HealthOptimize: Indiana Medicaid Pharmaceutical Expenditure Analysis

HealthOptimize Solutions: Combating Rising Drug Costs within Indiana Medicaid

Indiana Healthcare Program Request

Addressing the Problem

The escalating costs of pharmaceuticals present a pressing challenge for the Indiana Indiana Family and Social Services Administration (FSSA) and Indiana Healthcare Programs (IHCP). Despite efforts to contain expenses within the Indiana Medicaid program, the continuous rise in drug prices strains the budget allocated for healthcare services. This phenomenon not only threatens the financial sustainability of the program but also jeopardizes patient access to essential medications. To navigate these challenges effectively, the FSSA must implement new strategies that optimize managed care while prioritizing cost containment and patient welfare. Finding solutions that strike a balance between both of these obstacles is crucial to the long-term success of the Indiana Medicaid program.

Value in Stakeholders

The Indiana Family and Social Services Administration (FSSA) oversees Medicaid and needs to manage costs. Medicaid beneficiaries, who rely on the program for healthcare, are affected by changes in drug prices. Healthcare providers, like doctors and pharmacists, also have a role as they prescribe and dispense medications. Pharmaceutical companies, who make and sell drugs, influence prices, which impacts Medicaid spending. Managed care organizations (MCOs), contracted by FSSA, help manage healthcare services for beneficiaries. State government officials make decisions about Medicaid policies and budgets. Advocacy groups represent the interests of beneficiaries and providers. Finally, taxpayers fund Medicaid and care about how efficiently their money is used.

Data Sources and Preperation

Key Terminology

• NDC
  • National Drug Code
• NADAC
  • National Average Drug Acquisition Cost
  • Pricing Metric
• MCE
  • Managed Care Entity
• CPU
  • Cost per Unit

Data Source

When deciding the route we wanted to take to provide an anyltics solution, identifying the data that we had readily available was integral. We comprised a new table of data pulling together and cleaning. We sourced all data through Medicaid.Gov. The medicaid website provided us with utilization data, labeler information, product level details, units dispensed, the number of prescriptions, total amount reimbursed all on a quarterly basis. The availability of quarterly NADAC priced posed beneficial to our analysis, as we had total units reimbursed and NADAC per unit. This means we would be able to assess whether drugs are priced based on Medicaid guidelines. Having an awareness of the landscape surrounding the pharmaceutical was pivotal when deciding which fields that we had would be beneficial for analysis. In turn, we were able to pull in the fields needed and quickly identify how we can provide an analytics solution within the scope of the data available.

Data Preperation

In the initial phase of data preparation, the primary focus was on procuring the pertinent data from Medicaid.Gov, ensuring precision by filtering exclusively for Indiana-specific data. Next, ONLY the utilization data was imported into SQL. All of these steps were performed for each years data sets. The first step was splitting up the data between FFS and Managed Care utilization. Some restrictions imposed are only applicable to managed care, additionally, any NDCs that are not utilized for the managed care delivery system, and only FFS were removed the analysis. Generally speaking, most of these occurrences are going to be “carve-out” drugs. These are not considered or adjusted during Medicaid capitation rate setting and would be immaterial in policy changes that could occur. Other reasoning behind this is lower levels of utilization. Further refinement was employed for Managed Care data, with a stringent threshold set at a script count exceeding 500 for NDCs, ensuring the dataset’s manageable size without compromising granularity. The final utilization table contains the 4 years unioned together.

Due to the immense size of the NADAC and rebate files, the import functionality in Oracle was unable to process the import. The final utilization table was imported into R. Both the rebate and NADAC tables were filtered based on NDCs present in the utilization table for a given year and quarter. Since the NADAC data contains separate files by year, this was performed 4 times for each year, and a single table was created by unioning the 2019-2022 data sets. The NADAC and Rebate files were then imported into SQl. Data manipulation and cleaning are much more intuitive in SQL and easier to follow than R.

The next phase, was to identify the pharmaceutical company the individual NDCs are coming from. The rebate data table contains the pharmaceutical companies, the common columns were NDC, Year, and Labeler_Code. Due to changes in how the company names were formatted, manual updates were done to create single names for all individual labeler codes. NDC_Description, FDA_Drug_name were joined onto the table by NDC. The reabte flag was added to identify drugs that have a federal rebate. This was set to Y or N.

Repriced fields were created in SQL, more information regarding how these fields were calculated can be found in the data understanding tab

Analytical Approach

1. Preliminary Analysis

• assess pharmaceutical expenditure growth on a quarterly basis to evaluate the level of crisis
• assess the growth in unit cost
• explain and highlight the differences between units and prescriptions

2. Product Level Analysis

• Identify products with the largest potential of cost savings by using the average minimum cost per script over all time periods - 7 products will be assessed if the there are a significant difference in unit costs within the same NDC descriptions
• Based on these results, we will determine if utilization needs to be shifted to lower cost NDCs - The build of the shift in utilization will be determined after conducting all previous steps - More details regarding this modelling exercise will be provided later

3. NADAC Assessment

• The next phase will compare the unit cost to the NADAC unit cost
• Reprice to NADAC CPU by multiple total units dispensed by the NADAC unit cost
  • identify drugs with the highest cost differential
• There are additional pricing metrics, such as the State Maximum Allowable cost
  • The minimum will of total amount reimbursed and totl NADAC repriced will be populated in an Adjusted_repriced field
• Highlight the largest areas of concern

4. Generate Forecasts

• Four forecasts will be generated:
  • Utilizing actual expenditures
  • Utilizing adjusted expenditures after shifting utilization
  • Utilizing the adjusted NADAC repriced fields
  • Implementing both adjustments of utilization

5. Identify Methods to reduce total cost

• After assessing the results in prior phases suggestions to reduce the growth in expenditures a final solution will be provided
• the solution will be feasible and will explore multiple approaches

Preliminary Analysis

Preliminary Analysis

At this stage, our focus is on understanding the trajectory of expenditure growth over time. Our dataset covers the years 2019 to 2022, as we couldn’t incorporate the 2023 data due to timing constraints. We’ll also analyze the average unit cost trends during this period. While these analyses may not provide the complete picture, they will serve as valuable starting points. Given the anticipated increase in enrollment each year, it’s natural to expect overall expenditures to rise. However, our primary objective at HealthOptimize is to manage and potentially reduce this growth, particularly in light of the expanding Medicaid population.

Total Expenditures Increasing Rapidly

summarized_data <- Drug_Data %>%
  group_by(YEAR_QUARTER) %>%
  summarise(Sum_Total_Amount_Reimbursed = sum(TOTAL_AMOUNT_REIMBURSED))

suppressWarnings({
ggplot(summarized_data, aes(x = YEAR_QUARTER, y = Sum_Total_Amount_Reimbursed/1000000)) +
  geom_point(color = "steelblue", size = 2) +
  scale_y_continuous(labels = scales::dollar_format(prefix = "$", suffix = "M"), breaks = seq(0, max(summarized_data$Sum_Total_Amount_Reimbursed/1000000), by = 25)) +
  theme_minimal() +
  theme(
    plot.title = element_text(size = 12, hjust = 0.5, face = "bold"),
    axis.text = element_text(size = 9, color = "black"),  # Change color to black
    axis.text.x = element_text(size = 10, color = "black", angle = 45, hjust = 1),
    axis.title = element_text(size = 10, color = "black"),
    legend.position = "none"
  ) +
  labs(
    x = "Year Quarter",
    y = "Total Amount Reimbursed (Millions)",
    title = "Total Indiana Medicaid Pharmaceutical Expenditures"
  ) + 
geom_line(color = "steelblue", size = 1, group = 1)
})

The Medicaid spending depicted here offers insight into a specific segment of our data, capturing notable trends while recognizing certain exclusions. Notably, our analysis focuses on NDCs with a material amount of utilization of at least 500 scripts per quarter, and excludes FFS claims lacking managed care utilization. Despite these constraints, expenditures have surged by over $120 million since the first quarter of 2019.

