`
`Physician Awareness of Drug Cost:
`A Systematic Review
`
`G. Michael Allan1,2*, Joel Lexchin3,4, Natasha Wiebe5
`
`1 Department of Family Medicine, University of Alberta, Edmonton, Alberta, Canada, 2 Institute of Health Economics, Edmonton, Alberta, Canada, 3 Department of Family
`and Community Medicine, University of Toronto, Toronto, Ontario, Canada, 4 School of Health Policy and Management, York University, Toronto, Ontario, Canada,
`5 Department of Medicine, University of Alberta, Edmonton, Alberta, Canada
`
`Funding: This systematic review was
`funded from a $5,000 grant from the
`Institute of Health Economics. The
`funders had no role in study design,
`data collection and analysis, decision
`to publish, or preparation of the
`manuscript.
`
`Competing Interests: JL has been
`retained by lawyers acting for a
`Canadian generic company
`attempting to introduce a generic
`version of a product. Otherwise, the
`authors do not have any financial
`interests, relationships, or affiliations
`relevant to the subject matter of this
`manuscript.
`
`Academic Editor: Suzanne Hill,
`World Health Organization,
`Switzerland
`
`Citation: Allan GM, Lexchin J, Wiebe
`N (2007) Physician awareness of
`drug cost: A systematic review. PLoS
`Med 4(9): e283. doi:10.1371/journal.
`pmed.0040283
`
`Received: December 5, 2006
`Accepted: August 14, 2007
`Published: September 25, 2007
`
`Copyright: Ó 2007 Allan et al. This is
`an open-access article distributed
`under the terms of the Creative
`Commons Attribution License, which
`permits unrestricted use,
`distribution, and reproduction in any
`medium, provided the original
`author and source are credited.
`
`Abbreviations: GP, general
`practitioner
`
`A B S T R A C T
`
`Background
`
`Pharmaceutical costs are the fastest-growing health-care expense in most developed
`countries. Higher drug costs have been shown to negatively impact patient outcomes. Studies
`suggest that doctors have a poor understanding of pharmaceutical costs, but the data are
`variable and there is no consistent pattern in awareness. We designed this systematic review to
`investigate doctors’ knowledge of the relative and absolute costs of medications and to
`determine the factors that influence awareness.
`
`Methods and Findings
`
`Our search strategy included The Cochrane Library, EconoLit, EMBASE, and MEDLINE as well
`as reference lists and contact with authors who had published two or more articles on the topic
`or who had published within 10 y of the commencement of our review. Studies were included
`if: either doctors, trainees (interns or residents), or medical students were surveyed; there were
`more than ten survey respondents; cost of pharmaceuticals was estimated; results were
`expressed quantitatively; there was a clear description of how authors defined ‘‘accurate
`estimates’’; and there was a description of how the true cost was determined. Two authors
`reviewed each article for eligibility and extracted data independently. Cost accuracy outcomes
`were summarized, but data were not combined in meta-analysis because of extensive
`heterogeneity. Qualitative data related to physicians and drug costs were also extracted. The
`final analysis included 24 articles. Cost accuracy was low; 31% of estimates were within 20% or
`25% of the true cost, and fewer than 50% were accurate by any definition of cost accuracy.
`Methodological weaknesses were common, and studies of low methodological quality showed
`better cost awareness. The most important factor influencing the pattern and accuracy of
`estimation was the true cost of therapy. High-cost drugs were estimated more accurately than
`inexpensive ones (74% versus 31%, Chi-square p , 0.001). Doctors consistently overestimated
`the cost of inexpensive products and underestimated the cost of expensive ones (binomial test,
`89/101, p , 0.001). When asked, doctors indicated that they want cost information and feel it
`would improve their prescribing but that it is not accessible.
