OBJECTIVE

Repurposed medications for acute coronavirus disease 2019 (COVID-19) continued to be prescribed after results from rigorous studies and national guidelines discouraged use. We aimed to describe prescribing rates of nonrecommended medications for acute COVID-19 in children, associations with demographic factors, and provider type and specialty.

METHODS

In this retrospective cohort of children <18 years in a large United States all-payer claims database, we identified prescriptions within 2 weeks of an acute COVID-19 diagnosis. We calculated prescription rate, performed multivariable logistic regression to identify risk factors, and described prescriber type and specialty during nonrecommended periods defined by national guidelines.

RESULTS

We identified 3 082 626 COVID-19 diagnoses in 2 949 118 children between March 7, 2020 and December 31, 2022. Hydroxychloroquine (HCQ) and ivermectin were prescribed in 0.03% and 0.14% of COVID-19 cases, respectively, during nonrecommended periods (after September 12, 2020 for HCQ and February 5, 2021 for ivermectin) with considerable variation by state. Prescription rates were 4 times the national average in Arkansas (HCQ) and Oklahoma (ivermectin). Older age, nonpublic insurance, and emergency department or urgent care visit were associated with increased risk of either prescription. Additionally, residence in nonurban and low-income areas was associated with ivermectin prescription. General practitioners had the highest rates of prescribing.

CONCLUSIONS

Although nonrecommended medication prescription rates were low, the overall COVID-19 burden translated into high numbers of ineffective and potentially harmful prescriptions. Understanding overuse patterns can help mitigate downstream consequences of misinformation. Reaching providers and parents with clear evidence-based recommendations is crucial to children’s health.

What’s Known on This Subject:

Misinformation perpetuated the use of repurposed medications, such as hydroxychloroquine and ivermectin, for the treatment of acute coronavirus disease 2019 without supporting evidence for effectiveness and led to adverse outcomes and medication shortages.

What This Study Adds:

Using a large national claims database, we observed substantially higher rates of nonrecommended prescriptions for children with acute coronavirus disease 2019 in the south and with private insurance. These risk factors could be addressed through targeted messaging to key demographics.

There was intense pressure on clinicians and researchers to find therapies to mitigate severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection during the pandemic.1 ,2  Repurposed medications, including hydroxychloroquine (HCQ), an antirheumatic and antimalarial, and ivermectin, an antiparasitic, were considered potential treatments for acute coronavirus disease 2019 (COVID-19) until their effectiveness was disproven.

The US Food and Drug Administration granted an emergency use authorization (EUA) for chloroquine and HCQ on March 28, 2020 for adults or adolescents hospitalized with COVID-19 for whom a clinical trial was not available.3  However, the EUA was revoked June 15, 2020, given further studies questioning effectiveness and an inability to replicate previous findings.4  By September 12, 2020, consensus guidance endorsed by the Pediatric Infectious Diseases Society (PIDS) was published, recommending against use of HCQ outside a clinical trial.1 

Similarly, by late 2020, optimism for ivermectin mounted after clinical trials reported clinical effectiveness for COVID-19.5  However, meta-analyses using these trials were subsequently retracted given poor quality of underlying studies.5  Infectious Diseases Society of America (IDSA) guidelines published February 5, 2021 suggested against ivermectin use outside a clinical trial6  and recommended against use for ambulatory persons by July 2022.7 

Although the impact of such changes on medical practice and public perception are challenging to measure, they may have led to distrust of the medical and scientific establishment. Despite guidelines, use of these medications for acute COVID-19 continued and led to medication shortages.8 10  By August 2021, ivermectin prescriptions from outpatient retail pharmacies increased 24-fold compared with prepandemic baseline, and ivermectin-related calls to poison control centers increased 5-fold.11 ,12  One study reported ivermectin prescription rates in children 0 to 17 years as 1.5 per 10 000 patients from December 2020 to March 2021.13  No studies have assessed rates of other nonrecommended medications, such as hydroxychloroquine, nitazoxanide, colchicine, or lopinavir and ritonavir for acute COVID-19 or analyzed prescription risk factors. In this study, we aim to describe prescribing rates, demographic risk factors, distribution by provider type, and rates of adverse reactions to understand the problem and inform future health policies.

We performed a retrospective cohort study of children diagnosed with acute COVID-19 using a large US all-payer claims database, Komodo Healthcare Map.14  This anonymized database represents 330 million US patients and includes commercial, Medicaid, and Medicare claims, with complete medical and prescription encounters on 165 million patients. The Stanford Center for Population Health Sciences received a cohort from October 2015 onwards of patients with COVID-19-related International Classification of Diseases, 10th Revision (ICD-10) codes and quarterly data refreshes.15  The Institutional Review Board at Stanford University determined the study was exempt from review because of deidentified data. Data cleaning was conducted in Snowflake Structured Query Language,16  statistical models were implemented in R version 4.1.3,17  and graphics created in Python 3.8.5.18  Full details of our analysis plan were preregistered before conducting this analysis.19 

Children with acute COVID-19 were identified by ICD-10 code U07.1 from March 7, 2020 to December 31, 2022, at age <18 years. This ICD-10 code has been shown to have high positive predictive value (>90%) in identifying hospitalized patients with a positive polymerase chain reaction result for SARS-CoV-2.20 ,21  Additionally, decent predictive value was described in outpatient (78%) and emergency or urgent care (82%) settings for active disease when correlated with records 1 month before or after a diagnosis, although the most common reason for false positive documentation was personal history of COVID-19, for which ICD-10 code Z86.16 became effective near the end of the study.22  Furthermore, several previous studies have demonstrated general reliability of ICD codes in claims data.23 25 

COVID-19 diagnoses were excluded if there was a prior COVID-19 diagnosis within the previous 6 months, and reinfection rate was defined as repeat nonexcluded infections. Our cohort only included patients with continuous insurance coverage at least 1 year prior and at least 6 weeks after diagnosis, as well as a prescription drug plan. These inclusions were adjusted if a patient was born within 1 year of a COVID-19 diagnosis or died within 6 weeks afterward.

