Using prescription medication utilization trends to evaluate health system resilience and inform future emergency preparedness

Abstract

BACKGROUND:

Monitoring prescription medication utilization can serve as a powerful indicator of health system resilience and vulnerabilities during public health emergencies. By examining shifts in medication fills, policymakers and stakeholders can identify both strengths and weaknesses in access to care for vulnerable populations.

OBJECTIVE:

To evaluate health system responses during the COVID-19 pandemic and inform future preparedness strategies during public health crises using trends in prescription medication utilization.

METHODS:

Using data from the Colorado All Payer Claims Database (CO APCD), we conducted an interrupted time-series analysis of monthly prescription fills among insured adults from January 2019 to December 2021. Therapeutic categories included opioids, psychotropics, antibiotics, antivirals, cardiometabolic drugs, and oncology medications. Interventional autoregressive integrated moving average models assessed immediate and trend-level changes in utilization following the pandemic onset in March 2020. We separately evaluated prescriptions dispensed by retail pharmacies, including mail order, and physician-administered medications, highlighting differences in how each modality adapted to system-level disruptions.

RESULTS:

The pandemic led to an immediate decrease in prescription fills, with 3.3 fewer fills per 100 insured adults (95% CI = −0.049 to −0.016; P < 0.001). Retail pharmacy prescriptions rebounded over time, supported by telehealth and mail-order options, whereas physician-administered therapies faced sustained declines. Specific therapeutic classes showed varied responses. Opioid prescriptions decreased by 0.4 fills per 100 adults (95% CI = −0.0058 to −0.0026; P < 0.001), whereas psychotropic medication use increased by 0.8 fills per 100 insured adults (95% CI = 0.0037-0.0123; P < 0.001). Antibiotic and antiviral prescriptions declined significantly. Cardiometabolic and oncology medication utilization remained stable throughout the study period.

CONCLUSIONS:

The rebound in retail pharmacy prescriptions during the COVID-19 pandemic highlights the role of telehealth and mail-order services in mitigating care disruptions. However, the persistent declines in physician-administered therapies reveal structural vulnerabilities, particularly for populations requiring complex or injectable treatments. Policymakers should build on strengths such as telehealth expansion and existing successful overprescribing management programs for opioids and antibiotics and should also address gaps in access to safe in-person care, particularly for vulnerable populations. Emergency preparedness measures should also prioritize promoting mental health support to ensure comprehensive resilience in future public health crises. By incorporating prescription utilization surveillance into routine health system monitoring, stakeholders can respond proactively to emerging challenges and promote more equitable access to essential therapies during public health emergencies.

Plain language summary

Monitoring prescription medication use can provide important insights into how well health systems respond to emergencies like COVID-19. In Colorado, physician-administered therapies faced sustained disruption, whereas telehealth and mail-order strategies helped maintain retail drug utilization and address increased mental health medication needs. By tracking these shifts across various patient populations and treatment categories, policymakers can quickly identify and address emerging gaps, ultimately designing stronger, equitable health care systems prepared to sustain essential treatments during future emergencies.

Implications for managed care pharmacy

Our analysis demonstrates how closely monitoring prescription medication use can alert health care policymakers to critical disruptions during public health emergencies while highlighting system strengths that merit sustained support. Our results suggest that physician-administered therapies face more sustained declines, supporting the need for making in-person treatments accessible during public health crises and reinforce existing telehealth and mail-order options. Ensuring robust mental health support and antibiotic and opioid stewardship can help maintain care quality and resilience.

Prescription medication utilization is one of the most common ways individuals interact with the health care system, making it a valuable indicator of how well that system can adapt during public health emergencies and guide future pandemic preparedness efforts.¹ When the COVID-19 pandemic disrupted health care delivery worldwide, major questions arose about health care resilience and vulnerabilities in the face of such unprecedented systemic strain.² In the United States, many clinicians redirected efforts toward managing patients with COVID-19, whereas health care institutions scaled back or modified their care delivery modalities to reduce virus transmission.^3,4 At the same time, patients delayed or avoided routine care to minimize infection risk.⁵ Although media attention focused on postponed surgeries and delayed cancer screenings,^6–8 understanding the changes in prescription medication utilization offers more complete pictures of how both routine and specialized care were affected and could guide future emergency preparedness efforts.

