Trends and determinants of retail prescription drug costs

Abstract

Objective

To characterize price trends and variation for US generic and branded drugs at the retail level as they relate to pharmacy acquisition costs and local market factors.

Data Sources

Drug pricing data consisting of US pharmacy claims from 2014 to 2019 collected and licensed by GoodRx, an online tool for comparing drug prices.

Study Design

Time trends of median drug prices and coefficients of variation were measured for generic and branded drugs, including subgroups based on clinical condition (i.e., diabetes and cancer). Pharmacy competition was measured using the Herfindahl–Hirschman Index (HHI) at the zip‐code level. Multivariable linear regression analysis assessed the impact of local market‐level factors on drug prices and variation.

Data Collection

US drug pricing data consisting of claims filled through a mix of public and private insurance at 58,332 chain and independent pharmacies across 14,421 zip codes in all 50 states.

Principal Findings

From 2014 to 2019, pharmacy retail markets trended towards greater competition: average HHI by zip code decreased by 15.0% (p < 0.001). Median cash price increased significantly for both generic (6.58%, p < 0.001) and branded (84.10%, p < 0.001) drugs. When normalized to acquisition costs, cash prices for generic drugs rose 22.03% (p < 0.001) while those of branded drugs decreased by 2.31% (p < 0.001). Diabetes drugs showed higher baseline overall markup of cash prices relative to acquisition costs (10.54, Interquartile range (IQR) 3.28–18.43) than cancer drugs (1.88, IQR 1.36–3.08). Neither local pharmacy competition nor median income significantly predicted drug price or variation.

Conclusion

Measures of generic drug price and price variation are high despite decreased costs earlier in the pharmaceutical supply chain, defying expectations of what would happen in a competitive market. Efforts to bypass the pharmacy benefit model for generic drugs may offer consumers an opportunity for substantial savings.

What is known on this topic

• Previous studies on retail drug price and price variation have been published, but have been limited in terms of geography (e.g., based on state‐maintained websites publishing drug prices) or sample size (e.g., published on select drugs).

What this study adds

• Retail prices of generic drugs are increasing despite decreases in price at the manufacturer level and increases in local pharmacy competition.

• Patterns of both drug price at the retail level as well as drug price variation at the zip‐code level were mostly unexplained by pharmacy competition and local median income.

1. INTRODUCTION

Pharmaceutical therapies are the mainstay of modern medical practice, with each person in the country receiving an average of 11.6 prescriptions a year in 2019. ¹ In total, the retail pharmaceutical market reached $297 billion in 2019, up from $222 billion in 2010. ² Yet, despite the enormous size of this market, financial transactions at the pharmacy remain largely opaque to the public. This lack of transparency is the predominant feature of the US pharmaceutical market, and a striking contrast to the pharmaceutical markets in other countries.

Retail pharmacy prices for generic and branded drugs are not widely available to consumers, limiting the ability for individuals to shop for the best value from their medications. When prices are available, they represent cash prices, which may vary substantially across pharmacies and across products. Despite the enormous size of this market for consumers, little is known about retail drug prices in the market, and whether the retail pharmaceutical market functions effectively for consumers. Studies examining the determinants of drug prices at the point of sale are limited and varied, showing significant variation across sites with no clear correlation between drug prices and demographic or socioeconomic factors of consumers. ¹ , ² , ³ Lacking a transparent marketplace, previous research has often relied on sampling drug pricing reported in the few states that maintain websites that report drug prices; however, the information on these sites is often incomplete or out of date. ⁴ To date, there has been no large‐scale, national exploration on determinants of drug pricing at the point of sale.

