key: cord-0720632-xj03r432
authors: Chevinsky, Jennifer R; Tao, Guoyu; Lavery, Amy M; Kukielka, Esther A; Click, Eleanor S; Malec, Donald; Kompaniyets, Lyudmyla; Bruce, Beau B; Yusuf, Hussain; Goodman, Alyson B; Dixon, Meredith G; Nakao, Jolene H; Datta, S Deblina; Mac Kenzie, William R; Kadri, Sameer; Saydah, Sharon; Giovanni, Jennifer E; Gundlapalli, Adi V
title: Late conditions diagnosed 1–4 months following an initial COVID-19 encounter: a matched cohort study using inpatient and outpatient administrative data — United States, March 1–June 30, 2020
date: 2021-04-28
journal: Clin Infect Dis
DOI: 10.1093/cid/ciab338
sha: 919c86cf55eb49a84fef75d4e2259e6be687e09a
doc_id: 720632
cord_uid: xj03r432

BACKGROUND: Late sequelae of COVID-19 have been reported; however, few studies have investigated the time-course or incidence of late new COVID-19-related health conditions (post-COVID conditions) after COVID-19 diagnosis. Studies distinguishing post-COVID conditions from late conditions caused by other etiologies are lacking. Using data from a large administrative all-payer database, we assessed the type, association, and timing of post-COVID conditions following COVID-19 diagnosis. METHODS: Using the Premier Healthcare Database Special COVID-19 Release (PHD-SR) (release date, October 20, 2020) data, during March–June 2020, 27,589 inpatients and 46,857 outpatients diagnosed with COVID-19 (case-patients) were 1:1 matched with patients without COVID-19 through the 4-month follow-up period (control-patients) by using propensity score matching. In this matched-cohort study, adjusted odds ratios were calculated to assess for late conditions that were more common in case-patients compared with control-patients. Incidence proportion was calculated for conditions that were more common in case-patients than control-patients during 31–120 days following a COVID-19 encounter. RESULTS: During 31–120 days after an initial COVID-19 inpatient hospitalization, 7.0% of adults experienced at least one of five post-COVID conditions. Among adult outpatients with COVID-19, 7.7% experienced at least one of ten post-COVID conditions. During 31–60 days after an initial outpatient encounter, adults with COVID-19 were 2.8 times as likely to experience acute pulmonary embolism as outpatient control-patients and were also more likely to experience a range of conditions affecting multiple body systems (e.g. nonspecific chest pain, fatigue, headache, and respiratory, nervous, circulatory, and gastrointestinal system symptoms) than outpatient control-patients. Children with COVID-19 were not more likely to experience late conditions than children without COVID-19. CONCLUSIONS: These findings add to the evidence of late health conditions possibly related to COVID-19 in adults following COVID-19 diagnosis and can inform health care practice and resource planning for follow-up COVID-19 care.

M a n u s c r i p t 6 Late sequelae of coronavirus disease 2019 (COVID-19), the disease caused by SARS-CoV-2, have been reported (1) . Several large COVID-19 survivor advocacy groups are raising awareness of symptoms persisting after initial illness including shortness of breath, chest tightness, and fatigue (2) ; however, few studies have investigated the time-course or incidence of late new COVID-19-related health conditions (post-COVID conditions) after SARS-CoV-2 infection (3) (4) (5) (6) . Studies distinguishing late conditions associated with COVID-19 from conditions caused by other etiologies are lacking (7) .

There are multiple challenges with assessing post-COVID conditions. There is a need for following patients diagnosed with COVID-19 over time and an additional need to compare them with controls without COVID-19 to improve our understanding of conditions that may manifest after acute COVID-19 disease. Based on these needs, one challenge is that assessing post-COVID conditions requires longitudinal data. A second challenge is accessing a control population to be able to assess which conditions are associated with COVID-19 disease, rather than other factors such as age or care acuity. Initial prospective and observational studies have been uncontrolled (1) (2) (3) (4) (5) (6) . Large administrative databases with longitudinal data can be analyzed to identify a comparable control population using propensity score matching to provide information on late conditions possibly related to COVID-19. Using data from a large administrative all-payer database, we assessed the type, association, and timing of post-COVID conditions (1-4 months) following a COVID-19 diagnosis in inpatient and outpatient facility settings in a large group of patients. March 1-June 30, 2020. Index encounter date was defined as the hospital discharge date for an inpatient encounter or encounter date for an outpatient encounter. The discharge date was used as a reference point for inpatient encounters for two primary reasons: (1) discharge date is commonly used as a reference point to assess for complications after a hospitalization and therefore could be a clinically useful point of reference and (2) this approach could limit the inclusion of acute symptoms and conditions in the findings by establishing a baseline for all inpatients after the hospitalization. Because the point of reference is different for inpatients and outpatients, the timeline for inpatient case-patients may be compared to the timeline for inpatient control-patients and the timeline for outpatient case-patients may be compared to the timeline for outpatient control-patients, however the timeline for inpatients and outpatients may not align and therefore should not be directly compared.