`
`* To whom correspondence should
`be addressed. E-mail: michael.allan@
`ualberta.ca
`
`Conclusions
`
`Doctors’ ignorance of costs, combined with their tendency to underestimate the price of
`expensive drugs and overestimate the price of inexpensive ones, demonstrate a lack of
`appreciation of the large difference in cost between inexpensive and expensive drugs. This
`discrepancy in turn could have profound implications for overall drug expenditures. Much
`more focus is required in the education of physicians about costs and the access to cost
`information. Future research should focus on the accessibility and reliability of medical cost
`information and whether the provision of this information is used by doctors and makes a
`difference to physician prescribing. Additionally,
`future work should strive for higher
`methodological standards to avoid the biases we found in the current literature, including
`attention to the method of assessing accuracy that allows larger absolute estimation ranges for
`expensive drugs.
`
`The Editors’ Summary of this article follows the references.
`
`Exhibit 1100
`IPR2017-00807
`ARGENTUM
`
`PLoS Medicine | www.plosmedicine.org
`
`1486
`
`September 2007 | Volume 4 | Issue 9 | e283
`
`000001
`
`
`
`Introduction
`
`Financial constraints are a reality in almost all aspects of
`medicine. Pharmaceutical expenditure ranges from 8.5% to
`29.6% of health-care spending within Organisation for
`Economic Co-operation and Development countries and is
`increasing faster than other areas of health-care spending in
`almost all these countries [1]. For example,
`in Canada
`pharmaceutical spending increased from 9.5% of total health
`care costs in 1985 to over 16% in 2004, and its annual growth
`rate has exceeded that of all health expenditures in every year
`in that period [2]. Most countries struggle to reduce
`pharmaceutical spending [3,4] as escalating costs and limited
`resources threaten other budgetary priorities. While the
`policy makers in publicly funded systems and insurance
`agencies struggle to cope, strategies to shift costs, in part or
`whole, to the consumer are unavoidable. Unfortunately, these
`initiatives often shift costs to other areas of health care, result
`in worse patient outcomes, and are not cost-effective overall
`[5–10]. Initiatives that have targeted doctors to reduce
`pharmaceutical spending include guidelines, fund-holding,
`and others [11–13]. One way of helping to control drug costs
`would be for physicians to autonomously choose the least-
`costly medication when there are no substantial differences
`in safety and effectiveness between the least and most
`expensive. Price variations within drug classes [14,15] or
`between drug classes are common, and if physicians were to
`choose therapeutically equivalent but less-expensive drugs,
`large scale savings could be realized.
`In addition to budget concerns, doctors must consider drug
`costs to their patients. Increasing pharmaceutical costs
`negatively impacts patients in two ways. First, high direct
`expenses for those of limited resources may mean a choice
`between medicines and necessities such as food or clothing
`[16,17]. Alternatively, patients who do not take their medicine
`as directed or go without the potentially beneficial therapies
`entirely [16,17] often suffer negative health consequences [5–
`8,10]. Unfortunately, patients may be too embarrassed to tell
`their physicians when they cannot afford their medicines
`[18,19].
`
`Background: Drug Costs and Patient Expenses
`In the global market, the cost of drugs is highly variable and
`therefore obtaining accurate and relevant costs is often very
`complex. The situation in the United States (US) is likely the
`most complex, and multiple authors have attempted to distil
`the confusing and convoluted story of drug costs [20,21]. The
`often-quoted average wholesale price (the distributors’ price
`to pharmacies) can vary due to multiple factors such as
`demand, recent negotiations with pharmaceutical manufac-
`turers, and changes in coverage from large insurers. At the
`pharmacy, mark-up of the average wholesale price can be
`dramatic depending on the type of product (acute medicines
`have a larger mark-up) or the method of payment (cash
`customers often pay more). Alternatively, some high-use
`drugs may be marked down to draw customers in to the store.
`The amount the patient pays is based on his or her insurance,
`through private organizations such as managed care organ-
`izations and health maintenance organizations, government
`support (for example, Medicaid), or a combination of the
`above (families may have two or more providers). Insurers use
`a wide variety of strategies to control costs including
`
`Physician Awareness of Drug Cost
`
`copayment, tiered copay (the amount of shared payment
`varies with different drugs), and reference pricing for drugs,
`to name a few. Each insurer (private or government) covers
`different drugs and have different copay systems (flat fees,
`percent copay, or a mixture of the two).