Covariates were chosen based on availability in the Komodo database. Patient demographics were determined at the time of COVID-19 diagnosis. In the rare event of conflicting patient-level information, we used the value that appeared most frequently on the patient’s insurance claims. Sex was divided into female versus male or unspecified, and age was categorized based on natural breakpoints: 0 to 4 years, 5 to 12 years, and 13 to 17 years. Insurance type was classified into commercial health maintenance organization (HMO), commercial non-HMO, and public. Presence of a complex chronic condition was defined as any ICD-10 or Current Procedural Terminology code in the Feudtner classification that occurred in the year before COVID-19 diagnosis.26 ,27 

In terms of other sociodemographic information, patient race and ethnicity data were not available, but the Komodo database included 3-digit zip code as the smallest patient-specific geographic unit. This was classified into metropolitan statistical area status (urban, suburban, or rural), median household income quartile, and racial and ethnic diversity quartile by percent white, based on 2020 American Community Survey US census data,28  which was rolled up to the 3-digit-zip-code-level using population-weighted averages. Demographics per first COVID-19 infection were summarized by frequencies and percentages for categorical variables.

We analyzed prescriptions for medications not recommended for COVID-19 listed in the IDSA or National Institutes of Health (NIH) treatment guidelines. This included hydroxychloroquine, ivermectin, nitazoxanide, colchicine, and lopinavir and ritonavir. We did not analyze famotidine, azithromycin, or fluvoxamine because common indications made it difficult to determine whether the medication was intended to treat COVID-19. Prescriptions were counted if there was at least 1 nondenied pharmaceutical claim within 14 days after COVID-19 diagnosis. Prescription date was date written, or date filled when this was missing. We took the earliest date if there was more than 1 within the time frame, and the most frequent variables associated with a prescription if there was discrepancy because of redundancy. We identified relevant National Drug Codes (NDC 11) for each medication by matching ingredient names and excluding topicals. Patients were excluded if they had 1 or more ICD-10 codes in the year before COVID-19 diagnosis for conditions for which the medication was indicated (Appendix 1). These were determined by pediatric or adult indications in Lexicomp matched to ICD-10 codes using a reference,29 ,30  and ICD-10 codes associated with prescription of HCQ or ivermectin in our cohort from October 2015 onwards, irrespective of COVID-19 diagnosis (excluding COVID-19-related and nonspecific symptom codes).

We analyzed periods after which there were national guidelines recommending against using these medications outside of a clinical trial – termed “nonrecommended periods.” For HCQ, we chose a conservative date of publication of pediatric guidelines endorsed by PIDS (September 12, 2020),1  and for ivermectin, IDSA guidelines (February 5, 2021).6  Nonrecommended periods for nitazoxanide (July 8, 2021) and colchicine (December 16, 2021) were determined by NIH guidelines,31 ,32  and lopinavir and ritonavir (September 12, 2020) by PIDS guidelines.1 

Nonrecommended medication prescriptions were calculated as a proportion (ie, number of prescriptions for a specific medication divided by the number of COVID-19 diagnoses during each nonrecommended period). Inpatient prescription rates were also calculated. For nonrecommended medications with an overall prescription rate >0.01%, we reported the following: prescription rates over time, days from COVID-19 diagnosis to prescription, and variation by state. Statewide variation was displayed using forest plots with 95% Bernoulli confidence intervals (CI), with Wilson correction to accommodate very low rates. Results were only reported for states with >10 000 COVID-19 diagnoses in the cohort and prescriptions for medications of interest to minimize risk of deidentification.

Emergency department (ED) or urgent care (UC) visit or hospitalization within 14-days of a COVID-19 diagnosis was assessed as a proxy for COVID-19 severity, as well as death within 28-days of a COVID-19 diagnosis. Clinicians who diagnosed COVID-19 and those who prescribed the nonrecommended medication were matched by hashed National Provider Identifier to provider type and specialty, determined by National Uniform Claim Committee Health Care Provider Taxonomy codes selected by the provider. For this analysis, a patient was not limited to 1 prescription. Prescription rate per diagnosis was calculated. The number of prescriptions by the top 100 prescribers was determined, as well as type and specialty of the top 10 prescribers.

Adverse drug reactions (ADRs) were assessed using ICD-10 codes within 28 days of prescription. These were prespecified and adapted from published work on ICD-10 codes associated with adverse drug events (Appendix 2).33 

Logistic regression with state-level fixed effects and cluster-robust confidence interval at the county level was used to calculate odds ratios for the association between each demographic or geographic feature and prescription of a nonrecommended medication. Both univariate and multivariate analyses were performed for each drug. Results were interpreted using a Bonferroni-adjusted P value threshold of ≤.003, dividing the standard α of .05 by 16 comparisons.

We identified 3 082 626 COVID-19 diagnoses in 2 949 118 patients <18 years who met inclusion criteria (Fig 1). There were 113 508 COVID-19 diagnoses that met criteria for reinfection, for a rate of 4.5%. There were low mortality and hospitalization rates, but high rates of ED and UC visits within 14-days of COVID-19 diagnosis (Table 1). Most patients were older, had public insurance, and lived in a suburban zip code. After exclusions, during nonecommended periods there were 813 HCQ prescriptions for a rate of 0.03% per COVID-19 diagnosis, and 3602 ivermectin prescriptions for a rate of 0.14% per COVID-19 diagnosis (Appendix 3). Other nonrecommended medications did not meet our prespecified criteria of >0.01% for subsequent analysis. Most HCQ and ivermectin prescriptions occurred on the day of COVID-19 diagnosis (Supplemental Fig 4). Inpatient prescription rate was <0.03% for both HCQ and ivermectin (Appendix 3).

FIGURE 1

Consort diagram. COVID-19 diagnoses and distinct patients <18 years included and excluded from the cohort, March 7, 2020 to December 31, 2022.

FIGURE 1

Consort diagram. COVID-19 diagnoses and distinct patients <18 years included and excluded from the cohort, March 7, 2020 to December 31, 2022.