The majority of insured patients are shielded from the direct effects of retail prices; however, these prices are still an important topic in health policy. First, cash prices can indirectly impact insured patients by increasing premiums for private plans or the tax burden of publicly funded plans. For example, Blues plans in six states sued CVS for charging inflated cash prices to their health plans while offering lower prices to customers with CVS cash discount cards. ⁵ One recent study comparing generic drug prices at Costco to the rates reimbursed by Medicare and found that Medicare overpaid by $2.8 billion in 2018 alone. ⁶ Second, high retail prices are of concern to the uninsured and underinsured, and cash‐paying consumers still constitute 5% of total prescription spending. ⁷ Out‐of‐pocket payments for patients paying coinsurance have substantially increased from 2014 to 2018. ⁸ Thus, pricing in the retail pharmacy market is a relevant issue for examination for the clinical and policy communities.

This paper presents a national analysis of generic and branded drug price and variation in the United States. It will explore how market covariates, such as local income and pharmacy competition, relate to drug price. By better characterizing this market, we propose policies that improve the prescription drug retail market for consumers.

2. METHODS

2.1. Data

This study is an observational analysis of price pricing data in the United States. The study database consisted of generic and branded US pharmacy claims from 58,332 chain and independent pharmacies across 14,421 zip codes in all 50 states from 2014 through 2019 (See Table 1). The total number of US chain and independent pharmacies was 62,145 in 2019. ⁹ Thus, our study dataset represents approximately 94.67% of US pharmacies. Claims data included the year the claim was filed, the National Average Drug Acquisition Cost (NADAC), the cash price as well as drug‐ and pharmacy‐specific data. The NADAC is a national benchmark updated by the Centers for Medicare & Medicaid Services that reflects the average price at which pharmacies acquire a drug, based on actual invoices submitted by surveyed pharmacies. ¹⁰ Cash price, another common measure of drug price, is the price available at any retail pharmacy for consumers filling a prescription without using prescription drug coverage or any discount program. ¹ The cash price used has been independently validated by randomly sampling retail pharmacies and cross‐referencing the quote offered on the phone with the value in our database. ¹ Cash prices have also been shown to correlate strongly with other benchmarks used in negotiating out‐of‐pocket prices or pharmacy benefit manager (PBM) reimbursements such as list prices or average wholesale prices ¹¹ (See Data S1 (Section A). for further definitions of various drug prices). In these ways, higher cash prices ultimately, whether through out‐of‐pocket expenditures or raising premiums, correlate with higher costs for consumers. Prices are nominal and refer to the price of a prescription.

TABLE 1.

Characteristics of study sample

Variable	Generic (N a =71,393,034)		Brand (N a =12,858,837)
	N (millions)	%	N (millions)	%
Pharmacy class
Independent	10.3	14.43	2.0	15.74
Chain	61.1	85.57	10.8	84.26
Year
2014	9.7	13.61	2.3	17.65
2015	10.8	15.18	2.3	18.11
2016	12.0	16.81	2.4	18.71
2017	11.8	16.59	2.0	15.66
2018	13.0	18.23	2.0	15.48
2019	14.0	19.58	1.8	14.39
HHI
High competition	60.2	84.34	10.8	84.33
Medium competition	4.3	6.00	0.8	5.97
Low competition	1.7	2.39	0.3	2.32
Near monopoly	5.2	7.27	0.9	7.38
Median income
First quintile	7.8	10.94	3.4	11.87
Second quintile	13.6	19.04	2.4	19.21
Third quintile	13.4	18.80	2.4	19.18
Fourth quintile	17.6	24.71	3.1	24.76
Fifth quintile	18.9	26.51	3.2	24.97
Condition
Diabetes	1.6	2.20	1.9	14.75
Cancer	3.9	5.50	0.7	5.28

^a N is reported as the number of unique drug‐pharmacy combinations in a given year. Herfindahl–Hirschman Index (HHI) is reported in categories based on the following cutoffs: high competition (0–2500), medium competition (2501–5000), low competition (5001–8000), and near‐monopoly competition (8001–10,000). Median income by zip code is reported in quintiles based on the following cutoffs: first quintile (below $40,750), second quintile ($40,751–$49,848), third quintile ($49,849–$58,750), fourth quintile ($58,751–$73,245), fifth quintile (above $73,245).