`Elsewhere, the system is slightly less complex. In Canada,
`drug prices are believed to more closely parallel the wholesale
`price but are still subject to some of the variations and price
`competition found in the US. There is provincially based
`drug coverage for seniors and low-income individuals, but
`many provinces have some form of copay or reference -based
`pricing. The rest of the population pays for drugs out of
`pocket or has some form of insurance (which frequently has a
`copay component). In Europe, there are dramatic (.400%)
`differences in drug costs between neighboring countries
`[22,23]. Many countries have some elements of price
`competition (e.g., United Kingdom [UK] and Germany), but
`in some countries companies negotiated costs with regional
`(e.g., Spain) or federal (e.g., Italy) governments [24]. Many
`countries, including UK, France, Germany, Italy, The Nether-
`lands, Spain, Finland, Denmark, and Austria, have some form
`of copay [3,23]. The copay systems are often added to a mix of
`complementary insurance (e.g., France), reference-based
`pricing (e.g., The Netherlands), drug budgets for physicians
`(e.g., UK), price control (e.g., Italy) and combinations of them
`all with regional variation in some countries [3]. The systems
`are at times irrational. For example, fixed copayments in
`some countries can result in patients paying more for a
`prescription than the actual
`list price of the drug [23].
`Although many North Americans believe that drugs are free
`to patients in Europe, copayments have been shown to be a
`barrier for patients even in the UK [19]. Many other countries
`(e.g., Australia and Japan) also use a variety of copayment or
`cost-sharing schemes for prescriptions [24,25].
`Therefore, with global budgets a concern and the welfare of
`patients at risk, physicians need to consider drug cost when
`prescribing. If physicians are going to take costs into
`consideration they need to be cognizant of both the absolute
`drug cost and the relative differences between prices of
`products. However, in most places cost information is not
`easily available for doctors and even where it is, the large
`difference between inexpensive and expensive equivalents is
`not emphasized. To determine if it is necessary to enhance
`both physicians’ education about prices and the availability of
`that information, we undertook a systematic review to
`determine physicians’
`level of awareness of the cost of
`prescription drug products.
`
`Methods
`
`Templates for systematic review of survey studies are not
`well established, but QUOROM [26] (normally reserved for
`systematic reviews of randomized controlled trials) is a good
`guide for most systematic reviews and was used here wherever
`possible (Table S1).
`
`Search
`We searched the Cochrane Library (from 1966), EconLit
`(from 1969), EMBASE (from 1974), and MEDLINE (from 1950)
`up to 31 May 2005 using the search terms ‘‘physician’’,
`‘‘doctor’’,
`‘‘medical student’’,
`‘‘house staff’’,
`‘‘intern’’ or
`‘‘resident’’;
`‘‘medicine’’,
`‘‘medications’’,
`‘‘drug’’,
`‘‘therapeu-
`
`PLoS Medicine | www.plosmedicine.org
`
`1487
`
`September 2007 | Volume 4 | Issue 9 | e283
`
`000002
`
`
`
`tic’’, ‘‘test’’, ‘‘investigation’’ or ‘‘diagnostic test’’; ‘‘cost’’ or
`‘‘price’’; and ‘‘knowledge’’, ‘‘awareness’’ or ‘‘understanding’’.
`The original search attempted to capture all cost awareness
`studies including those in which doctors estimated the costs
`of investigations (knowledge of the cost of investigations will
`be presented in another publication). The titles and abstracts,
`where available, were independently screened by GMA and JL
`and if either investigator thought that the article would be
`potentially eligible, a complete copy was obtained. To
`identify additional studies, the reference list of any poten-
`tially eligible article was searched and authors with two or
`more publications in the area or who had published in the 10
`y preceding the start of our review were contacted.