A c c e p t e d M a n u s c r i p t 8 Clinical diagnoses established during January 2019 to the index encounter date in PHD-SR provided historical data on underlying conditions. Case-patients and control-patients were identified by using propensity score nearest-neighbor matching (10-12), a statistical technique for maximizing efficiency and for better isolating the effect of COVID-19 on the patient experiencing new conditions from the effect of other included variables. The match was based on propensity scores computed from patient demographics (age, sex, race, ethnicity, insurance status), clinical factors (number of previous inpatient encounters and conditions diagnosed before and at the index encounter), facility characteristics (urbanicity, region), and month of the index encounter. Inpatients and outpatients were matched separately. Outpatient encounters included the following facility settings: same day surgery, emergency, observation, diagnostic testing, recurring visits for services including dialysis, chemotherapy infusion and radiation, presurgical testing, and clinic. Inpatient encounters included exclusively a hospital facility setting. All other settings were excluded. Prior to matching, we excluded patients without at least one encounter preceding their index encounter in PHD-SR, who died during their index encounter, or who were pregnant at their index encounter. Potential control-patients who were diagnosed with COVID-19 during the four months after their index encounter were also excluded prior to matching.

ICD-10-CM codes recorded during encounters were classified to Clinical Classification Software Refined (CCSR) categories (13), which aggregates ICD-10-CM codes into clinically meaningful categories to form disease groupings. Diagnoses from encounters before (using the historical data from January 2019 to the index encounter date) and during the index encounter were classified as underlying or acute COVID-19 conditions. New persistent conditions (those newly starting during the index encounter and persisting after the index encounter) and exacerbations of underlying conditions (those starting prior to the index A c c e p t e d M a n u s c r i p t 9 encounter and worsening during or after the index encounter) were not assessed in this analysis because of challenges differentiating underlying conditions, acute conditions, and exacerbations in inpatient administrative data. Late conditions were defined as conditions not previously recorded as underlying or acute COVID-19 conditions during January 2019 through the index encounter date that occurred during 31-120 days (1 to 4 months) after the index encounter. Five CCSR categories were excluded from the late conditions analysis: pregnancy, perinatal, congenital malformations, external causes of morbidity, and factors influencing contact with health services (e.g. encounter for administrative purposes). Late conditions were identified using CCSR categories based on timing of occurrence after the index encounter date: 31-60 days, 61-90 days, and 91-120 days. The timeline was established using a variable that determined the days between each visit, allowing for a continuous timeline. Adjusted (for the matched variables with pairs as strata) odds ratios (aOR) and 95% confidence intervals (CI) were calculated using a conditional logit model for new conditions in case-patients compared with control-patients to identify post-COVID conditions that could be unique to patients with COVID-19, rather than searching for preestablished outcomes which could introduce additional bias. Among these statistically significant post-COVID conditions, the most common were selected for adult case-patients based on the highest incidence proportion to identify conditions that could be the most frequent new health conditions experienced 31-120 days after COVID-19 diagnosis.

A sensitivity analysis was conducted that restricted the control cohort to adult controlpatients with a respiratory CCSR category during the index encounter to examine if results were consistent with the larger study's findings. The larger analysis was not restricted to control-patients with a respiratory CCSR category during the index encounter because many respiratory illnesses, like influenza, have been less common during the pandemic (14) and A c c e p t e d M a n u s c r i p t 10 healthcare seeking patterns during the pandemic have been dissimilar to healthcare seeking patterns in previous years (15), potentially introducing bias when matching to patients with respiratory viruses in previous years. SAS (version 9.4; SAS Institute) was used for analyses. This activity was reviewed by CDC and was conducted consistent with applicable federal law and CDC policy, and was determined to be exempt from review from the Institutional Review Board (See e.g., 45 C. (Table 1) . Among the 27,589 inpatient match pairs, 305 match-pairs were in children (aged <18 years) and 27,284 match-pairs were in adults (aged ≥18 years). Among the 46,857 outpatient match pairs, 2,368 match-pairs were in children and 44,489 match-pairs were in adults.