Close modal
TABLE 1

Demographics at First COVID-19 Diagnosis (March 7, 2020–December 31, 2022)

Demographics at First COVID-19 DiagnosisPatients (n = 2 949 118), n (%)
28-d mortality 55 (0.002) 
Hospitalization within 14-d 46 272 (1.6) 
ED or UC visit within 14-d 2 527 401 (85.7) 
Female 1 463 040 (49.6) 
Age  
 0–4 541 478 (18.4) 
 5–12 1 092 638 (37.0) 
 13–17 1 315 002 (44.6) 
Complex chronic condition 181 958 (6.2) 
Insurance typea  
 Commercial HMO 551 972 (18.7) 
 Commercial non-HMO 544 500 (18.5) 
 Public 1 852 646 (62.8) 
3-digit zip code classification  
 Urban 287 362 (9.7) 
 Suburban 1 683 367 (57.1) 
 Rural 948 326 (32.2) 
 Unknown 30 063 (1.0) 
Income quartile of 3-digit zip code  
 Q1 (Median household income <$47 200) 838 097 (28.4) 
 Q2 ($47 200–$54 400) 789 653 (26.8) 
 Q3 ($54 400–$66 200) 631 450 (21.4) 
 Q4 (>$66 200) 659 966 (22.4) 
 Unknown 29 952 (1.0) 
Racial and ethnic diversity of 3-digit zip code  
 Q1 (% white <47.8%) 757 323 (25.7) 
 Q2 (47.8%–67.0%) 868 154 (29.4) 
 Q3 (67.0%–81.9%) 683 603 (23.2) 
 Q4 (>81.9%) 610 086 (20.7) 
 Unknown 29 952 (1.0) 
Demographics at First COVID-19 DiagnosisPatients (n = 2 949 118), n (%)
28-d mortality 55 (0.002) 
Hospitalization within 14-d 46 272 (1.6) 
ED or UC visit within 14-d 2 527 401 (85.7) 
Female 1 463 040 (49.6) 
Age  
 0–4 541 478 (18.4) 
 5–12 1 092 638 (37.0) 
 13–17 1 315 002 (44.6) 
Complex chronic condition 181 958 (6.2) 
Insurance typea  
 Commercial HMO 551 972 (18.7) 
 Commercial non-HMO 544 500 (18.5) 
 Public 1 852 646 (62.8) 
3-digit zip code classification  
 Urban 287 362 (9.7) 
 Suburban 1 683 367 (57.1) 
 Rural 948 326 (32.2) 
 Unknown 30 063 (1.0) 
Income quartile of 3-digit zip code  
 Q1 (Median household income <$47 200) 838 097 (28.4) 
 Q2 ($47 200–$54 400) 789 653 (26.8) 
 Q3 ($54 400–$66 200) 631 450 (21.4) 
 Q4 (>$66 200) 659 966 (22.4) 
 Unknown 29 952 (1.0) 
Racial and ethnic diversity of 3-digit zip code  
 Q1 (% white <47.8%) 757 323 (25.7) 
 Q2 (47.8%–67.0%) 868 154 (29.4) 
 Q3 (67.0%–81.9%) 683 603 (23.2) 
 Q4 (>81.9%) 610 086 (20.7) 
 Unknown 29 952 (1.0) 
a

Commercial designation included combined commercial and Medicaid plans, self-insured plans (employer funds cover claim costs), Health Insurance Marketplace plans, and Tricare and Veterans Affairs. HMO categorization was specified in the Komodo database. Public was inclusive of Medicaid, Children’s Health Insurance Program, and Medicare.

Prescribing rates peaked in April 2020 for HCQ and July to August 2021 for ivermectin (Fig 2). Prescribing for both medications continued to decrease by early 2022, despite the large spike in COVID-19 cases caused by the SARS-CoV-2 ο variant.34  Mean prescription rates by state during nonrecommended periods showed considerable variation, with the highest rates in states in the south US census region (Fig 3).35  Arkansas had the highest rate of HCQ prescription (0.13%, 95% CI 0.09%–0.21%), and Oklahoma the highest rate of ivermectin prescription (0.56%, 95% CI 0.46%–0.68%), both about 4 times the national average. From July 1 to Aug 31, 2021, the ivermectin prescription rate was 1.6% for the top 3 states (Oklahoma, Texas, Arkansas), and 1.9% for Oklahoma alone, about 13.6 times the national average.

FIGURE 2

Timelines of COVID-19 diagnoses and prescription rate per COVID-19 diagnosis by month. Solid lines represent prescription rate and shaded areas represent 95% CIs. Markers indicate dates after which the medication was not recommended outside a clinical trial for COVID-19.

FIGURE 2

Timelines of COVID-19 diagnoses and prescription rate per COVID-19 diagnosis by month. Solid lines represent prescription rate and shaded areas represent 95% CIs. Markers indicate dates after which the medication was not recommended outside a clinical trial for COVID-19.

Close modal
FIGURE 3

Mean prescription rate per COVID-19 diagnosis by state during nonrecommended periods. State abbreviations per postal codes. Solid dot represents mean prescription rate and line with bars represents 95% CI. Non-Recommended Periods: after Sept 12, 2020 for HCQ and after Feb 5, 2021 for ivermectin.

FIGURE 3

Mean prescription rate per COVID-19 diagnosis by state during nonrecommended periods. State abbreviations per postal codes. Solid dot represents mean prescription rate and line with bars represents 95% CI. Non-Recommended Periods: after Sept 12, 2020 for HCQ and after Feb 5, 2021 for ivermectin.

Close modal

Within states, patients who were older, had nonpublic insurance, or visited an ED or UC within 14 days of diagnosis had increased adjusted odds of HCQ prescription (Table 2, Appendix 4). Patients who were older, had nonpublic insurance, and lived in a nonurban and lowest income quartile 3-digit zip code had increased adjusted odds of ivermectin prescription. They were also more likely to have visited an ED or UC or be hospitalized within 14 days of diagnosis. The sex, complex chronic condition, and racial and ethnic diversity quartile of 3-digit zip were not significant in the multivariate model for either medication.