Moving forward, our analytical endeavors will delve into diverse methodologies aimed at mitigating this exponential expenditure growth.

Trend in Unit Cost Overtime

summarized_data <- Drug_Data %>%
  group_by(YEAR_QUARTER) %>%
  summarise(CPU = sum(TOTAL_AMOUNT_REIMBURSED)/sum(UNITS_REIMBURSED))

suppressWarnings({
ggplot(summarized_data, aes(x = YEAR_QUARTER, y = CPU)) +
  geom_point(color = "steelblue", size = 2) +
  theme_minimal() +
  theme(
    plot.title = element_text(size = 12, hjust = 0.5, face = "bold"),
    axis.text = element_text(size = 9, color = "black"),  # Change color to black
    axis.text.x = element_text(size = 10, color = "black", angle = 45, hjust = 1),
    axis.title = element_text(size = 12, color = "black"),
    legend.position = "none"
  ) +
  labs(
    x = "Quarter",
    y = "Average Cost per Unit",
    title = "Change in Average Unit Cost"
  ) + 
geom_line(color = "steelblue", size = 1, group = 1)
})

The dynamics of unit costs are rather complex as they can be influenced in a multitude of ways. Inflationary trends typically contribute to year-over-year increases in drug unit costs. However, while we might anticipate individual drug costs to rise in line with the inflation rate, other factors can drive unit cost increases as well.

These factors include drug shortages, shifts in market demand, and the introduction of new pharmaceuticals. For instance, when a branded drug faces competition from generic alternatives or similar products from different manufacturers, we often expect the unit cost of the branded drug to decrease. Yet, it’s essential to note that this isn’t always the case; sometimes, the opposite occurs.

Moreover, changes in risk pools further complicate the picture, adding another layer of complexity to the analysis of unit cost dynamics. This is related to different conditions of recipients and diagnosis leading to a change in the types of drugs being utilized.

Due to data limitations, we are unable to assess all reasons as to why unit costs are increasing at a dramatic rate. Yet, we can provide strategies to combat and slow the rise.

Total Units Reimbursed by year

Kaggle_table %>%
  kable("html", caption = "<span style='font-weight:bold; font-size:larger;'>Indiana Medicaid Yearly Drug Utilization</span>") %>%
  kable_styling(bootstrap_options = "striped", full_width = FALSE) %>%
  row_spec(0, bold = TRUE, color = "white", background = "navy")

Indiana Medicaid Yearly Drug Utilization

Year	Total Units	Total Scripts
2019	676,466,521	11,864,006
2020	690,254,158	12,389,338
2021	786,156,145	14,245,446
2022	888,627,963	15,934,403

This information is included to highlight the substantial difference between units and prescriptions. A unit could be a tablet, milliliters in a solution, or grams in an aerosol. That’s why the unit count can be much higher than the number of prescriptions. For example, a single prescription for a 90-day supply could contain multiple tablets or milliliters. Thus, looking at cost per script could be misleading depending on how the data is reported. So, in future analyses, we’ll focus on the cost for each unit instead of each prescription. Additionally, since the NADAC price is shown per unit, we’ll use that to compare Medicaid costs per unit.

Phase 1: Pharmaceutical Companies

Evaluating NDC Unit Costs

The initial phase of this analysis focused on evaluating the utilization of NDCs sharing the same NDC description. When the NDC description matches, it generally suggests that these products serve similar purposes and could be interchangeable. However, it’s important to note that while this assumption holds true for most drugs, it may not be applicable to all. Due to the unavailability of exact GPIs (Generic Product Identifiers), slight differences in formulary could exist, defining different drugs. GPIs are 14-digit numeric codes, where the first six digits delineate the drug subclass.

Regrettably, this level of detailed data wasn’t accessible for our analysis, necessitating the utilization of the highest available granularity. We identified NDC descriptions with more than one unique NDC on the market. Our aim was to pinpoint drugs exhibiting the greatest variance between the average minimum cost per unit across all time periods and the overall average cost per unit. These identified drugs underwent further scrutiny as they possess the potential to significantly impact expenditure if all utilization were to shift to the lowest-cost drug within the NDC description.

While transitioning all utilization to the least expensive drug within a given NDC description may not be feasible, forthcoming details will elucidate how HealthOptimize optimized their shifting metrics and restrictions.

The table above identifies the drugs that we elected to evaluate further after utilizing the methods described above. These drugs pose the greatest potential expenditure impact. Yet, whether or not this holds true will be evaluated further.

Explanation of Preliminary Assessments

Preliminary assessments were undertaken across all individual products to identify the various pharmaceutical companies supplying the drug and their activity over different time periods within the base data period. If there was limited utilization from pharmaceutical companies or irregular entry patterns into and out of the market, it signaled that the drug may not justify a shift or further analysis. Moreover, it’s important to note that the modeling conducted in this phase deliberately disregarded utilization. Instead, the focus was solely on determining whether there was a significant difference in the cost per unit among pharmaceutical companies. This approach allowed for a clearer understanding of the inherent cost discrepancies and potential implications for optimization strategies.

The next step in preliminary analysis was understanding the differences in average cost per unit for an NDC and determining if the product should be further analyzed. It’s worth emphasizing that these NDCs (National Drug Codes) encompass both Fee-for-Service (FFS) and Managed Care Organization (MCO) utilization. Therefore, minor variations in unit cost may arise if either FFS or Managed Care utilization is absent. To mitigate this, the average between MCO and FFS was adopted consistently in all visual representations and modeling exercises for the corresponding quarter of comparisons.

The drugs listed in the table provided in preface to phase 1 will be the drugs evaluated. Excluding Astrovaratin.

Product Analysis

Methylpenidate 36 MG

Methylphenidate is a central nervous system stimulant commonly used in the treatment of ADHD and narcolepsy. It works by increasing the levels of certain neurotransmitters in the brain. Recognizable products contain this substance include Ritalin, Concerta, and Daytrana to name a few. Methylpendiate 36 Mg is prescribed in the form of a tablet. This drug presented the greatest discrepancy in terms of cost per unit. An estimated $ 5.4 million could be saved if all utilization were to be shifted to the lowest cost NDC.

Pharmaceutical Company Comparison

Active Pharmaceutical Companies

From 2019 to 2022, Trigen Laboratories is the leading distributor of Methylphenidate, commanding over 50% of the market utilization during this period. Despite the entry of new pharmaceutical players such as Camber and Lanner into the landscape, Trigen maintained its stronghold as the primary supplier. The departure of Watson Pharma from the market in 2020 didn’t diminish Trigen’s dominance. Another distributor, Patriot, though not specifically labeled due to formatting constraints, entered and exited the market with minimal utilization within Indiana Medicaid.

It’s evident that Trigen effectively monopolizes the market. Even with the rise in Camber’s utilization following its entry in 2021, Trigen’s market share continued to expand at a notably faster pace. Their utilization shows no signs of interruption, affirming their dominant position in the distribution of Methylphenidate.

Comparison of Unit Costs

• Trigen Laboratories, LLC:
  • Unit costs range from $1.71 to $1.87.
  • Moderate pricing stability suggests strong market presence and consistent utilization within Indiana Medicaid.
• Patriot Pharmaceuticals, LLC:
  • Unit costs range from $0.88 to $0.97.
  • Lower prices observed, but utilization within Indiana Medicaid appears limited, indicating potentially smaller market share or less frequent prescriptions.
• Lannett Company, Inc.:
  • Unit costs range from $1.09 to $2.61.
  • Consistent pricing suggests stable utilization patterns within Indiana Medicaid, with likely steady market share over observed time periods.
• Camber Pharmaceuticals, Inc.:
  • Unit costs range from $2.87 to $3.19.
  • Higher costs observed, potentially indicating increasing utilization within Indiana Medicaid, considering upward pricing trend and potential market expansion.