Early in the pandemic, one survey estimated that 41% of US adults forwent medical care, including 8% of adults forgoing retail pharmacy medications.⁹ Examining shifts in prescription medication utilization across varied therapeutic categories is therefore essential for understanding the drivers behind these changes, their broader health implications, and how to best inform targeted emergency preparedness strategies for future crises. For instance, reductions in pediatric vaccine administration raised concerns about potential outbreaks of vaccine-preventable diseases across the nation,¹⁰ and nonadherence to medications for chronic conditions, such as antihypertensives, may have led to serious health consequences including increased risks of all-cause, cardiovascular, and cerebrovascular mortality.¹¹ Meanwhile, driven by heightened mental health stressors during the pandemic,¹² there was an increased demand for certain mental health medications, prompting questions about whether this surge in need was adequately addressed.

Although existing research has examined medication use during the COVID-19 pandemic within single systems, payers, or medication classes,^13–15 relatively few studies have adopted a broader perspective across large, diverse populations across therapeutic categories and payers. Such an approach is essential for identifying system-wide strengths and vulnerabilities, guiding targeted policy and preparedness measures for future crises. To address this gap, we analyze 3 years of prescription medication utilization data (2019-2021) using the Colorado All Payer Claims Database (CO APCD).

By evaluating where care disruptions were most pronounced and identifying areas of resilience, we aim to inform how health systems can better prepare for future crises and maintain equitable, uninterrupted access to essential medications during public health emergencies.

Methods

STUDY DESIGN AND SETTING

We conducted a cohort study with an interrupted time-series analysis of prescription medication use among residents of Colorado, United States, from January 1, 2019, to December 31, 2021, using the CO APCD. The CO APCD is administered by the Center for Improving Value in Health Care and is considered the state’s most comprehensive health care claims database representing most covered members and payers across commercial health insurance plans; Medicare, including both traditional Medicare (fee-for-service) and Medicare Advantage (managed care); and Health First Colorado (Colorado’s Medicaid program).¹⁶ This analysis did not include patients without insurance coverage as their data are not included in CO APCD. This study was deemed exempt by the Colorado Multiple Institutional Review Board (exemption #22-0559).

We divided our study period into 36 monthly segments, from January 1, 2019, and ending in December 31, 2021, creating a dynamic cohort of insured adults (aged ≥18 years). An individual could appear in 1 or more monthly segments, depending on changes in residence, insurance status, or mortality. This approach allowed us to capture a broad picture of health system performance and medication access throughout the study period.

ANALYTIC APPROACH

We compared demographic and insurance characteristics between the prepandemic and postpandemic cohorts using absolute standardized differences (ASDs), with a threshold of greater than 10% representing a meaningful difference; we used ASD rather than P values from 2-sample t tests (for means) and chi-square tests (for proportions) because ASD quantifies imbalance and is minimally influenced by sample size.¹⁷

We calculated the monthly number of pharmacy claims as a share of all individuals in the monthly eligible cohort (=the total number of drugs used per month/the total number of adults with health insurance coverage during the same period). Because the aim of the study is to isolate the effects of the pandemic on non-COVID-19–related medication use, we excluded claims related to COVID-19 vaccines and diagnosis kits. A detailed list of excluded COVID-19–related products is provided in Supplementary Table 1 ^{(303.2KB, pdf)} (available in online article).

We treated the month of March 2020 and all subsequent months as the pandemic period. We used interventional autoregressive integrated moving average (ARIMA) models to examine the association of the pandemic with changes in prescription medication utilization. Observation months from March 2020 and onward were defined as the pandemic period (n = 22) whereas earlier months were defined as the prepandemic period (n = 14). The selection of March 2020 as an index month was based on the time the World Health Organization declared COVID-19 a pandemic.¹⁸

We tested the immediate change in the level of monthly medication use after the onset of the pandemic by including a step function in our models. We also included a ramp function to test for a change in the slope before and after March 2020. Stationarity in our series was confirmed using augmented Dickey-Fuller and Philips-Perron tests. We used plots of the autocorrelation function and partial autocorrelation function to guide the selection of autoregressive and moving average terms into the ARIMA models. Details about the terms included in each model are provided in Supplementary Table 2 ^{(303.2KB, pdf)} .