`
`Eligibility
`Articles were included if: either doctors, trainees (interns
`or residents), or medical students were surveyed; there were
`more than ten survey respondents; costs of pharmaceuticals
`were estimated; results were expressed quantitatively; there
`was a clear description of how authors defined ‘‘accurate
`estimates’’; and there was a clear description of how the true
`cost was determined. Because costs are variable and complex,
`we felt it was only reasonable for doctors to have knowledge
`of the total costs of the prescription, whether that cost was
`borne partially or completely by the patient and/or the
`insurer (private or government),
`in their local practice
`environment. Therefore,
`‘‘true cost’’ was operationally
`defined as the actual cost the study authors verified from
`one or more locally relevant reliable sources for each drug in
`their study. This source would vary by location, but in the UK
`drug prices are more uniform so the British National
`Formulary would be a reasonable source, while in US quotes
`from local pharmacies (averaged from a broad sample) is
`most appropriate [21]. The definition of ‘‘accurate estimates’’
`was taken from the authors and typically fell within a defined
`‘‘accuracy range’’ (e.g., 625%) around the true cost. Articles
`were excluded if they were not published in English or if
`participants were asked to estimate costs within ranges or
`cost increments only (for example ‘‘please estimate which $20
`cost category/range is most appropriate for drug A’’). GMA
`and JL independently assessed each potential article for
`eligibility. Differences in decisions about inclusion and
`exclusion were resolved through consensus.
`
`Data Extraction
`From each eligible article GMA and JL independently
`extracted the following information: publication year; study
`country; response rate and number of participants, sample
`selection method (random, entire specified population,
`convenience); mode of survey administration (postal, hospital
`mail, meeting, face-to-face); participant level of training
`(medical student, intern, resident, qualified doctor); specialty;
`number of different drugs estimated; method of ascertain-
`ment of true cost (from formulary, acquisition cost, amount
`billed to patient, survey of retail pharmacies, wholesale price);
`method of assessing accuracy of cost estimate (within a
`specified percent or dollar range of true cost); and estimation
`accuracy (percent of respondents with accurate estimate,
`percent above and below true cost, median percent error of
`estimations). Primary quality measures were method of
`sample selection, mode of survey administration, and
`response rate, as well as errors or unclear description of
`
`Physician Awareness of Drug Cost
`
`calculations (e.g., incorrect method of calculating estimation
`error). This selection was based on our understanding of the
`places where the greatest biases can occur in survey studies.
`Where data were not reported in a way that allowed
`extraction in one of our categories, we attempted to calculate
`the information from available data (e.g., number of
`respondents calculated from the number of surveys distrib-
`uted multiplied by the response rate). Comparisons within
`studies, such as differences between medical student and
`resident accuracy, were extracted when available. Qualitative
`information, such as surveys of physicians’ opinions, was also
`extracted when available. Authors were contacted for further
`data where necessary. After each investigator independently
`extracted the above information, the results were compared
`and differences resolved by consensus.
`
`Data Analysis
`The studies were too diverse to pool meta-analytically (e.g.,
`different therapies, different cost estimation procedures,
`different groups of physicians), but we did examine accu-
`racies by grouping studies with nonparametric summaries.
`Mean accuracy (expressed as the percent of physicians who
`correctly estimated drug costs) for each study was calculated
`by averaging the accuracy from each participant group or
`drug estimated with weighting for the number of estimation
`attempts. For example, if accuracy was 30% for drug A (n ¼
`100) and 50% for drug B (n ¼ 80), the average accuracy would
`be 39% ([(0.30 3 100) þ (0.5 3 80)] 4 180). We calculated
`nonparametric summaries (median and ranges [minimum –
`maximum]) for the following outcomes: average cost accuracy
`(within defined percent margins of error), average percent of
`estimates over and under true cost, average percent of
`estimates over and under the margins of error (as defined by
`the original authors) around the true costs, and average
`percent error (jestimate – true costj/true cost).
`Percent error is the statistic used to demonstrate the
`degree of estimation error. To be reliable, each estimate
`error (the amount above or below the true cost) must be
`converted to an absolute value. If it is not, high estimates will
`be positive numbers and low estimates will be negative
`numbers, and when summed will partially cancel each other
`giving a lower value and a false impression of accuracy. For
`example, if the true cost of a drug is $100 a month and two
`doctors estimate $50 and $150 respectively, the correct
`percent error would be 50%. However, if absolute values
`were not used, the percent error of the high estimation error
`would be 50% and the low would be 50%. This would make
`the combined percent error 0%,
`indicating no error in
`estimation and yielding a false representation of perfect
`accuracy.