ANOVA Comparison of Means - All time Periods

Comprehensive Comparison

In our analysis, we took measures to ensure comprehensive comparison and complete data representation. Specifically, we imputed data for Watson Pharma by employing an effective decrease year or a year similar to that of all companies. This approach was adopted to maintain consistency across the dataset and enable a more accurate comparison with other pharmaceutical companies. Additionally, for Camber Pharmaceuticals, we employed a method of imputation that utilized the average for all pharmaceutical companies. By applying a similar effective percentage increase going back, we aimed to maintain coherence in the dataset while providing a basis for comparison with other entities in the market.

Statistical Significance

The ANOVA critical values indicate that there are statistically significant differences in the distribution of Methylphenidate ER 36 mg tablets among the pharmaceutical companies (F(4, 75) = 5.69, p < 0.001). The between-groups variance (130.97) significantly exceeds the within-groups variance (431.32), suggesting that the differences observed in the average tablet distribution across the companies are not likely due to random variation but are instead attributable to genuine disparities between the groups.

Comparison of Mean Values

The data on the distribution of Methylphenidate ER 36 mg tablets among different pharmaceutical companies reveals notable variations in the average distribution of tablets. While Trigen Laboratories, LLC, Patriot Pharmaceuticals, LLC, Watson Pharma, and Lannett Company, Inc. all show relatively similar average distributions, ranging from 2.93 to 3.42 tablets, Camber Pharmaceuticals, Inc. stands out with a significantly higher average distribution of 6.41 tablets. With that information, it would be expected that Camber Pharmaceuticals would have the lowest market share. But as presented, this does not hold true. Camber comprises the second largest in market share for this drug. Additionally, Patriot has the lowest average across all time periods, but is the least utilized company with active utilization from 2021 through 2022.

ANOVA Comparison of Means - 2021 through 2022

We also elected to perform this testing exclusively within overlapping periods from 2021 to 2022, with limited imputation. Focusing on recent periods also enables us to gain insights into current market dynamics and pharmaceutical company performance. Moreover, comparing results with and without imputation allows us to assess the impact of imputation on the analysis outcomes without undermining the validity of earlier imputation efforts. By conducting the analysis in this manner, we ensure that the comparison is based on the most complete and reliable data available, thereby enhancing the robustness and accuracy of our findings.

Statistical Significance

The ANOVA results indicate significant differences in the distribution of Methylphenidate among different groups (F(3, 28) = 6.78, p = 0.001). The between-groups variance (9.14) exceeds the within-groups variance (12.59), suggesting that the observed differences in distribution are unlikely due to random variation but are instead attributed to genuine disparities between the groups.

Suggested Methods for Improvement

We would recommend shifting away from Camber Pharmaceuticals due to the wide range of distribution observed in the ANOVA results, indicating potential variability or inconsistency in their distribution patterns. This variability may introduce uncertainty or unpredictability in medication access and utilization, which could impact patient care and outcomes. On the other hand, Lannett Company, Inc. and Patriot Pharmaceuticals, LLC demonstrate more stable and consistent distribution patterns, as indicated by their narrower distribution ranges. Encouraging additional utilization of Lannett and Patriot could help mitigate potential risks associated with wide distribution ranges, ensuring more reliable access to Methylphenidate for patients within the Medicaid program.

It makes sense that Trigen Laboratories, LLC emerges as the primary distributor of Methylphenidate given its stable and consistent distribution patterns observed across different analyses and time periods. Trigen’s reliability in maintaining a relatively consistent average distribution indicates a well-established and efficient distribution system. This consistency suggests that Trigen has likely developed strong relationships with healthcare providers and institutions, ensuring consistent access to medication for patients within the Medicaid program. At least for this specific drug.

Comparison of Mean Values

Upon comparing the current analysis covering 2021-2022 with the previous examination across all time periods, the dominance of Camber Pharmaceuticals, Inc. as the highest-cost pharmaceutical company for this drug remains evident. However, there has been a noticeable decrease in the average cost, indicating a potential shift in pricing strategies within the market.

Meanwhile, Trigen Laboratories, LLC maintains a consistent average distribution between the two analyses, reflecting stability in their distribution patterns over time. Similarly, both Patriot Pharmaceuticals, LLC and Lannett Company, Inc. demonstrate stable average distributions, raising questions as to why they have not emerged as primary distributors of Methylphenidate within the state Medicaid program despite their consistent performance.

Epinepherine 0.3 MG Autoinject

Epinephrine 0.3 mg autoinjectors are devices used for treating severe allergic reactions quickly. They contain a dose of epinephrine, a medication that helps to open up airways and improve breathing during emergencies like anaphylaxis. These autoinjectors are easy to use and are often carried by people who have allergies that could cause severe reactions.

Pharmaceutical Company Comparison

Active Pharmaceutical Companies

Impax Generics, Mylan Specialty, and Teva Pharmaceuticals are the main distributors of Epinephrine 0.3 mg autoinjectors. Impax Generics stands out as the primary distributor, consistently accounting for over half of the utilization from 2022 and maintaining stable presence over time. This suggests that Impax Generics likely offers a generic version of the product when considering the literal name of the company. For example, a generic version means that it contains the same active ingredient as the brand-name product but is typically more affordable. While Teva Pharmaceuticals is the least utilized of the three, followed by Mylan, the presence of all three companies across all time periods allows for comprehensive comparison without the need for imputation of values.

Comparison of Unit Costs

• MYLAN SPECIALTY L.P.:
  • Unit costs range from $61.39 to $67.14.
  • Moderate pricing stability suggests strong market presence and consistent utilization within the observed time period.
• TEVA PHARMACEUTICALS USA, INC.:
  • Unit costs range from $480.84 to $494.84.
  • Wide range of costs indicates variability and potential for differing market strategies or product offerings.
• IMPAX GENERICS:
  • Unit costs range from $38.02 to $44.85.
  • Relatively stable pricing with lower costs suggests consistent utilization and competitive positioning within the market.

ANOVA Comparison of Means

Statistical Significance

The ANOVA output reveals significant differences between groups (F(2, 45) = 9104.96, p < 0.001), indicating substantial variability in the data. The between-groups variance (2133778.93) significantly outweighs the within-groups variance (5272.95), suggesting that the observed differences are not due to random variation but rather reflect genuine disparities between the groups. This suggests that the groups being compared are significantly distinct from each other in terms of the variable being analyzed. The extremely low p-value (2.07e-59) further supports the notion of significant differences between groups.

Generally, lower average costs per unit, such as that of Mylan Specialty L.P. and Impax Generics, suggest that their products are more affordable compared to those of Teva Pharmaceuticals USA, Inc. This affordability has contributed to their higher utilization rates when compared to Teva Pharmaeuticals. It further poses the question as to why Teva has any presence. Yet, this could be atrributed to contractual agreements, additional rebates, or considerations such as prior authorization. Meaning, that the patient was specifically approved for this drug.

When directly comparing these averages to the market presence, it does appear that cost considerations were explored when deciding which drug is preferred over another. Yet, there is still reason to begin moving increasing utilization from Teva to both IMpax and Mylan.

ARIPIPRAZOLE 15 MG TABLET

Aripiprazole is used to treat specific mood disorders and mental health diagnosis. One common disease treated with this drug is schizophrenia, or bipolar disorder.

Pharmaceutical Company Comparison

Active Pharmaceutical Companies

Early on the the data, Accord healthcare was the only pharmaceutical company in the market for Indiana providing this specific drug. Meaning it is a single source drug. With limited data for Torrent Pharma because their entry only occurred in the start of 2022, it would not pose a significant benefit to conduct statistical analysis for periods prior to 2022. Imputation over a large period of time would not provide reliable enough results to make suggestions in a utilization shift. Yet we can look at the smaller subset of data from 2022Q1 and 2022Q4. Some assumptions can be made in terms of the average cost.

Comparison of Unit Costs

After looking over the unit costs for dates within CY 2022, it is apparent that Accord Healthcare has a much higher cost per unit than Torrent Pharma. Although $1.50 does not seem like a substantial difference between the two drugs, with growing utilization of 60k+ for later quarters, the difference adds up quick. As mentioned, Accord has been the primary distributor. In 2022 we saw increasing utilization for Torrent, this logically makes sense as the drug is lower cost. Additionally, it is likely Torrent Pharma entered the market space for this drug because the units needed increased significantly entering 2022. We see total units for 2022 nearly double from the previous.