Claims data were provided by GoodRx, a company that tracks prescription drug prices and offers coupons for discounts. The dataset includes a representative list of de‐identified pharmacy claims filled in the United States, representing a diversity of channels and providers. Drugs were included in our study only if tablet formulations (i.e., alternative formulations such as oral solutions and topical creams were excluded). It includes a mix of sources from the public domain and sources that are externally licensed. It is not limited to internal claims filled with GoodRx coupons. A full description of their dataset and its validation has been previously published. ¹²

Pharmacy information at the zip‐code level was provided through the National Council for Prescription Drug Programs database, which includes pharmacy class (i.e., chain or independent). Socioeconomic data (i.e., median income) used in regression analysis were collected from the American Community Survey (2014–2018 5‐year estimate) and linked to claims data at the zip‐code level. Median income by zip code is reported in quintiles based on the following cutoffs: first quintile (below $40,750), second quintile ($40,751–$49,848), third quintile ($49,849–$58,750), fourth quintile ($58,751–$73,245), and fifth quintile (above $73,245).

Local pharmacy competition was measured using the Herfindahl–Hirschman Index (HHI) at the zip‐code level, which is the preferred measure of competition by the US Department of Justice. ¹³ It is calculated by summing the squares of the individual firms market share. ¹⁴ Our study defined market share as volume of drug sales (i.e., number of prescriptions filled) rather than by revenue. Similar to previous studies examining the effect of competition on generic pricing, ¹⁵ we defined HHI both as a continuous variable (for time trends) and in categories (for regression analysis): an HHI of 0–2500 was defined as a high competition (i.e., at least four pharmacies of roughly equal market share); an HHI of 2501–5000 was defined as medium competition; an HHI of 5001–8000 was defined as low competition, and an HHI of 8001–10,000 was defined as near‐monopoly.

Prescription claims were aggregated by drug, pharmacy, and year before subsequent analysis. That is, the prices of multiple claims for the same drug that were filled at the same pharmacy and in the same year were averaged. Each drug was delineated based on generic name rather than at the NDC level. Prices for the same drug offered in multiple quantities (e.g., 30 vs. 60 tablets) were adjusted to the price of a typical quantity and analyzed together. In order to quantify price variation between pharmacies for each drug in a given area, variation in cash price was measured at the zip‐code level using coefficient of variation (i.e., the ratio of the standard deviation of cash price to the mean cash price).

2.2. Trends in competition, price and variation

Time trends were calculated both for HHI, cash‐price variation, absolute cash price and NADAC separately for generic and branded medications. All reported prices in figures represent median prices and are reported in the text with interquartile range (IQR), as drug prices were not normally distributed. Time trends reflect changes in drug prices independent of changes in prescription patterns. This was accomplished by holding frequency weights for each drug constant over the study period. Frequency weights were determined by total claim count for each drug over the 2014–2019 period. To prevent influence of new drugs entering the market on observed trends, only drugs represented in the entire study period were included. The ratio of cash price to NADAC was reported as overall pharmacy markup when summarizing data at the national level. Condition subgroups (i.e., diabetes and cancer) were determined using a list of indications by drug molecule maintained by GoodRx. (See Data S1 (Section B). for complete list of drugs used for each indication). These categories were included in our analysis both because they cause substantial financial burden to the US health care system and because they differ in important variables. Namely, the financial burden of cancer drugs tends to be concentrated in more rarely prescribed drugs of high cost, whereas the financial burden in diabetes drugs is more in the high volume of lower cost drugs.

2.3. Statistical analysis

For each outcome, we used a multiple linear regression model to assess the predictive value of the following independent variables on the outcomes of interest: local (i.e., zip‐code level) pharmacy competition (i.e., HHI), local median income, and pharmacy class (i.e., independent vs. chain). The primary outcome of this study was cash price, which was normalized to NADAC (See Table 3). The secondary outcome of this study examined local price variation, measured using coefficient of variance of cash price calculated between pharmacies at the zip‐code level. Pharmacy class was applied only to the regression examining cash price as the primary outcome, as it was not relevant to the calculation of variation, which was determined at the zip‐code level. Each regression was run separately for generic and branded drugs. Both models were run using fixed effects, with drug name and year as panel and time variables.