TABLE 2

Association of Demographic and Geographic Features With Prescription During Nonrecommended Periods With State-Level Fixed Effects

Demographics per COVID-19 InfectionHydroxychloroquine Prescription (after September 12, 2020)Ivermectin Prescription (after February 5, 2021)
Univariate ModelsMultivariate ModelUnivariate ModelsMultivariate Model
OR [95% CI]PaOR [95% CI]POR [95% CI]PaOR [95% CI]P
Female 1.09 [0.94–1.26] .25 1.01 [0.87–1.17] .91 1.07 [1.00–1.15] .06 0.99 [0.92–1.07] .89 
Age (years) 
 0–4 Reference  Reference  Reference  Reference  
 5–12 16.02 [4.11–62.41] <.001 15.19 [3.91–59.03] <.001a 4.89 [3.49–6.85] <.001 4.82 [3.52–6.61] <.001a 
 13–17 180.47 [44.31–735.04] <.001 151.27 [37.43–611.43] <.001a 21.12 [12.58–35.47] <.001 19.62 [12.43–30.96] <.001a 
Complex chronic condition 1.22 [0.93–1.60] .15 1.05 [0.81–1.37] .70 1.31 [1.14–1.50] <.001 1.09 [0.97–1.23] .16 
 Insurance type 
 Public Reference  Reference  Reference  Reference  
 Commercial HMO 3.56 [2.79–4.55] <.001 3.04 [2.41–3.83] <.001a 1.79 [1.24–2.61] .002 1.81 [1.46–2.25] <.001a 
 Commercial non-HMO 5.24 [4.14–6.63] <.001 4.03 [3.17–5.11] <.001a 2.77 [1.82–4.20] <.001 2.50 [1.88–3.33] <.001a 
3-digit zip code classification 
 Urban Reference  Reference  Reference  Reference  
 Suburban 1.99 [1.24–3.21] .004 1.27 [0.68–2.36] .45 3.18 [1.90–5.33] <.001 2.52 [1.49–4.25] <.001a 
 Rural 3.45 [2.19–5.46] <.001 1.99 [1.19–3.31] .01 5.60 [4.00–7.84] <.001 3.49 [2.28–5.35] <.001a 
Income quartile of 3-digit zip code 
 Q1 (median household income <$47 200) 0.96 [0.53–1.73] .88 1.20 [0.70–2.08] .51 1.96 [1.30–2.98] .001 2.10 [1.28–3.44] .003a 
 Q2 ($47 200–$54 400) 0.63 [0.34–1.17] .14 0.93 [0.47–1.85] .83 0.80 [0.49–1.32] .39 0.99 [0.60–1.62] .96 
 Q3 ($54 400–$66 200) 0.59 [0.37–0.94] .03 0.67 [0.41–1.08] .10 0.83 [0.57–1.19] .30 0.87 [0.61–1.24] .44 
 Q4 (>$66 200) Reference  Reference  Reference  Reference  
Racial and ethnic diversity of 3-digit zip code 
 Q1 (% white <47.8%) Reference  Reference  Reference  Reference  
 Q2 (47.8%–67.0%) 2.32 [1.25–4.29] .007 1.71 [0.97–3.00] .06 1.12 [0.54–2.34] .76 0.89 [0.55–1.45] .65 
 Q3 (67.0%–81.9%) 2.57 [1.32–5.02] .006 1.62 [0.83–3.15] .16 1.76 [0.88–3.50] .11 1.26 [0.82–1.95] .29 
 Q4 (>81.9%) 4.05 [1.94–8.46] <.001 2.00 [0.96–4.14] .06 2.06 [1.02–4.14] .04 1.01 [0.60–1.69] .98 
Hospitalization within 14-d 0.96 [0.52–1.79] .89 2.23 [1.10–4.51] .03 0.73 [0.53–1.00] .05 2.30 [1.54–3.45] <.001a 
ED or UC visit within 14-d 2.06 [1.49–2.84] <.001 2.39 [1.69–3.39] <.001a 2.74 [2.12–3.52] <.001 3.26 [2.40–4.43] <.001a 
Demographics per COVID-19 InfectionHydroxychloroquine Prescription (after September 12, 2020)Ivermectin Prescription (after February 5, 2021)
Univariate ModelsMultivariate ModelUnivariate ModelsMultivariate Model
OR [95% CI]PaOR [95% CI]POR [95% CI]PaOR [95% CI]P
Female 1.09 [0.94–1.26] .25 1.01 [0.87–1.17] .91 1.07 [1.00–1.15] .06 0.99 [0.92–1.07] .89 
Age (years) 
 0–4 Reference  Reference  Reference  Reference  
 5–12 16.02 [4.11–62.41] <.001 15.19 [3.91–59.03] <.001a 4.89 [3.49–6.85] <.001 4.82 [3.52–6.61] <.001a 
 13–17 180.47 [44.31–735.04] <.001 151.27 [37.43–611.43] <.001a 21.12 [12.58–35.47] <.001 19.62 [12.43–30.96] <.001a 
Complex chronic condition 1.22 [0.93–1.60] .15 1.05 [0.81–1.37] .70 1.31 [1.14–1.50] <.001 1.09 [0.97–1.23] .16 
 Insurance type 
 Public Reference  Reference  Reference  Reference  
 Commercial HMO 3.56 [2.79–4.55] <.001 3.04 [2.41–3.83] <.001a 1.79 [1.24–2.61] .002 1.81 [1.46–2.25] <.001a 
 Commercial non-HMO 5.24 [4.14–6.63] <.001 4.03 [3.17–5.11] <.001a 2.77 [1.82–4.20] <.001 2.50 [1.88–3.33] <.001a 
3-digit zip code classification 
 Urban Reference  Reference  Reference  Reference  
 Suburban 1.99 [1.24–3.21] .004 1.27 [0.68–2.36] .45 3.18 [1.90–5.33] <.001 2.52 [1.49–4.25] <.001a 
 Rural 3.45 [2.19–5.46] <.001 1.99 [1.19–3.31] .01 5.60 [4.00–7.84] <.001 3.49 [2.28–5.35] <.001a 
Income quartile of 3-digit zip code 
 Q1 (median household income <$47 200) 0.96 [0.53–1.73] .88 1.20 [0.70–2.08] .51 1.96 [1.30–2.98] .001 2.10 [1.28–3.44] .003a 
 Q2 ($47 200–$54 400) 0.63 [0.34–1.17] .14 0.93 [0.47–1.85] .83 0.80 [0.49–1.32] .39 0.99 [0.60–1.62] .96 
 Q3 ($54 400–$66 200) 0.59 [0.37–0.94] .03 0.67 [0.41–1.08] .10 0.83 [0.57–1.19] .30 0.87 [0.61–1.24] .44 
 Q4 (>$66 200) Reference  Reference  Reference  Reference  
Racial and ethnic diversity of 3-digit zip code 
 Q1 (% white <47.8%) Reference  Reference  Reference  Reference  
 Q2 (47.8%–67.0%) 2.32 [1.25–4.29] .007 1.71 [0.97–3.00] .06 1.12 [0.54–2.34] .76 0.89 [0.55–1.45] .65 
 Q3 (67.0%–81.9%) 2.57 [1.32–5.02] .006 1.62 [0.83–3.15] .16 1.76 [0.88–3.50] .11 1.26 [0.82–1.95] .29 
 Q4 (>81.9%) 4.05 [1.94–8.46] <.001 2.00 [0.96–4.14] .06 2.06 [1.02–4.14] .04 1.01 [0.60–1.69] .98 
Hospitalization within 14-d 0.96 [0.52–1.79] .89 2.23 [1.10–4.51] .03 0.73 [0.53–1.00] .05 2.30 [1.54–3.45] <.001a 
ED or UC visit within 14-d 2.06 [1.49–2.84] <.001 2.39 [1.69–3.39] <.001a 2.74 [2.12–3.52] <.001 3.26 [2.40–4.43] <.001a 