Beyond all drug claims combined, we separately analyzed retail pharmacy (including mail order) vs physician-administered medications, as well as 6 therapeutic classes: (1) opioids, (2) psychotropics, (3) antibiotics, (4) antivirals, (5) cardiometabolic drugs, and (6) oncology treatments. The rationale for selecting these groups stems from our interest in capturing the broader impact of the pandemic on medication utilization and the nuances related to patient and provider behavior in the management of acute vs chronic conditions. We used the Uniform System of Classification to identify the National Drug Code numbers of included products in each group. Supplementary Table 3 ^{(303.2KB, pdf)} summarizes the therapeutic classes included in each group.

We used Akaike statistics, the autocorrelation function, and partial autocorrelation function plots of the models’ residuals to examine the models fit and an alpha level of 0.05 was considered statistically significant.

Analytic files were generated using SAS software. The data were then analyzed using STATA18.

Results

The study cohort consisted of a total of 3,919,327 eligible Coloradans in the pre–March 2020 period and 4,403,129 members in the post–March 2020 period. We report on the demographic characteristics of the study cohort in Table 1. In the pre–March 2020 cohort and the post–March 2020 cohort, 51.5% and 51.2% of the members were women, respectively. Most of the population in both cohorts had commercial insurance followed by Medicaid, traditional Medicare, and then Medicare Advantage. The mean age decreased slightly from 51.7 years in the pre–March 2020 period to 50.5 years in the post-March period. Most individuals resided in urban areas, and more than half had missing race and ethnicity data.

TABLE 1.

Demographic and Insurance Characteristics of the Study Cohort Before and After March 2020

	Pre–March 2020 cohort (N = 3,919,327)	Post–March 2020 cohort (N = 4,403,129)	Standardized differences, %
Insurance type, n (%)
Traditional Medicare	617,021 (15.7)	622,592 (14.1)	4.5
Medicaid	1,085,570 (27.7)	1,117,180 (25.4)	5.3
Commercial	1,888,629 (48.2)	2,300,293 (52.2)	8.1
Medicare Advantage	328,107 (8.4)	363,064 (7.8)	2.3
Sex, n (%)
Female	2,017,248 (51.6)	2,253,737 (51.2)	0.6
Age, mean (SD), years	51.68 (20.7)	50.50 (20.1)	5.8
Age categories, n (%)
18–29 years	676,230 (17.3)	781,016 (17.7)	1.3
30–49 years	1,245,254 (31.8)	1,473,299 (33.5)	3.6
50–64 years	750,814 (19.2)	852,851 (19.4)	0.5
65–79 years	867,634 (22.1)	939,993 (21.4)	1.9
80+ years	379,395 (9.7)	355,970 (8.1)	5.6
Race, n (%)
White	930,454 (23.7)	1,019,257 (23.2)	1.4
Black/African American	70,942 (1.8)	77,287 (1.8)	0.4
Asian	27,321 (0.7)	31,096 (0.7)	0.1
Native Hawaiian or other Pacific Islander	2,756 (0.07)	3,220 (0.1)	0.0
American Indian/Alaska Native	308,256 (0.4)	466,958 (0.3)	1
Other race	609,948 (15.6)	613,118 (13.9)	4.6
Unknown	2,261,700 (57.7)	2,644,722 (60.1)	4.8
Urbanicity, n (%)
Rural	412,856 (11.2)	451,983 (10.3)	2.9
Urban	3,334,905 (85.1)	3,772,332 (85.7)	1.6
Frontier	92,887 (2.4)	100,347 (2.3)	0.6
Unknown	78,679 (2)	78,467 (1.8)	1.7

The distribution of demographic and insurance characteristics remained relatively stable between the prepandemic and pandemic periods. All standardized differences between these periods were below 10%, indicating minimal variation in the cohort characteristics over time. A plot of the monthly number of eligible individuals across the study period is in Supplementary Figure 1 ^{(303.2KB, pdf)} .

