Real-world retrospective analysis of patient characteristics, healthcare resource utilization, costs, and treatment patterns among unvaccinated adults with COVID-19 diagnosed in outpatient settings in the United States

Abstract

Aims

This retrospective analysis of the Optum Clinformatics Data Mart database evaluated US patient characteristics, healthcare resource utilization (HCRU), costs, and treatment patterns among unvaccinated adults with outpatient-diagnosed COVID-19 to quantify US economic burden.

Materials and methods

The index event was the earliest outpatient diagnosis of confirmed COVID-19 from May 1 to December 10, 2020. Patients had 12 months’ continuous enrollment before and were followed for ≥60 days after index date until insurance dis-enrollment or study end.

Results

236,589 patients had outpatient-diagnosed COVID-19 (7,692 with and 228,897 without subsequent COVID-19-related inpatient admission >48 h post-diagnosis). The median age was 51 years (≥65 years, 30.0%); 72.4% had ≥1 risk factor. Patients with versus without subsequent inpatient admission were more often male, older, Black/Hispanic, and had comorbidities/risk factors. With a median follow-up of 162 days, patients had a median of 1 COVID-19-related outpatient visit (with inpatient admission, 5 outpatient visits). Those with inpatient admission had a median of 1 COVID-19-related inpatient visit (median length of stay [LOS], 6 days), 33.3% were admitted to intensive care (median LOS, 8 days), 8.4%, 7.1%, and 13.3% received invasive mechanical ventilation, noninvasive mechanical ventilation, and supplemental oxygen, respectively; 13.5% experienced readmission. Inpatient mortality was 6.0% (0.3% for nonhospitalized patients). Antithrombotic therapy, antibiotics, corticosteroids, and remdesivir use increased among patients with inpatient admission versus without. Median total COVID-19-related non-zero medical costs were $208 for patients without inpatient admission (with inpatient admission, $39,187).

Limitations

Results reflect the circulating SARS-CoV-2 and treatment landscape during the study period. Requirements for continuous enrollment could have biased the population. Cost measurements may have included allowed (typically higher) and charge amounts.

Conclusions

Given the numbers of the US population who are still not fully vaccinated and the evolving epidemiology of the pandemic, this study provides relevant insights on real-world treatment patterns, HCRU, and the cost burden of outpatient-diagnosed COVID-19.

Keywords:

• Real-world evidence
• HCRU
• healthcare costs
• treatment patterns
• COVID-19
• SARS-CoV-2

JEL Classification Codes:

• I10
• I1
• I
• I11

Introduction

Coronavirus disease 2019 (COVID-19) is caused by the novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The disease was first reported to the World Health Organization at the end of 2019 and spread globally within a few months¹. The Centers for Disease Control and Prevention (CDC) reported the first case of COVID-19 in the United States on January 21, 2020². As of December 15, 2021, more than 50.3 million cases and 799,850 deaths have been reported in the United States³.

The clinical presentation of COVID-19 ranges from asymptomatic infection to critical illness requiring intensive care⁴. Most cases are asymptomatic or mild, with approximately 7% requiring hospitalization as of December 15, 2021⁵. Nonhospitalized adult patients can be managed largely with supportive care⁶. SARS-CoV-2 antivirals and anti–SARS-CoV-2 monoclonal antibody products have received emergency use authorization (EUA) for nonhospitalized patients at high risk of disease progression (defined by EUA criteria)^4,7,8.

A majority of the research on patient characteristics, clinical outcomes, healthcare resource utilization (HCRU), and costs have focused on individuals hospitalized for COVID-19^9–18. A publication on the health outcomes and economic burden among patients hospitalized for COVID-19 in the United States using the Premier Healthcare COVID-19 database¹⁸ reported that of 173,942 patients hospitalized for COVID-19 from April to October 2020, 22% were admitted to the intensive care unit (ICU) and 17% received IMV. The median hospital length of stay (LOS) was 5 days, in-hospital mortality was 14%, median hospital charges were $43,986, and median costs were $12,046, all of which increased with ICU admission, IMV use, and age¹⁸.

Much less is known about outcomes, HCRU, and economic burden among patients diagnosed with COVID-19 in outpatient settings. Studies have shown that approximately 10% of these patients had a COVID-related urgent medical or emergency department (ED) visit, and 4% to 8% were hospitalized within 30 days of diagnosis^19,20. Even patients with mild to moderate infection may have persistent symptoms^21–25, and infection may exacerbate underlying chronic medical conditions, thus further impacting HCRU^26,27. Currently, few if any studies addressing the economic burden of COVID-19 diagnosed in outpatient settings or that differentiate between HCRU in the acute versus post-acute phases of the disease have been published. With the rapidly evolving pandemic landscape, for example, preliminary reports of the Omicron variant point to a less severe infection²⁸, the outpatient burden continues to be relevant.

This retrospective analysis of real-world data complements existing studies conducted in the outpatient²⁹ and inpatient settings¹⁸ by evaluating HCRU and costs among unvaccinated patients diagnosed with COVID-19 in the outpatient setting, to quantify the economic burden of disease in the United States.

Methods

Study design and data source

This study is a descriptive retrospective analysis that used the Optum Clinformatics Data Mart (CDM) database to evaluate patient characteristics, HCRU, costs, and treatment patterns among unvaccinated adults with a confirmed outpatient diagnosis of COVID-19. The Optum CDM database includes deidentified patient-level information derived from a database of administrative health claims for members of large commercial and Medicare Advantage health plans submitted for all medical and pharmacy healthcare services with information related to healthcare costs and resource utilization. The population is geographically diverse, spanning all 50 states.

This study was deemed exempt from Institutional Review Board (IRB) review pursuant to the terms of the US Department of Health and Human Service's Policy for Protection of Human Research Subjects at 45 C.F.R. 46.104(d); category 4 exemption (Sterling IRB, Boston, Massachusetts waived ethical approval for this work).

The index event was defined as the earliest outpatient diagnosis of confirmed COVID-19 during the index identification, with the corresponding service end date as the index date. The outpatient diagnosis was defined as a diagnosis made during any outpatient healthcare visit (e.g. physician office, ED, outpatient clinic), as opposed to during an inpatient hospital admission.