aOR, adjusted odds ratio; OR, odds ratio.

a

Significant in multivariate model using Bonferroni-adjusted P value of ≤.003 given multiple comparisons.

Pediatricians most frequently diagnosed COVID-19 infections but had the lowest prescribing rates among provider categories (Table 3). The top 100 prescribers by National Provider Identifier were responsible for over half of HCQ and over one-fourth of ivermectin prescriptions during nonrecommended periods. Specialty for the top 10 prescribers did not include Rheumatology or Infectious Diseases, as might be associated with indicated reasons for HCQ or ivermectin prescriptions.

TABLE 3

Medication Prescription Rate per COVID-19 Diagnosis by Provider Type and Specialty During Nonrecommended Periods

Hydroxychloroquine (after September 12, 2020)Ivermectin (after February 5, 2021)
Provider Type and Specialty, n (%)Clinician Who Diagnosed COVID-19 (n = 2 048 002 casesa), n (%)Clinician Who Prescribed Nonrecommended Medication (n = 777 prescriptionsa), n (%)Rate of Prescriptions per Diagnosisb, %Clinician Who Diagnosed COVID-19 (n = 1 826 406 casesa), n (%)Clinician Who Prescribed Nonrecommended Medication (n = 3588 prescriptionsa), n (%)Rate of Prescriptions per Diagnosisb, %
Pediatrics 625 027 (30.5) 39 (5.0) 0.01 563 082 (30.8) 300 (8.4) 0.05 
Nurse practitioner 415 769 (20.3) 187 (24.1) 0.04 368 463 (20.2) 1295 (36.1) 0.35 
Family practice 217 717 (10.6) 179 (23.0) 0.08 187 335 (10.3) 576 (16.1) 0.31 
Physician assistant 168 296 (8.2) 46 (5.9) 0.03 150 765 (8.3) 391 (10.9) 0.26 
Emergency medicine 276 268 (13.5) 101 (13.0) 0.04 249 671 (13.7) 380 (10.6) 0.15 
Internal medicine 33 951 (1.7) 40 (5.1) 0.12 29 456 (1.6) 126 (3.5) 0.43 
General practice 22 109 (1.1) 34 (4.4) 0.15 19 304 (1.1) 91 (2.5) 0.47 
Other 288 865 (14.1) 151 (19.4) 0.05 258 330 (14.1) 429 (12) 0.17 
Hydroxychloroquine (after September 12, 2020)Ivermectin (after February 5, 2021)
Provider Type and Specialty, n (%)Clinician Who Diagnosed COVID-19 (n = 2 048 002 casesa), n (%)Clinician Who Prescribed Nonrecommended Medication (n = 777 prescriptionsa), n (%)Rate of Prescriptions per Diagnosisb, %Clinician Who Diagnosed COVID-19 (n = 1 826 406 casesa), n (%)Clinician Who Prescribed Nonrecommended Medication (n = 3588 prescriptionsa), n (%)Rate of Prescriptions per Diagnosisb, %
Pediatrics 625 027 (30.5) 39 (5.0) 0.01 563 082 (30.8) 300 (8.4) 0.05 
Nurse practitioner 415 769 (20.3) 187 (24.1) 0.04 368 463 (20.2) 1295 (36.1) 0.35 
Family practice 217 717 (10.6) 179 (23.0) 0.08 187 335 (10.3) 576 (16.1) 0.31 
Physician assistant 168 296 (8.2) 46 (5.9) 0.03 150 765 (8.3) 391 (10.9) 0.26 
Emergency medicine 276 268 (13.5) 101 (13.0) 0.04 249 671 (13.7) 380 (10.6) 0.15 
Internal medicine 33 951 (1.7) 40 (5.1) 0.12 29 456 (1.6) 126 (3.5) 0.43 
General practice 22 109 (1.1) 34 (4.4) 0.15 19 304 (1.1) 91 (2.5) 0.47 
Other 288 865 (14.1) 151 (19.4) 0.05 258 330 (14.1) 429 (12) 0.17 

All provider type and specialties listed are physicians except nurse practitioner and physician assistant. For nurse practitioner and physician assistant, practice setting not specified. Other includes mostly physicians, and a few other provider types. Emergency Medicine is separate from Pediatric EM for COVID-19 diagnoses, although Emergency Medicine type was not available for prescriptions.

a

Among clinicians with known provider type and specialty by National Uniform Claim Committee Health Care Provider Taxonomy code.

b

Second column divided by first column, by provider type and specialty.

There were no medical claims with a prespecified ICD-10 code for ADR within 28 days of medication prescription during nonrecommend periods. In our cohort, there were only 6 instances of an ADR code, and all occurred long before the nonrecommended period.