The total number of claims included in the study was 161,412,224, including 145,745,664 (90.3%) retail pharmacy drug claims and 15,666,566 (9.7%) physician-administered drug claims. The total number of prescriptions filled over the study period was 2,799,299 for antibiotics, 636,343 for antivirals, 3,679,902 for opioids, 14,444,510 for psychotropics, 13,007,690 for cardiometabolic system medication, and 860,235 for oncology products. In the Supplementary Materials ^{(303.2KB, pdf)} , we present a plot of the overall trends in monthly prescription medication utilization over the study time for all drugs combined, retail drugs, and physician-administered drugs (Supplementary Figure 2 ^{(303.2KB, pdf)} ) and per the selected therapeutic groups (Supplementary Figure 3 ^{(303.2KB, pdf)} ).

CHANGES IN OVERALL MEDICATION UTILIZATION

Following March 2020, we observed a significant immediate decrease in all prescription fills, with 3.3 fewer fills per every 100 insured Coloradan adults (β = −0.0329; 95% CI = −0.0494 to −0.0164; P = 0.000) (Table 2). However, usage then rebounded at a rate of 0.13 additional fills per 100 adults per month (β = 0.0013; 95% CI = 0.0001-0.0025; P = 0.041). The highest overall utilization rate was observed in March 2020, with each adult filling an average of 1.5 prescriptions. The lowest utilization rate was recorded in April 2020, with each adult filling almost 1.2 prescriptions (Figure 1).

TABLE 2.

Autoregressive Integrated Moving Average Models Results

Medications	Change in level		Change in slope
	Parameter estimate (95% CI)	P value	Parameter estimate (95% CI)	P value
All	−0.0329 (−0.0494 to −0.0164)	<0.001	0.0013 (0.0001 to 0.0025)	0.041
Retail	−0.0202 (−0.0388 to −0.0016)	0.034	−0.0026 (−0.0042 to −0.0010)	0.736
Physician-administered	−0.0118 (−0.0136 to −0.0100)	<0.001	0.0016 (0.0014 to 0.0018)	<0.001
Antibiotics	−0.0022 (−0.0044 to −0.00004)	0.034	0.00007 (−0.0001 to 0.0002)	0.315
Antivirals	−0.0008 (−0.0014 to −0.0002)	0.004	−0.00005 (−0.0001 to 0.00001)	0.143
Opioids	−0.0042 (−0.0058 to −0.0026)	<0.001	−0.00004 (−0.0001 to 0.00002)	0.911
Psychotropics	0.008 (0.0037 to 0.0123)	<0.001	−0.00009 (−0.0005 to 0.0003)	0.596
Cardiometabolic system	0.0029 (−0.0047 to 0.0105)	0.465	−0.0001 (−0.0007 to 0.0005)	0.669
Oncology	0.0003 (−0.00009 to 0.0007)	0.107	0.00004 (0.00002 to 0.00006)	0.010

FIGURE 1.

Observed Monthly Proportions of All Prescriptions, Retail Drugs Prescriptions, and Physician-Administered Prescriptions Filled by Colorado Residents Compared With Predicted Use, By Analysis Group

Panels represent March 2019 through December 2021 monthly proportions of prescription medication fills per 100 insured Colorado adults for (A) all prescriptions, (B) retail-dispensed medications (including mail order), and (C) physician-administered drugs. In each panel, observed values are shown as gray dots, model-predicted values from the interventional autoregressive integrated moving average without a COVID-19 interruption are shown as a solid black line, and model-predicted values with a COVID-19 interruption are shown as a solid yellow line. A vertical dashed blue line marks March 2020 (pandemic onset). The y axes are scaled to highlight changes in the observed-predicted differences after March 2020. Monthly denominators reflect the dynamic insured adult cohort, COVID-19 vaccines and diagnostic products are excluded, and days supply is not accounted for.

CHANGES IN RETAIL VS PHYSICIAN-ADMINISTERED MEDICATIONS UTILIZATION

Similarly, we observed a significant decrease in the utilization of retail drugs with approximately 2 fewer fills per every 100 adults (β = −0.0202; 95% CI = −0.0388 to −0.0016; P = 0.034). However, the slope change was not significant.

