Background

Post-COVID conditions are characterized by symptoms of COVID-19 that extend beyond the initial recovery from acute illness [1,2,3]. Presentation is highly variable across patients and may include residual symptoms that persist after acute illness, persistent organ dysfunction, or new symptoms or syndromes that develop after initial recovery from COVID-19 [1,2,3]. Post-COVID conditions often affect multiple organ systems, persist for several months, and have a substantial impact on daily functioning and productivity [4]. The clinical diagnosis, termed post-acute sequelae of COVID-19 (PASC), was officially defined and assigned an International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM), code (U09.9) in October 2021 [5, 6], but a widely accepted case definition and associated symptom timeframe is still under development.

Multiple studies have identified post-COVID conditions as a frequent result of COVID-19 illness [1, 7,8,9,10], but prevalence estimates vary. The Centers for Disease Control and Prevention (CDC) recently reported that 38.2% of patients diagnosed with COVID-19 experienced ≥ 1 post-COVID condition from 30 to 365 days following diagnosis [10]. The World Health Organization reports a range from 10% to 20% of patients diagnosed with COVID-19 [2], and other studies have reported rates of up to 57% 6 months after diagnosis [8, 11,12,13,14]. Heterogeneous cohorts likely contribute to varying results given that certain characteristics, such as female sex and older age, are associated with developing long-term sequelae [15, 16].

Although the risk of post-COVID conditions has been associated with more severe cases of acute illness, it can occur across all levels of COVID-19 severity [1, 15, 17]. In a systematic review of > 2000 studies conducted worldwide among patients with varying disease severity, more than half of patients who recovered from acute illness experienced long-term sequelae for ≥ 6 months after diagnosis [8]. In 2 studies from Germany and the Faroe Islands of mainly nonhospitalized patients with asymptomatic or mild to moderate acute illness, approximately 28% to 53% of patients experienced persistent symptoms of COVID-19 months later [11, 13].

The potential for post-COVID conditions to develop in patients with mild or moderate illness is particularly relevant as new variants emerge, such as the highly transmissible Omicron variant, which is associated with milder illness compared with earlier strains [18,19,20,21]. In the context of these variants, immunity from natural infection or vaccination may provide better protection against severe disease than against overall infection [21]. Moreover, evidence suggests that risk of developing long-term symptoms may vary by circulating strain [22, 23]. The effects of these changes on population-level incidence of post-COVID conditions remain to be determined, but it is likely that the true burden of COVID-19 in terms of reduced quality of life, loss of productivity, strains on healthcare systems, and economic impact reaches far beyond acute infection.

Emerging data on long-term COVID-19 impact indicate a substantial healthcare burden. In 2 studies using the US Department of Veterans Affairs database, patients with COVID-19 compared with a control cohort had increased healthcare resource and medication use, as well as abnormalities across multiple organ systems, during the year after diagnosis [24, 25]. Other studies in patients with COVID-19 have identified increases in COVID-19─related healthcare provider visits, emergency department (ED) visits, and readmissions for up to 7 months after diagnosis [26,27,28]. However, comparison studies are often limited by the small numbers of patients and high variability between patient characteristics.

The aim of this retrospective analysis was to describe morbidity, healthcare resource use, and costs associated with the post-acute phase of COVID-19 among patients with underlying medical conditions or characteristics associated with higher risk of severe COVID-19 (hereafter referred to as “high risk”) in the United States [29]. This population was selected based on immediate relevance of the data to emerging COVID-19 treatments, which are authorized first for high-risk patients. To maintain focus on descriptive results and the guiding of hypothesis generation for future research, no formal comparisons were planned.

Methods

Study design and data source

This was a descriptive, retrospective, cohort study in patients diagnosed with COVID-19 between April 1 and May 31, 2020, in which each patient served as their own control for evaluation of diagnoses, medications, healthcare utilization, and costs before versus after acute COVID-19. Patients were identified using administrative health claims from Optum’s de-identified Clinformatics® Data Mart Database (CDM), which contains de-identified patient-level information derived from administrative healthcare claims from commercial and Medicare Advantage health plan members in all 50 states. Claims encompass medical and pharmacy healthcare services and include information regarding healthcare costs and resource utilization.