T-Test Comparison of Means

Considering that there are only two pharmaceutical companies, a t-test is the most fitting statstical method to assess whether there is a signficant difference between the two distributors. As mentioned, we will not impute values as Torrent was only present for CY 2022. A comparison will be made for all quarters within 2022.

Based on the t-test results for the Aripiprazole 15 mg tablet without imputation, there is a notable difference between the mean prices of the medication from Accord Healthcare and Torrent Pharma. Specifically, the mean price for Accord Healthcare is $4.06, while for Torrent Pharma, it stands at $2.51.

If we conduct a one-tailed test, we are specifically testing if the price of Aripiprazole 15 mg tablets from Torrent Pharma is significantly lower than that of Accord Healthcare. Our null hypothesis is that there is no difference or that Torrent Pharma’s price is equal to or higher than Accord Healthcare’s price. The P-value for the one sided test is .0548. Yes, this is above .05 which is widely used significance level across analysis, given that there are only 4 observations for each company, raising the level of significance is reasonable. This will reduce the risk of ignoring type II errors (Prof. Bouzar class notes) such as the failure to reject a null hypothesis. The higher level of significance allows for more flexibility in the analysis.

Suggested Method for Improvement

Begin shifting additional utiization to Torrent Pharma in the upcoming years. Torrent is a large pharmaceutical company meaning the shift is a feasible option if they are granted the time to increase production of this drug. To facilitate the change, the IHCP could alter the preferred market for the given NDC, meaning switching the NDC to preferred for Torrent’s version and non-preferred for Accord. T

Loratadine 10 MG Tablets

Loratadine is a substance that is used to alleviate the symptoms of allergies. It can also assist in the prevnetion of hives due to an allergic reaction, so slightly stronger than your OTC zyrtec.

Pharmaceutical Company Comparison

Active Pharmaceutical Companies

Perigo NY dominated the market by distributing over 80% of the drug. However, in the third quarter of 2022, Rising Pharmaceuticals quickly emerged and ended up with the highest utilization by the end of the year. Northstar remained consistently utilized across all time periods, while Ohm’s presence faded away after 2020. With this information, multiple ANOVAs will be performed with different methodologies that will be explained at each indivdual step.

Average Unit Cost

Northstar and Perigo have the highest unit cost in comparison to other companies. Major Pharmaceuticals has one of the lowest unit costs, but is one of the least utilized companies that are dispensing the drug. We also saw the extreme increase in utilization for Rising Pharmaceuticals mid-year 2022. During this time period, their unit cost is the lowest of all distributors and their quick emergence would suggest the desire to utilize the lower cost NDCs.

ANOVA Comparison of Means

Imputation - All time periods

For the first round of testing, we elected to impute missing values for all pharmaceutical companies. There is enough data present thoughout all time periods that we thought it was reasonable to assume that the missing companies would be within range of the others. We did employ different techinques if there was limited unit costs. For example, Bluepoint Labs was not present from 2019-2021Q2. We used the average of the most utilized pharmaceutical companies to populate their unit costs for missing date. Major Pharamceuticals only had 4 missing observations, so we just utilized their average cost per units over all time periods to populate. We followed a similar methodology for Ohm pharamceuticals because for over half of all time periods they were utilzed.

Statistical Signifiance

This ANOVA table shows the analysis of Loratadine 10 mg tablet unit cost among different pharmaceutical companies. The the between group variation indicates the differences in unit cost among these companies, which is statistically significant with an F-value of 7.64 and a very low p-value. This suggests that there are significant differences in unit costs among the companies.

Perigo NY: - Consistently low pricing implies competitive pricing strategies or efficient production methods, likely leading to high utilization rates.

• Rising Pharmaceuticals:
  • Swift entrace and possesses the lowest average unit cost
• Northstar RX LLC:
  • Stable and low pricing indicates competitive positioning and possibly high utilization rates.
• Ohm Pharmaceuticals, Inc.:
  • Higher unit costs may lead to lower utilization rates - we saw their exit in 2020
• Bluepoint Laboratories:
  • Moderate pricing stability suggests competitive positioning in the market with potential for consistent utilization rates.

The interesting part regarding this portion, is that overall time periods, Perigo NY had one of the lowest unit costs, yet in 2022 for a majority of the year they had the second highest cost amongst pharmaceutical companies. It appears that the entrance of Rising Pharmaceuticals was related to Perigo NYs increasing drug prices as they became the majority distributor later in the year.

Suggested Improvements

Based upon our analysis of this specific drug, their seems to be reasonable utilization shifting away from the higher cost NDCs. Thus, it would be suggested that the IHCP/FSSA continue with the suggested utilization patterns and using Rising Pharmaceuticals as a primary distributor for Loratadine 10 mg.

Oseltamivir

Oseltamivir is primarily used to treat flu like symptoms as an anti-viral medication.

Pharmaceutical Company Comparison

Looking closely at how this drug is used, it’s clear it’s not a great fit for statistical analysis. Alvogen is a primary distributor of this drug and it seems to be predominately used during COVID times as it is an anti-viral. We see a large spike during 2020 where utilization rose to over 2 million units, whereas prior to the pandemic, it hovered around 100k. Utilization is also higher during later quarters of the year (flu season). Thus, it is likely that this drug possesses seasonality. Even so, we can visually assess the differences in unit costs and determine if we could suggest a favored pharmaceutical company for the . Performing statistical analysis with so many missing values would not pose any validity as we either have to impute 80% of the data, or just observe two quarters of data. Thus, visualize analysis will be conducted. A potential conclusion here would be to suggest a favored company, as utilization across all remain at relatively similar levels from 2022Q3 to 2023 Q4. This will be using only data from 2022 as it is more complete.

Average Unit Cost

There is not a drastic difference in the unit costs for this drug. In addition, the lower and higher cost drugs seem to change quarterly. This drug likely appeared in the largest fiscal impact if utilizing 100% of the low cost drug because of the great amount of utilization in 2020 and limited availability due to the cranking up of production. Hence shifting to the cost drug, even if the difference was small, would present a substantial impact since we were using the overall average across all time periods. During 2020 we saw highest drug prices, and greatest utilization.

Suggested

Continue evaluating the tendencies in pharmaceutical pricing, make attempts to utilize the lower cost drug for the quarter given the seasonlity of utilization.

Albuterol

Albuterol is a substance that is found inhalers, this drug is primarily used to treat the symptoms of asthmatic episodes.

Pharmaceutical Company Comparison

Active Pharmaceutical Companies

There are two key players in the distribution of Albuterol Sulfate, Prasco Laboratories and Teva Pharmaceuticals both exert a consistent and dominant presence and maintain a primary hold on the market across all observed time periods. However, alongside these six other pharmaceutical companies emerge with smaller utilization levels. Despite the formidable market presence of Prasco Laboratories and Teva Pharmaceuticals, the existence of these other companies signify a diversification in the supply chain, this likely enhances patient access to these drugs amidst high demand.

Comparison of Unit Costs

There is a wide range of unit costs across 2022. The visual displays the unit costs for all companies with utilization for 2022. We see that Prasco Labs has one of the lowest cost throughout, and remains relatively consistent in pricing throughout the year. Hence, it makes sense that this is one of the most utilized companies. On the otherhand, Teva rounds out at one of the higher cost NDCs within the description. We see great variability in unit costs for most other drugs so their lower levels of utilization tie out to expectations. Another unusual feature is the growth of Cipla in 2022 and their cost. Based on the visuals, it would be suggested to shift away from both Cipla and Teva.

ANOVA - Comparison of Means

ANOVA 2020Q2 - 2022 Q4

We elected to create a subset of data for this product because there are fewer pharmaceutical companies with utilization in 2022. Considering the shifting methofology explained at a later stage, CY 2022 dates of service are the primary concern for evaluation. Thus, we utilized the data for these periods for the four companies with utilization from 2020Q2 and 2022Q4.