`Additionally, a priori-defined subgroups, such as year of
`publication (divided by median year of publication of
`studies), location of study, training level of participants, and
`specialty were examined to determine if these variables
`influenced the accuracy of the cost estimation. We also
`examined the influence of study quality on estimation
`accuracy by separating studies with a similar accuracy range
`into those of high, mid, and low quality. For this analysis, we
`used weaknesses of response rate (50% or unclear),
`sampling method (convenience or unclear), and survey
`distribution (unclear) as markers of quality. While there is
`no defined adequate response rate, low response rates can
`
`PLoS Medicine | www.plosmedicine.org
`
`1488
`
`September 2007 | Volume 4 | Issue 9 | e283
`
`000003
`
`
`
`Physician Awareness of Drug Cost
`
`a total of 2,954 studies, 24 were included in the systematic
`review (Table S2 provides the list of articles excluded after
`full review and the reason for exclusion). Disagreement
`between reviewers was rare (2% in eligibility and 6% in data
`extraction).
`
`Study Characteristics
`The main characteristics and methodological aspects of
`each study are provided in Table 1. Studies were conducted
`from 1978 to 2004 in six countries, with the US (nine studies),
`UK (eight), and Canada (four) predominating. Eleven studies
`included licensed physicians only, two involved house staff
`only, and 11 included a mixture of participants. Eight studies
`involved general practitioners (GPs) alone, seven specific
`specialists groups, six a mix, and three were unclear as to the
`specialty of the doctors.
`Hospital-based studies in Canada, Denmark, Italy, India,
`and the UK defined true costs with formulary lists from the
`hospital [30], government formularies [31,32], wholesale costs
`paid by the hospital [33,34], or total cost to the hospital [35–
`37]. The only US study of in-hospital physicians used hospital
`charges as true costs [38]. Most US outpatient studies [39–45]
`used the averaged prices from surveys of local pharmacies to
`determine true costs, but one [46] used the average wholesale
`price. Most of the remaining outpatient studies from Canada
`and the UK, where price varies little from the single-payer
`agreed reimbursement, determined true cost from single
`sources such as the wholesale costs [47,48] or the British
`National Formulary [49–52]. One Canadian study used a
`pharmacy survey for outpatient prescribing and cost to the
`hospital for inpatient prescribing [53].
`The majority of studies (79%) selected drugs based on the
`common drugs for that specialty. The others picked agents
`based on specific representative mixes of generic/branded
`medications [41,50–52] or based on cost impact by frequency
`and expense [35]. Only two studies specifically identified the
`percent of generics (30% [46] and 39% [52]), but it appears
`the proportion in studies overall was approximately 50%.
`
`Study Quality
`Quality and methodological reporting were poor in most of
`the studies. The method of survey distribution was unclear in
`seven studies, and sampling was convenience or unclear in 12
`studies. The response rates were 50% or unknown in seven
`studies. Only seven (29%) of 24 studies [30,39,45,47,50–52] did
`not have any of these three weaknesses. In addition, of 12
`studies attempting to quantify the degree of estimation error
`(for example percent error), nine used average estimations
`without regard for signage (that is, averaging overestimates
`with underestimates) or inadequately described the calcu-
`lation. In total, 19 (79%) of the 24 studies had one or more of
`these four weaknesses, and only five trials [30,39,47,50,51]
`were without substantial weaknesses. There was also a large
`variation in study design; five methods were used to
`determine true costs, and reasonable accuracy was defined
`nine different ways.
`
`Estimation Accuracy
`Table 2 summarizes cost estimation accuracy outcomes. In
`general, average estimation accuracy was less than 50%,
`decreasing with tighter definitions of accuracy. Overestima-
`tion tended to be more frequent than underestimation, and
`
`Figure 1. Study Identification and Selection Process
`doi:10.1371/journal.pmed.0040283.g001
`
`bias surveys [27,28] and we felt 50% was generous. Non-
`probability sampling, such as convenience sampling, can bias
`studies because the sample is not representative of the
`population. Different modes of questionnaire administration
`have different inherent biases, and while there is no clearly
`superior method [29], we felt the information was important
`in reviewing surveys. High-quality studies had none of these
`weaknesses, mid-quality studies had a single weakness, and
`low-quality had two or more weaknesses. In post hoc analyses,
`where studies reported potential within-study factors influ-
`encing the accuracy of cost estimation (e.g., cost of drug), we
`used the binomial test to combine ‘‘votes’’ across studies.