TABLE 3.

Effect of local competition, median income and pharmacy class on generic and brand drug price, and price variation

	Generic			Brand
	Coef.	95% CI	p‐value	Coef.	95% CI	p‐value
A. Cash price
HHI
High competition	—	—	—	—	—	—
Medium competition	−6.65	±0.61	***	−9.31	±0.60	***
Low competition	−2.57	±0.61	***	−1.97	±0.60	***
Near monopoly	−5.67	±0.57	***	−2.15	±0.56	***
Median income
First quintile	—	—	—	—	—	—
Second quintile	−1.63	±0.25	***	−3.27	±0.25	***
Third quintile	2.22	±0.25	***	0.33	±0.25	***
Fourth quintile	1.37	±0.25	***	1.60	±0.25	***
Fifth quintile	2.68	±0.25	***	1.93	±0.25	***
Pharmacy class
Independent	—	—	—	—	—	—
Chain	70.39	±0.19	***	11.35	±0.19	***
R2	0.14			0.03
N	71,247,007			12,834,307
B. Price variation
HHI
High competition	—	—	—	—	—	—
Medium competition	−0.02	±0.00	***	0.01	±0.00	***
Low competition	0.01	±0.00	***	0.00	±0.00	0.004
Near monopoly	−0.03	±0.00	***	−0.01	±0.00	***
Median income
First quintile	—	—	—	—	—	—
Second quintile	0.01	±0.00	***	0.00	±0.00	***
Third quintile	−0.01	±0.00	***	−0.01	±0.00	***
Fourth quintile	0.00	±0.00	***	0.00	±0.00	0.036
Fifth quintile	−0.05	±0.00	***	−0.02	±0.00	***
Constant	0.41			0.11
R‐squared	0.01			0.01
N	71,247,007			12,834,307

Note: Results of fixed‐effects multivariable linear regression pharmacy markup on listed independent variables: Herfindahl–Hirschman Index (HHI), median income, and pharmacy class. Cash price was normalized to acquisition costs at the drug level, using national average drug acquisition cost (NADAC) as a standard. Coefficients for cash price are reported as percentage changes in price relative to the reference value (e.g., price for generic drugs in areas of near monopoly competition is 5.67% lower than in areas of high competition). Price variation was calculated as the coefficient of variation (i.e., the ratio of the standard deviation to the mean) of drug prices within a zip code; as such, pharmacy class was excluded from this model. Regression coefficients for variation are reported as changes in the coefficient of variation relative to the reported constant (e.g., for generic drugs, near monopoly conditions are associated with a 0.03 lower value for coefficient of variation than areas of high competition). Generic drug name and year were included as panel and time variables. HHI is reported in categories based on the following cutoffs: high competition (0–2500), medium competition (2501–5000), low competition (5001–8000), and near‐monopoly (8001–10,000). Median income by zip code is reported in quintiles based on the following cutoffs: first quintile (below $40,750), second quintile ($40,751–$49,848), third quintile ($49,849–$58,750), fourth quintile ($58,751–$73,245), fifth quintile (above $73,245).

^*** p < 0.001.

All analyses were performed using STATA 16 on the Stanford Computing Sherlock Cluster.

2.4. IRB approval

This study was considered exempt research by the IRB.