Statistical Significance

This ANOVA table shows the analysis of ALBUTEROL HFA 90 MCG INHALER unit cost among different pharmaceutical companies. The the between group variation indicates the differences in unit cost among these companies, which is statistically significant with an F-value of 2.839 and a very low p-value. This suggests that there are significant differences in unit costs among the companies.

We see that Prasco laboratories has the absolute lowest unit cost across all companies and the lowest variance. This would mean not only is the cost low, but the price of the drug for this company has remained stable throughout all quarters of utilization. When compared to Cipla, we see a unit cost that is literally double, and a very high variance in comparison to the unit cost. One thing to note about the unit costs, is the variance of the primarily used companies are the lowest. Meaning, Teva’s utilization may be related to needed to supplement the market. Yet, encouraging the use of a lower cost NDC lik from Lupin would contribute to lower spend and is suggested.

T-Test Prasco v Teva

This analysis was performed because Teva and Prasco Labs are the primary distributors Teva posseses a unit cost that is near double the average unit cost when compared to Prasco Laboratories. Based upon this information it will be suggested that additional utilization shifts to Prasco labs.

Suggested Method for Improvement

Encourage the continuity and trust in Prasco Laboraties and make additional efforts to shift utilization away from Teva into Lupin and Prasco Labs.

Duloxetine

Duloxetine is a substance that is commonly used to treat anxiety and depression. It can also be used to help relieve nerve pain.

Pharmaceutical Company Comparison

Active Pharmaceutical Companies

There are three primary distributors for this drug, and 4 that are present post 2019. Breckenridge holds a majority of the market when compared to others, followed by Citron and Ajanta. They are quite stable in their utilization meaning none are increasing or decreasing their presence.This would indicate that Breckenridge is the favored product to be used, and the others are present to increase the supply chain.

Unit Cost Comparisons

During 2022, Breckenridge is the lowest cost NDC of all within the market. Given this an their presence, it does not pose benefit to analysis to assess these values statistically. Other pharmaceutical companies posess relatively the same unit cost quarter over quarter and the cost is stable for all opharmaceutical companies. Based upon the visuals for this particular product, that utilization be shifting accordingly throughout quarters to the lower cost of Ajanta and Citron. Thus this would mean that the shifting methodology would have to be conducted on a quarterly basis rather than comprehensive time periods like a calendar year. Quarterly shifts are reasonable as the drug utilization review board meets on a quarterly bases to determine which drugs should be preferred and non - preferred. Hence this would be a reasonable decision to be made regarding this drug.

Further Analysis

Due to the validity of the statistical exercises, at this stage we can confirm that pharmaceutical companies are NOT providing prescription drugs at the same pricing level. Although this pattern was not present for Duloxetine, there are still lower cost NDCs available, but it was not a primary distributor. Through these findings, it encouraged a shift in methodology in the shifting metrics. Meaning we will not shift all the utilization to the lowest cost NDC because there are occasions, like Duloxetine where the lowest cost NDC is not utilized.

We have completed the statistical analysis for the remaining products, but based upon the previous results, it is not necessary to include the findings. We can move forward with suggested methodologies and observe the remaining products visually. The project submission contains the modelling conducted on these products.

Shifting Utilization

Subsetting and Analysis of Pharmaceutical Data for 2022

In this analysis, we aim to stratify and optimize pharmaceutical utilization data for the year 2022. The goal is to redistribute utilization among NDCs (National Drug Codes) to optimize costs while ensuring that total units remain constant.

1. Subsetting Data

• We first isolate data for the year 2022 and stratify it by Quarter, NDC, FFS/MCO, and Unit cost.
• NDC descriptions with only one unique NDC are excluded from further analysis to focus on meaningful variations.

2. Sorting and Grouping

• Data is sorted and grouped by NDC Description, Quarter, Utilization Type, and Unit Cost.
• This facilitates ranking of unit costs from smallest to largest within each NDC Description and Quarter combination.

3. Identifying Key Metrics

• Minimum, Median, and Maximum unit cost rankings are identified for each NDC Description.
• For instance, if an NDC Description has 5 unique NDCs, the Minimum would be 1 (lowest unit cost), Median would be 3, and Maximum would be 5.

4. Determining Market Weight

• Market weight for each NDC within its NDC Description is calculated by summing total units for the unique NDC and dividing by the total for that NDC Description in the quarter.

5. Normalizing Values

• Utilization for NDCs with unit costs at or below the median is normalized.
• Total weight for NDCs with lower unit costs is summed, and each NDC’s weight is divided by this total to determine the percent of utilization to shift.

6. Reducing High Cost Utilization

• Utilization for NDCs with unit costs above the median is reduced by 90%.
• This reduction is justified by the assumption that utilization can be shifted to lower-cost NDCs within the same NDC Description.

7. Adding Utilization

• Utilization percentages obtained from normalization are used to redistribute utilization among NDCs below the median unit cost.

Key Considerations

• Total units across 2022 remain constant throughout the analysis. Any changes would indicate an increase in utilization.
• While these methods optimize utilization at a granular level, they may not be perfect due to data limitations and lack of higher-level information such as GPI-14.
• Median unit cost values may closely align with script-level averages, enhancing the reliability of our analysis.

Through this comprehensive approach, we aim to efficiently optimize pharmaceutical utilization, ensuring cost-effectiveness without compromising on quality or availability.

Example of shifts

The below table described the methods taken. There are a few intermediary calculations to go between steps, but the methods highlighted above.

CLOPIDOGREL 75 MG TABLET Shift - This example provides only the MCO utilization. - Utilization was only removed from the highest cost NDCs within the group description. - Utilization is shifted into the lower cost NDCs by their weight within their group

Drugs with Highest Fiscal Impact

After conducting our analysis, we observed that only five of the drugs we initially anticipated to have the highest fiscal impact actually made it to the top 10 after the utilization shift. Surprisingly, the impact wasn’t as significant as we had initially projected.

How do these results relate to out prior analysis and comparison of means? We found that, in most cases, both the maximum and minimum cost scripts were not utilized as frequently as those drugs hovering around the median cost. Therefore, when redistributing utilization away from higher-cost drugs, there wasn’t as much movement toward the absolute minimum cost per unit NDC. This discrepancy emphasizes a critical factor: pharmaceutical companies may not be able to immediately ramp up production if their presence in the market is limited. Consequently, any transition from higher or mid-level costs to lower-cost NDCs would need to occur gradually, allowing for the opportunity to increase production capacity.

Furthermore, we noticed less variance in price for the mid-level NDCs. This could explain their more prominent presence and reliability, instilling greater confidence among providers and managed care entities. Their established reputation and reliability likely contribute to their increased utilization despite not being the lowest-cost options available.

By recognizing these nuances, we can refine our approach to utilization shifting, ensuring a more realistic and effective strategy that acknowledges the complexities of pharmaceutical market dynamics. This understanding will ultimately lead to more informed decisions and greater cost savings in the long run.

Shift of Initial Top Drugs of Utilization

Epinephrine .3 MG Auto Inject

As previously analyzed, Teva Pharmaceuticals stood out with the highest cost, reaching nearly $400 per unit, while Impax Generics maintained a more moderate unit cost around $67. Given the lower levels of utilization for this drug, a strategic decision was made to reallocate resources away from both Mylan and Teva Pharmaceuticals. A significant 90% of their utilization was redirected towards Impax Generics. This shift is not only feasible but also strategically advantageous, considering Impax Generics’ established dominance in the market. Consequently, the anticipated increase in utilization for Impax Generics is expected to be comfortably accommodated in the upcoming quarters.

Methylpenidate 36 MG

For Methylphenidate 36 mg, it became evident that Camber Pharmaceuticals significantly led in terms of unit cost. Considering the nature of this drug and the associated costs, a redistribution strategy was implemented among the next three lowest-cost pharmaceutical companies, aligning with their current market share. Trigen Laboratories experienced the most substantial increase, given their already significant presence and primary distribution role for this product, essentially holding a monopoly. Their consistent pricing stability year over year and minimal variance establish them as a dependable and preferred source for utilization. Despite Patriot boasting lower overall costs, their current market presence doesn’t warrant a complete transition to their product.