In this large retrospective cohort of children with acute COVID-19, we found overall rates of HCQ and ivermectin prescriptions during nonrecommended periods of <1%. Although this proportion may seem reassuringly low, the overall incidence of COVID-19 translates into ineffective and potentially harmful prescriptions for many children. As of September 13, 2023, there were over 17.1 million COVID-19 cases in children <18 years in the US per Centers for Disease Control and Prevention data.36  At prescribing rates in our study of 0.03% for HCQ and 0.14% for ivermectin, this would translate into over 5000 HCQ and almost 24 000 ivermectin prescriptions. Prescribing rates were even higher depending on the state and period. Low mortality and hospitalization rates in our cohort suggest these infections were less likely to be serious, further underscoring the risks of an unnecessary prescription against any perceived benefits.

Within states, we found the adjusted odds of HCQ or ivermectin prescription increased with age, with teenagers substantially more likely to receive prescriptions than younger children. This was not surprising, given ivermectin prescription rate roughly increased with each decade of age in a prior study.13  We hypothesize younger children may have milder illness, providers may be more hesitant to prescribe medications, or they may have been more likely to receive care from pediatricians, who had lower rates of prescribing than nonpediatricians.

Patients with commercial insurance were more likely to receive a prescription for HCQ or ivermectin than those with public insurance. For ivermectin, sociodemographic risk factors were more notable and retained significance in our multivariate model. Children with 3-digit zip codes in rural areas had the highest odds of prescription – almost 3.5 times those in urban areas. Children living in 3-digit zip codes in the lowest income quartile also had twice the risk compared with those living in areas with the highest income. These findings align with several prior studies showing increased risk for nonrecommended medication overuse in pediatric patients with commercial insurance and in rural settings.37 ,38  This emphasizes the importance of further exploring underlying causes and ensuring guidance reaches parents and providers in low-income rural areas with commercial insurance. States in the south US census region had the highest HCQ and ivermectin prescribing rates, consistent with a Centers for Disease Control and Prevention study of ivermectin prescriptions from a large outpatient retail prescription database.12  This suggests messaging could be regionally targeted given complex geographic and sociopolitical factors.

Children who were hospitalized within 14-days of a COVID-19 diagnosis were more likely to be prescribed ivermectin. Utilizing this metric as a proxy for illness severity, we suspect these children may be more vulnerable to unproven therapies, given worried parents and clinicians. An UC or ED visit within 14-days was also associated with HCQ or ivermectin prescription, presumably for similar reasons; however, an UC or ED may have been the only available option for rapid access to testing and evaluation given the high proportion of children with these visits in our cohort. Interestingly, inpatient prescribing rates for nonrecommended medications in our cohort were low, suggesting guidelines were more likely to be followed in hospitalized patients. This was consistent with a previous study of children hospitalized with COVID-19 that found low rates of HCQ use after EUA revocation and use of ivermectin only in a few critically ill patients in summer 2021.2 

Nonpediatricians were far more likely to prescribe nonrecommended medications, with general practitioners having the highest prescribing rates of HCQ at 15 times and ivermectin at 9 times the rate of pediatricians. This disparity emphasizes how educational efforts might be best-directed toward nonpediatric-focused providers to reduce low-value care for children. Additionally, relatively few clinicians were responsible for large numbers of prescriptions, echoing the “pill mills” where individual clinicians have written tens of thousands of HCQ and ivermectin prescriptions through large telehealth operations.39  Further research is needed to determine whether a lack of pediatricians in low-income rural areas leads to nonevidence-based prescribing.

Nonrecommended prescriptions tapered off in 2022, especially for ivermectin, despite a subsequent increase in COVID-19 cases. Perhaps more mild illness with the ο SARS-Cov-2 variant circulating in 2022 decreased perceived need for treatment40 42  or targeted messaging and media attention against nonrecommended medications was effective.43 ,44  Also, retractions of meta-analyses on ivermectin effectiveness that relied on potentially biased or fraudulent studies had been published in the interim.5  Outside of HCQ and ivermectin, it was reassuring that prescription rates of other nonrecommended medications for COVID-19 were very low.

We did not find any ICD-10-coded ADRs within 28 days of a HCQ or ivermectin prescription in our cohort, which is plausible if these nonrecommended medications were prescribed at standard dosing for other indications. However, it is also possible these specific ICD-10 codes were underutilized or side effects not recognized while also experiencing COVID-19 symptoms.

There are several limitations of our study. With a claims database, we cannot account for SARS-COV-2 infections not reported to a healthcare provider, such as those only diagnosed with home tests. Providers may have assigned a COVID-19 ICD-10 code to a patient not acutely infected with SARS-CoV-2, such as for SARS-CoV-2 testing or a history of COVID-19. However, this would lead to an underestimate of prescribing for those with acute infections. We could not account for medications from other sources, such as veterinary preparations or prescriptions from family members. These may have had a higher chance of ADR, given nonstandard dosing. We were limited in our ability to detect adverse drug events with ICD-10 codes. We did not search for nonspecific symptom ICD-10 codes that could suggest possible ADRs because of difficulty with attribution. There is also a possibility the medication was prescribed for another indication than acute COVID-19; however, we hope this would be minimized by extensive exclusions for medication indications and narrow prescription window after diagnosis. Finally, although we captured prescriptions, we cannot determine whether the medication was taken as prescribed.

Children were prescribed ineffective and potentially harmful medications for acute COVID-19 despite national clinical guidelines. This was most notable in children of older age and with commercial insurance; however, those in southern states, low-income and nonurban areas also had increased risk. Understanding the consequences of misinformation is the first step in reducing low-value care and preventing downstream effects such as unintended shortages during future pandemics. Ensuring timely and clear evidence-based recommendations are disseminated and adhered to by all types of providers will reduce disparities, ensure quality and cost-effective pediatric care, and should be prioritized by policymakers.

The authors thank Nidia Rodriguez-Ormaza, MD, PhD, MPH, Quantitative Sciences Unit, Department of Medicine, Stanford University School of Medicine and Thomas J. Horan, MEng, MMOAS, Air Force Institute of Technology for statistical mentorship.

