reimbursement for drugs that are covered by drug insurance. Included in this category are pre‐authorisation for individual patients and general restrictions, for example based on medical specialty, diagnostic requirements, prior use of alternative treatments. If restrictions are not followed, the reimbursements for the patients will be reduced. (Green 2010)
9. Policies on price and purchasing: Policies that determine the price that is paid for drugs. Included in this category are price control, maximum prices, price negotiations, rebates, reference pricing, index pricing, volume‐based pricing, and procurement policies. (Aaserud 2006)
10. Co‐payment and caps: Policies that regulate out‐of‐pocket payments for drugs by patients, including increases and decreases in the amount paid directly by patients, limits on the amount paid by patients, and limits on the amount reimbursed. Included is fixed or relative co‐payments (based on income or age) ,prescription caps, deductibles and benefits. (Austvoll‐Dahlgren 2008)
11. Patient information: System‐wide requirements for drug information or patient education regarding drug use.
12. Multi‐component policies: Policies that include multiple components that cut across the above categories of policies.
Objectives
To determine the effects of pharmaceutical policies on rational drug use.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCT), non‐randomised controlled trials (CCT), interrupted time series (ITS) analyses, including repeated measures (RM) designs, and controlled before‐after studies (CBA).
We will use the EPOC definition of RCT, ITS and CBA studies. For ITS studies the definition is: "The study must have a clearly defined time of intervention AND must have at least three data points before and three data points after the intervention." We will also include designs where there was a control ITS group. Control ITS designs are conceptually similar to CBA studies, but the addition of multiple time points pre and post intervention decreases the likelihood of secular change bias.
ITS studies usually have only one data item at each point in time (eg number of hospitalised cases). However, in these reviews there are studies that have multiple ITS of many individual patients (i.e. each individual contributes a data item to every point in time). Such designs usually produce more reliable evidence than a simple ITS because between patient variability can be modelled as well as within patient variability, resulting in a study with substantially higher power than a simple ITS. The criteria for protection against bias are the same as for an ITS study, except that the appropriate methods of analysis differ (e.g. repeated measures anova, generalised estimating equations, random effects models). We have chosen to call such a study a repeated measures (RM) design. As with an ITS design with a control group ITS, the RM design could also have a control group consisting of many control patients being repeatedly observed over time.
Types of participants
Health care consumers and providers within a large jurisdiction or system of care. Jurisdictions can be regional, national or international. Studies within organisations, such as health maintenance organisations, will be included if the organisation is multisited and serves a large population.
Types of interventions
Pharmaceutical policies, defined as laws, rules, financial and administrative orders made by governments, non government organisations or private insurers that are intended to directly affect the use or cost of drugs. Categories and examples of included interventions are presented in the methods section.
Interventions at the level of a single facility, evaluations of one‐off continuing education and quality improvement interventions, and policies that are not specifically targeted at prescribing will be excluded.
Types of outcome measures
To be included a study must include an objective measure of at least one of the following outcomes:
‐Drug use
‐Healthcare utilization
‐Health outcomes
‐Costs
Search methods for identification of studies
We have developed and conducted the search strategy in an iterative fashion without language restrictions. The following databases have been searched:
-
Effective Practice and Organisation of Care Group Register, Idealist database searched 22/08/03
-
MEDLINE Ovid, 1966 to June Week 1 2003, searched 18/06/03
-
EMBASE Ovid, 1980 to 2003 Week 23, searched 18/06/03
-
ISI Web of Science, searched 08/09/05 for cited key references
-
CENTRAL, The Cochrane Central Register of Controlled Trials, Ovid, searched 15/10/03
-
CSA Worldwide Political Science Abstracts from 1975‐present, searched 21/10/03
-
EconLit WebSPIRS from 1969‐present, searched 23/10/03
-
SIGLE, System for Information on Grey Literature in Europe, WebSPIRS from 1980‐2003/06, searched 12/11/03
-
INRUD, International Network for Rational Use of Drugs, searched 21/11/03
-
PAIS International, Public Affairs Information Service, WebSPIRS from 1972‐2003/07, searched 23/03/04
-
International Political Science Abstracts, WebSPIRS from 1989‐2003/12, searched 09/01/04
-
NHS EED, National Health Services Economic Evaluation Database, CRD, searched 20/02/04
-
PubMed searched 25/02/04 for relevant journals not indexed in MEDLINE
-
NTIS, National Technical Information service from 1964‐present, searched 03/03/04
-
IPA, International Pharmaceutical Abstract, WebSPIRS from 1970‐2003/12, searched 22/04/04
The Health Management Information Consortium (HMIC) database has been tested and found not to be useful for this review.