Methylpenidate 54 MG

Methylphenidate presents a scenario where several pharmaceutical companies offer comparable products, with significant fiscal implications. Once more, Camber emerges as one of the costlier providers. Consequently, the utilization previously allocated to Camber was redirected to Trigen.

Largest Pharmaceutical Company Shift

The following assessment examines the most substantial decreases and increases in utilization by pharmaceutical companies. A significant decline in utilization for a particular company suggests a considerable presence of NDCs with high costs per prescription. Teva Pharmaceuticals, Zydus, Cipla, Rising, and Breckenridge were all previously highlighted in our analysis of unit costs, representing drugs with high associated costs that underwent further scrutiny. If future rate-setting periods prioritize pharmaceutical budgeting, it would be prudent to significantly reduce the market share of these companies within the state of Indiana. Alternatively, another potential solution involves encouraging Managed Care Organizations to negotiate better rates that align more closely with median drug prices.

Notable Decreases

Conversely, the shift towards higher post-shift units among these pharmaceutical companies signifies a positive trend towards lower-cost providers. Notably, Citron Pharmaceuticals and Mylan Pharmaceuticals, Inc. were both present in prior analyses, representing lower-cost NDCs at a product level. This observation aligns with the significant increase in post-shift units for these companies. The absolute and percent differences illustrate substantial increases in utilization, indicating a strategic shift towards more cost-effective options. This shift highlights a potential strategy to optimize pharmaceutical budgeting, with an emphasis on maximizing cost savings while maintaining quality patient care through feasible methods.

Notable Increases

After applying our shifting methods, we identified several pharmaceutical companies that experienced the most significant increases in unit distribution. This suggests that these companies likely offer lower-cost options for the medications in question. It would be prudent to consider fostering closer relationships with these pharmaceutical companies. This could involve establishing regular communication between the front office staff within the IHCP and FSSA. Strengthening these connections can facilitate better negotiation opportunities and ensure timely access to cost-effective medications, ultimately benefiting both patients and healthcare providers.

Phase 2: NADAC Assessment

Methodology

In this stage of the analysis, we’ll streamline our approach for a simpler assessment. A key consideration is the mandatory requirement of paying at a maximum of the NADAC as per Medicaid guidelines within the state of Indiana. Actually there are more pricing metrics that could be considered, but the NADAC is representative of the National Average Drug Acquisition Cost, which tends to be lower than both the State Maximum Allowable Cost and the Federal Upper Limit (which are also repricing metrics). Information has been provided to us from Optum, that INdiana seems to be closely following the INMac but limited analysis has been conducted in how these fields compare to the NADAC. (The comment in these parenthesis will be removed, but I am actually conducting an analysis on this at work at a more granular level. Which has been interesting to see some of this CAPstone come to life, we have the ability for a higher level of granulrity and additional ways to group drugs, but.. )

However, it’s important to note that we lack information on whether dispensing fees are included in the total paid amounts. Consequently, the repricing impacts we observe here might be slightly inflated, although this inflation would likely be minimal. This is particularly relevant since our analysis operates at a unit level rather than a prescription-based level. While dispensing fees may vary by drug type, they generally constitute only a small portion of the overall prescription cost. Total units are nearly 400% higher than that of total prescriptions.

Description of Method

This analysis involved a straightforward calculation and summarization of the highest fiscal impacts. By multiplying the NADAC price, provided per unit, by the total amount of utilization, we obtained the total NADAC Repriced Field. Subsequently, another addition field was generated, containing the minimum value between the NADAC and the total paid amount. Typically, these values are expected to be fairly proximate, reflecting the inherent balance between the NADAC pricing and the actual expenditure. Yet, there are specific products that contain a much higher total paid, than what should have been paid.

Repricing Impacts: Top Products

The table below presents the drugs with the most notable contrast between the total amount reimbursed and the NADAC repriced. Interestingly, five of these ten products coincide with those demonstrating the largest difference between the minimum and maximum cost per unit. This observation suggests a potential link, indicating that pharmaceutical companies providing lower unit costs for a specific drug tend to converge more closely with the NADAC amount during the given time frame. This hypothesis merits additional scrutiny and exploration.

Average CPU versus Average NADAC for 2022

The stark contrast between the Paid CPU and the NADAC CPU for these products indicates that Indiana Medicaid is overpaying significantly. The percentage differences, ranging from -52% to -97%, illustrate the extent of the discrepancy, with some costs being less than 95% of what Indiana Medicaid is currently paying. Such substantial variations emphasize the need for a thorough review and potential renegotiation of pricing agreements to ensure cost-effectiveness and optimal allocation of resources within the healthcare system.

Phase 3: Generation of Forecasts

Methodology

As promised, we are predicting future costs for the next year with multiple forecasts. Although this data is from sometime in the past, it will highlight how changes in policy and approach to dispensing habits can change the landscape for years to come and provide more sustainable future costs. At this stage we will compare costs after incorporating the shift in utilization, incorporating the NADAC repriced field, and adjusted total expenditures after incorporating these adjustementts. In total, four forecast will be generated utilizing the auto.arima function in R. This function is a reliable approach because it automatically selects the best ARIMA model based on statistical criteria, making it well-suited for time series forecasting tasks like ours. We are not using additional fields to predict future costs, solely the total dollars from previous years. ARIMA models can capture a wide range of patterns, and do not require the mean and or variance of the data to remain constant overtime. The auto.arima function automatically selects the best model by evaluating multiple models based on AIC and also BIC. Although, this is a simplistic way, the ultimate goal is not to say your expenditures will be this much

Understanding Fiscal Impacts at all Stages

The following analysis compares the original total reimbursed amount with the adjusted utilization paid amounts, the NADAC repriced amount, and finally, the combined utilization and repriced NADAC. This comparison offers insights into how each adjustment affects the overall reimbursement amounts.

Utilization Shift: Impact on Paid dollars

The shift in utilization had a minimal impact on total paid dollars, with the analysis conducted solely over the course of 2022. The points annotated on this visualization represent mid-year 2022. The difference between the original paid amount and the adjusted paid amount falls just short of $2 million. If this utilization shift had been implemented at the beginning of 2022, total savings would have ranged between $3 and $4 million. While these figures may seem surprisingly low, it’s important to note, as mentioned in prior analyses, that completely shifting all utilization to lower-cost NDCs lacking market presence is not feasible. However, the application of this principle, alongside the NADAC repricing adjustment, promises to yield intriguing results.

NADAC Repricing Comparison

Implementing the repricing guidelines mandated by the state results in a significant decrease in the total amount reimbursed. The adjusted paid amount through 2022Q2 stands at $296 million, which is nearly $28 million less than the total reimbursed at that point within the year. With a clear understanding of the total fiscal impacts, it becomes evident that more efforts are needed to assess the contracted amounts and the variation between the NADAC unit cost and the amount that the Indiana Medicaid program is currently paying for drugs. Adhering to these metrics for pricing all drugs year over year could alleviate much of the financial burden caused by rising pharmaceutical expenditures.

Combined Adjustment Impact

Incorporating the new utilization model for 2022 reveals the potential for an additional $13 million in savings. This emphasized the effectiveness of implementing both the repricing guidelines mandated by the state and the utilization shifts towards drugs with lower unit costs. The adjusted paid amount through 2022Q2 is now $283 million, which significantly lower than the total reimbursed amount at that point in the year or $324. This significant decrease in total reimbursement emphasizes the need to assess contracted amounts and the variance between NADAC unit costs and current payments by the Indiana Medicaid program for drugs. The observation that pharmaceutical companies charge different amounts for the same drugs, coupled with the lower NADAC unit costs for drugs with shifted utilization, highlights the significance of implementing both methodologies. Doing so could lead to substantial cost savings and again alleviate the financial burden caused by rising pharmaceutical expenditures.