Study population

Adults (≥18 years old) were included if they had a primary or secondary International Classification of Diseases, Tenth Revision, Clinical Modification (ICD–10–CM) diagnosis code for COVID-19 (U07.1) in a claim for an outpatient encounter between May 1, 2020, and December 10, 2020. This cutoff date was selected based on the EUA date for the first COVID-19 vaccine (December 11, 2020); patients would not have been vaccinated during this timeframe. Patients were required to have 12 months of continuous healthcare enrollment before the index date (baseline period) and ≥60 days of continuous healthcare enrollment after the index date (follow-up period). The follow-up period varied by patient, began after the index date, and continued until insurance dis-enrollment, or the end of the study period, whichever was earliest (Figure 1). Monthly gaps were allowed for calculating continuous enrollment (months with ≥1 enrolled day are treated as wholly covered when determining continuous intervals, but the ends of the newly formed intervals reflect the actual start and end dates). Patients were excluded if they were <18 years old at the index date; had any COVID-19 diagnoses within 30 days before the index date (washout period); were admitted to a hospital for COVID-19 within 48 h of their diagnosis in the outpatient setting (i.e. considered to have severe disease at diagnosis); were admitted into a skilled nursing facility, hospice, or inpatient rehabilitation facility during the baseline or follow-up period; or had <60 days of available follow-up time. Patients were stratified by baseline risk of severe COVID-19 infection (higher risk vs standard risk) and age at index date (18–29 vs 30–49 vs 50–64 vs 65–74 vs ≥75 years).

Figure 1. Study scheme. The follow-up period varied by patient, began after the index date, and continued until insurance dis-enrollment, or the end of the study period, whichever was earliest.

Higher risk subgroup

The higher risk subgroup included patients with ≥1 characteristic or underlying medical condition associated with an increased risk of developing severe illness from COVID-19 as defined by the CDC (May 13, 2021)³⁰ and assessed during the 12 months before the index date: ≥65 years of age (age was assessed on the index date), history of cancer, chronic kidney disease, chronic lung diseases, dementia or other neurologic conditions, diabetes (type 1 or type 2), Down syndrome, cardiovascular disease, HIV infection, hypertension, immune deficiencies/immunocompromised state, liver disease, obesity, pregnancy and recent pregnancy, sickle cell disease or thalassemia, smoking (current or former), solid organ or blood stem cell transplant, stroke or cerebrovascular disease, and substance use disorders. To define these 18 conditions, sentinel code lists³¹ were used where available and refined to align with the CDC criteria for each condition. All codes were reviewed by sponsor personnel with medical and epidemiological expertise (FD, JA, NB), and discrepancies were resolved by consensus. To be included in this subset, patients needed a diagnosis code (ICD-10-CM) procedure code (ICD-10-PCS, CPT, HCPCS) or drug code (NDC) for ≥1 of the 18 conditions in the 12 months before COVID-19 diagnosis (the index event) or ≥65 years of age on the index date.

Study objectives

The primary objective was to assess clinical and demographic characteristics, HCRU and associated costs, and treatment patterns among patients diagnosed with COVID-19 in outpatient settings. The secondary objective was to assess the primary objective measures with further stratification by key patient characteristics.

Study outcomes/variables

Patient demographics were assessed during the index event and included age, sex, geographic region, payer type, index event point of service, and index event date (year and month). Comorbidities and risk factors of severe COVID-19 were evaluated during the baseline period and follow-up period (index event combined with follow-up period). Variables associated with HCRU and cost included those associated with outpatient encounters, hospitalizations, ICU use, ventilation, costs, provider specialty, and COVID-19 tests (Supplementary Table 1). All-cause and COVID-related HCRU and costs were evaluated during the index event/follow-up period (index event combined with follow-up period) and included all claims with COVID-19 as either a primary or secondary diagnosis code. Cost is modeled by Optum and not taken directly from claims. The standard cost variable was used because it estimated the allowed amount paid to providers. The non-zero cost was where the standard cost was >0; if a patient only had costs = 0, then that patient was excluded from that cost analysis. Time to the first COVID-related hospitalization, readmissions any time after the COVID-19 inpatient discharge date as well as readmission within 30 days of the index event, and outpatient visits (e.g. ED, urgent care) were determined during the follow-up period. Treatment variables included those associated with COVID-19 tests, ICU admission, treatments provided during admission, and discharge status (Supplementary Table 1). COVID-19 treatment-related patterns and 30-day mortality were evaluated during the index event and follow-up period. The number of COVID-19 tests performed during follow-up and in the first 30 days following the index event was also evaluated. Progression to outcomes of interest was evaluated overall and during specific periods following diagnosis (3–7, 8–14, 15–21, and ≥22 days). All COVID-coded visits had a confirmed COVID diagnosis code on the claim. Mortality data were recorded as month and year; exact day calculation could not be performed.

The mortality data are sourced from a reputable third-party source that includes public and private obituary data. Optum death data (MM/YYYY) within the CDM are deterministically linked using personal identifiable information (PII) to ensure an accurate match with our patient population. A variety of PII exists to provide an accurate match regardless of social security number, including but not limited to, first and last name, date of birth, and the social security number. Death dates linked to social security numbers are flagged accordingly.

Death data were refreshed monthly to integrate the most recently reported deaths into the data, retroactive to 2007, and a complete restatement of death data was available in each quarterly delivery. Optum continually quality checks the data to ensure an accurate match at the patient level with available death information. However, death data may be depressed if there is a contributing source not able to capture the death data (e.g. patient fall off) or due to the extended lag on the Social Security Administration death master file. Although the absence of a death month and year do not indicate the member is living, when business parameters (e.g. claims runout) are applied to the research, there is a high probability that the member is still living.

Data analysis

Descriptive statistics were used to summarize patient demographic and clinical characteristics, HCRU and costs, and COVID-related treatment patterns. Counts and percentages were reported for categorical variables. Continuous variables were summarized using means, standard deviations, medians, and Q1 and Q3 quartiles. Stratified analyses were generated to assess the relationship between patients diagnosed in an outpatient setting, patients at higher risk of progressing to severe COVID, and the outcomes of interest while preserving the descriptive nature of the analysis. Data were also stratified by follow-up time period (3–7, 8–14, 15–21, and ≥22 days) and age at index date (18–29 vs 30–49 vs 50–64 vs 65–74 vs ≥75 years).