The index period was April 1 to May 31, 2020, and the index date was defined as the date of the first COVID-19 diagnosis. Individual patient data were described during the 12 months before the index date (baseline phase) and during the 12 months after the end of the 30-day acute phase [3] of COVID-19 illness (Fig. 1). For patients whose hospital stays spanned across study phases, total numbers of events and associated costs were calculated per hospitalization day and were attributed to each phase based on the number of days falling within that phase.

Fig. 1
figure 1

Study design. The date of first COVID-19 diagnosis was defined as the index date, and the index period was April 1 through May 31, 2020. For each patient, the baseline phase included the 12 months before the index date, and the post-acute phase included the 12 months after the end of the acute phase of COVID-19 illness. aThe acute phase, defined as the 30-day period after the index date, was excluded from analysis

Full size image

Participants

Patients were eligible for inclusion if they had ≥ 1 ICD-10 diagnosis code for confirmed COVID-19 (U07.1) during the index period. Patients were required to have continuous enrollment (gaps of ≤ 45 days were permitted) in Optum CDM over the 12 months before and 13 months after COVID-19 diagnosis, to be aged ≥ 18 years on the index date, and to have ≥ 1 high risk condition per CDC definition as of October 14, 2021 [29]. Sentinel code lists [30] were used to define these conditions when available, and all codes were reviewed by sponsor personnel with medical expertise (FD, JCA, NB, MLNFV). Criteria for immunocompromised individuals were developed from a recent literature review [31]. Inclusion criteria specified having a diagnosis code (ICD-10-CM), procedure code (ICD-10-Procedure Coding System [ICD-10-PCS], Current Procedural Terminology [CPT®], Healthcare Common Procedures Coding System [HCPCS]), or National Drug Code (NDC) for ≥ 1 of the listed conditions within the 12 months before the index date or being aged ≥ 65 years at the index date. Patients were excluded if they were hospitalized for ≥ 5 consecutive days during the baseline phase; spent any time at a long-term care facility, skilled nursing facility, inpatient rehabilitation, or hospice during baseline or at index date; had an ICD-10 code for confirmed COVID-19 before the index period; or died during the acute phase of COVID-19.

Descriptive analysis

Outcomes of interest included ICD-10 diagnosis codes (other than confirmed COVID-19); medication use; outpatient visits and laboratory tests; ED visits; inpatient hospitalizations, including length of stay (LOS), intensive care unit (ICU) visits/LOS, ventilator use, and 30-day readmissions; and both standard and nonzero healthcare-associated costs. Standard costs were calculated based on all patients with a related visit or service, and nonzero costs were calculated based on all patients with a cost > 0 associated with that visit or service. The top 500 individual diagnosis codes were aggregated to their chapter and broader diagnostic categories and used for further analysis. Medications were analyzed via AnalySource® (Fayetteville, NY, USA) according to the Uniform System of Classification class. Diagnoses and medications were calculated as dichotomous occurrences during the 12 months before the index date (baseline phase) and the 12 months after the 30-day acute phase (post-acute phase); those with a prevalence of < 2% within the overall population during the baseline phase were excluded. Biologics were also excluded because the category primarily consisted of incompletely captured vaccine data. No adjustments were made for patients who died during the 12-month post-acute phase; all deaths that occurred during the post-acute phase were accounted for and reported. Continuous measures were totaled for each period. For each outcome, absolute and relative change from baseline to the post-acute phase was calculated using frequency counts.