`We also performed two sensitivity analyses. To minimize
`the heterogeneity inherent to comparing studies with multi-
`ple different drugs, we compared the average cost accuracies
`for specific drugs common among three or more studies.
`When data cannot be combined and nonparametric statistics
`such as medians and ranges must be used, there is a concern
`that larger studies are weighted equally with smaller ones. To
`determine the potential influence of ‘‘weighting,’’ we per-
`formed sensitivity analyses where the median nonparametric
`statistic was selected based on the number of therapies in
`each study, the number of physicians in each study, or the
`total number of estimates in each study.
`Ethics approval was not required as the research involved
`publicly available material.
`
`Results
`
`Literature Search and Study Selection
`A study flow diagram is provided in Figure 1. Eleven
`authors were contact to identify possible studies and six
`responded, to yield two previously unidentified studies. From
`
`PLoS Medicine | www.plosmedicine.org
`
`1489
`
`September 2007 | Volume 4 | Issue 9 | e283
`
`000004
`
`
`
`Physician Awareness of Drug Cost
`
`Table 1. Study Characteristics
`
`Author
`
`Publication
`Year
`
`Country Participants’
`Training
`Level
`
`Specialty Number of
`Therapies
`
`Number
`Responding
`(Response
`Rate as %)
`
`Sample
`Selection
`
`Mode of
`Survey
`Administration
`
`True Cost
`Determined
`from
`
`Allan 1 [47]
`Allan 2 [48]
`Bailey [31]
`Beringer [39]
`Conti [30]
`Dresnick [38]
`Fairbass [33]
`Fink [40]
`Glickman [41]
`Innes [35]
`Mills [34]
`Mishra [32]
`Oppenheim [42]
`Perrine [43]
`Ringenberg [44]
`Rowe [49]
`Ryan 1 [50]
`Ryan 2 [51]
`Safavi [45]
`Schlunzen [36]
`Silcock [52]
`Walzak [46]
`Weber [53]
`Wynick [37]
`
`2002
`2004
`1993
`1984
`1998
`1979
`1988
`1978
`1994
`2000
`1993
`1998
`1981
`1982
`1988
`1986
`1990
`1992
`1992
`1999
`1997
`1994
`1986
`1985
`
`Canada
`Canada
`UK
`US
`Italy
`US
`UK
`US
`US
`Canada
`UK
`India
`US
`US
`US
`UK
`UK
`UK
`US
`Denmark
`UK
`US
`Canada
`UK
`
`HS
`Lic
`Lic, HS
`Lic, HS
`Lic, HS
`Lic, HS, MS
`Lic
`Lic
`Lic
`Lic
`Lic
`Lic, HS
`Lic, HS
`Lic, HS
`HS
`Lic, HS
`Lic
`Lic
`Lic, HS
`Lic
`Lic
`Lic, HS
`Lic, HS, MS
`Lic
`
`GP
`GP
`Anesthesia
`Neurology
`Mix
`Mix
`Anaesthesia
`Mix
`Mix/GP
`Emergency
`Anaesthesia
`ns
`GP
`GP
`GP
`Mix/GP
`GP
`GP
`Mix/GP
`Anaesthesia
`GP
`Mix/GP
`Pediatrics
`ns
`
`20
`20
`15
`39
`15
`3
`17
`22
`14
`20
`17
`6
`10
`15
`18
`15
`21
`21
`5
`19
`31
`20
`6
`2
`
`82 (85)
`283 (47.2)
`40 (100)
`24 (75)
`60 (100)
`427 (ns)
`20 (100)
`114 (31.2)
`132 (80a)
`75 (100)
`20 (100)
`42 (ns)
`152 (68.2)
`58 (48b)
`65 (72)
`50 (100)
`281 (76.6)
`244 (61)
`188 (71)
`47 (92.2)
`627 (62.7)
`137 (50)
`71 (90)
`82 (48)
`
`Meeting/mail
`Entire
`Random
`Face to face
`Convenience
`Entire
`Face to face
`Random
`Unclear
`Unclear
`Unclear
`Unclear
`Random
`Convenience Meeting
`Convenience
`Face to face
`Convenience
`Face to face
`Unclear
`Unclear
`Mix
`Unclear
`Unclear
`Unclear
`Convenience Meeting
`Unclear
`Unclear
`Mix
`Random
`Entire