Adjusted Paid Forecast

Total <- NADAC_Correction %>%
  group_by(YEAR_QUARTER) %>%
  summarise(Sum_Total_Amount_Reimbursed = sum(TOTAL_AMOUNT_REIMBURSED))

Total$YEAR_QUARTER <- as.yearqtr(Total$YEAR_QUARTER)

total_ts <- ts(Total$Sum_Total_Amount_Reimbursed, frequency = 4, start = c(2019, 1), end = c(2022, 4))


fit <- auto.arima(total_ts)


forecast_result <- forecast(fit, h = 4)

print(forecast_result)

##         Point Forecast     Lo 80     Hi 80     Lo 95     Hi 95
## 2023 Q1      346243045 329544481 362941608 320704796 371781293
## 2023 Q2      354048285 330432950 377663620 317931747 390164823
## 2023 Q3      361853526 332930766 390776286 317619981 406087070
## 2023 Q4      369658766 336261639 403055893 318582269 420735263

The table provides point forecasts along with lower and upper bounds at 80% and 95% confidence intervals for the year 2023, broken down by quarters (Q1, Q2, Q3, and Q4).

For example, in Q1 2023, the point forecast for the total amount reimbursed is approximately $346.24 million, with an 80% confidence interval ranging from $329.54 million to $362.94 million and a 95% confidence interval ranging from $320.70 million to $371.78 million.

Similarly, for Q4 2023, the point forecast is approximately $369.66 million, with an 80% confidence interval ranging from $336.26 million to $403.06 million and a 95% confidence interval ranging from $318.58 million to $420.74 million.

This is quit a large range of values, but this is expected. As we look at the trend year over year, this is a large amount of variance in the total paid. Although it is consistently trending upwards, there are signficant drops and rises quarter over quarter. But we can confidently say that expenditures will continue to rise.

autoplot(forecast_result) +
  ggtitle("Forecast of Total Amount Reimbursed") +
  xlab("Year Quarter") +
  ylab("Total Amount Reimbursed") +
  scale_y_continuous(labels = scales::label_number_si(scale = 1e-6, suffix = "M"), expand = c(0, 0))

## Warning: `label_number_si()` was deprecated in scales 1.2.0.
## ℹ Please use the `scale_cut` argument of `label_number()` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.

The above visual displays the forecasted amounts through 2023. As expenditures are predicted to increase to up 420 million dollars for 2023, it is pertinent that adjustment are made with the current pharmaceutical budgeting, by both adjusting utilization and addressing the disparities between the NADAc unit cost, and the Medicaid paid unit cost.

Adjusted Utilization

Adj_Paid <- NADAC_Correction %>%
  group_by(YEAR_QUARTER) %>%
  summarise(`Adjusted Paid` = sum(Adjusted_Paid))

Adj_Paid$YEAR_QUARTER <- as.yearqtr(Adj_Paid$YEAR_QUARTER)

total_ts2 <- ts(Adj_Paid$`Adjusted Paid`, frequency = 4, start = c(2019, 1), end = c(2022, 4))

fit <- auto.arima(total_ts2)

forecast_result2 <- forecast(fit, h = 4)

print(forecast_result2)

##         Point Forecast     Lo 80     Hi 80     Lo 95     Hi 95
## 2023 Q1      344515523 328013863 361017184 319278411 369752636
## 2023 Q2      352212794 328875921 375549666 316522127 387903460
## 2023 Q3      359910064 331328349 388491779 316198104 403622025
## 2023 Q4      367607334 334604013 400610656 317133110 418081559

Comparing the new forecasted values to the previous ones, it’s evident that adjusting utilization by shifting to NDCs with lower unit costs has resulted in lower projected total reimbursed amounts for each quarter in 2023.

• In Q1 2023, the point forecast decreased from approximately $346.24 million to $344.52 million.
• In Q2 2023, the point forecast decreased from approximately $354.05 million to $352.21 million.
• In Q3 2023, the point forecast decreased from approximately $361.85 million to $359.91 million.
• In Q4 2023, the point forecast decreased from approximately $369.66 million to $367.61 million.

Overall, these adjustments have led to a reduction in the projected total reimbursed amounts across all quarters, contributing to cost savings for the Indiana Medicaid program, but still limited to what was initially expected when pursuing this analysis.

autoplot(forecast_result2) +
  ggtitle("Forecast of Impact of Adjusted Utilization on Total Paid") +
  xlab("Year Quarter") +
  ylab("Total New Amount Reimbursed") +
  scale_y_continuous(labels = scales::label_number_si(scale = 1e-6, suffix = "M"), expand = c(0, 0))

As mentioned, there was a limited impact on total paid dollars decrease. Yet, we do see a slight decrease in the range of confidence intervals displayed. There is a little less variation in total expenditures when adjusting the utilization to lower cost NDCs. This is likely due to less severe increase in total expenditure from the end of 2021Q4 to the beginning of 2022Q1.

Repricing to NADAC

Adj_NADAC <- NADAC_Correction %>%
  group_by(YEAR_QUARTER) %>%
  summarise(`NADAC Repriced` = sum(Adjusted_NADAC_MIN))

Adj_NADAC$YEAR_QUARTER <- as.yearqtr(Adj_NADAC$YEAR_QUARTER)

total_ts3 <- ts(Adj_NADAC$`NADAC Repriced`, frequency = 4, start = c(2019, 1), end = c(2022, 4))

fit <- auto.arima(total_ts3)

forecast_result3 <- forecast(fit, h = 4)

print(forecast_result3)

##         Point Forecast     Lo 80     Hi 80     Lo 95     Hi 95
## 2023 Q1      317108742 301214141 333003344 292800047 341417437
## 2023 Q2      323093089 300614728 345571450 288715402 357470775
## 2023 Q3      329077435 301547178 356607693 286973540 371181331
## 2023 Q4      335061782 303272579 366850985 286444391 383679172

• In Q1 2023, the original repriced point forecast was approximately $346.24 million. With the NADAC CPU repricing, it decreased to $317.11 million.
• In Q2 2023, the original repriced point forecast was approximately $354.05 million. With the NADAC CPU repricing, it decreased to $323.09 million.
• In Q3 2023, the original repriced point forecast was approximately $361.85 million. With the NADAC CPU repricing, it decreased to $329.08 million.
• In Q4 2023, the original repriced point forecast was approximately $369.66 million. With the NADAC CPU repricing, it decreased to $335.06 million.

Utilizing the NADAC CPU to reprice claims has consistently resulted in lower projected total reimbursed amounts compared to the original repriced values across all quarters. This underscores the effectiveness of leveraging accurate pricing information to optimize budget allocation and minimize expenses for the Indiana Medicaid program.

autoplot(forecast_result3) +
  ggtitle("Forecast of NADAC Repriced") +
  xlab("Year Quarter") +
  ylab("Total New Amount Reimbursed") +
  scale_y_continuous(labels = scales::label_number_si(scale = 1e-6, suffix = "M"), expand = c(0, 0))

We see that the maximum dollar amount present in the forecast does not exceed 400 million dollars. Meaning thus far, this adjustment is far more impactful than the utilization. Meaning if one aspect of this analysis was focused upon, it would be ensuring that all pharmaceutical companies and Managed care organizations are doing there part in contracting lower costs for these drugs that are in line with state Medicaid guidelines. One method is by performing more rigorous analysis. There is an assumption at OptumRx, which is the states pharmacy benefit manager, that the Indiana maximum allowable cost is the standard and are set at the lowest rate. Yet, after completing this analysis, it is proven that this assumption does not hold true. Rigorous comparison of NADAC and INMAc have become a necessity year over year.