For physician-administered medications, we found a significant immediate level decrease in their utilization with approximately 1 fewer fill per every 100 adults (β = −0.0118; 95% CI = −0.0136 to −0.0100; P = 0.000) followed by a significant monthly increase of 0.16 more fills per every 100 insured adults following the pandemic (β = 0.0016; 95% CI = 0.0014-0.0018; P = 0.000). The lowest utilization rate of physician-administered medications was observed in April 2020 when only 8.5 physician-administered medications were administered per every 100 insured adults (Figure 1).

THERAPEUTIC CLASS–SPECIFIC FINDINGS

Antibiotics and Antivirals

We found a significant immediate reduction with 0.22 fewer antibiotic prescription fills per every 100 insured adults after the onset of COVID-19 (β = −0.0022; 95% CI = −0.0044 to −0.00004; P = 0.034). The highest rate of antibiotic utilization was recorded in January 2020, when 3.1 antibiotic prescriptions were filled per every 100 insured adults.

Similarly, we found a significant immediate reduction in the level of antiviral utilization (β = −0.008; 95% CI = −0.0014 to −0.00004; P = 0.034) and no significant change in the trend over time (β = −0.00005; 95% CI = −0.0001 to 0.00001; P = 0.143). The highest antiviral utilization rate occurred 2 months before the start of the pandemic, in January 2020, with 0.7 fills per every 100 insured adults (Figure 2).

FIGURE 2.

Observed Monthly Proportions of Antibiotics, Antivirals, Opioids, Psychotropics, Cardiometabolic System, and Oncology Prescriptions Filled by Colorado Residents Compared With Predicted Use

Panels represent March 2019 through December 2021 monthly proportions of prescription medication fills per 100 insured Colorado adults for (A) antibiotics, (B) antivirals, (C) opioids, (D) psychotropics, (E) cardiometabolic drugs, and (F) oncology medications. In each panel, observed values are shown as gray dots, model-predicted values from the interventional autoregressive integrated moving average without a COVID-19 interruption are shown as a solid black line, and model-predicted values with a COVID-19 interruption are shown as a solid yellow line. A vertical dashed blue line marks March 2020 (pandemic onset). The y axes are scaled to highlight changes in the observed-predicted differences after March 2020. Monthly denominators reflect the dynamic insured adult cohort, COVID-19 vaccines and diagnostic products are excluded, and days supply is not accounted for.

Opioids

We observed a significant immediate level change of 0.4 fewer fills per every 100 insured adults (β = −0.0042; 95% CI = −0.0058 to −0.0026; P < 0.001). The slope change was not statistically significant (β = −0.00004; 95% CI = −0.0001 to 0.00002; P = 0.911). The lowest opioids utilization rate occurred in April 2020 (2.5 fills per every 100 insured adults) (Figure 2).

Psychotropics

In contrast, psychotropics utilization experienced a significant increase with 0.8 more fills per every 100 insured Coloradan adults (β = 0.008; 95% CI = 0.0037-0.0123; P = 0.000). The highest utilization rate occurred in March 2020 with 13.1 fills per every 100 insured adults (Figure 2).

Cardiometabolic System Drugs

The utilization of cardiometabolic system medications remained relatively stable throughout the entire study period. The results from our ARIMA model showed a nonsignificant immediate level change (β = 0.0029; 95% CI = −0.0047 to 0.0105; P = 0.465) and a nonsignificant change in slope (β = −0.0001; 95% CI = −0.0007 to 0.0005; P = 0.669) (Figure 2).

Oncology Medications

Finally, oncology medication utilization showed a nonsignificant immediate level change after the pandemic onset (β = 0.0003, 95% CI = −0.0009 to 0.0007; P = 0.107) but a significant change in slope (β = −0.00004; 95% CI = 0.00002-0.00006; P = 0.010) equivalent to 0.004 additional fills per 100 insured adult per month following the pandemic onset (Figure 2).

Discussion

To our knowledge, this study offers the first broad state-level examination of any and all prescription medication utilization following the onset of the COVID-19 pandemic. Our analysis reflects the broader resilience and vulnerabilities of Colorado’s health system during a high-stress event. One key strength of our research lies in the use of a population-based cohort that includes patients with varying insurance types, areas of residence, clinical needs, and medication access preferences. Employing ARIMA modeling further enabled us to account for time-series patterns in autocorrelated data allowing for more accurate trend analysis. By examining both overall medication utilization and specific therapeutic groups, our study provides insights into how the pandemic affected a wide spectrum of treatments, from chronic disease management to acute conditions.