Supportive analyses

A sensitivity analysis was performed using the proportion of patients who were alive after 30 days after the index date. The patient selection criteria for the sensitivity population were otherwise the same as the default study population. All other study measurements, including index event and study period definitions for the primary study analysis, were employed for the sensitivity analysis as well. The sensitivity analysis included both primary and secondary objective measures, and the follow-up time periods used for analysis were 3–7, 8–14, 15–21, and ≥22 days.

Results

Study population

Of the 18,432,067 individuals in the Optum CDM database from May 1, 2020, to December 10, 2020, 236,589 (1.28%) met the inclusion criteria for this analysis (Figure 2).

Figure 2. Patient disposition. COVID-19, coronavirus disease 2019.

Patient demographic and baseline clinical characteristics

Among the overall outpatient population, the median age was 51 years and 30.0% of patients were ≥65 years (Table 1). Most patients were women and White. The most prevalent comorbidities were gastrointestinal (GI) conditions, obesity, and chronic lung diseases; 72.4% of patients had ≥1 characteristic or underlying medical condition associated with a higher COVID-19 risk, the most common being immune deficiencies/immunocompromised state, hypertension, and age ≥65 years (Table 1).

Table 1. Patient demographic and baseline clinical characteristics.

	All patients (N = 236,589)		Patients not requiring inpatient admission (n = 228,897)		Patients requiring inpatient admission anytime during follow-up (n = 7,692)
Characteristic	n	%	n	%	n	%
Age, years
Median, Q1–Q3	51	35–68	50	34–67	68	56–76
Mean, SD	50.95	18.97	50.48	18.91	65.00	14.88
Age group, years
18–29	40,412	17.08	40,209	17.57	203	2.64
30–49	71,408	30.18	70,393	30.75	1,015	13.20
50–64	53,863	22.77	52,019	22.73	1,844	23.97
65–74	43,372	18.33	40,927	17.88	2,445	31.79
75+	27,534	11.64	25,349	11.07	2,185	28.41
Sex
Female	125,601	53.09	122,037	53.32	3,564	46.33
Male	110,985	46.91	106,857	46.68	4,128	53.67
Unknown	3	0.00	3	0.00	0	0.00
Race
White	121,621	51.41	118,141	51.61	3,480	45.24
Hispanic	37,453	15.83	36,022	15.74	1,431	18.60
Black	18,643	7.88	17,853	7.80	790	10.27
Asian	6,246	2.64	6,091	2.66	155	2.02
Unknown	52,626	22.24	50,790	22.19	1,836	23.87
Division
South Atlantic	48,381	20.45	46,412	20.28	1,969	25.60
West South Central	45,449	19.21	43,732	19.11	1,717	22.32
East North Central	35,035	14.81	33,877	14.80	1,158	15.05
West North Central	30,428	12.86	29,601	12.93	827	10.75
Mountain	25,944	10.97	25,313	11.06	631	8.20
Mid-Atlantic	20,377	8.61	20,000	8.74	377	4.90
Pacific	14,214	6.01	13,795	6.03	419	5.45
East South Central	9,969	4.21	9,559	4.18	410	5.33
New England	6,538	2.76	6,355	2.78	183	2.38
Unknown	254	0.11	253	0.11	1	0.01
Insurance type
Commercial	160,396	67.80	157,763	68.92	2,633	34.23
Medicare	76,177	32.20	71,118	31.07	5,059	65.77
Unknown	16	0.01	16	0.01	0	0.00
Antithrombotic therapy at baseline
Any antithrombotic therapy	12,981	5.49	11,937	5.22	1,044	13.57
Low–molecular-weight heparin^a	6,071	2.57	5,537	2.42	534	6.94
Factor Xa inhibitor^b	5,949	2.51	5,503	2.40	446	5.80
Warfarin	2,033	0.86	1,875	0.82	158	2.05
Thrombin inhibitor^c	542	0.23	506	0.22	36	0.47
Comorbidities at baseline^d
Gastrointestinal condition	82,839	35.01	78,942	34.49	3,897	50.66
Obesity	63,562	26.87	60,386	26.38	3,176	41.29
Chronic lung diseases	14,769	6.24	13,709	5.99	1,060	13.78
COPD, emphysema, and chronic bronchitis	8,986	3.80	8,353	3.65	633	8.23
Rheumatic disease	8,720	3.69	8,261	3.61	459	5.97
Asthma (moderate or severe)	3,056	1.29	2,859	1.25	197	2.56
Pulmonary hypertension	2,480	1.05	2,224	0.97	256	3.33
Interstitial lung disease, pulmonary fibrosis	1,861	0.79	1,687	0.74	174	2.26
Solid organ or blood stem cell transplant	198	0.08	166	0.07	32	0.42
Cystic fibrosis	38	0.02	38	0.02	0	0.00
≥1 Characteristic or underlying medical condition  associated with a higher risk of COVID-19	171,252	72.38	164,069	71.68	7,183	93.38
Immune deficiencies/immunocompromised state	121,939	51.54	116,405	50.85	5,534	71.94
Hypertension	88,229	37.29	82,952	36.24	5,277	68.60
≥65 years of age	70,906	29.97	66,276	28.95	4,630	60.19
Obesity	63,562	26.87	60,386	26.38	3,176	41.29
CVD combined	43,686	18.46	40,539	17.71	3,147	40.91
Diabetes (type 1 or 2)	41,186	17.41	38,075	16.63	3,111	40.44
Current or former smoker	24,809	10.49	23,310	10.18	1,499	19.49
Cancer history	21,718	9.18	20,381	8.90	1,337	17.38
Chronic kidney disease	18,397	7.78	16,632	7.27	1,765	22.95
Chronic lung diseases	14,769	6.24	13,709	5.99	1,060	13.78
Stroke or cerebrovascular disease	9,628	4.07	8,928	3.90	700	9.10
Dementia or other neurologic condition	5,952	2.52	5,537	2.42	415	5.40
Pregnancy and recent pregnancy	4,679	1.98	4,476	1.96	203	2.64
Substance use disorder	4,649	1.97	4,420	1.93	229	2.98
Liver disease	1,402	0.59	1,297	0.57	105	1.37
HIV infection	777	0.33	756	0.33	21	0.27
Sickle cell disease or thalassemia	452	0.19	426	0.19	26	0.34
Solid organ or blood stem cell transplant	198	0.08	166	0.07	32	0.42
Down syndrome	68	0.03	63	0.03	5	0.07

Abbreviations. COPD, chronic obstructive pulmonary disease; COVID-19, coronavirus disease 2019; CVD, cardiovascular disease; Q1, first quartile; Q3, third quartile.