For adults, the incidence of post-COVID conditions was predominantly in the 31-60-day range rather than in the 61-90-or 91-120-day ranges. Adults with an initial inpatient COVID-19 encounter were significantly more likely to experience the following diagnoses in the 31-60 days after discharge compared to hospitalized adults without COVID-19: non-specific chest pain (aOR = 1.3; 95% CI = 1.0 -1.7), respiratory system symptoms (aOR = 1.4; 95% CI = 1.1 -1.8), circulatory system symptoms (aOR = 1.3; 95% CI = 1.1 -1.7), and nervous system symptoms (aOR = 1.3; 95% CI = 1.1 -1.6) ( Table 2 ). Among 27,284 inpatient adult case-patients, 7.0% A c c e p t e d M a n u s c r i p t 11 newly experienced one or more of five identified most common post-COVID conditions during 31-120 days: respiratory symptoms (e.g. shortness of breath), nervous system symptoms (e.g. altered mental status), urinary tract infections, circulatory symptoms (e.g. tachycardia), and nonspecific chest pain (Table 3 ). Outpatient adult case-patients were more likely to experience a range of diagnoses corresponding to multiple body systems compared to outpatient adult control-patients ( Table 2 ). During 31-60 days, adults with an outpatient index encounter for COVID-19 were more likely than outpatient control-patients to experience acute pulmonary embolism (aOR = 2.8; 95% CI = 1.3 -6.0). During 31-120 days, 7.7% of 44,489 adults with an initial outpatient encounter for COVID-19 newly experienced one or more of ten identified post-COVID conditions: respiratory symptoms (e.g shortness of breath), abdominal pain and other digestive/abdominal symptoms (e.g. diarrhea), nonspecific chest pain, nervous system symptoms (e.g. altered mental status), headache (including migraine), circulatory symptoms (e.g. tachycardia), fluid and electrolyte disorders (e.g. hypokalemia), malaise and fatigue, nausea and vomiting, and urinary tract infections (Table 3) . Among 44,489 adult case-patients with an outpatient index encounter, 1,222 (2.8%) were later hospitalized during 31-120 days with the most common diagnoses including pneumonia and fluid and electrolyte disorders (Table 4) .

Children with COVID-19 were not more likely to experience new diagnoses than children without COVID-19. The results of the sensitivity analysis that restricted the control cohort to adult control-patients with a respiratory CCSR category during the index encounter were consistent with the study findings with identification of new diagnoses in multiple body systems for adult case-patients. 

Among 27,284 inpatient adults and 44,489 outpatient adults who had a diagnosis of COVID-19, 7.0% and 7.7%, respectively, were newly diagnosed with one or more identified post-COVID conditions (31-120 days following their initial COVID-19 encounter as defined above) in a large administrative all-payer database. Children with COVID-19 were not more likely to experience post-COVID conditions than children without COVID-19. Because this study compared COVID-19 case-patients with control-patients who did not have COVID-19, it is probable that the identified post-COVID conditions in adults are related to COVID-19 rather than to other factors such as age or care setting. Furthermore, the findings of a sensitivity analysis suggest excess risk for adult patients with COVID-19 for experiencing conditions in multiple body systems, compared to adults with other respiratory diseases.

Other researchers have found evidence of continued COVID-19 illness or of a post-acute COVID-19 syndrome (2) (3) (4) (5) , with conditions that affect multiple body systems (16-18). A proposed population-based framework defined acute SARS-CoV-2 infection during days 0-14 after symptom onset, post-acute hyperinflammatory illness during days 14-30 after symptom onset, and late sequelae at >30 days from symptom onset (6) . This study supports the existence of post-COVID conditions that might start after 30 days among some adults diagnosed with COVID-19.

Hypercoagulability and thromboembolic disorders have been reported following COVID-19

(1). In this study, inpatient adult case-patients with COVID-19 were not more likely to experience acute pulmonary embolism than control-patients after 31 days. During 31-60 days after the index encounter, non-pregnant adult patients with outpatient COVID-19 encounters were 2.8 times as likely as outpatient control-patients to experience acute A c c e p t e d M a n u s c r i p t 13 pulmonary embolism; acute pulmonary embolism continued to be more than two times as likely during 61-90 days. Most adults with outpatient COVID-19 encounters did not progress to moderate/severe acute disease (requiring subsequent inpatient care), although they were more likely than control-patients to experience a range of additional conditions involving multiple body systems (e.g. nonspecific chest pain, fatigue, headache, and respiratory, nervous, circulatory, and gastrointestinal system symptoms). Patients with an index inpatient encounter experienced a more limited list of new conditions compared to control-patients. This could be due to post-COVID conditions emerging during the initial hospitalization or due to similarities in new conditions that may have been experienced by both inpatient case-patients and inpatient control-patients following hospital discharge.