3. RESULTS

We analyzed 71,393,034 unique drug‐pharmacy pairs for generic drugs and 12,858,837 pairs for branded drugs (See Table 1). The majority (84.3%) of these claims were filled in highly competitive markets (i.e., zip codes where at least four pharmacies held roughly equal share). A significant number of claims (i.e., 7.3%) were filled in near‐monopoly markets, where one pharmacy holds at least 90% of the market share. From 2014 to 2019, markets trended to get more competitive: HHI decreased by 15.01% (p < 0.001) for generic drugs and 16.55% (p < 0.001) for branded drugs (See Table 2). Cash‐price variation between pharmacies within the same zip code remained relatively constant over the study period, with generic drugs showing a 5.79% (p < 0.001) increase and branded drugs showing a 6.41% increase (p < 0.001). Importantly, variation was significantly higher for generic drugs, which had a coefficient of variation of 0.41 (IQR 0.24–0.65) in 2019, than for branded drugs, which had a coefficient of variation of 0.08 (IQR 0.05–0.13) in 2019.

TABLE 2.

Summary of trends in drug and pharmacy characteristics from 2014 to 2019

	Generic			Brand
	2014	2019	Change (%) ***	2014	2019	Change (%) ***
HHI	1600	1359	−15.06	1606	1340	−16.56
Markup
Overall	7.82	10.03	22.03	1.30	1.27	−2.31
Diabetes	10.54	12.98	23.14	1.25	1.25	0.00
Cancer	1.88	3.08	63.83	1.60	1.42	−11.25
Coef. of variance	0.41	0.43	5.79	0.07	0.08	6.41
Diabetes	0.69	0.65	−6.01	0.08	0.07	−1.73
Cancer	0.43	0.40	−7.44	0.12	0.15	18.34
Cash price
Overall	37.99	40.49	6.58	238.99	439.99	84.10
Diabetes	23.99	26.00	8.38	379.99	567.99	49.47
Cancer	16.49	18.53	12.37	41.20	49.66	20.53
NADAC
Overall	4.44	3.08	−30.63	183.51	337.36	83.84
Diabetes	1.99	1.74	−12.56	301.56	433.19	43.65
Cancer	9.96	8.77	−11.95	25.89	35.69	37.85

Note: Herfindahl–Hirschman index (HHI) was first calculated at the zip‐code level and reported here as the national mean. Markup, coefficient of variation, and price values are reported here as median values of all claims with claim frequencies for specific drugs fixed by year, controlling for changes in prescribing patterns. Markup is calculated as ratio of cash price to NADAC and reported as a year‐specific, frequency‐weighted average across all drugs. Coefficient of variance is calculated for each drug at the zip‐code label and reported here as a year‐specific, frequency‐weighted average across all drugs for the corresponding year. Prices are nominal and refer to the price of a prescription.

Abbreviations: HHI, Herfindahl–Hirschman index. NADAC, National Average Drug Acquisition Cost.

^*** p < 0.001 for all calculated percent changes.

Median cash price increased significantly for both generic (6.58%, p < 0.001) and branded (84.10%, p < 0.001) drugs. Cash prices increased from $37.99 (IQR 18.39–85.00) to $40.49 (IQR 18.75–90.95) for generic drugs and $238.99 (IQR 128.41–354.99) to $439.99 (IQR 214.61–596.09) for branded drugs. NADAC decreased for generic drugs (−30.63%, p < 0.001), dropping from $4.44 (IQR 1.74–11.22) to $3.08 (IQR 1.68–8.04), and NADAC increased for branded drugs (83.84%, p < 0.001) from $183.51 (IQR 97.35–279.92) to $337.36 (IQR 171.41–450.17). For generics, the increase in cash price and decrease in NADAC drove a 22.03% increase in overall markup over the study period (p < 0.001), from 7.82 (IQR 2.76–20.79) to 10.03 (IQR 4.56–24.56). For branded drugs, the proportional increase in cash price and NADAC led to a minimal overall markup decrease of 2.31%, from 1.30 (IQR 1.23–1.39) to 1.27 (IQR 1.24–1.40).