^aEnoxaparin, dalteparin, and heparin.

^bApixaban, fondaparinux, rivaroxaban, and edoxaban.

^cBivalirudin, argatroban, dabigatran, and antithrombin III.

^dIncludes baseline COVID-19 risk conditions.

Data by time to admission are in Supplementary Table 2.

Among patients who did not require later COVID-related inpatient hospital admission, the median age was 50 years, and 29.0% of patients were ≥65 years (Table 1). Most patients were women and White. The most prevalent comorbidities were GI conditions, obesity, and chronic lung diseases; 71.7% of patients had ≥1 characteristic or underlying medical condition associated with a higher COVID-19 risk, the most common being immune deficiencies/immunocompromised state, hypertension, and age ≥65 years (Table 1).

Most patients who later required COVID-related inpatient hospital admission were men older than 65 years and had Medicare. The proportions of Black and Hispanic patients were slightly higher in the hospitalized group compared with the nonhospitalized group, as were the proportions of patients who had comorbidities at baseline, notably GI conditions, obesity, chronic lung diseases, and chronic obstructive pulmonary disease (COPD). The proportions of patients who had GI conditions, chronic lung diseases, and COPD were higher still in patients who were admitted ≥22 days after outpatient diagnosis than in those who did not require admission for COVID. The proportion of patients with a higher risk of severe COVID-19 was also higher among hospitalized patients compared with patients who did not require admission for COVID (Supplementary Table 2).

COVID-19 testing patterns

At baseline, COVID-19 tests were documented in 38.3% of patients in this analysis overall, 38.4% of patients without inpatient hospital admission for COVID-19 anytime in follow-up, and 35.2% of those with inpatient admission for COVID-19 (Supplementary Figure 1, Supplementary Table 3). When stratified by age (the inpatient group), the mean numbers of tests documented at baseline ranged from 0.5 to 0.76, decreasing by age (Supplementary Table 4). Results were available in very few patients: 10.3%, 10.4%, and 8.0%, respectively. The median number of days from COVID-19 test result date to outpatient diagnosis (index date) was 17.0 (interquartile range [IQR], 6.0–68.0) days (Supplementary Table 3).

During follow-up, COVID-19 tests were documented in 84.7% of patients overall and without inpatient admission and in 84.9% of those with admission (Supplementary Table 5). When stratified by age (the inpatient group), the mean numbers of tests documented during follow-up ranged from 1.67 to 1.99, generally decreasing by age (Supplementary Table 4). Again, results were available in very few patients: 16.7%, 16.8%, and 14.7%, respectively. Among those who later required inpatient hospital admission, the median number of days from COVID-19 test result date during follow-up to first COVID-19 inpatient visit was 6.0 (IQR, 4.0–9.0) days (Supplementary Table 6).

COVID-19–related HCRU

The median number of COVID-19–related outpatient visits at any time during follow-up (median [IQR] available follow-up data, 162 [132–237] days) was 1 (IQR, 1–2) in both the overall population and patients without COVID-19–related inpatient admissions. Median visit number was also generally 1 across age groups (2 for those aged ≥75 years; Supplemental Table 4) and 5 (IQR, 3–10) among those with COVID-19–related inpatient admissions (median [IQR] available follow-up data, 156 [129–234] days). Very few COVID-related telehealth, urgent care, or noninpatient ED visits were reported (Table 2). For reference, the median number of outpatient visits for any reason was 8 (IQR, 4–14) in the overall population, 7 (IQR, 4–15) in patients without COVID-19–related inpatient admissions (and ranged from 5 to 13 by age group, increasing with age; Supplementary Table 4), and 18 (IQR, 10–31) among those with COVID-19–related inpatient admissions. Very few telehealth, urgent care, or inpatient ED visits for any reason were reported (Supplementary Table 7).

Table 2. COVID-19–related outpatient HCRU.

	All patients (N = 236,589)		Patients not requiring inpatient admission (n = 228,897)		Patients requiring inpatient admission anytime during follow-up (n = 7,692)
Number of outpatient visits
Patients, n (%)	236,589	(100)	228,897	(100)	7,692	(100)
Median, Q1–Q3	1	1–2	1	1–2	5	3–10
Mean, SD	2.27	3.32	2.08	2.78	7.83	8.79
Number of telehealth visits
Patients, n (%)	236,589	(100)	228,897	(100)	7,692	(100)
Median, Q1–Q3	0	0–0	0	0–0	0	0–0
Mean, SD	0.03	0.27	0.03	0.24	0.11	0.73
Number of urgent care visits
Patients, n (%)	236,589	(100)	228,897	(100)	7,692	(100)
Median, Q1–Q3	0	0–0	0	0–0	0	0–0
Mean, SD	0.06	0.29	0.06	0.29	0.12	0.39
Number of ED visits
Patients, n (%)	236,589	(100)	228,897	(100)	7,692	(100)
Median, Q1–Q3	0	0–0	0	0–0	0	0–1
Mean, SD	0.20	0.48	0.19	0.46	0.58	0.72

Abbreviations. COVID-19, coronavirus disease 2019; ED, emergency department; HCRU, healthcare resource utilization; Q1, first quartile; Q3, third quartile.

Data by time to admission are in Supplementary Table 7.

The median number of COVID-19–related outpatient visits during the first 30 days was 1 (IQR, 1–2) in both the overall population and patients without COVID-19–related inpatient admissions (median number of visits was 1 across age groups; Supplementary Table 4) and 3 (IQR, 2–6) among those with COVID-19–related inpatient admissions. The median number of COVID-related noninpatient telehealth, urgent care, or ED visits was 0 (Supplementary Table 8).