Because older age is a risk factor for both increased COVID-19 severity and post-COVID conditions [21], data were separated into categories of patients aged 18 to 64 years or aged ≥ 65 years at the index date. To better understand the relationship between post-COVID conditions and acute COVID-19 severity, data were also stratified according to patient disposition during acute COVID-19 illness: not hospitalized, hospitalized without ICU admission, or admitted to the ICU. All variables were presented descriptively using mean ± SD or median (quartile 1 [Q1]; quartile 3 [Q3]) for continuous variables and frequencies and percentages for categorical variables. No statistical inference tests were conducted. Analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results

Patient population

The full cohort included 19,558 patients with a median (Q1; Q3) age of 66 (51; 74) years (Table 1). A slight majority of patients were female (55.4%), White (52.3%), enrolled in Medicare (54.7%), and aged ≥ 65 years (52.0%). When categorized by acute COVID-19 severity, 15,457 patients (79.0%) did not require hospitalization, 2916 patients (14.9%) were hospitalized without ICU admission, and 1185 patients (6.1%) were admitted to the ICU. Male patients, Black patients, and those aged ≥ 65 years were observed in higher proportions among the cohorts that were hospitalized (with or without ICU admission) for COVID-19.

Table 1 Baseline demographic characteristics
Full size table

The majority of patients (81.5%) had ≥ 2 high-risk conditions, and 7.7% of patients had ≥ 8 such conditions (Table 2). The most common high-risk conditions in the overall population were immunocompromised state (71.2%), hypertension (60.5%), and age ≥ 65 years (52.0%). Most conditions were more common among older patients and among those who were hospitalized (with or without ICU admission) for acute COVID-19.

Table 2 Baseline medical characteristics, including presence of high-risk conditionsa
Full size table

Most patients (82.7%) were diagnosed with COVID-19 in the outpatient setting (Table 2). Of those who were hospitalized at any time during the acute phase, 143 patients (0.7%) had hospital stays spanning > 30 days after the index date and therefore overlapping the acute and post-acute phases. Few patients (0.5% of the overall cohort) died during the post-acute phase. A full list of reasons for exclusion from the analysis is shown in Table S1.

Diagnoses

Between the baseline and post-acute phases, the frequency of individual ICD-10-CM diagnosis codes increased within several chapters (Fig. 2). The greatest percentage increase was observed in the ICD-10 chapter of “diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism” (+ 16.3%), followed by “diseases of the nervous system” (+ 11.1%), “external causes of morbidity and mortality” (+ 7.9%), and “mental and behavioral disorders” (+ 7.7%).

Fig. 2
figure 2

Percentage change from baseline to post-acute phase in ICD-10 diagnoses in the overall population (N = 19,558). Diagnosis codes (shown in parentheses) include those with a prevalence of ≥ 2% in the baseline population and with an increase from baseline to post-acute phases. Code ranges, raw values at baseline and post-acute phases, and diagnoses decreasing from baseline to post-acute phases are shown in Table 3. ICD-10, International Classification of Diseases, Tenth Revision

Full size image

Frequencies of other disorders increased more modestly or decreased (Table 3); the greatest percentage decrease was observed in the chapter of “diseases of the respiratory system” (–18.3%). Within this chapter, the decrease was driven specifically by a drop in frequency of the subchapters acute upper and lower respiratory infections (–56.6% and –69.3%, respectively), which outweighed increases in frequency of other respiratory conditions, such as “other respiratory diseases principally affecting the interstitium” (+ 59.8%; Table S2 and Fig. S1).

Table 3 Diagnoses during the baseline and post-acute phases in the overall population (N = 19,558)a
Full size table

Medication use

Prescription frequencies across several medication classes increased between baseline and the post-acute phase (Fig. 3). The greatest percentage increases were observed for vitamins (+ 47.5%), miscellaneous preparations (+ 38.0%), blood factors (+ 31.6%), and hemostatic modifiers (+ 29.9%). Among medication classes with increases of ≥ 10% in the overall population, increases in prescription frequency were observed across all ages and severities (Tables 4 and 5).

Fig. 3
figure 3

Percentage change from baseline to post-acute phase in prescription frequency in the overall population (N = 19,558). Medication classes shown are according to USC designations; included classes are those prescribed to ≥ 2% of the baseline population and with a percentage change from baseline to post-acute phase of ≥ 10%. Percentage changes from baseline to post-acute phase according to age group and disease severity are shown in Table 4, and raw values during baseline and post-acute phases for all medication classes regardless of percentage change are shown for the overall population in Table S3. USC, Uniform System of Classification

Full size image
Table 4 Medications with frequency Increases ≥ 10% from the baseline to the post-acute phasea by age group
Full size table
Table 5 Medications with frequency increases ≥ 10% from baseline to post-acute phasea by disposition during acute COVID-19
Full size table

Other prescription frequencies increased marginally or decreased (Table S3). The greatest overall decreases were within the classes of cough/cold/flu preparations (− 73.7%), antimalarials (–54.5%), and antivirals (–40.4%).