In addition the following web sites and databases have been searched:
-
OECD (Organisation for Economic Co‐operation and Development) Publications & Documents, searched 30/08/05
-
SourceOECD, searched 30/08/05
-
World Bank Documents & Reports, searched 30/08/05
-
World Bank e‐Library, searched 04/05/05
-
JOLIS, The Library Network serving the World Bank Group and IMF, searched 22/08/05
-
Global Jolis, online catalogue for the World Bank Country Office PIC/Libraries, searched 22/08/05 and 23/08/05
-
WHO (World Health Organisation), browsed 25/08/05
-
WHOLIS, the WHO library database, searched 29/08/05
A Medline search strategy (Appendix 1) has been developed using the reviews cited in the Background section and references from these reviews. It includes terms for the following categories of interventions:
‐Regulation and classification (licensing) policies
‐Patent and profit policies
‐Marketing policies
‐Policies that regulate the provision of drug insurance
‐Policies that determine which drugs are reimbursed
‐Restrictions on reimbursed drugs
‐Prescribing policies
‐Policies on price and purchasing
‐Regulation of sales
‐Co‐payment and caps
‐Patient information
We have used a modified version of the EPOC search strategy methodology filter to limit the MEDLINE strategy to randomized trials, controlled trials, time series analyses and controlled before‐after studies. Search strategies for most of the other databases have been developed on the basis of the MEDLINE strategy. The full search strategies can be obtained by sending an email to the contact author.
We will screen the reference lists of all of the relevant reports that we retrieve. We will also search the Science Citation Index for articles citing key references. Authors of relevant papers, relevant organizations, and discussion lists will be contacted to identify additional studies, including unpublished and ongoing studies. Finally, key original database search strategies will be revised based on the yield of the above searches and updated for each of the reviews.
Data collection and analysis
Two of the review authors will review all of the search results and reference lists of relevant reports. The full text of potentially relevant reports will be retrieved and two (of the same) authors will assess the relevance of those studies, assess the quality of included studies and extract data from included studies independently. Disagreements will be resolved by discussion, when necessary including another person.
STUDY LIMITATIONS
We will use the standard criteria recommended by EPOC to assess the methodological limitations of studies (protection against bias) included in EPOC reviews (EPOC 2006).
The criteria for RCTs and CCTs are:
1. Concealment of allocation
2. Baseline measurement of outcomes
3. Follow‐up of professionals
4. Follow‐up of patients
5. Intention‐to‐treat analysis
6. Blinded assessment of primary outcomes
7. Reliable primary outcomes measures
8. Other risk of bias
The criteria for CBA studies are:
1. Baseline measurement of outcomes
2. Baseline characteristics of studies using second site as control
3. Follow‐up of professionals
4. Follow‐up of patients
5. Reliable primary outcomes measures
6. Blinded assessment of primary outcomes
7. Protection against contamination
8. Other risk of bias
Based on experience with two previous systematic reviews (Davey 2005; Grilli 2002), we (including the EPOC statistical editor) have revised the EPOC criteria for ITS and RM studies. These changes are minimal. They include defining reanalysed studies as meeting the 'analysed appropriately' criterion and allowing studies that had at least 12 monthly data points pre and post to meet the 'reason for number of data points' criterion. These criteria more accurately reflect the chance of bias in the study effect sizes. We will use the following criteria:
1. The intervention was independent of other changes (protection against secular changes). This was "MET" if there were compelling arguments that the intervention occurred independently of other changes over time and the outcome was not influenced by other confounding variables/historic events during study period.
2. Data were analysed appropriately. This was "MET" if autoregressive integrated moving average (ARIMA) models were used OR time series regression models were used to analyse the data and serial correlation was adjusted/tested for OR reanalysis performed.