Dr Burns conceptualized and designed the study, conducted the primary analysis and interpreted the data, drafted the initial manuscript, and revised the manuscript; Dr Dahlen conceptualized and designed the study, conducted the primary analysis and interpreted the data, and reviewed and revised the manuscript; Dr Schroeder conceptualized and designed the study, supervised the primary analysis and interpreted the data, and reviewed and revised the manuscript; Drs Bio, Chamberlain, Bassett, Schwenk, and Teufel substantially contributed to study design, interpreted the data, and critically reviewed and revised the manuscript for important intellectual content; Ms Ramaraj substantially contributed to study design, analyzed the data, and critically reviewed and revised the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

Disclaimer: Data for this project were accessed using the Stanford Center for Population Health Sciences Data Core. The PHS Data Core is supported by a National Institutes of Health National Center for Advancing Translational Science Clinical and Translational Science Award (UL1TR003142) and from Internal Stanford funding. The content is solely the responsibility of the authors and does not necessarily represent the official views to NIH. 

FUNDING: Drs Burns and Dahlen were supported in part by the Stanford Maternal and Child Health Research Institute. No other external funding. Stanford Maternal and Child Health Research Institute had no role in the design and conduct of the study.

CONFLICT OF INTEREST DISCLOSURES: The authors have no conflicts of interest relevant to this article to disclose.