Generic diabetes medications showed high starting overall markups and increased 23.14%, from 10.54 (IQR 3.29–18.43) to 12.98 (IQR 4.45–22.89). Meanwhile, generic cancer medications exhibited a relatively low initial overall markup at 1.88 (IQR 1.36–3.08) and grew to 3.08 (IQR 1.50–6.13) by 2019, a 63.83% growth. This difference was largely explained by a larger increase in cash price of cancer drugs (11.63%), from $16.49 (IQR 11.99–29.98) to $18.53 (IQR 11.99–31.99). Branded diabetes and cancer drugs largely followed overall trends for branded medications, with more conservative overall markups but higher starting prices: In 2019, those for branded diabetes drugs were 1.25 (IQR 1.23–1.34) and those for branded cancer drugs were 1.42 (IQR 1.30–1.73).

Keeping prescribing frequency constant over the study period, we observed a modest increase in cash price and a significant decrease in NADAC for generic drugs (See Figure 1). This consistency indicates two disparate trends for generic and branded drug prices: On one hand, retail generic drug prices are rising despite lowering input costs from earlier in the supply chain. On the other hand, the rise in branded drug prices appears to be driven by prices set before the pharmacy level. We observe an increasing overall markup for generic drugs, while markups for branded drugs are relatively stagnant (See Figure 2).

FIGURE 1.

Time trends in generic and branded drug prices from 2014 to 2019. National Average Drug Acquisition Cost (NADAC) values for generic and branded drugs were normalized to their respective 2014 costs. Cash price for both generic and branded drugs were normalized to 2014 NADAC values plus an additional $9 dispensing fee. Frequency‐weighting of drugs was held constant over the study period to isolate trends in drug pricing and exclude the effects of changes in prescribing patterns. Prices are nominal

FIGURE 2.

Time trends in generic and branded drug price markup from 2014 to 2019. Markup, reported as the median value, is calculated as the ratio of cash price to National Drug Acquisition Cost (NADAC). Frequency‐weighting of drugs was held constant over the study period to isolate trends in drug pricing and exclude the effects of changes in prescribing patterns. Prices are nominal. A list of medications included in each indication subgroup is included in Data S1 (Section B) [Color figure can be viewed at wileyonlinelibrary.com]

Regressing cash price on HHI showed no meaningful association even when comparing very competitive markets to near‐monopoly markets: Generic drugs were 5.67% cheaper (95% CI −6.24 to −5.10) and branded drugs were 2.15% cheaper (95% CI −2.71 to −1.59) in near‐monopoly markets, with no consistent relationship between HHI and cash price (See Table 3A). The effect of changes in median income on cash prices was small: Zip codes in the fifth quintile of median income had cash prices 2.68% (95% CI 2.43–2.93) higher than those in the first quintile (p < 0.001), and there was no consistent dose–response relationship between median income and cash prices. Chain pharmacies had greater cash prices for both generic (70.39%, 95% CI 70.10–70.58) and branded (11.35%, 95% CI 11.16–11.54) drugs (p < 0.001). Differences in competition, median income and pharmacy type explained variation in cash prices more completely for generic drugs (R ² = 0.14) than for branded (R ² = 0.03) drugs.

Multivariable linear regression of the coefficient of variation of cash price on HHI and median income appeared to show no significant association (See Table 3B), as the strength of these associations was weak for both generic (R ² = 0.01) and branded (R ² = 0.01) drugs, suggesting that the examined variables fail to meaningfully lower price variation.

4. DISCUSSION

Our analysis revealed substantially divergent market trends for generic and branded medications. Higher cash prices in the branded drug market were characterized by proportional increases in acquisition costs, indicating that rising costs for branded drugs largely reflect higher prices set by manufacturers, distributers, or wholesalers. By contrast, the generic drug market is characterized by decreasing pharmacy acquisition costs but increasing cash prices, suggesting that higher prices to consumers reflect factors later in the supply chain. Variation of generic drug cash prices was significantly higher than that of branded drugs, potentially indicating divergent pricing strategies between different pharmacies. It is important to note that these divergent trends in generic prices or average markup are not explained away by considering a $9 dispensing fee, as was observed in Figure 1. ¹⁶

Rising consumer‐facing costs—trends concentrated in high‐volume and diabetes drug categories—are deeply concerning, especially given the well documented high out‐of‐pocket costs of insulin and other diabetes‐related supplies. ¹⁷ One previous study showed that competition among generic manufacturers helped to lower (or maintain low) prices. ¹⁵ Our study provided evidence of significantly decreased prices at the manufacturing level (i.e., as reflected in a down‐trending NADAC); however, the divergent trends for NADAC and cash price observed in our study suggest further opportunity for ensuring that savings accomplished early in the supply chain reach the consumer.