3. Reasons for number of data points were given. This was "MET" if data for 12 months (or more) pre‐ and post‐intervention was used OR reason for the number and spacing of data points is given OR sample size calculation performed.
4. Shape of the intervention effect was pre‐specified. This was "MET" if point of analysis was the point of intervention OR a rational explanation for the shape of intervention effect was given by the author(s). Where appropriate, this should include an explanation if the point of analysis was NOT the point of intervention.
5. Intervention unlikely to affect data collection (protection against detection bias). This was "MET" if it is reported that intervention itself was unlikely to affect data collection (for example, sources and methods of data collection were the same before and after the intervention).
6. Blinded assessment of primary outcome(s). This was evaluated as protection against detection bias. This was "MET" if the authors stated explicitly that the primary outcome variables were assessed blindly OR the outcome variables were objective, e.g. length of hospital stay, drug levels as assessed by a standardised test.
7. Completeness of data set. This was "MET" if the data set covered 80‐100% of total number of participants or episodes of care in the study.
8. Reliable primary outcome measure(s). This was "MET"if two or more raters with at least 90% agreement or kappa greater than or equal to 0.8 OR the outcome was obtained from some automated system e.g. length of hospital stay, drug levels as assessed by a standardised test.
9. Other risk of bias.
For CITS (controlled ITS) and CRM (controlled RM) studies, the time series part of the studies will be assessed independently from the control part, using the above described criteria for ITS and RM studies. The control series part of the study will be assessed using the CBA criteria above. If the control part has a high risk of bias, it will not be included and the study will be classified as ITS or RM, otherwise the control data will be used as a control in the review.
Overall limitations for each main outcome within each study will be assessed by each of the data extractors using the following guidelines:
No serious limitations = low risk of bias = all criteria scored as 'met'
Some limitations = moderate risk of bias = one or two criteria scored as 'not clear' or 'not met'
Serious limitations = high risk of bias = more than two criteria scored as 'not clear' or 'not met'
Fatally flawed: Untrustworthy results
Some setting dependant judgement may be necessary when assessing overall limitations. Where setting dependent judgement are used, the explanations will be shown in a table summarising all of the judgements that are made for each criterion for all of the included studies in each of the reviews.
DATA EXTRACTION
The following additional information will be extracted from included studies using a standardised data extraction form:
‐Type of study (randomised trial, interrupted time series, controlled before‐after)
‐Study setting (country, key features of the healthcare system and concurrent pharmaceutical policies)
‐Characteristics of the participants (consumers, physicians, practices, hospitals, etc.)
‐Characteristics of the policies
‐Main outcome measures and study duration
‐The results for the main outcome measures
‐The sponsors of the study
If a study presents results for more than one main outcome in each of the four outcome groups (drug use, health, health care utilization and costs), we will choose what we consider to be the most important outcome in each group, either as specified by the authors or based on discussions among ourselves. In cases where additional main outcomes in the four outcome groups are thought to provide important insight, we will also include them.
A table will be prepared for each category of intervention including the following information: study identification, characteristics of the intervention, drug use, healthcare utilization, health outcomes, and resource utilization (costs). The primary analyses will be qualitative analyses based on these tables and will include an analysis of the mechanisms through which the policies were intended to affect drug use and postulated mechanisms for other affects, both intended and unintended. Cost data will be appraised using standard criteria when relevant. We will summarise what is known about the effects of alternative policies within each category, including important policy options for which no evaluations are found. Our confidence in the available estimates of effects will be graded using the approach recommended by the GRADE Working Group (GRADE 2004) with one modification. When grading the quality of evidence, we will start out grading ITS and RM studies as moderate quality, and CBA studies as low quality. This reflects our impression that the results of ITS and RM studies are more compelling (more likely to be correct) than those of CBA studies.
If there are sufficient numbers of comparisons for similar outcomes across studies, we will use graphical displays (bubble and whisker plots) to visually explore heterogeneity of the results across studies. The following potential explanatory factors will be considered: differences in the characteristics of the policies, differences in the settings and differences in study quality. These visual analyses will be supplemented with multivariate statistical analyses (meta‐regression), if appropriate, to examine how the size of observed effects are related to characteristics of the policies, differences in settings and differences in study quality.