Medical care and hospitalizations

Increases in total and per-patient inpatient and outpatient visits and procedures were observed during the post-acute phase compared with baseline (Tables 6 and 7). The greatest percentage increases were observed for LOS (including total days [+ 165.6%] and mean days per patient [+ 132.9%]), inpatient lab tests (+ 105.1%), and all-cause readmissions within 30 days (+ 249.9%). For these 3 variables, increases from baseline to the post-acute phase were greatest among patients who were admitted to the ICU during acute COVID-19, but were substantial even among those who did not require hospitalization for acute COVID-19. The number of patients with ICU visits was relatively consistent from baseline to the post-acute phase in the overall population but increased by 43.2% among those who were admitted to the ICU during acute COVID-19 illness. Among all age and severity categories, the numbers of patients visiting the ED decreased. Regarding hospital discharge status, percentages of all discharges to another hospital department or facility (such as hospice care or a skilled nursing facility) were low during the baseline phase owing to exclusion criteria but substantially increased among both age groups (up to a 13,000% increase in intra-institution transfers among patients aged ≥ 65 years; Table S4).

Table 6 Healthcare resource use during baseline and post-acute phasesa in the overall population and by age
Full size table
Table 7 Healthcare resource use during baseline and post-acute phasesa stratified by disposition during acute COVID-19
Full size table

Healthcare costs

In the overall population, total medical costs (including prescription, inpatient, and outpatient costs) increased by 23.0% from the baseline to the post-acute phase (Table 8). Increased percentages were observed across all categories of age and acute COVID-19 severity but were greater among patients aged ≥ 65 years (+ 27.2%) versus younger patients (+ 16.7%) and were greater among patients admitted to the ICU during the acute phase of COVID-19 (+ 70.6%) versus those who were not hospitalized (+ 14.3%) or who were hospitalized without ICU admission (+ 27.9%; Tables 8 and 9).

Table 8 Medical costs during baseline and post-acute phasesa in the overall population and by ageb
Full size table
Table 9 Medical costs during baseline and post-acute phasesa stratified by disposition during acute COVID-19b
Full size table

Inpatient and outpatient costs also increased in the overall population and across all age and severity categories (Tables 8 and 9). Percentage increases in inpatient costs were driven primarily by cost increases among patients of any age who were admitted to the ICU during acute COVID-19 (+ 267.4%). Much of the increase in inpatient costs within this patient subpopulation was due to an 896% increase in 30-day all-cause readmissions (Table 7), resulting in a 1227.7% cost escalation during the post-acute phase. Outpatient costs increased more modestly and were greatest among older patients and those who were hospitalized (with or without ICU admission) during acute COVID-19. Overall outpatient cost increases were observed despite decreases associated with ED visits across all age and severity categories.

Discussion

In this retrospective analysis encompassing > 2 years of healthcare data from nearly 20,000 high-risk individuals diagnosed with COVID-19, resource use and costs were substantially higher in the year after acute COVID-19 illness compared with the previous year. Consistent with previous observations [15], increases were greatest among older individuals and those who required hospitalization for acute COVID-19; however, notable changes were observed even among younger patients and those who had less severe acute disease.