`Hospital mail
`Unclear
`Hospital mail
`Random
`Random
`Unclear
`Mail/face to face
`Entire
`Unclear
`
`Wholesale
`Wholesale
`Formulary
`Survey
`Acquisition
`Billing
`Acquisition/wholesale
`Survey
`Survey
`Survey
`Wholesale
`Wholesale
`Survey
`Survey
`Survey
`Formulary
`Acquisition
`Acquisition
`Survey
`Survey
`Formulary
`Wholesale
`Acquisition/survey
`Acquisition
`
`HS, house staff; Lic, licensed physicians; Mix, mixture of specialized physicians; MS, medical students; ns, not specified.
`aExact response rate and number surveyed not given (‘‘.80%’’ stated)
`bExact number surveyed unclear (‘‘approximately 120’’) so response rate approximate.
`doi:10.1371/journal.pmed.0040283.t001
`
`percent error was very large (well over 200%). In the
`sensitivity analyses of the most commonly used margin of
`error (620% or 625%), the number of therapies, the number
`of physicians, or the number of estimates from each study did
`not change the median cost accuracy by more 2%. This
`finding demonstrates that weighting would not have influ-
`enced the final result and that the median accuracy is very
`similar to a weighted mean if the data could have been
`combined.
`Table 3 presents nonparametric summaries for subgroups
`
`using the most commonly used margin of error (620% or
`625%). Results were similar using the 650% or 50%–200%
`margin of error (unpublished data). While dramatic differ-
`ences were not apparent, the quality of the studies may play a
`role in reporting the accuracy of cost estimation. By
`comparison,
`the highest-quality studies had a median
`accuracy of 29% (range 16%–33%) while the lowest quality
`studies had a median accuracy of 38% (range 27%–45%).
`There is large estimation variability within studies (percent
`error), and between studies accuracy varied widely (for
`
`Table 2. Cost Accuracy Summaries
`
`Definition
`
`Number of Studies
`
`Median (Range), %
`
`References
`
`Within 610% or 615%
`Within 620% or 625%
`Within 650% or 50%–200%
`Overestimationa
`Underestimationa
`Above 120/125%b
`Below 75/80%c
`Percent errord
`
`4
`12
`9
`4
`4
`6
`6
`3
`
`15 (9–26)
`31 (16–51)
`44 (32–60)
`68 (60–86)
`32 (14–40)
`39 (16–49)
`23 (18–48)
`243 (218–301)
`
`[32,38,39,43]
`[30,32,35,38,40,43,47,48,50–53]
`[31,33,34,36,37,44,47–49]
`[33,34,47,48]
`[33,34,47,48]
`[32,40,50–53]
`[32,40,50–53]
`[35,47,48]
`
`aPercent of estimates over or under the true cost.
`bPercent of estimates above the 20% or 25% acceptable margin of error around the true cost.
`cPercent of estimates below the 20% or 25% acceptable margin of error around the true cost.
`dCalculated as jestimate – true costj/true cost.
`doi:10.1371/journal.pmed.0040283.t002
`
`PLoS Medicine | www.plosmedicine.org
`
`1490
`
`September 2007 | Volume 4 | Issue 9 | e283
`
`000005
`
`
`
`Physician Awareness of Drug Cost
`
`Table 3. Between-Study Comparisons in Cost Accuracy (for Studies Using 20% or 25% Margins of Error for Accuracy)
`
`Variable
`
`Country
`
`Training
`
`Year study published
`
`Physician
`
`Qualitya
`
`aSee text for definition of quality.