Meanwhile, trends for branded medication reflected price increases earlier in the supply chain, as reflected by up‐trending NADAC values. Our findings for branded medications may reflect broader market dynamics where higher list prices for prescription drugs (i.e., the prices set by manufacturers) are associated with higher prices set by the manufacturer and higher rebate payments from manufacturers to intermediaries in the market (PBMs and distributors). A recent study documented the rapid rise of list prices of 8.1% per annum from 2014 to 2019 for a sample of manufacturers, with an increase in manufacturer payments (manufacturer rebates and chargeback payments) from 38.4% to 67.4% of net revenues over this period. ¹⁸ The change in pharmacy acquisition cost we observed for branded drugs is similar to the growth rate of gross revenue of these manufacturers, and our observations may reflect market dynamics characterized in other studies, where increasing manufacturer prices and rebate payments are strongly associated with increases in drug prices for branded drugs observed at the retail level. ¹⁸ , ¹⁹ Further, without passing these rebates to patients, people with coinsurance plans have been facing significantly higher out‐of‐pocket expenses as a direct result of these rising list prices. ⁸

Our analysis characterized a significant amount of price variation for the same drug between different pharmacies. Variation was higher for generic drugs than branded drugs. Consistent with previously published literature, these levels of variation were not adequately explained by competition. ²⁰ Large price variation independent of competition can be explained by two factors. First, price variation is classically the result of high search costs brought on by opaque pricing, where the costs of searching are high or the utility of searching is unknown. ¹⁵ Many insured patients are unable to receive copayment information before transferring a prescription to a specific pharmacy, and many cash‐paying patients may not know of price comparison tools such as GoodRx. Thus, pharmacies are able to operate as pseudo‐monopolies, charging prices unaffected by local competition. Second, price variation may reflect the trend for pharmacies to inflate cash prices in order to negotiate higher reimbursement rates from PBMs. ²¹ Cash prices are set with the insurance benefit rather than out‐of‐pocket costs in mind. With either explanation, cost‐sharing forces many patients to participate in the negotiation of drug prices without giving them adequate tools to do so.

4.1. Policy relevance

When drug coverage was first introduced to the employer market and to Medicare Part D, the concern was financing the purchase of brand name pharmaceuticals which were, and continue to be, relatively expensive purchases for consumers. However, the market itself has shifted dramatically over the last two decades with the expiration of patents on many effective small molecule drugs. As a result, almost 90% of the prescriptions written today are for generic medications. While the structure of the pharmaceutical benefit may be necessary for ensuring access to high‐cost, branded drugs, it may be an inefficient mechanism for financing high‐volume, lower‐cost generic drugs. Nonetheless, the pharmacy benefit structure remains unchanged.

There are three categories of pharmaceuticals in the market: (1) over‐the‐counter medications that are non prescription, multi‐source generic drugs covered mostly with self‐payments, (2) generics that are prescription, multi‐source drugs most often covered by third‐party payments, and (3) branded drugs that are prescription, single source, and covered most often by third‐party payments. Based on this simple stratification, there are two generic drug markets existing in parallel, the over‐the‐counter and prescription market. In the over‐the‐counter market, we see price transparency, efforts to assure consumers of quality, and competition between branded and un‐branded (store‐brand) generic products. In contrast, in the retail pharmacy market for prescription generic medications, we see aggressive price increases, a lack of price transparency, and a lack of focus on quality. ⁶ , ¹⁸ , ¹⁹

If there were an effective retail market for generic prescription pharmaceuticals, it is easy to imagine that we could construct a cash‐pay alternative to insurance for generic medications (with or without an employer subsidy). Unfortunately, our results suggest that at the present time, this option is largely closed off due to a lack of price transparency and competition at the local level in the market.