OVERALL PRESCRIPTION MEDICATION UTILIZATION AS A PROXY FOR HEALTH SYSTEM RESILIENCE

Our finding of the 3.3 percentage point decline in overall prescription utilization is consistent with other studies describing the disruption in health care utilization.⁹ This decrease reflects how factors such as initial lockdowns, canceled appointments, and patient concerns about contracting COVID-19 infection interrupted routine care. In Colorado, legislative orders^19,20 restricting nonessential medical procedures and the subsequent drop in health care visits likely contributed to the decline in medication use.²¹ On the national level, similar patterns emerged in various systems and populations, often driven by fear of infection.²² Another driver of this decline is the “acute stress”²³ on the domestic pharmaceutical supply chains through introducing unprecedented changes in the demand for certain drugs exceeding manufacturers’ production capacity.²⁴ Uncertainties about the sourcing of pharmaceutical products from global suppliers created additional stresses.^23,25 Interestingly, we observed the highest utilization rate in March 2020 (with 146.6 prescriptions filled per every 100 insured Coloradans), coinciding with the initial months of the pandemic, which suggests early hoarding behaviors in response to the pandemic.²⁶

Retail medications experienced a temporary 2 percentage point decline in utilization reflecting the broader reduction in routine medical visits during the lockdown. However, this rebounded by the end of the study period, likely because of the rapid expansion of mail order and telehealth accessibility.²⁷ By contrast, physician-administered medications (eg, parenteral medications) were harder to sustain. Because these therapies require in-person visits, they were more susceptible to access restrictions that were imposed and the temporary closure of infusion centers during the pandemic.^19,20 This disruption reveals a clear health system vulnerability that emerged under pandemic constraints.

STRENGTHS AND VULNERABILITIES REVEALED BY THERAPEUTIC CATEGORIES

Opioids utilization dropped, potentially reflecting existing regulatory and clinical efforts to curb opioid prescribing, alongside fewer surgeries that often trigger new opioids prescriptions.²⁸ Meanwhile, psychotropic medication usage surged, mirroring intensified mental health challenges such as social isolation, job loss, and feelings of fear and uncertainty.^29–31 This increase could also indicate that many Coloradans were able to successfully receive mental health care via telehealth solutions.

Antibiotics and antivirals saw declines following the onset of the COVID-19 pandemic in Colorado. This finding stands in contrast to other studies reporting spikes in antibiotic and antiviral use^32–34 likely because of the increase of the off-label use of drugs like azithromycin, chloroquine, hydroxychloroquine, and lopinavir/ritonavir in patients infected with the SARS-CoV-2 virus during the early phases of the pandemic.^35–38 Colorado’s comparatively lower SARS-CoV-2 infection rates,³⁹ the success of public health measures such as social distancing and increased hygiene practices,⁴⁰ and ongoing antimicrobial stewardship initiatives⁴¹ may have moderated antibiotic and antiviral use and explain this divergence in results from other studies. Other factors, such as supply chain constraints, changes in health-seeking behavior, and evolving prescribing guidelines, likely also contributed to this trend.^23,42,43

Cardiometabolic and oncology medications remained stable, suggesting that patients with chronic and critical conditions were relatively well-served by existing telehealth and extended medication supply.^44,45 Similarly, oncology treatments remained stable in Colorado, despite national reports or reductions in cancer care.^7,46 This resilience is promising, given the risks of not treating cardiovascular disease and cancer, and implies that some continuity structures (eg, specialty clinics and remote monitoring) were effective.

IMPLICATIONS FOR EMERGENCY PREPAREDNESS

Our findings suggest several actionable strategies for future public health crises:

Sustain telehealth and mail order: National data show that rapid scaling of remote care and delivery services helped maintain prescription medication utilization. Maintaining these modalities can support continuity of medication access during future public health crises.^47,48

Reinforce in-person therapeutic access: Physician-administered treatments remain a clear vulnerability. Ensuring adequate protective measures and streamlined in-person service availability during future crises is vital for patients who cannot pivot to remote care options.⁴⁹