ADR

adverse drug reaction

CI

confidence interval

COVID-19

coronavirus disease 2019

ED

emergency department

EM

Emergency Medicine

EUA

Emergency Use Authorization

HCQ

hydroxychloroquine

HMO

health maintenance organization

ICD-10

International Classification of Diseases, 10th Revision

IDSA

Infectious Disease Society of America

NIH

National Institutes of Health

PIDS

Pediatric Infectious Diseases Society

UC

urgent care

1
Chiotos
K
,
Hayes
M
,
Kimberlin
DW
, et al
.
Multicenter initial guidance on use of antivirals for children with coronavirus disease 2019/severe acute respiratory syndrome coronavirus 2
.
J Pediatric Infect Dis Soc
.
2020
;
9
(
6
):
701
715
2
Burns
JE
,
Thurm
C
,
Antoon
JW
, et al
.
Medications and adherence to treatment guidelines among children hospitalized with acute COVID-19
.
Pediatrics
.
2022
;
150
(
3
):
e2022056606
3
Hinton
D
.
Letter to Dr Rick Bright re: request for emergency use authorization for use of chloroquine phosphate or hydroxychloroquine sulfate supplied from the strategic national stockpile for treatment of 2019 coronavirus disease
. Available at: https://www.fda.gov/media/136534/download. Accessed August 31, 2023
4
Hinton
D
.
Letter to Dr. Gary Disbrow re: request for revoking emergency use authorization for use of chloroquine phosphate or hydroxychloroquine sulfate supplied from the strategic national stockpile for treatment of 2019 coronavirus disease
. Available at: https://www.fda.gov/media/138945/download. Accessed August 31, 2023
5
Hill
A
,
Mirchandani
M
,
Pilkington
V
.
Ivermectin for COVID-19: addressing potential bias and medical fraud
.
Open Forum Infect Dis
.
2022
;
9
(
2
):
ofab645
6
Bhimraj
A
,
Morgan
R
,
Shumaker
A
, et al
.
Infectious diseases society of america guidelines on the treatment and management of patients with COVID-19. Version 3.8.0
. Available at: https://www.idsociety.org/COVID19guidelines#UpdateHistory. August 31, 2023
7
Bhimraj
A
,
Morgan
R
,
Shumaker
A
, et al
.
Infectious diseases society of america guidelines on the treatment and management of patients with COVID-19. Version 10.0.0
. Available at: https://www.idsociety.org/COVID19guidelines#UpdateHistory. Accessed August 31, 2023
8
Frellick
M
.
COVID-19-related HCQ shortages affected rheum patients worldwide
. Available at: https://www.medscape.com/viewarticle/940483#vp_1. Accessed September 23, 2023
9
Cornet
A
,
Andersen
J
,
Tani
C
,
Mosca
M
.
Hydroxychloroquine availability during COVID-19 crisis and its effect on patient anxiety
.
Lupus Sci Med
.
2021
;
8
(
1
):
e000496
10
Woo
E
.
How covid misinformation created a run on animal medicine
. Available at: https://www.nytimes.com/2021/09/28/technology/ivermectin-animal-medicine-shortage.html. Accessed September 23, 2023
11
Centers for Disease Control and Prevention
.
Rapid increase in ivermectin prescriptions and reports of severe illness associated with use of products containing ivermectin to prevent or treat covid-19
. Available at: https://emergency.cdc.gov/han/2021/han00449.asp. Accessed September 6, 2023
12
Lind
JN
,
Lovegrove
MC
,
Geller
AI
,
Uyeki
TM
,
Datta
SD
,
Budnitz
DS
.
Increase in outpatient ivermectin dispensing in the US during the COVID-19 pandemic: a cross-sectional analysis
.
J Gen Intern Med
.
2021
;
36
(
9
):
2909
2911
13
Chua
KP
,
Conti
RM
,
Becker
NV
.
US insurer spending on ivermectin prescriptions for COVID-19
.
JAMA
.
2022
;
327
(
6
):
584
587
14
Komodo Health, Inc.
Healthcare map
. Available at: https://www.komodohealth.com/technology/healthcare-map. Accessed September 21, 2023
15
Stanford Center for Population Health Sciences
.
Komodo Health COVID-19 (Version 0.1)
.
Redivis
;
2022
16
Snowflake, Inc.
Snowflake data cloud
. Available at: https://www.snowflake.com/en/. Accessed September 21, 2023
17
R Core Team
.
R: a language and environment for statistical computing
. Available at: https://www.R-project.org/. Published online 2022
18
Python
.
Version 3.8.5. Python Software Foundation; 2021–2023
. Available at: https://www.python.org. Accessed October 4, 2023
19
Dahlen
A
.
OSF registries: prescribing patterns of non-recommended medications for children with acute COVID-19
. Available at: https://osf.io/se6qk. Accessed September 6, 2023
20
Blatz
AM
,
David
MZ
,
Otto
WR
,
Luan
X
,
Gerber
JS
.
Validation of International Classification of Disease-10 code for identifying children hospitalized with coronavirus disease-2019
.
J Pediatric Infect Dis Soc
.
2021
;
10
(
4
):
547
548
21
Kadri
SS
,
Gundrum
J
,
Warner
S
, et al
.
Uptake and accuracy of the diagnosis code for COVID-19 among US hospitalizations
.
JAMA
.
2020
;
324
(
24
):
2553
2554
22
Lynch
KE
,
Viernes
B
,
Gatsby
E
, et al
.
Positive predictive value of COVID-19 ICD-10 diagnosis codes across calendar time and clinical setting
.
Clin Epidemiol
.
2021
;
13
:
1011
1018
23
Cullen
MR
,
Vegso
S
,
Cantley
L
, et al
.
Use of medical insurance claims data for occupational health research
.
J Occup Environ Med
.
2006
;
48
(
10
):
1054
1061
24
Tessier-Sherman
B
,
Galusha
D
,
Taiwo
OA
, et al
.
Further validation that claims data are a useful tool for epidemiologic research on hypertension
.
BMC Public Health
.
2013
;
13
:
51
25
Thyagarajan
V
,
Su
S
,
Gee
J
, et al
.
Identification of seizures among adults and children following influenza vaccination using health insurance claims data
.
Vaccine
.
2013
;
31
(
50
):
5997
6002
26
Feudtner
C
,
Feinstein
JA
,
Zhong
W
,
Hall
M
,
Dai
D
.
Pediatric complex chronic conditions classification system version 2: updated for ICD-10 and complex medical technology dependence and transplantation
.
BMC Pediatr
.
2014
;
14
:
199
27
Feudtner
C
,
Christakis
DA
,
Connell
FA
.
Pediatric deaths attributable to complex chronic conditions: a population-based study of Washington State, 1980–1997
.
Pediatrics
.
2000
;
106
(
1 Pt 2
):
205
209
28
U.S. Census Bureau
. American community survey 5-year data. Available at: https://data.census.gov. Accessed September 21, 2023
29
Lexicomp, Inc.
Ivermectin (systemic): drug information. Hydroxychloroquine: drug information. Lopinavir and ritonavir: drug information. Nitazoxanide: drug information. Colchicine: drug information
. Available at: www.uptodate.com. Accessed May 4, 2023
30
ICD-10Data.com
.
The web’s free 2023 ICD-10-CM/PCS medical coding reference
. Available at: https://www.icd10data.com. Accessed May 4, 2023
31
COVID-19 Treatment Guidelines Panel
.
National Institutes of Health coronavirus disease 2019 (COVID-19) treatment guidelines
. Available at: https://www.covid19treatmentguidelines.nih.gov/about-the-guidelines/guidelines-archive/. Accessed September 8, 2023
32
COVID-19 Treatment Guidelines Panel
.
National Institutes of Health coronavirus disease 2019 (COVID-19) treatment guidelines
. Available at: https://www.covid19treatmentguidelines.nih.gov/about-the-guidelines/guidelines-archive/. Accessed September 8, 2023
33
Hohl
CM
,
Karpov
A
,
Reddekopp
L
,
Doyle-Waters
M
,
Stausberg
J
.
ICD-10 codes used to identify adverse drug events in administrative data: a systematic review
.
J Am Med Inform Assoc
.
2014
;
21
(
3
):
547
557
34
Lambrou
AS
,
Shirk
P
,
Steele
MK
, et al
;
Strain Surveillance and Emerging Variants Bioinformatic Working Group
;
Strain Surveillance and Emerging Variants NS3 Working Group
.
Genomic surveillance for SARS-CoV-2 variants: predominance of the delta (B.1.617.2) and omicron (B.1.1.529) variants - United States, June 2021–January 2022
.
MMWR Morb Mortal Wkly Rep
.
2022
;
71
(
6
):
206
211
35
United States Census Bureau
.
Geographic levels
. Available at: https://www.census.gov/programs-surveys/economic-census/guidance-geographies/levels.html. Accessed September 15, 2023
36
Centers for Disease Control and Prevention
.
COVID-19 data tracker: demographic trends of COVID-19 cases and deaths in the US reported to CDC
. Available at: https://covid.cdc.gov/covid-data-tracker/#demographics. Accessed September 18, 2023
37
Wolf
ER
,
Sabo
RT
,
Lavallee
M
, et al
.
Overuse of reflux medications in infants
.
Pediatrics
.
2023
;
151
(
3
):
e2022058330
38
Wolf
ER
,
Richards
A
,
Lavallee
M
, et al
.
Patient, provider, and health care system characteristics associated with overuse in bronchiolitis
.
Pediatrics
.
2021
;
148
(
4
):
e2021051345
39
D’Ambrosio
A
.
Stella Immanuel highest U.S. prescriber of ivermectin and HCQ—controversial physician has doled out tens of thousands of scripts for both drugs
. Available at: https://www.medpagetoday.com/special-reports/exclusives/103353?xid=nl_mostcommented_2023-03-10&mh=07f73d99b2b546c079b0b5a2312d87b1. Accessed September 22, 2023
40
Ma
KC
,
Shirk
P
,
Lambrou
AS
, et al
.
Genomic surveillance for SARS-CoV-2 variants: circulation of omicron lineages - United States, January 2022–May 2023
.
MMWR Morb Mortal Wkly Rep
.
2023
;
72
(
24
):
651
656
41
Lewnard
JA
,
Hong
VX
,
Patel
MM
,
Kahn
R
,
Lipsitch
M
,
Tartof
SY
.
Clinical outcomes associated with SARS-CoV-2 Omicron (B.1.1.529) variant and BA.1/BA.1.1 or BA.2 subvariant infection in Southern California
.
Nat Med
.
2022
;
28
(
9
):
1933
1943
42
Menni
C
,
Valdes
AM
,
Polidori
L
, et al
.
Symptom prevalence, duration, and risk of hospital admission in individuals infected with SARS-CoV-2 during periods of omicron and delta variant dominance: a prospective observational study from the ZOE COVID Study
.
Lancet
.
2022
;
399
(
10335
):
1618
1624
43
Why You Should Not Use Ivermectin to Treat or Prevent COVID-19. U.S. Food & Drug Administration: consumer updates
. Available at: https://www.fda.gov/consumers/consumer-updates/why-you-should-not-use-ivermectin-treat-or-prevent-covid-19. Accessed September 23, 2023
44
Goldberg
E
.
Demand surges for deworming drug for Covid, despite scant evidence it works
. https://www.nytimes.com/2021/08/30/health/covid-ivermectin-prescriptions.html. Accessed September 23, 2023

Supplementary data