The most common specialties of outpatient healthcare providers seen by patients at the index visit in this analysis were family medicine physicians, clinical medical laboratories, emergency medicine physicians, and internal medicine physicians (Supplementary Table 9). The type of provider was unknown for 30,196 patients (12.7%).

Patients who required a COVID-19 inpatient visit had a median of 1 (IQR, 1–1) visit across the entire follow-up time and across all time periods evaluated. Median time from diagnosis is shown in Figure 3, Table 3, and Supplementary Table 2. Median LOS ranged between 5 to 6 days across all time points. Among all patients with inpatient hospital admission at any time, 33.3% were admitted to the ICU, which was consistent across all time periods. Median ICU LOS ranged between 5 to 8 days across all time points (Table 3). For reference, the median number of inpatient visits for any reason was 0 (IQR, 0–0) in both the overall population and patients without COVID-19–related inpatient admissions and 1 (IQR, 1–1) among those with COVID-19–related inpatient admissions.

Figure 3. COVID-19–related HCRU among patients with inpatient hospital admission(s). Abbreviations. COVID-19, coronavirus disease 2019; ICU, intensive care unit; LOS, length of stay; Q1, first quartile; Q3, third quartile.

Table 3. COVID-19–related inpatient HCRU.

	All patients (N = 236,589)		Patients not requiring inpatient admission (n = 228,897)		Patients requiring inpatient admission anytime during follow-up (n = 7,692)
Number of inpatient visits
Median, Q1–Q3	0	0–0	0	0–0	1	1–1
Mean, SD	0.04	0.21	0.00	0.00	1.10	0.35
Days from outpatient diagnosis to inpatient admission
Patients, n (%)	7,692	(3.3)	0	(0)	7,692	(100)
Median,  Q1–Q3	7	5–11	–	–	7	5–11
Mean, SD	18.98	38.26	–	–	18.98	38.26
Inpatient LOS
Patients, n (%)	7,692	(3.3)	0	(0)	7,692	(100)
Median, Q1–Q3	6	4–9	–	–	6	4–9
Mean, SD	7.86	7.21	–	–	7.86	7.21
Inpatient non-ICU LOS
Patients, n (%)	5,131	(2.2)	0	(0)	5,131	(66.7)
Median, Q1–Q3	5	4–7	–	–	5	4–7
Mean, SD	6.31	4.29	–	–	6.31	4.29
Inpatient ICU LOS
Patients with ICU admission, n (%)	2,561	(1.1)	–	–	2,561	(33.3)
Median, Q1–Q3	8	5–13	–	–	8	5–13
Mean, SD	10.96	10.24	–	–	10.96	10.24

Abbreviations. COVID-19, coronavirus disease 2019; HCRU, healthcare resource utilization; ICU, intensive care unit; LOS, length of stay; Q1, first quartile; Q3, third quartile.

Data by time to admission are in Supplementary Table 10.

Invasive mechanical ventilation (including extracorporeal membrane oxygenation) was used in 8.4% of all hospitalized patients and ranged between 5.9% to 10.3% across time points (Table 4 and Supplementary Table 4). Noninvasive mechanical ventilation was used in 7.1% of all hospitalized patients and ranged between 5.3% to 7.9% across time points. Supplemental oxygen was used in 13.3% of all hospitalized patients and ranged between 9.8% to 14.3% across time points. The most common discharge status (excluding those discharged to further care) among all hospitalized patients was ‘discharge to home or self-care (routine discharge),’ followed by ‘discharged to home under the care of home health service organization.’ The rate of readmission among all hospitalized patients was 13.5% and ranged between 12.7% to 16.4% across time points. Mortality rates (keeping the limitations as discussed previously in mind) were 0.5% in the overall population, 0.3% among patients without inpatient admission (ranging from 0.01% to 1.81% across age groups, increasing by age; Supplementary Table 4), 6.0% following discharge among all hospitalized patients, and as high as 13.7% among those with inpatient admission ≥22 days after outpatient diagnosis (Supplementary Table 12).

Table 4. COVID-19 treatment characteristics.

	All patients (N = 236,589)		Patients not requiring inpatient admission (n = 228,897)		Patients requiring inpatient admission anytime during follow-up (n = 7,692)
Treatment characteristic	n	%	n	%	n	%
Invasive mechanical ventilation during COVID-19–related hospitalization^a	646	0.27	0	0	648	8.42
Noninvasive mechanical ventilation during COVID-19–related hospitalization	546	0.23	0	0	546	7.10
Supplemental oxygen during COVID-19–related hospitalization	1,022	0.43	0	0	1,022	13.29
Death after discharge from COVID-19–related hospitalization	461	0.19	0	0	461	5.99
All-cause inpatient readmission after discharge from COVID-19–related hospitalization	1,041	0.44	0	0	1,041	13.53
Mortality during follow-up	1,246	0.53	785	0.34	461	5.99
Discharge status from COVID-19–related hospitalization
Discharged to home or self-care (routine discharge)	5,189	2.19	0	0.00	5,189	67.46
Discharged to home under care of home health service organization	1,357	0.57	0	0.00	1,357	17.64
Discharged/transferred to a short-term general hospital for inpatient care	135	0.06	0	0.00	135	1.76
Discharged/transferred to long-term care hospital	46	0.02	0	0.00	46	0.60
Discharged to home hospice	33	0.01	0	0.00	33	0.43
Discharged to medical hospice facility	30	0.01	0	0.00	30	0.39
Discharged/transferred to skilled nursing facility with Medicare	30	0.01	0	0.00	30	0.39
Left against medical advice or discontinued care	30	0.01	0	0.00	30	0.39
Transferred to intermediate care facility	15	0.01	0	0.00	15	0.20
Not discharged or expected to return for outpatient services	14	0.01	0	0.00	14	0.18
Discharged/transferred to an inpatient rehabilitation facility inc [sic] hospital	11	0.00	0	0.00	11	0.14
Discharged/transferred within institution	7	0.00	0	0.00	7	0.09
Reserved for national assignment	6	0.00	0	0.00	6	0.08
Discharged/transferred to a psychiatric hospital or dist pt [sic]	4	0.00	0	0.00	4	0.05
Discharged/transferred to another type of healthcare institution	4	0.00	0	0.00	4	0.05
Discharge/transferred to a critical access hospital	2	0.00	0	0.00	2	0.03
Discharged/transferred to a nursing facility certified under Medicare	2	0.00	0	0.00	2	0.03
Admitted as inpatient to this hospital	1	0.00	0	0.00	1	0.01
Patients treated in an outpatient setting only	228,897	96.75	228,897	100.00	0	0.00
Not available from the source	416	0.18	0	0.00	416	5.41
Unknown discharge status	360	0.15	0	0.00	360	4.68

Abbreviation. COVID-19, coronavirus disease 2019.