Reinforce mental health care: Increased psychotropic use highlights the system’s ability to adapt to mental health needs but also signals an increase in demand that must be met with sufficient workforce capacity, funding, and outreach efforts.⁵⁰

Build on antibiotic and opioid stewardship: Although it is important to ensure that reduced antibiotic and antiviral use did not stem from barriers to necessary care, these declines may indicate fewer infections and more judicious prescribing. To maintain this positive trend, health care stakeholders can strengthen antimicrobial stewardship programs and continue public health campaigns promoting the importance of hygiene and vaccination. Similarly, the observed decrease in opioid use may reflect both fewer surgeries and the effectiveness of existing opioid-management strategies. Sustaining these gains in future crises requires continued vigilant prescribing, enhanced pain management alternatives, and robust patient education on responsible opioid use.

Enhance supply chain transparency: To reduce drug hoarding and panic-driven demand, health systems and policymakers should establish clear communication channels about medication availability, invest in reliable domestic production, and implement contingency plans for manufacturing bottlenecks.

Prioritize health care equity: Addressing persistent disparities in health care access, particularly in rural areas and marginalized communities disproportionately impacted during crises, remains critical to fostering a resilient system.⁵¹

LIMITATIONS

Although our study design offers significant strengths, it also has several limitations that are important to consider. First, our outcome measure (proportion of monthly fills per eligible cohort) provides an overall estimate of the volume of all types of prescription medications fills per month but does not identify actual medication use through the number of days dispensed or supplied. For example, a single prescription fill could represent different quantities or days supply (eg, a 30-day or 90-day supply). Some providers might have switched to prescribing larger quantities to avoid patients’ direct interaction with the health care system during the pandemic. Similarly, we did not disaggregate mail order from community retail across all payers; as a result, channel-specific patterns may therefore be obscured.

Another limitation stems from the nature of the CO APCD, which excludes data on the uninsured population. During the pandemic, many Coloradans lost their jobs and health insurance coverage; although some transitioned to exchange coverage or Medicaid, others became uninsured. The omission of this group may lead to underestimating the pandemic’s impact on medication utilization, particularly among vulnerable populations who faced additional barriers to health care and medication access. Furthermore, the CO APCD lacks detailed and complete data on race and ethnicity, limiting the capacity to analyze disparities in medication access and utilization among groups disproportionately affected by the pandemic.^52,53

Additionally, the experience of Coloradans during the pandemic may not mirror that of residents in other states, given differences in public health policies, pandemic severity, and health care infrastructures across the country.³⁹ These differences may limit the generalizability of our findings and the conclusions drawn from this study should be contextualized within Colorado’s unique public health landscape.

Moreover, grouping medications into broad therapeutic categories may conceal important differences in utilization patterns of specific products or within particular disease areas, which may be more evident at a finer level of analysis. In particular, we did not disaggregate oncology by route of administration; oral agents (retail pharmacy) and infused agents (physician-administered) were analyzed within their respective aggregate series, which may mask route-specific utilization patterns.

Furthermore, because this was a population-level analysis, we did not fit age-stratified interrupted time-series models; therefore, any age-specific differences in level or trend may be obscured. Finally, beyond ARIMA identification and diagnostic checks, we did not perform additional post hoc sensitivity analyses (eg, alternative model specifications).

Conclusions

By examining prescription medication utilization patterns throughout the COVID-19 pandemic, we identified critical vulnerabilities, particularly for in-person, physician-administered therapies, and notable strengths such as relative stability and rebound in retail medications. To prepare for future public health emergencies, policymakers should maintain telehealth expansion, reinforce antibiotic and opioid stewardship programs, and ensure consistent access to safe in-person care, especially for populations requiring complex treatments. Robust mental health support must also be a priority, given the psychological burdens exacerbated during crises. These findings illustrate how medication utilization data can illuminate areas where health systems need strengthening under acute stress. By embedding ongoing prescription utilization surveillance into routine health system monitoring, policymakers can detect emerging disruptions and intervene more rapidly, thereby constructing a more resilient, equitable, and adaptive health care landscape for future public health emergencies.

Disclosures

This research was funded by the University of Colorado Skaggs School of Pharmacy and Pharmaceutical Sciences Seed Grant Program for COVID-19 Research and supported by a Colorado All Payer Claims Database scholarship. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funders.