To support this notion of a potential alternative to insurance for these products, we have seen substantial price reductions for consumers in bypassing the complexity of the financing system. For example, the Walmart $4 prescription cash payment model offers lower prices than the co‐payment required through many different insurance plans. ²⁰ Further, one recent study estimates that if Medicare programs reimbursed PBMs at the rate that Costco charges members of the public, they would have saved $2.8 billion in 2018 alone. ⁶

Ultimately, high retail drug prices, both for cash‐paying underinsured and cost‐sharing insured populations, rely on a pharmaceutical market structure corrupted by opaque consumer‐facing prices and mis aligned incentives for the entities initially formed to lower prices. In other words, third‐party payment allows for price inflation and variation in the market for prescription generic drugs, with consequences for all of us.

Transparency of prices in the health care market is an area of increased focus for policy makers. In an effort to “make it easier for consumers to shop and compare prices across hospitals and estimate the cost of care before going to the hospital,” Centers for Medicare & Medicaid Services has enacted a new regulation on price transparency for hospitals, which requires hospitals to make pricing information available in consumer‐accessible language and formats. ¹⁸ Additionally, the Department of Health and Human Service's Transparency in Coverage ruling (CMS‐9915‐F) is scheduled to begin implementation on July 1, 2022. ²² This policy will work towards increasing health care price transparency more broadly, including with pharmaceutical prices.

4.2. Limitations

Results were limited in that they were based on cash price, which do not directly reflect actual payments. To date, no study breaks down the portion of patients paying the full cash price, an insurance‐negotiated price or a coupon‐discounted price. While the cash price is a common proxy of drug spending used in previous similar studies, ²³ cash prices may not reflect actual acquisition prices for patients purchasing their medications through a health plan.

Additionally, our measure of competition was based on the number of pharmacies in 2019, which may have changed over the 2014–2019 period. Also, our models assume that people shop for drugs only within their own zip code, which may fail to capture complexities such as prices being driven by the characteristics of nearby communities. The pharmacy markup includes the drug acquisition costs and the pharmacy's amortized expenses related to the costs of drug dispensing, which we could not observe (i.e., pharmacy labor and overhead, inventory costs, space, and services such as patient counseling).

Further, NADAC is a national average and may vary in how well it approximates actual acquisition costs. While this would not impact our analysis of price trends, it may influence the validity of our regression analysis examining cash price, as NADAC may systematically vary with examined independent variables.

Lastly, 52% of price variation remained unexplained for generic drugs and 59% of price variation remained unexplained for branded drugs in our analysis. Some of this is likely due to pharmacy‐specific characteristics not available in our dataset such as pharmacy ownership for chain stores.

5. CONCLUSION

Our analysis revealed markedly different patterns in the markets for generic and branded drugs. Cash prices for generic products rose despite a significant fall in pharmacy acquisition costs and increases in pharmacy competition. Meanwhile, cash prices of branded drugs rose significantly and proportionally with acquisition costs, reflecting increases in the pharmacy acquisition costs. Except in extreme cases (i.e., near‐monopoly conditions), increases in competition at the retail level are largely unrelated to retail drug prices and failed to suppress price variation. Given the increase in cash‐paying patients and inefficiencies of managing generic drugs under a pharmacy benefit, greater transparency of pricing in the pharmaceutical market may allow for more widespread uptake of low‐cost cash‐paying options, offering all consumers an opportunity for substantial savings.

Supporting information

Data S1. Supporting Information.

Teasdale B, Nguyen A, van Meijgaard J, Schulman KA. Trends and determinants of retail prescription drug costs. Health Serv Res. 2022;57(3):548-556. doi: 10.1111/1475-6773.13961