^aIncludes extracorporeal membrane oxygenation.

Data by time to admission are in Supplementary Table 12.

Drug usage

Among the overall population, the most frequently used therapies were antibiotics and corticosteroids. The proportions of patients who used antithrombotic therapy, antibiotics, corticosteroids, and remdesivir were higher among patients who required inpatient admission compared with those who did not require admission. These proportions were highest among patients with inpatient admission between 8 to 21 days after diagnosis, with the exception of remdesivir, which was highest among patients with admission 3 to 7 days after diagnosis (Supplementary Table 13). The proportion of patients who were alive after 30 days after the index date generally followed the same pattern of drug usage as their respective overall patient populations (Supplementary Table 14).

Outpatient and inpatient costs associated with COVID-19 infection

Median non-zero standard costs of COVID-19–related outpatient encounters were $199 (IQR, $118–$494) in the overall population (236,431/236,589; 99.9%), $194 (IQR, $118–$438) among patients who did not require inpatient admission (228,741/228,897; 99.9%), and $2,276 (IQR, $618–$4,944) among those who required admission (7,690/7,692; 99.9%; Table 5). Median non-zero standard costs of COVID-19–related inpatient encounters were $36,460 (IQR, $32,891–$38,386) among all patients with inpatient admissions (7,687/228,897; 3.4%). Median standard non-zero costs of COVID-19–related prescriptions were $16 (IQR, $8–$36) among patients without inpatient admissions (79,074/228,897; 34.5%) and $34 (IQR, $13–$313) among patients with inpatient admissions (5,900/7,692; 76.7%). Median total COVID-19–related medical non-zero standard costs (including inpatient and outpatient encounters and prescriptions) were $220 (IQR, $124–$611) among the overall population (236,493/236,589; 99.9%), $208 (IQR, $121–$509) among those without inpatient admissions (228,891/228,897; 99.9%), and $39,187 (IQR, $36,071–$44,994) among those with admissions (7,692/7,692; 100%; Table 5). Non-zero costs for outpatient and inpatient visits for any reason other than COVID-19 tended to be much higher than those only for COVID-19 (Supplementary Table 16). Non-zero costs for the proportion of patients who were alive after 30 days after the index date generally were similar to the primary analysis groups (Supplementary Table 17).

Table 5. COVID-19–related outpatient and inpatient costs.

Cost	All patients (N = 236,589)		Patients not requiring inpatient admission (n = 228,897)		Patients requiring inpatient admission anytime during follow-up (n = 7,692)
Standard cost of outpatient encounters, N, missing value	236,589	0	228,897	0	7,692	0
Median, Q1–Q3	$198	$118–$493	$194	$117–$437	$2,274	$618–$4,943
Mean, SD	$1,077	$6,993	$901	$4,272	$6,300	$30,542
Standard cost of telehealth visits, N, missing value	236,589	0	228,897	0	7,692	0
Median, Q1–Q3	$0	$0–$0	$0	$0–$0	$0	$0–$0
Mean, SD	$3	$30	$3	$26	$12	$84
Standard cost of urgent care visits, N, missing value	236,589	0	228,897	0	7,692	0
Median, Q1–Q3	$0	$0–$0	$0	$0–$0	$0	$0–$0
Mean, SD	$15	$85	$14	$83	$32	$130
Standard cost of outpatient ED visits	236,589	0	228,897	0	7,692	0
Median, Q1–Q3	$0	$0–$0	$0	$0–$0	$0	$0–$1,762
Mean, SD	$337	$1,119	$313	$1,066	$1,055	$2,043
Standard cost of inpatient encounters, N, missing value	236,589	0	228,897	0	7,692	0
Median, Q1–Q3	$0	$0–$0	$0	$0–$0	$36,459	$32,866 – $38,382
Mean, SD	$1,407	$10,945	$0	$0	$43,289	$43,261
Standard cost of prescription claims, N, missing value	236,589	0	228,897	0	7,692	0
Median, Q1–Q3	$0	$0–$9	$0	$0–$8	$19	$3–$99
Mean, SD	$79	$460	$71	$441	$319	$822
Total standard medical costs (prescription claims, inpatient, and outpatient encounters), N, missing value	236,589	0	228,897	0	7,692	0
Median, Q1–Q3	$219	$124–$611	$208	$121–$509	$39,187	$36,071–$44,994
Mean, SD	$2,563	$13,969	$972	$4,309	$49,908	$55,970
Non-zero standard cost of outpatient encounters, N, missing value	236,431	158	228,741	156	7,690	2
Median, Q1–Q3	$199	$118–$494	$194	$118–$438	$2,276	$618–$4,944
Mean, SD	$1,077	$6,995	$902	$4,273	$6,302	$30,546
Non-zero standard cost of telehealth visits, N, missing value	5,670	230,919	5,140	223,757	530	7,162
Median, Q1–Q3	$85	$60–$128	$84	$60–$126	$126	$81–$212
Mean, SD	$123	$151	$118	$132	$180	$267
Non-zero standard cost of urgent care visits, N, missing value	13,013	223,576	12,234	216,663	779	6,913
Median, Q1–Q3	$194	$128–$318	$194	$128–$311	$241	$158–$384
Mean, SD	$272	$246	$269	$244	$318	$275
Non-zero standard cost of outpatient ED visits, N, missing value	40,485	196,104	36,871	192,026	3,614	4,078
Median, Q1–Q3	$1,635	$1,029–$2,489	$1,608	$1,011–$2,466	$1,844	$1,131–$2,700
Mean, SD	$1,967	$2,027	$1,940	$1,974	$2,245	$2,492
Non-zero standard cost of inpatient encounters, N, missing value	7,687	228,902	0	228,897	7,687	5
Median, Q1–Q3	$36,460	$32,891–$38,386	–	–	$36,460	$32,891–$38,386
Mean, SD	$43,317	$43,261	–	–	$43,317	$43,261
Non-zero standard cost of prescription claims, N, missing value	84,974	151,615	79,074	149,823	5,900	1,792
Median, Q1–Q3	$17	$8–$39	$16	$8–$36	$34	$13–$313
Mean, SD	$219	$748	$205	$732	$416	$917
Total non-zero standard medical costs (prescription claims, inpatient, and outpatient encounters), N, missing value	236,493	96	$228,801	96	7,692	0
Median, Q1–Q3	$220	$124–$611	$208	$121–$509	$39,187	$36,071–$44,994
Mean, SD	$2,564	$13,971	$972	$4,310	$49,908	$55,970