We found increases in incidence of blood-related, neurologic, and psychiatric disorders, all of which are consistent with previous reports on post-COVID conditions [32,33,34]. We observed a 16.3% increase in the ICD-10 diagnostic codes comprising “diseases of the blood and blood-forming organs” along with 31.6% and 29.9% increases in blood factor prescriptions and hemostatic modifier prescriptions, respectively. This is consistent with the potential for COVID-19 to cause persistent changes in the mechanisms underlying coagulation and hemostasis [33]. An 11.1% increase in “diseases of the nervous system” and a 7.7% increase in “mental and behavioral disorders” were also observed alongside increased prescribing of neurologic/neuromuscular disorder treatments and psychotherapeutic drugs (+ 12.7% and + 11.1%, respectively). The magnitude of increases in mental health–related prescriptions were relatively similar among patients who were not hospitalized during acute COVID-19 illness compared with the overall cohort, supporting previous reports that identified long-term impairments in mood, anxiety, and cognitive functioning that were unrelated to COVID-19 severity or hospitalization [32, 34, 35]. Our results are further corroborated by a longitudinal study of UK Biobank participants, in which cognitive declines observed > 3 months after COVID-19 diagnosis were significant even among nonhospitalized cases and were associated with structural changes in the brain [35]. In contrast with recent CDC data [10], we did not observe an overall increase in respiratory conditions because acute respiratory infections decreased. CDC defined respiratory conditions only as acute pulmonary embolism, asthma, or respiratory symptoms.

The percentages of outpatient and inpatient medical service use were substantially higher during the post-acute phase compared with baseline, including a 165.6% increase in total days spent in the hospital and a 249.9% increase in 30-day all-cause readmissions. This increase in healthcare utilization was directly reflected in observed medical costs. Total costs increased by 23.0% in the overall population; increases were greater among patients aged ≥ 65 years compared with the younger population and ranged from 14.3% among patients who were not hospitalized for acute COVID-19 to 70.6% among those admitted to the ICU.

Mean overall per-patient medical cost during the year after the acute COVID-19 phase was approximately $27,000 (whereas the baseline cost was approximately $22,000). Although there are no direct comparisons that can be made with previous studies, cost estimates of post-COVID conditions based on similar diagnoses (eg, myalgic encephalomyelitis/chronic fatigue syndrome) were nearly $9000 per person per year [36]. The higher values observed in our study may reflect a greater impact of COVID-19 than similar conditions but could also be a result of our high-risk patient population. Notably, studies comparing post-COVID conditions with long-term sequelae after seasonal influenza found a greater symptom, diagnosis, and healthcare resource burden with COVID [12, 24].

Substantial decreases were observed in the percentages of acute upper and lower respiratory infections, prescriptions for cough/cold/flu preparations, and ED visits. These data are consistent with early reports in the United States that show a sharp decline in both influenza rates and ED visits after the onset of the pandemic, including ED visits specifically for non-COVID upper respiratory infections [37,38,39]. This decline was likely due to a combination of reduced influenza circulation because of COVID-19 mitigation strategies and the reluctance of patients to seek treatment because of the perceived risk of contracting COVID-19 in a healthcare setting [37, 39].

A main strength of this study was that all patients served as their own control, which inherently adjusts for potential confounders, including patient demographics and stable characteristics, such as healthcare-seeking behavior. Our study had some limitations, including that the population was limited to commercially insured individuals who were diagnosed early in the pandemic and who survived the acute phase of COVID-19. There was also potential for incomplete data capture (due to nonbillable diagnoses) and surveillance bias by which those who contracted COVID-19 were under higher medical scrutiny following their diagnosis. It is important to note that the study was conducted during a time of low preexisting immunity (ie, before vaccination and previous infection) and before the emergence of and SARS-CoV-2 variants of concern. Such factors could limit the generalizability of the findings to the current landscape. Additionally, baseline assessments were performed during the pre-pandemic period, whereas post-acute COVID-19 assessments were performed during a public health emergency that altered healthcare practices, access, and considerations; these changes may have affected clinical burden and healthcare costs during the post-acute COVID-19 period. Because this study included only high-risk individuals, a companion report describes results from a separate cohort of patients without high risk of developing severe COVID-19.

Conclusion

The health and economic burden of post-COVID conditions among high-risk US adults is substantial. Although the greatest impacts were observed among patients aged ≥ 65 years and those who were admitted to the ICU for acute COVID-19, increases in most outcomes were apparent even among younger individuals and those who did not require COVID-19 hospitalization. These results improve our understanding of post-COVID conditions and associated costs, as well as support hypothesis generation for future work characterizing the COVID-19 impact on individuals and society.