In addition, we will identify important factors that should be taken into consideration by anyone contemplating implementing any of the policy alternatives, including: possible trade‐offs (of the expected benefits versus harms and costs), the quality of the available evidence, possible differences in baseline risk (e.g. levels of inappropriate prescribing) and other important factors that might affect the translation of the available evidence into practice in specific settings.
ITS AND RM STUDIES
The preferred analysis method for ITS and RM studies is either a regression analysis with time trends before and after the intervention, which adjusted for autocorrelation and any periodic changes, or ARIMA analysis. The results for the outcomes should be presented as changes along two dimensions: Change in level and change in slope. Change in level is the immediate effect of the policy and is measured as the difference between the fitted value for the first post intervention data point (one month after the intervention) minus the predicted outcome one month after the intervention based on the pre‐intervention slope only.
Change in slope is the change in the trend from pre to post intervention, reflecting the "long" term effect of the intervention. Since the interpretation of change in slope can be difficult, we will present the long‐term effects similar to the way we calculated and present the immediate effects. We will present the effects after half a year as the difference between the fitted value for the sixth month post intervention data point (half a year after the intervention) minus the predicted outcome six months after the intervention based on the pre‐intervention slope only. The effects after one year and two years, if available, will be measured similarly. For drug expenditures we will also calculate the savings after a half year, one year and two years as the area between the predicted expenditures curves and the actual expenditures.
Given that policy changes are often announced some months prior to official implementation, we will define a transition phase as the six months from official announcement. If the included ITS and RM studies state a different transition phase, we will use the study's definition. All results will exclude the transition phase data.
If papers with ITS design do not provide an appropriate analysis or reporting of results, but present the data points in a scannable graph or in a table, we will reanalyse the data using methods described in Ramsay 2003. The following segmented time series regression model will be used: Y(t) = B0 + B1*Preslope + B2*Postslope + B3*intervention + e(t)
where Y(t) is the outcome in month t. Pre slope is a continuous variable indicating time from the start of the study up to the last point in the pre intervention phase and coded constant thereafter. Post slope is coded 0 up to and including the first point post intervention and coded sequentially from 1 thereafter. Intervention is coded 0 for pre intervention time points and 1 for post intervention timepoints. In this model, B1 estimates the slope of the pre intervention data, B2 estimates the slope of the post intervention data and B3 estimates the change in level of outcome as the difference between the estimated first point post intervention and the extrapolated first point post intervention if the pre intervention line was continued into the post intervention phase. The difference in slope is calculated by B2‐B1. The error term e(t) was assumed to be first order autoregressive. Confidence intervals (95%) will be calculated for all effect measures.
In a repeated measures design, the data are repeated outcome measures from many individual patients. If a study does not report appropriate results we will not reanalyse the data from the summary graphs, because no estimate of within patient variability can be obtained from the summary graphs and any reanalysis would underestimate or overestimate the standard error of the effect sizes. Therefore, for RM studies we will present the results reported in the original papers only.
CBA STUDIES
For CBA studies we will report relative effects. For dichotomous outcomes we will report, if possible, the relative risk, adjusted for baseline differences in the outcome measures; i.e. the relative risk post intervention / relative risk pre intervention. For continuous variables we will report, if possible, the relative change, adjusted for baseline differences in the outcome measures; i.e. (the absolute post‐intervention difference between the intervention and control groups ‐ the absolute pre‐intervention difference between the intervention and control groups) / the post‐intervention level in the control group.
OVERVIEW OF REVIEWS
We will prepare a table summarising the main findings across reviews based on the GRADE evidence profiles from each review. Because we anticipate few direct comparisons among policies both within and across categories of policies, the analysis and discussion in the overview will rely primarily on indirect comparisons among policies. These comparisons will be made cautiously and will focus on identifying alternative policies within and across reviews that might be used to achieve rational prescribing and circumstances in which different policies are most likely to be effective and useful. In addition, we will summarise generalisable lessons for practice and research that can be drawn across the 13 reviews.