Abbreviations. COVID-19, coronavirus disease 2019; ED, emergency department; Q1, first quartile; Q3, third quartile.

Data by time to admission are in Supplementary Table 14.

Clinical characteristics of patients hospitalized and not hospitalized with COVID-19 infection

For all patients and those not requiring an inpatient admission, approximately one-third had hypertension or GI conditions and one-fifth were obese. For those who did require an inpatient admission, 78% (range throughout follow-up, 78%) had hypertension, 66% had GI conditions (range throughout follow-up, 64%–73%), 60% had cardiovascular disease (range throughout follow-up, 58%–70%), and approximately 46% had diabetes (range throughout follow-up, 43%–47%) or were obese (50%; range throughout follow-up, 39%–51%) (Table 6 and Supplementary Table 19). For the proportion of patients who were alive after 30 days after the index date, the characteristics followed a similar ranking. However, the percentages of each characteristic were lower in this subpopulation (Supplementary Table 20).

Table 6. Clinical characteristics of patients hospitalized and not hospitalized with COVID-19 infection.

	All patients (N = 236,589)		Patients not requiring inpatient admission (n = 228,897)		Patients requiring inpatient admission anytime during follow-up (n = 7,692)
Characteristic	n	%	N	%	n	%
Hypertension	80,092	33.85	74,079	32.36	6,013	78.17
Gastrointestinal condition	68,675	29.03	63,625	27.80	5,050	65.65
Obesity	48,755	20.61	44,941	19.63	3,814	49.58
Cardiovascular disease	41,914	17.72	37,270	16.28	4,644	60.37
Diabetes	39,118	16.53	35,559	15.53	3,559	46.27
Cancer history	18,258	7.72	16,803	7.34	1,455	18.92
Chronic kidney disease	16,741	7.08	14,567	6.36	2,174	28.26
Chronic lung disease	12,171	5.14	10,497	4.59	1,674	21.76
Cerebrovascular disease	7,532	3.18	6,701	2.93	831	10.80
Rheumatic disease	7,136	3.02	6,681	2.92	455	5.92
COPD	6,281	2.65	5,454	2.38	827	10.75
Neurologic disease	5,559	2.35	4,961	2.17	598	7.77
Asthma (moderate or severe)	2,641	1.12	2,394	1.05	247	3.21
Pulmonary hypertension	2,332	0.99	1,908	0.83	424	5.51
Interstitial lung disease, pulmonary fibrosis	2,248	0.95	1,745	0.76	503	6.54
Chronic liver disease	1,271	0.54	1,119	0.49	152	1.98
HIV infection	707	0.30	682	0.30	25	0.33
Sickle cell/anemia disease	364	0.15	333	0.15	31	0.40
Organ/stem cell transplantation	148	0.06	116	0.05	32	0.42
Cystic fibrosis	33	0.01	30	0.01	3	0.04

Abbreviations. COPD, chronic obstructive pulmonary disease; COVID-19, coronavirus disease 2019.

Data by time to admission are in Supplementary Table 19.

Discussion

This retrospective analysis of the Optum CDM database describes patient characteristics, HCRU, costs, and treatment patterns among unvaccinated adults with a confirmed outpatient COVID-19 diagnosis. The patient population included here was largely similar to that from previously published analyses of patients diagnosed with COVID-19 in outpatient settings^19,20 and to that from the Premier study on hospitalized patients¹⁸, except for the percentage of patients who were White (51.4% vs 77.5% reported by Wei et al.²⁰ and 32.7% reported in the Premier study¹⁸) It is also important to note that the prospective study from the Johns Hopkins Health System included only 118 patients¹⁹. The proportion of patients with COVID-19 risk factors in our analysis was similar to that in the study by Wei et al.²⁰; however, our analysis included a lower percentage with obesity (20.6% vs 43.5%).

The percentage of patients who later required inpatient hospital admission in our analysis (3.3%) was similar to those in the previous analyses (7.6% and 3.8%)^19,20. These patients were older, had a higher incidence of key comorbidities or COVID-19 risk factors, and were slightly more often Black or Hispanic compared with those who did not require inpatient admission^19,20. Prospective studies are needed to understand the predictors of outpatient COVID-19 progressing to severe disease.

The median number of outpatient visits was increased among patients with inpatient hospital admissions, and most who were admitted had 1 inpatient visit 7 days after outpatient diagnosis with a LOS of 5 to 6 days, similar to the Premier study (5 days)¹⁸. Late inpatient admissions occurred at a median of 67 days after diagnosis. Similar to the hospitalized patients included in the Premier study¹⁸, approximately one-third of admitted patients received ICU care, with a median LOS between 5 to 8 days. IMV and noninvasive mechanical ventilation each were used in <10% of admitted patients, and oxygen supplementation was used in 13% of admitted patients. Among the 128,063 patients in the Premier study who did not require ICU admission or IMV, 31% received supplemental oxygen, 4% received noninvasive ventilation, and 2% utilized a high-flow nasal cannula during their stay. Among the 16,496 who were admitted to the ICU but did not require IMV, 53% received supplemental oxygen, 18% received noninvasive ventilation, and 5% utilized a high-flow nasal cannula during their stay¹⁸, although oxygen supplementation was higher than in our analysis. Use of antithrombotic therapy, antibiotics, corticosteroids, and remdesivir increased among patients in our analysis who later required inpatient admission compared with those who did not.