Combined Impacts

Combined <- NADAC_Correction %>%
  group_by(YEAR_QUARTER) %>%
  summarise(`Combined Impact` = sum(Adjusted_NADAC_NEW_UTIL))

Combined$YEAR_QUARTER <- as.yearqtr(Combined$YEAR_QUARTER)

total_ts4 <- ts(Combined$`Combined Impact`, frequency = 4, start = c(2019, 1), end = c(2022, 4))

fit <- auto.arima(total_ts4)

forecast_result4 <- forecast(fit, h = 4)

print(forecast_result4)

##         Point Forecast     Lo 80     Hi 80     Lo 95     Hi 95
## 2023 Q1      295005079 277115477 312894681 267645293 322364865
## 2023 Q2      295005079 269705361 320304797 256312499 333697659
## 2023 Q3      295005079 264019379 325990779 247616540 342393618
## 2023 Q4      295005079 259225874 330784284 240285508 349724651

It’s intriguing to note that after incorporating the new utilization by shifting to lower-cost NDCs, the predicted costs remain relatively flat for 2023, with minimal variance within the confidence intervals. This stability is quite remarkable and suggests a certain level of predictability in the forecasted expenses.

However, achieving such a goal would undoubtedly be challenging, as resources need to be allocated across various priorities. As a result, the impacts of such adjustments are likely to occur more gradually over time. Nevertheless, this observation highlights the potential for significant cost savings and the importance of ongoing efforts to optimize utilization and pricing strategies within the healthcare system.

autoplot(forecast_result4) +
  ggtitle("Forecast of NADAC Repriced") +
  xlab("Year Quarter") +
  ylab("Total New Amount Reimbursed") +
  scale_y_continuous(labels = scales::label_number_si(scale = 1e-6, suffix = "M"), expand = c(0, 0))

The stability observed in the forecasted values after implementing shifts in utilization patterns and adhering to NADAC repricing guidelines presents a promising potential for substantial cost savings. In fact, this stability could potentially result in savings of up to $120 million in a single calendar year. However, it’s essential to acknowledge that immediate shifts in utilization and complete repricing of all NDCs to the NADAC may not be entirely feasible, particularly if the INMAc exceeds this metric.

Nevertheless, the opportunity to achieve a significant reduction in expenditures by as much as 25% is undeniable. This becomes especially crucial amidst the backdrop of rising medical costs, increased drug utilization, and the incorporation of additional covered services within the Medicaid program. While realizing these savings may require phased implementation and careful resource allocation, the potential for substantial cost containment underscores the importance of continued efforts to optimize spending habits.

The Analytics Solution

Restating the Business Problem

The escalating costs of pharmaceuticals pose a significant challenge for the Indiana FSSA and IHCP. Despite efforts to contain expenses within the Medicaid program, the continuous rise in drug prices strains the allocated budget and threatens patient access to essential medications. To address these issues effectively, the FSSA must implement new strategies that prioritize cost containment and patient welfare while optimizing managed care. Finding solutions that strike a balance between these obstacles is crucial for the long-term success of the Indiana Medicaid program.

The HealthOptimize Solution

Utilization Shift: Shocking Revelations

We at HealthOptimize, have evaluated the rising pharmaceutical costs from various standpoints. Our first method was assessing these costs a product level basis. Our goal was to determine two things here, one if there us utilization of higher cost NDCs. Although the answer to this question is yes, there is higher cost NDC utilization across nearly every drug. Yet, what we found is that higher cost NDC use is not always a bad thing. Hence we elected to shift most of the utilization away from the highest cost NDC, but to ensure the validity and sanctity of our results, we shifted the utilization away from these drugs based upon the market share of the drugs that fell at or below the median cost. In turn, we found that a utilization shift away from these higher cost NDCs would have a limited impact on total expenditures. The total dollar amount saved in a calendar year was minimal. This was definitely a shock, after preliminary analysis provided a number that could reduce pharmaceutical expenditures by nearly 50 million dollars in a single CY.

The primary driver for this was the inconsistency of market presence for the lower cost NDCs. This could be related to the size of the pharmaceutical company and that they are only present to offset the chance of market shortages for a specific drug. Furthermore, upon conducting an analysis of the variance in unit costs and utilization patterns over time, a notable trend emerged: drugs that exert significant market influence tend to exhibit stable pricing across all time periods. In essence, while utilization rates may increase over time, these key market players demonstrate a consistent growth trajectory in overall costs.

NADAC Repriced: Not Quick, but Integral

When first beginning the analysis, it was unknown the impact that repricing to the NADAC per unit would have on total expenditures. One aspect, that is important to note when we begin implementing a solution, is that pharmacy unit costs have never been repriced. The reasoning behind this is the fact that MCEs have full disclosure and to contract rates with the pharmaceutical companies for these drugs. That is why we will see NADAC repriced dollars that are sometimes greater than that of average unit costs. Which is why when conversing with our counterparts, we were shocked to learn that the validity of unit costs had never been assessed. Hence, moving forward, HealthOptimize suggests that the IHCP and FSSA provide MCEs with a deadline to meet the unit cost requirements or penalties will be assessed. This would be a similar penalty structure that MCEs would face if they do not comply with the preferred drug list. We would not expect total expenditures to decline by 70 million for the next calendar year, but we would expect them to be reduced year over year, and slow the increasing trend in pharmaceutical spend.

Combinatory Approach: -

AS mentioned, it will be a very daunting task to incorporate both a shift in utilization and ensuring that drugs are priced according to state guidelines at the same time. Although it may be somewhat of a shock, the shifting of utilization to various products with lower unit costs will be the easiest to handle. We did not have access to the preferred drug list and exact NDCs contained on the file, but it is easily accessible for members of the FSSA and IHCP. HealthOptimize modelling could could be used to supplement how these lists are decided. Essentially, the goal of the PDL is to manage costs effectively and also ensure patient access to drugs. These files are updated a weekly basis and also submitted to healthcare providers and dispensers of drugs on the same timeline. Meaning these changes could happen rather quickly. Moving the below median cost drugs and altering their status to preferred, and the maximum cost drug to Non - Preferred would ensure the desired utilization shift would happen. We saw that incorporating the utilization shift and again repriced to NADAC netted an additional 13 million dollars saved in a calendar year. Although too many changed cannot be made too soon, implementing all utilization shifts within 2-3 months would be a reasonable expectation.

Limitations

Limitations of Analysis

There are limitations to the analysis that was provided. The first of which being the lack of availability of a preferred drug list. The preferred drug list is a method in which states use to encourage prescribing habits in a direction that they desire. If this was available, we would first have a better understanding of which drugs are actually preferred and additionally being able to make suggestions to change specific products from preferred to non-preferred.

There were uncertainties regarding what is included in the total amount reimbursed. For example, we are unable to determine the number of claims that contain third party liability. In simplest terms, third party liability indicates that there is another insurer paying for a portion of the claim. In most cases, when this occurs, Medicaid is not the primary payer, they only pay the remaining balance of the claim after the primary insurer pays for this claim. This does have the potential to be imapctful for analysis, because these claims are not included in other pharmaceutical companies.

Another limitation of this analysis, as mentioned in previous steps, was the availability of Medi-Span data. Medi-Span produced another identifier for drugs called the Generic Product Identifier, which is a 14 digit code that describes different the different componets of the drug. The first two digits identify the drug group, the first 4 are the class, the next 6 are a subclass. Each step becomes more descriptive. If we had this information, we would be able to identify specific classes, and groups of drugs with the largest fiscal impact instead of assessing individual products. This would have broadened the scope of the analysis. Additionally, it is not guaranteed that two NDCs that possess the same NDC description are the exact same drug. Having the additional GPI-14 is another means of confirmation that these drugs are comparable and utilization can be shifted. Although in most cases this is the case, but for complete accuracy in shifting model a GPI-level comparison is preferred.

Improvements

The preferred drug list is available through the Indiana FSSA website, but the NDC level descriptions are not available. For a given NDC description there could be multiple products and only certain drugs within that description would be preferred. Thus, requesting this information directly from the state would be the first step in improving this analysis. This would allow us to suggest certain drugs to be moved to the preferred drug list if they are not on the list already, which would directly impact the dispensing habits immediately. These files are updated on a weekly basis and MCEs are expected to send these to providers each week.

Medi-Span data is available to be purchased online. Given the budgeting constraints for this project, HealthOptimize was unable to obtain this data. If provided the opportunity to continue this analysis, the purchase of this data would be a necessity. It would open up the door to understand the dispensing habits presence additional scenarios such as differentiating between brand and generic utilization.