`doi:10.1371/journal.pmed.0040283.t003
`
`Category
`
`Number of Studies
`
`Median (Range), %
`
`References
`
`UK
`Canada
`US
`Licensed
`Mixed
`1990
`1991
`Generalist
`Specialist
`High
`Mid
`Low
`
`3
`4
`3
`6
`5
`5
`7
`6
`5
`5
`4
`3
`
`29 (26–33)
`23 (16–51)
`33 (27–38)
`31 (23–33)
`38 (16–51)
`33 (27–51)
`23 (16–45)
`28 (16–33)
`33 (16–51)
`29 (16–33)
`28 (23–51)
`38 (27–45)
`
`[50–52]
`[35,47,48,53]
`[38,40,43]
`[35,40,48,50–52]
`[30,32,38,43,53]
`[33,40,43,50,53]
`[30,32,35,47,48,51,52]
`[43,47,48,50–52]
`[30,35,38,40,53]
`[30,47,50–52]
`[35,40,48,53]
`[32,38,43]
`
`studies using a 20% or 25% margin, average accuracy ranged
`from 16% to 51%). When some heterogeneity is reduced by
`focusing on estimation accuracy for the same drugs, the
`variability between studies persists (Figure 2).
`
`program may have limited the utility of cost information, and
`the higher awareness may be due to selection bias because
`physicians chose to be fund holders and likely had a prior
`interest in costs [52].
`
`Factors Influencing Estimation Pattern and Accuracy
`Table 4 summarizes the results of subgroup comparisons
`within included studies. Very few variables impacted the
`estimation of cost. Two studies [39,42] of three found cost
`estimations of nonacademic physicians more accurate than
`academic physicians. The most consistent factor influencing
`the pattern of estimation was the true cost of the therapy. All
`11 studies that examined the influence of drug price on
`estimation patterns found that expensive drugs are consis-
`tently underestimated and inexpensive drugs are consistently
`overestimated. This finding was reinforced in the five studies
`[35,47,48,50,51] that provided enough data (true cost and the
`percentage of high/low estimations for each drug) to examine
`the effect of drug cost on the estimation pattern for
`individual drugs. For 89 of the 101 drugs in these studies,
`doctors consistently overestimate the cost of inexpensive
`drugs and underestimate the cost of expensive drugs
`(binomial test, 89/101, p , 0.0001).
`Six studies [35,43,47,48,50,51] provided enough data (true
`cost and estimation accuracy for each drug) to examine the
`implication of the true drug cost on the estimation accuracy
`for individual drugs. Expensive drugs are generally estimated
`more accurately: compared to the mean estimation accura-
`cies for the studies, only 23 (31%) of 74 inexpensive drugs had
`a higher estimation accuracy while 32 (74%) of 43 of
`expensive drugs had a higher estimation accuracy (Chi-
`square, p , 0.001).
`The influence of physician membership in a health
`maintenance organization or managed care organization is
`uncertain because US outpatient studies involved large
`communities and did not specifically examine or identify
`physicians within these organizations. One UK community
`study [52] compared estimation accuracy of fund holders to
`non-fund holders and users with desktop computer cost
`information to those without and found no difference except
`for a slightly (2%, p ¼ 0.01) improved awareness among fund
`holders for inexpensive and very inexpensive drugs. However,
`the authors acknowledge that weaknesses in the computer
`
`Qualitative Information
`Many of the studies collected additional qualitative cost
`information. When asked, doctors rated their cost awareness
`as low in five of five studies [39,40,45,46,48], rated their
`previous cost education as low or absent in five of five studies
`[30,39,42,44,48], and stated that costs are important in eight
`of eight studies [41,42,45–48,50,52]. In four of four studies
`doctors reported that cost information is not easily accessible
`[41,46,48,50] but that they wanted more cost information
`[45,46] and that it would change their prescribing [48,50,52]
`without negatively impacting patient care [50,52], or would
`improve patient care [46].
`
`Discussion
`
`Physicians’ awareness of the cost of therapeutics is poor.
`With only 31% of estimates within 20% or 25% of the true
`drug cost and the media