Mortality rates were low in the overall population: 6.0% among hospital-admitted patients and 13.7% among those with late admissions (≥22 days after outpatient diagnosis). Patients in the Premier study had a similar in-hospital mortality rate (13.6%)¹⁸.

Median non-zero standard costs of COVID-19–related outpatient encounters in our analysis were $199 in the overall population, and median non-zero standard costs of COVID-19–related inpatient encounters were $30,546 among hospitalized patients. It is important to note that these cost measurements may have included both allowed and charge amounts and pre-adjudicated charge amounts are typically higher than the final reimbursed amount. In the Premier study, median hospital charges were $43,986 and costs were $12,046¹⁸. As expected, median costs of COVID-19–related outpatient encounters, prescriptions, and total COVID-19–related medical costs in our analysis were higher in patients who later required inpatient hospital admission compared with those who did not. COVID-19–related inpatient encounters in this analysis are generally 2‒4 times higher than previously published estimates of seasonal influenza hospitalization costs from high-income countries^32–34.

This study evaluated costs associated with COVID-19 diagnosed in outpatient settings using real-world data, and it provides more robust data on patient characteristics, HCRU, and treatment patterns in this unvaccinated patient population. While clinical trials are controlled and not representative of real-world practice, the data used in this study are therefore more representative of factors such as the patient population, patient histories, and prescribing patterns, given the controlled nature of clinical trials. The comprehensive nature of the CDM data allowed for precise determination of COVID-19 risk status during baseline. The CDM data also allowed for the determination of treatment profiles that reflect real-world prescribing practices, to provide an understanding of the real-world treatments used in the management of outpatient COVID-19. The large sample size included in the CDM database provided a robust view of the economic burden and treatment patterns associated with outpatient-diagnosed COVID-19.

The index identification period for this analysis ended before EUA of the first SARS-CoV-2 vaccine on December 11, 2020, thereby excluding any patients who may have received vaccination for this disease, because vaccination status may not be reliably captured in the database (neither insurance coverage nor electronic health record tracking was required during the vaccine rollout).

The first limitation of this analysis is the absence of a comparator group because, in our scientific opinion and modeling experience, we think the data is not yet evolved to the steady state needed to provide clinically meaningful matches (propensity score and other probabilistic methods). These study results reflect the circulating SARS-CoV-2 and treatment landscape present during the study period and may not be applicable to newer variants. The study focused on the first wave of COVID-19, before vaccination in the context of the US healthcare system; COVID-19 impacted different countries at different time points and the patient pathway in other countries is not generalizable to the US patients; we, therefore, have not made direct comparisons with studies from other countries. Patients were included in this analysis if they had a confirmed COVID-19 diagnosis after May 1, 2020, because no specific diagnosis code for COVID-19 existed before April 1, 2020. Clinical trial participation may not be reliably recorded in the data source. The requirement for continuous enrollment, particularly in the post-index period during the pandemic, could have biased the population toward healthier, higher-income patients because they would likely have been less impacted by the increased unemployment (and consequent loss of health insurance) during the pandemic. Several limitations of this analysis are associated with the nature of the claims database (e.g. incomplete medical histories, virtual visits/telephone/email-based healthcare encounters not captured, inpatient pharmacy claims not captured, billing/coding errors that would not have been identified because the deidentified records do not allow for validation). Medicaid, VA and uninsured patients are not cited by the vendor as being included in the Optum CDM database. Testing data in this database were incomplete; therefore, we expect the measured testing patterns in our analysis to be underestimated. An additional limitation was that a confirmed COVID ICD code did not require a positive laboratory test. Also, it was possible that the testing results were received after the COVID diagnosis code was entered, as the testing landscape in the United States at the time of data collection was experiencing slow turnaround times. As mentioned earlier, cost measurements may have included both allowed and charge amounts. Additionally, indirect costs associated with COVID-19 infection were not available. There are other societal costs related to the overall burden that was not captured in this analysis, including items such as loss of productivity, which would be expected to be substantial in the outpatient-only group as most were within the working-age range of 18–64 years. Finally, mortality data were reported monthly; therefore, the exact dates of death are unknown.

Conclusions

The results of this retrospective claims database study complement previous analyses of hospitalized patients and patients diagnosed in the outpatient setting and provide relevant insights into the real-world treatment patterns, HCRU, and cost burden of outpatient-diagnosed COVID-19. Given the numbers of the US population who are still not fully vaccinated, and the evolving epidemiology of the pandemic, such as the potential lesser severity of the Omicron variant²⁸, more research is needed to evaluate HCRU and cost in the acute and post-acute³⁵ phases of COVID-19 for this outpatient-diagnosed patient population.

Transparency

Declaration of funding

This work was supported by Pfizer Inc.

Declaration of financial/other interests

All authors are employees of Pfizer Inc and may hold stock or stock options.

Peer reviewers on this manuscript have received an honorarium from JME for their review work but have no other relevant financial relationships to disclose.

Author contributions

All authors were involved in the conception and design or analysis and interpretation of the data, the drafting of the paper or revising it critically for intellectual content, and the final approval of the version to be published. All authors agree to be accountable for all aspects of the work.

Acknowledgements

The authors thank Manuela di Fusco, Michael Benigno, Jennifer Nguyen, Joanna Atkinson, and Wajeeha Ansari, of Pfizer Inc, for their support in the study design and their contributions to this paper. Programming support and expertise were provided by Ewa Śleszyńska-Dopiera (Quanticate) and Ziphora Sam of Eliassen Group (Somerset, NJ). This study was sponsored by Pfizer Inc. Editorial/medical writing support was provided by Erin O’Keefe, PhD, and Rozena Varghese, PharmD, CMPP, of ICON plc (Blue Bell, PA), and was funded by Pfizer Inc.

Data availability statement

The data that support the findings of this study are available from the corresponding author, AMS, upon reasonable request.