Hospitalized adults might experience other persistent and new conditions (4) after hospital discharge that might last for months (3) . Of note, both patients with an index inpatient or outpatient COVID-19 encounter were more likely to be diagnosed with nonspecific chest pain and neurological, circulatory, and respiratory symptoms in the post-acute period, though the timelines for experiencing these new symptoms may not align between inpatients and outpatients because of the different index encounter date reference points. * Propensity score matching is a statistical technique used to achieve an even distribution of patient characteristics among case and control groups in order to compare two groups in an observed (non-randomized) population. In this analysis, propensity score matching was used to better isolate * New condition is defined as any CCSR category recorded 1 to 120 days after the index encounter that was not recorded in any preceding health care encounter during January 2019 through the index encounter. All case-patients and control-patients included in this analysis had at least one encounter preceding their index encounter recorded in the large administrative all-payer database. Adjusted odds ratios calculated from the presence of a new condition in 20 or fewer case-patients are not presented and instead are denoted by a symbol (--). † Timeline was established using a variable that determined the days between each visit, allowing for a continuous timeline. § Among case-patients, an index encounter was defined as the initial encounter with a COVID-19 diagnosis during March-June 2020. Among control-patients, an index encounter was defined as the patient's propensity matched encounter that was used for comparison, during March-June 2020. For an inpatient encounter, the hospital discharge date was assigned as time 0. For an outpatient encounter, the encounter date was assigned as time 0. ¶ Adjusted odds ratios and 95% confidence intervals were calculated using a conditional logit model for new conditions in case-patients compared with control-patients. The odds ratios were adjusted for patient demographics (age, sex, race, ethnicity, insurance status), clinical factors (number of previous inpatient encounters and diagnoses before and at the index encounter), facility characteristics (urbanicity, region), and month of the index encounter.

Conditions that are statistically significantly more common in case-patients than control-patients are bolded. ** Post-COVID condition is defined as any CCSR category newly recorded 31-120 days after the index encounter that was not recorded in any preceding health care encounter from January 2019 through the index encounter.

A c c e p t e d M a n u s c r i p t 30 Table 3 : Most common post-COVID conditions* 31-120 days after an initial COVID-19 encounter by incidence proportion in adult (aged ≥18 years) case-patients, stratified by index encounter facility setting, from a large administrative all-payer database -United States, March 1-June 30, 2020

Incidence proportion (31-120 days) M a n u s c r i p t 31 preceding encounter. N, therefore, reflects the eligible adult population for a given post-COVID condition. Casepatients can be represented in more than one row if they experienced more than one of the post-COVID conditions. § The top five respiratory symptoms ICD-10-CM codes, by frequency, include shortness of breath, cough, hypoxemia, dyspnea, and pleurodynia. ¶ The top five nervous system symptoms ICD-10-CM codes, by frequency, include headache, altered mental status, disorientation, abnormalities of gait and mobility, and unspecified difficulty in walking. 

Long-term Health Consequences of COVID-19

What does COVID-19 recovery actually look like? An analysis of the prolonged COVID-19 symptoms survey by Patient-Led Research Team. Patient Led Research for COVID-19

Gemelli Against COVID-19 Post-Acute Care Study Group. Persistent symptoms in patients after acute COVID-19

Postdischarge symptoms and rehabilitation needs in survivors of COVID-19 infection: A cross-sectional evaluation

Symptom duration and risk factors for delayed return to usual health among outpatients with COVID-19 in a multistate health care systems network-United States

A proposed framework and timeline of the spectrum of disease due to SARS-CoV-2 infection: illness beyond acute infection and public health implications

Facing up to Long COVID

Abbreviations: COVID-19 = coronavirus disease

CCSR = Clinical Classification Software Refined Categories

COVID condition is defined as any CCSR category newly recorded 31-120 days after the index encounter that was not recorded in any preceding health care encounter during

For an inpatient encounter, the hospital discharge date was assigned as time 0. For an outpatient encounter, the encounter date was assigned as time 0. The incidence proportion was calculated for the conditions that were statistically significantly more common among case-patients than among matched control-patients without COVID-19 during 31-120 days

represents the number of adult case-patients of the total 27,284 inpatient or 44,489 outpatient adult case-patients who were not previously diagnosed with the given condition in their index encounter or in a

Abbreviation: COVID-19 = coronavirus disease 2019.* Most common new conditions following an inpatient or outpatient index encounter with a COVID-19 diagnosis were determined by selecting up to 10 of the conditions with the highest incidence proportion with a statistically significant adjusted odds ratio (compared with control-patients) in each time-period. † Among case-patients, an index encounter was defined as the initial encounter with a COVID-19 diagnosis during March-June 2020. Among control-patients, an index encounter was defined as the patient's propensity matched encounter that was used for comparison, during March-June 2020. For an inpatient encounter, the hospital discharge date was assigned as time 0. For an outpatient encounter, the encounter date was assigned as time 0. § Timeline was established using a variable that determined the days between each visit, allowing for a continuous timeline. ¶ The top five respiratory symptoms ICD-10-CM codes, by frequency, include shortness of breath, cough, hypoxemia, dyspnea, and pleurodynia. ***** The top five other nervous system disorders ICD-10-CM codes, by frequency, include metabolic encephalopathy, type 2 diabetes with diabetic neuropathy, toxic encephalopathy, unspecified encephalopathy, and other encephalopathy.